Subtype diversification and synaptic specificity of stem cell ...

Subtype diversification and synaptic specificity of stem cell-derived spinal inhibitory interneurons

Phuong Thi Hoang

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy under the Executive Committee of the Graduate School of Arts and Sciences

COLUMBIA UNIVERSITY

2017

© 2017

Phuong Thi Hoang

All Rights Reserved

Abstract

Subtype diversification and synaptic specificity of stem cell-derived spinal inhibitory interneurons

Phuong Thi Hoang

During nervous system development, thousands of distinct neuronal cell types are generated and

assembled into highly precise circuits. The proper wiring of these circuits requires that developing

neurons recognize their appropriate synaptic partners. Analysis of a vertebrate spinal circuit that controls

motor behavior reveals distinct synaptic connections of two types of inhibitory interneurons, a ventral V1

class that synapses with motor neurons and a dorsal dI4 class that selectively synapses with

proprioceptive sensory neuron terminals that are located on or in close proximity to motor neurons. What

are the molecular and cellular programs that instruct this remarkable synaptic specificity? Are only

subsets of these interneurons capable of integrating into this circuit, or do all neurons within the same

class behave similarly?

The ability to answer such questions, however, is hampered both by the complexity of the spinal

cord, where many different neuronal cell types can be found synapsing in the same area; as well as by

the challenge of obtaining enough neurons of a particular subtype for analysis. Meanwhile, pluripotent

stem cells have emerged as powerful tools for studying neural development, particularly because they

can be differentiated to produce large amounts of diverse neuronal populations. Mouse embryonic stem

cell-derived neurons can thus be used in a simplified in vitro system to study the development of specific

neuronal cell types as well the interactions between defined cell types in a controlled environment. Using

stem cell-derived neurons, I investigated how the V1 and dI4 cardinal spinal classes differentiate into

molecularly distinct subtypes and acquire cell type-specific functional properties, including synaptic

connectivity.

In Chapter Two, I describe the production of lineage-based reporter stem cell lines and optimized

differentiation protocols for generating V1 and dI4 INs from mouse embryonic stem cells, including

confirming that they have molecular and functional characteristics of their in vivo counterparts.

In Chapter Three, I show that a well-known V1 interneuron subtype, the Renshaw cell, which

mediates recurrent inhibition of motor neurons, can be efficiently generated from stem cell differentiation.

Importantly, manipulation of the Notch signaling pathway in V1 progenitors impinges on V1 subtype

differentiation and greatly enhances the generation of Renshaw cells. I further show that sustained

retinoic acid signaling is critical for the specific development of the Renshaw cell subtype, suggesting that

interneuron progenitor domain diversification may also be regulated by spatially-restricted cues during

embryonic development.

In Chapter Four, using a series of transplantation, rabies virus-based transsynaptic tracing, and

optogenetics combined with whole-cell patch-clamp recording approaches, I demonstrate that stem cell-

derived Renshaw cells exhibit significant differences in physiology and connectivity compared to other V1

subpopulations, suggesting that synaptic specificity of the Renshaw cell-motor neuron circuit can be

modeled and studied in a simplified in vitro co-culture preparation.

Finally, in Chapter Five, I describe ongoing investigations into molecular mechanisms of dI4

interneuron subtype diversification, as well as approaches to studying their synaptic specificity with

proprioceptive sensory neurons.

Overall, my results suggest that our stem cell-cell based system is well-positioned to probe the

functional diversity of molecularly-defined cell types. This work represents a novel use of embryonic stem

cell-derived neurons for studying inhibitory spinal circuit assembly and will contribute to further

understanding of neural circuit formation and function during normal development and potentially in

diseased states.

i

Table of Contents List of Figures ......................................................................................................................................... iii Acknowledgments ................................................................................................................................... v CHAPTER 1: Introduction ....................................................................................................................... 1

A. Mechanisms of neuronal diversity ................................................................................................. 5 1. Regionalization of the nervous system 2. Temporal identity specification 3. Intradomain diversification 4. Notch signaling in neuronal diversification

B. Synaptic specificity in the developing nervous system ............................................................... 15 1. Models of synaptic specificity 2. Molecular cues guiding synaptic specificity 3. In vitro recapitulation of synaptic specificity

C. Spinal cord neurogenesis ........................................................................................................... 22 1. Rostrocaudal spinal cord patterning 2. Dorsoventral spinal cord patterning 3. TGFβ signaling pathway 4. Development of the dorsal spinal cord 5. Sonic hedgehog signaling 6. Ventral spinal cord patterning 7. Motor neuron subtype diversity

D. Monosynaptic stretch reflex circuit .............................................................................................. 35 1. Proprioceptive sensory neurons 2. GABApre interneuron development 3. V1 interneuron subtype diversity 4. Renshaw cell development and function

E. Approach ..................................................................................................................................... 51 F. Figures ........................................................................................................................................ 54

CHAPTER 2: Directed differentiation of spinal inhibitory interneurons from stem cells ............... 63

A. Introduction ................................................................................................................................. 63 B. Results ........................................................................................................................................ 64

1. Derivation of V1 and dI4 lineage reporter stem cell lines 2. Optimized differentiation of V1 interneurons 3. Immunohistochemical characterization of ES-V1 INs 4. Optimized differentiation of dI4 interneurons 5. Immunohistochemical characterization of ES-dI4 INs 6. Effects of ActivinA on dorsal spinal patterning 7. RNA-seq gene expression profiling 8. Transplant of ESC-derived interneurons into the developing neural tube

C. Discussion ................................................................................................................................... 77 D. Figures ........................................................................................................................................ 85

CHAPTER 3: Renshaw cell subtype specification .............................................................................. 99

A. Introduction ................................................................................................................................. 99 B. Results ...................................................................................................................................... 101

1. Molecular heterogeneity of ESC-derived V1 interneurons 2. Characterization of ESC-derived Renshaw cells 3. Inhibition of Notch signaling enhances Renshaw cell differentiation

ii

4. Sustained retinoid signaling is required for specific generation of Renshaw cells 5. Confluence of Notch and RA signaling to specify Renshaw cells…...

C. Discussion ................................................................................................................................. 111 D. Figures ...................................................................................................................................... 119

CHAPTER 4: Synaptic specificity of ESC-derived Renshaw cells .................................................. 131

A. Introduction ............................................................................................................................... 131 B. Results ...................................................................................................................................... 132

1. Co-culture of stem cell-derived spinal interneurons and motor neurons 2. Monosynaptic rabies virus tracing reveals subtype-specific inputs onto MNs 3. Differential VAChT contacts on V1 interneuron subtypes 4. Physiological signature of ESC-derived Renshaw cells 5. Optogenetic approach to studying MN-RC synaptic specificity in vitro


CHAPTER 5: Specification of dI4 interneuron subtypes: in search of the GABApre .................... 154

A. Introduction ............................................................................................................................... 154 B. Results ...................................................................................................................................... 155

1. Molecular differentiation of stem cell-derived dI4 IN subtypes 2. Synaptic connectivity of ESC-derived dI4 INs in vitro 3. Programming proprioceptive sensory neurons from stem cells


CHAPTER 6: General discussion and future directions .................................................................. 175

A. Summary ................................................................................................................................... 175 B. In vitro modeling of spinal interneuron subtype specification .................................................... 177 C. Cell-intrinsic programs directing interneuron subtype-specific synaptogenesis ........................ 185 D. Subtype-specific synaptic connectivity of spinal inhibitory interneurons ................................... 189 E. Implications for studying and treating neurological disease ...................................................... 191 F. Conclusion ................................................................................................................................ 194 G. Figures ...................................................................................................................................... 195

Experimental procedures .................................................................................................................... 198 References ........................................................................................................................................... 210

iii

List of figures Introduction Fig 1.1 Distinct origins of inhibitory inputs modulating the monosynaptic stretch reflex .......................... 54 Fig 1.2 Morphogen signaling during neural tube development ................................................................ 55 Fig 1.3 Expression of RA-synthesizing enzyme Raldh2 in developing vertebrate spinal cord ................ 56 Fig 1.4 Dorsoventral patterning of the developing spinal cord ................................................................. 57 Fig 1.5 Hox expression patterns in the spinal cord underlie MN subtype diversity .................................. 58 Fig 1.6 Molecular diversity of dorsal inhibitory interneurons .................................................................... 59 Fig 1.7 V1 interneuron subtype diversity ................................................................................................. 60 Fig 1.8 Renshaw cell neurogenesis ......................................................................................................... 61 Fig 1.9 Directed differentiation of spinal motor neurons from ESCs ........................................................ 62

Chapter 2 Fig 2.1 Derivation of lineage reporter ESC lines for V1 and dI4 interneuron differentiations .................. 85 Fig 2.2 Directed differentiation of V1 inhibitory spinal interneurons from mouse ESCs .......................... 86 Fig 2.3 Molecular development of V1 interneurons in vivo and in vitro ................................................... 88 Fig 2.4 ESC-derived V1 interneurons recapitulate in vivo molecular developmental programs .............. 89 Fig 2.5 Directed differentiation of ESCs to dI4 spinal interneurons ......................................................... 91 Fig 2.6 ActivinA induces Ptf1a expression in differentiating ESCs .......................................................... 92 Fig 2.7 Molecular development of in vitro-generated dI4 interneurons .................................................... 93 Fig 2.8 Comparison of different TGFß family members on dorsal spinal patterning ............................... 94 Fig 2.9 ActivinA treatment generates distinct cell types, including glia .................................................... 95 Fig 2.10 RNA-seq gene expression profiling of ESC-derived spinal interneurons .................................. 96 Fig 2.11 Distinct migration and axonal projections of transplanted ES-V1 and dI4 interneurons ............ 98 Chapter 3 Fig 3.1 Enrichment of V1 interneuron subtype-specific transcription factors ......................................... 119 Fig 3.2 Subtype diversity of ESC-derived V1 interneurons .................................................................... 120 Fig 3.3 Calbindin-expressing V1 interneurons acquire Renshaw cell properties ................................... 122 Fig 3.4 BrdU birthdating of ESC-derived V1 interneurons ..................................................................... 123 Fig 3.5 Notch inhibition promotes the formation of Calbindin-expressing V1 interneurons ................... 124 Fig 3.6 RNA-seq expression profiling of DAPT-treated V1 interneurons ............................................... 125 Fig 3.7 DAPT treatment upregulates MafA, MafB and OC2 expression ................................................ 127 Fig 3.8 Sustained retinoid signaling is required for Renshaw cell specification ..................................... 128 Fig 3.9 Motor neuron-V1 interneuron co-cultures .................................................................................. 129 Fig 3.10 Raldh2-expressing motor neurons rescue Renshaw cell loss ................................................. 130 Chapter 4 Fig 4.1 Differential interactions of ESC-derived interneurons with motor neurons ................................ 144 Fig 4.2 Monosynaptic rabies virus tracing of V1 interneuron-motor neuron connectivity ...................... 145 Fig 4.3 RABV tracing reveals V1 interneuron subtype-specific connectivity with motor neurons .......... 146 Fig 4.4 Differential VAChT-immunoreactive inputs on V1 and dI4 inhibitory interneurons in vitro ........ 147 Fig 4.5 Stem cell-derived Renshaw cells preferentially receive VAChT+ cholinergic inputs ................. 148 Fig 4.6 ESC-derived Renshaw cells exhibit distinctive passive membrane properties .......................... 150 Fig 4.7 Active membrane properties of ESC-derived Renshaw cells .................................................... 151 Fig 4.8 Selective motor neuron cholinergic inputs onto ESC-derived Renshaw cells ........................... 152 Fig 4.9 Monosynaptic motor neuron connections onto Renshaw cells in vitro ...................................... 153

Chapter 5 Fig 5.1 Ascl1-dependent and independent dI4 interneuron subpopulations .......................................... 169 Fig 5.2 Molecular and spatially distinct subsets of dI4 interneurons in vivo and in vitro ........................ 170

iv

Fig 5.3 Transplanted dI4 interneuron subsets migrate into distinct dorsal horn laminae ....................... 171 Fig 5.4 Monosynaptic RABV tracing reveals premotor connections of dI4 interneurons ....................... 172 Fig 5.5 Transcriptional programming of proprioceptive sensory neurons from ESCs ........................... 173 Fig 5.6 Morphology and molecular maturation of induced proprioceptive sensory neurons .................. 174 General Discussion Fig 6.1 Model for Notch and retinoid signaling regulation of Renshaw cell specific development ......... 195 Fig 6.2 Renshaw cell distribution along the rostrocaudal axis of e12.5 mouse spinal cord ................... 196 Fig 6.3 GCaMP6-expressing V1 interneurons for recording subtype-specific activity ........................... 197

v

Acknowledgments I would like to thank my PhD advisor, Hynek Wichterle, for his encouragement and guidance on

my project, and for instilling the importance of rigor, creativity, and enthusiasm in all scientific endeavors.

I would also like to thank the members of my thesis committee: Wesley Grueber, George Mentis,

and Thomas Jessell, who were all very generous with their time, feedback and support. I want to also

acknowledge Carol Mason, who served on my qualifying exam committee, and Gordon Fishell, for being

my external examiner.

Thank you to everyone in the Wichterle lab, past and present. Stephane Nedelec, Chris Tan, and

Michael Closser were especially helpful to me during my time as a rotation student and in my PhD.

I also want to thank the people who made this project possible: Joshua Chalif and George Mentis

for the electrophysiology collaboration; Jay Bikoff and Julia Kaltschmidt for helping to establish the ESC

reporter lines and providing antibodies; Thomas Reardon for generous provision of rabies virus; Joriene

de Nooij for sensory neuron programming collaboration; Lora Sweeney for discussion of retinoic acid

signaling and limb-specific interneuron development; Susan Brenner-Morton for providing antibodies and

plasmids; James Caceido and John Smerdon for primary astrocyte cultures; and Mercedes Fissore-

O’Leary, an undergraduate student who helped with the retinoic acid studies.

This work would not have been possible without administrative and financial support from the

Columbia Neurobiology and Behavior PhD Program, Columbia MSTP, and NIH NINDS.

Lastly, I would like to thank my friends and family for their support over the years.

1

Chapter 1: Introduction

The ability of the brain to accurately encode information and produce meaningful behavioral

output relies on the precise organization of neurons into complex neural circuits. During nervous system

development, many thousands of different neuronal cell types are generated and these acquire a

multitude of specialized characteristics, including gene expression profiles, cellular morphology, and

physiological properties. Once specified, developing neurons must be able to migrate correctly from their

site of origin, project axons along correct trajectories, and form functional synapses with their correct

cellular partners while avoiding the wrong targets. How neuronal diversity is created and how developing

neurons become properly assembled into functionally distinct neural circuits are fundamental, interlocked

questions central to the study of neuroscience.

The vertebrate spinal cord is an excellent model system for studying questions of neuronal

specification and synaptic connectivity. Pioneering work over the past several decades has identified key

signaling molecules involved in patterning the developing spinal cord as well as genetic programs

specifying the cardinal neuron types – spinal motor neurons (MN), which project out of the spinal cord to

innervate muscle in the periphery; and spinal interneurons (IN), which comprise diverse classes of

inhibitory and excitatory neurons that fine-tune sensorimotor activity (Jessell, 2000; Goulding, 2009). To

establish functional motor circuits, the MN progenitor domain generates dozens of MN subtypes with

distinct molecular identity, target muscle connectivity, and physiological properties. Diffusible patterning

signals, intrinsic transcription factor (TF) programs, and cell-cell interactions have all been shown to

contribute to specification of MN subtype identity and assembly of motor circuits (Shirasaki & Pfaff, 2002;

Dasen, 2009; Kanning et al, 2010). While we have increasing insight into the programs that specify spinal

MN identity and synaptic connectivity, we know significantly less about the developmental programs that

specify spinal INs. Indeed, whether similar diversification occurs in spinal IN progenitor domains and

whether analogous molecular mechanisms operate to diversify spinal INs and establish their specific

patterns of synaptic connectivity is not currently known.

In order to better understand the development of IN identity and connectivity, I focused my

studies on a well-established monosynaptic stretch reflex microcircuit in the ventral spinal cord that

controls motor activity (Fig 1.1) (Sherrington, 1906). At the most basic level, motor behavior is regulated

2

by feedback information provided by proprioceptive sensory neurons (pSN) onto MNs along a

monosynaptic reflex arc. When a muscle in the periphery is stretched, group Ia and II pSNs, which

innervate the primary and secondary endings of muscle spindle fibers, respectively, detect the change in

muscle spindle length and transmit this signal via afferents back into the spinal cord onto MNs whose

axons innervate the same muscle. Excitatory inputs from pSN terminals evoke MN activity, which is then

transmitted to the muscle, leading to a muscle contraction that resists the stretch, thereby restoring the

position of the muscle and providing joint and postural stability (Hultborn, 2006).

Sensorimotor circuits such as the one described can be further modulated by local spinal INs

releasing excitatory or inhibitory neurotransmitters that act to either reinforce or dampen MN activity in

response to sensory-evoked stimulation (Brownstone et al, 2010). For example, certain types of

excitatory INs, including V2a INs, receive input directly from pSNs and form direct synaptic contacts with

MNs. These excitatory INs amplify the sensory signal to MNs, resulting in increased and/or prolonged MN

excitability (Ni et al, 2014; Kiehn, 2016). Alternatively, inhibitory INs can intervene in the reflex arc by

acting to either directly inhibit MN activity (otherwise known as postsynaptic inhibition), or by restricting

excitatory sensory signals to MNs, thereby acting to provide indirect inhibition (also known as presynaptic

inhibition) (Jankowska and Puczynska, 2008; Betley et al, 2009). In the monosynaptic stretch reflex

circuit, both levels of inhibition are observed: postsynaptic inhibitory INs synapse directly onto MNs while

presynaptic inhibitory INs synapse with pSN terminals that are located on or in close proximity to MNs.

Importantly, while presynaptic inhibitory INs can form physical contacts with MNs early in development,

their synaptic differentiation and maturation only occurs upon contact with the SN afferent terminal. Most

impressively, when SNs are removed from the circuit, presynaptic INs do not aberrantly synapse with

MNs and even withdraw their axons from the ventral spinal cord (Betley et al, 2009).

How is this specificity encoded? Recently, sensory-derived recognition signals have been

implicated in directing presynaptic inhibitory INs to their appropriate sensory targets (Betley et al, 200;

Ashrafi et al, 2014; Mende et al, 2016). Despite remarkable progress in understanding the formation of

this specific spinal circuit, the relationship between neural identity and target selection remains poorly

understood. Previous studies have revealed that the two types of inhibitory INs in this circuit have distinct

genetic identities and developmental origins (Betley et al, 2009). Postsynaptic inhibition is mediated by V1

3

INs, a population of inhibitory INs that arise from a domain in the ventral spinal cord that transiently

expresses the TF En1 (Engrailed-1) (Sapir et al, 2004). In contrast, INs mediating presynaptic inhibition

are dI4 INs, GABAergic neurons that have dorsal origins and arise from a progenitor domain defined by

the TF Ptf1a (Pancreas specific transcription factor, 1a) (Glasgow et al, 2005). Given their distinct

molecular development and synaptic choices, V1 and dI4 INs represent compelling models to study the

link between molecular identity and synaptic connectivity.

Further complicating the matter is that both classes of spinal INs can be subdivided into

molecularly distinct subpopulations. Dorsally-derived dI4 INs can be segregated into at least two major

subsets, an early-born group (dI4) that provides presynaptic inhibition to pSN afferents while a later-born

group (dILA) preferentially inhibits cutaneous sensory afferents targeting more superficial laminae

(Glasgow et al, 2005; Wildner et al, 2006). Although we know that these broad subsets can be

distinguished by their birth order, definitive molecular markers for dI4 subtypes are lacking and the

precise developmental programs distinguishing them are currently unknown.

Compared to dI4 INs, the V1 IN class can be partitioned into over fifty discrete subpopulations

based on combinatorial expression of nineteen different TFs (Bikoff et al, 2016; Gabitto et al, 2016). While

the majority of these subtypes have not yet been characterized, it is proposed that V1 subpopulations are

analogously heterogeneous in their functional properties and thus differentially recruited to distinct spinal

microcircuits depending on the biophysical requirements of the individual motor unit (Bikoff et al, 2016).

Among the few well-characterized spinal IN subtypes is the V1-derived Renshaw cell (RC), which has the

unique role of providing recurrent inhibition of MNs, in which RCs receive excitatory inputs from MN axon

collaterals and provide feedback inhibition to the same MNs (Renshaw, 1946; Eccles et al, 1954; Alvarez

& Fyffe, 2007). RCs are the only V1 IN subtype for which there is substantial evidence of their synaptic

preferences; yet how RCs are uniquely specified from other V1 INs is not well understood. Furthermore,

whether RCs are uniquely specified among V1 INs to acquire their distinct functional properties, including

physiological signature, settling position and role in recurrent inhibition of MNs; or whether these

properties are a result of the cellular milieu and a response to spatial constraints has not been addressed

(Sapir et al, 2004). By examining the formation and specificity of the RC circuit, we can begin to infer the

mechanism and logic of the extensive subtype diversity within the V1 IN class.

4

Apart from RCs, the molecular programs that differentiate V1 and dI4 IN subtypes are virtually

unknown and detailed analyses of the synaptic connections made by these different subtypes are

severely lacking. Furthermore, while we know that molecular codes established early during development

underlie the differentiation of apparently homogenous progenitors into different neural cell types, the role

of transcriptional identity in informing decisions about functional maturation and synaptic connectivity later

during development is much less understood. We can thus ask the following questions: Given the

apparent molecular heterogeneity in these spinal IN classes, is there a role for cell-intrinsic programs in

directing different inhibitory IN subtypes towards their preferred synaptic partners? Alternatively, is

matching between pre and postsynaptic partners in this circuit principally determined by cell non-

autonomous signaling or interactions? Finally, are all or only certain subsets of V1 and dI4 INs capable of

integrating into the monosynaptic reflex circuit? By examining these issues, we can gain insights into how

neuronal cell type identity is translated into specific patterns of synaptic connectivity.

The ability to answer these questions, however, is hampered by the complexity of the spinal cord,

where many different neuronal cell types can be found synapsing in the same small area. Furthermore,

there is the challenge of obtaining cells of a particular subtype for analysis, both because there are not

yet methods to isolate specific V1 and dI4 subtypes from spinal cord and because some of these

subpopulations likely contain only very small numbers of cells. Recently, pluripotent stem cells have

emerged as powerful tools for studying mechanisms of neural development, in particular because they

can be differentiated to produce large amounts of diverse neuronal cell types for in-depth study (Petros et

al, 2011). Indeed, our lab has shown that embryonic stem cells (ESC) can be directed to differentiate into

spinal MNs in a process that recapitulates normal MN development in vivo (Wichterle, 2002; Peljto et al,

2011; Tan et al, 2016). Taking this innovation one step further, I will propose in this thesis that ESC-

derived INs can also be efficiently generated and used in a simplified in vitro co-culture system with MNs

to examine as well as manipulate the interactions between defined cell types in order to learn about the

construction of neural circuits.

In this introductory chapter, I will first describe general mechanisms for establishing neuronal

diversity and synaptic specificity in different areas of the central nervous system across different species.

I will then explore these processes more deeply through investigation of the development of an inhibitory

5

reflex microcircuit in the mammalian spinal cord that exhibits striking synaptic specificity. Finally, I will

describe my approach to studying this reflex circuit through the use of neurons generated from rational

differentiation of embryonic stem cells.

A. Mechanisms of neuronal diversity

Our modern conception of the basic organization of the nervous system relies on advances made

by neuroanatomists over a century ago, in particular the recognition that individual neurons are the basic

building blocks of the brain (Cajal, 1911). The mammalian brain contains tens of billions of neurons,

including thousands of molecularly distinct neuronal cell types (Luo et al, 2008). Over the past several

decades, studies in model organisms ranging from simple invertebrates such as worms and flies to more

complex mammals such as mice have revealed important developmental principles underlying neuronal

diversity. A fundamental discovery is that vast numbers of neurons can be generated from a limited

number of progenitor pools that are themselves uniquely specified by their spatial position and temporal

identity in the developing neuroepithelium. Indeed, the regionalization of the developing neural tube into

distinct compartments by secreted morphogens from patterning centers along the dorsoventral and

rostrocaudal axes of the developing embryo is essential for the generation of distinct progenitor domains

giving rise to the basic neuronal cell types (Skeath, 1999; Jessell, 2000). In addition to spatial patterning

of developing neural tissue, in many regions of the CNS neural progenitors are also progressively

restricted in their ability to generate different cell types over time (Pearson & Doe, 2004).

Regionalization of the nervous system

Neural progenitors acquire distinct identities and different fates depending on their positions along

the rostrocaudal and dorsoventral axes of the neural tube. These positional coordinates are established

by morphogenetic signaling centers, oriented so that cells situated closer to the signaling source are

exposed to higher concentrations of the signal, leading them to interpret the signal differently than cells

located further away. The differential progenitor response to the secreted morphogen leads to the

regulation of genes, especially TFs, which constitute a neural code directing the differentiation of neural

6

stem cells towards particular cell fates. Such gene regulatory cascades often involve cross-repressive

interactions that help to define and maintain the boundaries of these different fates (Jessell, 2000; Briscoe

et al, 2000).

The phenomenon of positional designation of neural progenitors is especially well-established in

the Drosophila embryonic CNS, which develops from a small set of neural stem cells called neuroblasts

(NBs) arising from the developing neuroectoderm in a remarkably stereotyped pattern. In Drosophila,

morphogen gradients lead to expression of segment-polarity and columnar genes governing the

acquisition of different fates of NBs developing in distinct rostrocaudal and dorsoventral domains of the

neuroectoderm, respectively. Indeed, NB formation and fate specification are essentially identical in all

hemisegments of the developing Drosophila embryo, with each NB developing in the same relative

position in any hemisegment acquiring the same cell fates, including producing an invariant and unique

sequence of neurons and/or glia according to their spatial identity (Skeath, 1999).

Spatial mechanisms are also used to pattern progenitors in the mammalian CNS. During early

mouse embryonic development, the formation of rhombomere 1 (r1), from which the cerebellum arises, at

the boundary between the midbrain and hindbrain is regulated by morphogenetic signals, especially

fibroblast growth factors (FGF) (Hidalgo-Sanchez et al, 1999; Wingate & Hatten, 1999; Butts et al, 2014).

In particular, Fgf8 functions to define r1 territory by repressing Otx2 expression at the anterior boundary

of r1. Conversely, Gbx2 expressed at the posterior boundary acts to maintain Fgf8 expression, while also

engaging in cross-repressive interactions with Otx2 (Simeone, 2000; Sato et al, 2004; Sato & Joyner,

2006). Further evidence suggests that Fgf8 acts together with other morphogenetic signaling molecules,

including Wingless (WNT), Sonic Hedgehog (Shh) and transforming growth factor (TGF)-ß family

members at the midbrain-hindbrain boundary to specify two distinct cerebellar neuroeithelial zones which

generate different types of cerebellar neurons: GABAergic inhibitory-type neurons produced from the

ventricular zone (VZ), while the more dorsally-positioned rhombic lip produces glutamatergic excitatory

neurons of the cerebellum (Hatten et al, 1997; Martinez et al, 2013; Hoshino et al, 2005; Machold &

Fishell, 2005; Wang et al, 2005; Yamada et al, 2014). Thus, spatially restricted extrinsic signaling cues

and gene expression interact to pattern cerebellar progenitors to produce distinct cerebellar cell-types

according to their site of origin.

7

Similarly, in the mouse neocortex, neural progenitors that are the source of excitatory cortical

projection neurons are heterogeneous and produce discrete neuronal lineages according to their

positional identity. Early during development, morphogenetic signals FGFs, WNTs, and bone

morphogenetic protein (BMP) family members are secreted in orthogonal orientations along the

rostrocaudal and medial-lateral (ML) boundaries of the dorsal telencephalon, inducing the graded

expression of TFs that define the positional coordinates of progenitors in the ventricular zone (VZ) and

endow them with specific area identities. Indeed, manipulations of morphogen gradients, especially Fgf8,

or TF expression results in significant changes in the size and position of cortical areas. Each re-specified

area subsequently expresses molecular markers consistent with their new areal identity, suggesting that

positional cues established in progenitors dictate the ultimate identity adopted by postmitotic neurons

(Greig et al, 2013).

The other major neuronal population in the cortex are GABAergic inhibitory INs, which are born in

subpallial regions of the developing cortex, specifically the ganglionic eminences (Wichterle et al, 2001).

The three eminences are partitioned based on their spatial position within the subventricular zone (SVZ) –

lateral (LGE), medial (MGE) and caudal (CGE). FGF and Sonic hedgehog (Shh) morphogen gradients

induce expression of genes such as Nkx2-1, Gsx2, and Pax6, which establish the independent progenitor

cell populations in MGE, LGE and CGE; and interactions between these three genes define the

boundaries between the different progenitor zones (Corbin et al, 2003). The distinction between these

three zones is crucial given that each produces a different assemblage of cell types, with LGE progenitors

producing medium spiny striatal projection neurons and olfactory bulb INs; MGE progenitors being the

primary source of cortical INs; and CGE progenitors producing cortical INs that are distinct from MGE-

derived neurons. The results of heterotopic transplantation and in vitro culture studies indicates that

progenitors in each ganglionic eminence has intrinsic ability to generate these distinct IN cell types,

providing further evidence that the positional allocation of progenitors is an important mechanism for

establishing neural diversity (Nery et al, 2002; Butt et al, 2005; Corbin & Butt, 2011).

8

Temporal identity specification

In addition to spatial patterning, temporal identity specification is another important mechanism

for establishing neuronal diversity. Indeed, stem and progenitor cells often produce different cell types in

stereotyped birth order during development. In general, temporal identity may be regulated intrinsically or

extrinsically, depending on the organism and cell types under consideration. In the intrinsic model of

temporal identity regulation, progenitors first acquire unique spatial identity and subsequently produce

progeny in fixed birth order, without additional influences from spatial patterning cues. Alternatively,

extrinsic regulation of temporal identity allows for progenitor responsiveness to changes in spatial

patterning cues over time such that different progeny are generated at different times depending on the

cellular environment (Pearson & Doe, 2004).

In the Drosophila embryo, NBs generate an invariant sequence of neurons through asymmetric

divisions that maintain the self-renewing NB progenitor while generating a ganglion mother cell (GMC)

that undergoes symmetric divisions to generate two sister neurons. As they are generated, younger

GMCs push older GMCs and neurons deeper into the developing embryo, resulting in a laminar

organization in which older neurons reside in deep layers while newly born neurons inhabit superficial

layers (Isshiki et al, 2001; Schmid et al, 1999). Genetic studies have provided strong evidence that NBs

acquire their temporal identities through sequential expression of TFs that define their temporal

competence, including Hunchback, Krüppel, Pdm and Castor (Isshiki et al, 2001). Ectopic expression of

Hunchback leads to acquisition or maintenance of early temporal identity so that later-born GMCs are

forced to differentiate into early-born neuron types. Conversely, overexpression of the later temporal

identity factors in progenitors causes skipping of earlier lineages and induces the formation of neurons

that are temporally matched to the lineage appropriate to each TF (Isshiki et al., 2001; Novotny et al.,

2002; Pearson and Doe, 2003). The sequential expression of temporal identity factors appears to be

intrinsically coded since isolated NBs grown in vitro still progress through the same TF cascade and

generate normal numbers of GMCs and neurons (Grosskortenhaus et al, 2005; Broadus & Doe, 1997;

Pearson & Doe, 2004). Thus, temporal specification of NBs enables the generation of different neuronal

cell types in fixed birth order and only within specific developmental time windows, providing an additional

9

layer of cellular complexity. As a result, neuronal diversity is significantly enhanced despite starting from a

relatively restricted set of progenitors.

In the developing mammalian brain, especially in layered structures such as the mouse retina and

neocortex, neuroepithelial progenitors have also been observed to progress through a series of

competence states in which they generate distinct neuronal cell types in stereotyped birth order. During

retinal development, six types of neurons (and one type of glia) are generated in a conserved order, with

retinal ganglion cells generated first, followed by rod and cone photoreceptors, bipolar cells, and Müller

glia (Livesey & Cepko, 2001). Retinal progenitors are multipotent and pass through a series of

competence states, during each of which progenitors are competent to produce a subset of retinal cell

types. Heterochronic transplantation experiments in which retinal progenitors from different stages of

development were transplanted into either young or aged hosts demonstrated that early progenitors

generate retinal neuron types consistent with their age regardless of the age of the environment,

indicating that the competence state of retinal progenitors is principally intrinsically determined (Pearson

& Doe, 2004; Cepko, 2014).

Meanwhile, in the developing neocortex, different progenitor types in the VZ and SVZ also

generate distinct PNs in sequential waves. Newly born neurons migrate out from their neurogenic zone

and settle in the cortical plate in an inside-out manner such that early-born neurons reside in deeper

layers (VI, then V) and later-born neurons migrate past them to populate more superficial layers (IV, then

II and III) (Greig et al, 2013). Milestone transplantation experiments performed in the 1990s, as well as

recent in vitro studies of dissociated primary and stem cell-derived cortical progenitors, suggest that not

only can cortical progenitors cell-autonomously recapitulate the sequential generation of PN subtypes, but

that they also become progressively restricted in their ability to produce different types of PNs (McConnell

& Kaznowski, 1991; Greig et al, 2013).

It is clear from these selected examples in flies and mice that considerable progress has been

made in the last few decades towards understanding how spatial and temporal factors converge to

pattern distinct progenitor domains giving rise to the most basic neuronal cell types. Nevertheless, the few

progenitor domains established during development cannot account for the immense diversity of neuronal

10

subtypes present in the mature nervous system. Indeed, more recent studies suggest that intradomain

diversification of progenitors and postmitotic neurons also significantly contributes to neuronal diversity.

Intradomain diversification

In many areas of the CNS, the establishment of principal progenitor domains is only the first step

in generating the appropriate complement and diversity of neurons. These principal classes of neurons

are often further partitioned into even more specialized neuronal subtypes (Rockhill et al, 2002;

Molyneaux et al, 2007; Dasen, 2009; Peljto et al, 2010). The generation of cellular diversity within a single

progenitor domain can be achieved at both the progenitor and posmtitotic neuron stages. Unsurprisingly,

spatial and temporal patterning mechanisms also contribute to intradomain diversification.

As described earlier, each progenitor NB in the developing Drosophila embryo produces an

invariant repertoire of neurons. Although initial spatial and temporal specification is essential for each NB

to acquire a unique lineage identity, progenitor lineages themselves are subsequently further diversified

through binary cell fate decisions between sister neurons generated from the same GMC, as well as

through acquisition of distinct temporal identities (Skeath, 1999). In regards to the second mechanism, it

has been proposed that the temporal identity factors can be reused to specify many more neuron types

within a single progenitor lineage (Pearson & Doe, 2003). For example, in the NB 5-6T lineage, a single

temporal identity factor, Castor, triggers a series of transcriptional cascades that act to establish distinct

sequential competence time windows, ultimately leading to the cell fate diversification of four lineage-

related neurons that all express the LIM-homeodomain TF Apterous (Ap) (Baumgardt et al, 2007;

Baumgardt et al, 2009). The reuse of the same temporal identity factor within the same progenitor

lineage, together with changes in progenitor competence, allows for expression of more temporal

identities than there are temporal identity factors. Thus, cellular diversification in the Drosophila CNS

appears to rely not only on the initial specification of unique progenitor lineages through spatial and

temporal patterning mechanisms, but also through subsequent and iterative transformations of

progenitors within each of these lineages.

Intradomain diversification of progenitors has also been observed in the mouse CNS. In the

cerebellum, the bHLH TFs Ptf1a and Atoh1 are expressed in the VZ, which produces GABAergic INs, and

11

the rhombic lip (RL), which generates glutamatergic neurons, respectively. Indeed, Ptf1a and Atoh are

both required and sufficient for specification of GABAergic versus glutamatergic cell fate in the

cerebellum. Ptf1a-expressing progenitors in the cerebellum produce Purkinje, Golgi, Lugaro,

candelabrum, basket, and stellate cells of the cerebellar cortex, as well as two types of inhibitory

projection INs of the cerebellar nuclei (CN-GABA-ION and CN-GABA) (Hoshino et al, 2005; Sudarov et al,

2011). By contrast, Atoh-expressing neuroepithelial cells generate granule cells of the cerebellar cortex,

as well as large glutamatergic projection neurons in the cerebellar nuclei (CN-Glu) (MacHold & Fishell,

2005; Wang et al, 2005).

Birthdating studies in the cerebellum reveals that different cell types are generated at different

times during cerebellar neurogenesis both in the VZ and the RL, with projection neurons preceding the

formation of local INs (Sekerková et al, 2004a,b). In the VZ, Ptf1a-derived progenitors sequentially

generate different types of inhibitory neurons in an inside-out fashion, with deep nuclei projection neurons

generated first, then Purkinje cells, followed by Golgi and Lugaro cells in the granular layer, candelabrum

cells, and finally basket and stellate cells in the molecular layer (Schilling, 2000; Leto et al, 2006; Sudarov

et al, 2011). Previous studies using heterochronic and heterotopic transplantations suggest that

progenitors from the early cerebellar primordium are capable of differentiating into all major cerebellar

neuronal cell types, while postnatal cerebellar tissues only differentiate into Pax2-positive INs, suggesting

that cerebellar progenitors become restricted in their developmental potential over time (Gao & Hatten,

1994; Alder et al, 1996; Jankovski et al, 1996; Carletti et al, 2002). However, a recent study using similar

transplantation approach proposes that while different types of GABAergic INs likely originate from a

common pool of progenitors, these maintain their developmental potentials until the end and adopt

mature neuronal identities based on local instructive cues. Meanwhile, projection neurons from the VZ

(e.g. CN-GABA-ION and Purkinje cells) are committed to their fate early on and adopt their mature

identities largely through cell-intrinsic mechanisms (Leto et al, 2006; Leto et al, 2009; Leto & Rossi 2012).

Nevertheless, these results indicate that diversification of the cerebellar Ptf1a domain to generate

different inhibitory IN cell types likely arises from the confluence of both spatial and temporal information.

As described earlier, the majority of cortical inhibitory INs are produced from the MGE, which is

initially patterned by FGF and sonic hedgehog (Shh) signaling to induce the expression of Nkx2.1, a TF

12

that is required for MGE development (Fishell, 2007). Recent studies suggest that Shh signaling is used

again for intra-MGE patterning, such that a higher level of Shh promotes the generation of somatostatin-

expressing inhibitory INs from MGE precursors at the expense of the parvalbumin-expressing subtype

(Xu et al, 2010). In addition to spatial patterning, progenitors in the MGE are also thought to produce

different IN subtypes over time, which may account for the startling diversity of inhibitory cell types

observed in cortical circuits in the adult. Using a genetic fate-mapping strategy to label MGE progenitors

at different time points, Miyoshi et al (2007) showed that physiologically distinct cortical IN subtypes are

generated in a defined temporal sequence, which may explain their restricted laminar distribution in the

cortex. Another recent study examined the development origins of chandelier cells, a highly distinct

subclass of cortical IN, showing that they derive from Nkx2.1-lineage progenitors in the MGE later during

mouse embryonic development compared to other IN subtypes (Taniguchi et al, 2013). These studies

demonstrate that spatial and temporal specification of progenitors early in development can have a strong

influence on the subsequent migration, connectivity, and function of INs.

Intradomain diversification of neurons can also occur at postmitotic stages. In the mouse retina,

photoreceptors are born with generic identity but are subsequently specialized at stages through the

activation of TFs that control the selection of one of three subtype-specific identities: rod, M cone, or S

cone (Oh et al, 2007; Swaroop et al, 2010). Similarly, in the developing mouse neocortex, the

specification of upper layer cortical PNs is controlled by a TF that is expressed predominantly in young

postmitotic upper layer neurons but not in progenitors. In the absence of Satb2 TF expression , upper

layer neurons lose their subtype-specific molecular and functional identity and ectopically activate other

neuronal programs, including those normally expressed by deep layer neurons (Britanova et al, 2008).

In summary, intradomain neural diversification can be enacted by mechanisms specifying cell fate

in both progenitors and postmitotic cells, including spatial and temporal programs reminiscent of those

involved in initial patterning of broader progenitor domains. In the next section, I will introduce and review

a specific signaling pathway that is involved in controlling patterning and cell fate in neurons.

13

Notch signaling in neuronal diversification

The Notch signaling pathway represents one of the best known, evolutionarily conserved

molecular mechanisms for generating intradomain cellular diversity in the CNS. Signals exchanged by

neighboring cells through Notch receptors control many critical developmental processes, including the

interdependent programs controlling cell-cycle exit and cell-fate specification (Artavanis-Tsakonas et al,

1999). Although Notch signaling is often associated with alternative cell fate specification, especially

through asymmetric division and lateral inhibition, these same cellular mechanisms underlie the ability of

Notch signaling to control the timing of cell cycle exit and differentiation to affect the temporal

specification of different cell types (Tan et al, 2016). Here, I will discuss several systems in which Notch

signaling has been shown to be involved in generating intradomain neuronal diversity.

While many genes are implicated in Notch signaling, a small subset are widely recognized as

forming the core of the pathway. In Drosophila, there is one Notch receptor and two ligands (Delta and

Jagged). Mammals have four Notch receptors (Notch 1-4), three Delta ligands (Delta 1,3,4), and two

Jagged ligands (Jag 1,2) (Gazave et al, 2009). During vertebrate neural development, Notch receptors

and ligands are primarily expressed by undifferentiated neural progenitor cells within the proliferative VZ

(Imayoshi et al, 2010). The interaction between the extracellular domain of the Notch receptor on one cell

and its ligand expressed on an adjacent cell leads to gamma secretase-mediated cleavage and release of

the intracellular domain of the Notch receptor (NICD) from the plasma membrane into the cytoplasm.

NICD translocates to the nucleus, where it interacts with the DNA binding complex CSL (CBF1/RBPJ,

Suppressor of hairless, Lag1) to recruit transcriptional co-activators such as Mastermind1 (Maml1), which

activates transcription of Notch target genes, including Hes1 and Hes5 (Kimble and Simpson, 1997;

Greenwald, 1998; Wilson & Kovall, 2006).

One of the best-characterized roles of Notch signaling is lateral inhibition, a process by which a

single cell within a field of uniform progenitors acquires a specific cell fate. Lateral inhibition was first

extensively studied in the Drosophila peripheral nervous system (Simpson, 1991). Here, uniform

neuroepithelial progenitors all have the ability to differentiate into either neuroblasts or epithelial cells, but

active Notch signaling inhibits their differentiation. All progenitors initially express low levels of both Delta

ligand and Notch receptor but due to stochastic variations, a single progenitor will escape Notch signaling

14

and differentiate into an epithelial cell. As this cell differentiates, it also upregulates expression of Delta

ligand, inducing Notch signaling in neighboring cells and therefore inhibiting its neighbors from

differentiating into additional epithelial cells (Heitzler and Simpson, 1993). This Notch-dependent system

of lateral inhibition is widely used to pattern embryonic tissues emerging from homogeneous progenitors,

including in the vertebrate CNS (Greenwald & Rubin, 1992). At the molecular level, this is achieved

through the regulation of basic-helix-loop-helix (bHLH) TFs that function to specify neuronal identity and

differentiation. High Notch activity maintains neural progenitors by upregulating TFs that repress bHLH

genes. When Notch activity is low, bHLH TFs are de-repressed, leading to progenitor exit from cell cycle

and cellular differentiation (Heitzler et al, 1996; Artavanis-Tskakonas et al, 1999; Huang et al, 2014).

Notch-mediated lateral inhibition also regulates neurogenesis and cell fate specification in the

vertebrate CNS. In the mouse CNS, Notch receptor mutants misexpress proneural bHLH TFs and

undergo precocious neurogenesis throughout the CNS, resulting in an overall decrease in neurons as a

result of progenitor pool depletion. Conversely, ectopic Notch expression maintains cells in the progenitor

state and prevents neuronal differentiation (Yoon et al, 2004; Yoon & Gaiano, 2005). In the retina

specifically, disruption of Notch signaling by genetic removal of either the Notch1 receptor or Notch

downstream effectors during mouse retinal development leads to premature cell cycle exit and neuronal

differentiation (Jadhav et al, 2006a; Jadhav et al, 2006b; Yaron et al, 2006). These results are consistent

with the idea that Notch signaling normally inhibits neuronal differentiation and promotes maintenance of

progenitor identity.

Importantly, Notch control of progenitor maintenance indirectly leads to control of cell fates

generated during temporally patterned neurogenesis. The role of Notch in temporal patterning has been

especially well-studied in the mouse retina, where progenitors produce different retinal cell types in

stereotyped birth order. Inhibition of Notch signaling early during retinal neurogenesis results in

precocious neuronal differentiation and enrichment of early-born retinal subtypes at the expense of later-

born neurons (Livesey & Cepko, 2001). Conversely, later inhibition of Notch bypasses the generation of

early-born subtypes and increases production of neuronal subtypes normally born at those later time

points (Nelson et al, 2007). Thus, Notch appears to control cell identity by forcing progenitors to

differentiate into temporally-relevant neuronal cell types.

15

Interestingly, Notch inhibition in chick retinal progenitors leads to the exclusive increase in RGCs,

while Notch signaling is required in postmitotic retinal neurons for acquisition of non-rod photoreceptor

identities (Austen et al, 1995; Mizeracka et al, 2013). Therefore, in addition to having a permissive role in

regulating timing of neurogenesis in the retina, Notch also appears to have an instructive in specifying

distinct cell fates at both the progenitor and postmitotic stages of retina development. A role for Notch-

mediated instructive signaling in intradomain cell fate diversification has also been firmly established in

the mammalian cochlea. During development of the inner ear, sensory hair cells and their associated

non-sensory support cells differentiate from common progenitors. Experimental ablation of hair cells

results in the activation and differentiation of support cells, suggesting that support cells normally receive

inhibitory Notch signals from neighboring hair cells to prevent their differentiation. Furthermore, genetic

deletion of Jag2 results in a significant increase in production of sensory hair cells at the expensive of

support cells, providing evidence that Notch signaling is required for proper development of the mosaic of

hair and support cells in the vertebrate cochlea (Lanford et al, 1999).

Thus, in the developing nervous system of both invertebrates and vertebrates, diversity is first

established at the progenitor level by intersection of spatial and temporal patterning cues. This cellular

diversity is further elaborated by the differential response of progenitors and newly postmitotic cells to

signals such as Notch, the ultimate result of which is the richness of nerve cell types found in the mature

nervous system. Despite progress in understanding how neuronal diversity is established, the extent to

which molecularly distinct neuronal subtypes diverge in their physiological properties and patterns of

synaptic connectivity remains relatively unknown. In the next section, I will discuss our current

understanding of how emerging neural cell types assemble into functionally diverse neuronal circuits

through selective synaptic connections with their appropriate cellular targets.

B. Synaptic specificity in the developing nervous system

One of the main expressions of a neuron’s unique identity is the formation of highly specific

synaptic connections. The nervous system uses multiple mechanisms to ensure that developing axons

reach their designated target areas and choose appropriate synaptic partners among many potential

16

partners in the vicinity. These molecular mechanisms include complementary cell-adhesion labels on pre-

and postsynaptic partners that can elicit attraction or repulsion, guidepost cells, and synaptic

differentiation or pruning (Shen & Scheiffele, 2010). In this section, I will first discuss several models of

synaptic specificity development, including examples of molecular mechanisms driving the assembly and

remodeling of synaptic connectivity. I will then detail some experiments that have examined whether

synaptic specificity can be recapitulated in dissociated neuronal cultures as an introduction to my own

attempts to model the synaptic specificity of a specific spinal microcircuit in vitro.

Models of synaptic specificity

The striking specificity of connectivity seen in many neural circuits suggests that this process is

highly regulated. According to one theory, the development of such highly precise synaptic connections is

established through genetically hardwired developmental programs such that neurons are highly specified

and their connections are fully determined at specification (Jacobson, 1969; Sanes & Yamagata, 2009).

In the late 1800s, J.N. Langley observed that sympathetic preganglionic neurons synapse on distinct

populations of ganglion cells in the mammalian autonomic nervous system according to their rostrocaudal

identity (1892). Decades later, Roger Sperry proposed the existence of “cytochemical affinities,”

genetically predetermined factors that would endow each pair of neurons with complementary chemical

identities that would facilitate proper recognition (Sperry, 1963).

Embedded in this deterministic theory of synaptic specificity is the distinction between absolute

versus hierarchical specificity. In the former, more extreme variant, neurons form synapses only on a

restricted set of targets that are selectively recognized by specificity factors shared between the pre- and

postsynaptic partners. This is exemplified in the Drosophila larval neuromuscular system, in which the

specificity of MN innervation to muscle fibers persists even when certain muscle fibers are genetically

duplicated or deleted (Kurusu et al, 2008; Rose & Chiba, 2000).

Hierarchical synaptic specificity, on the other hand, proposes that neurons are able to form

functional synapses with secondary targets, particularly in cases where the preferred target is removed or

unavailable (Cash et al, 1992; Shen & Scheiffele, 2010). Hierarchical, rather than stringent, specificity is

more commonly observed in neural circuits. In C. elegans, guidepost epithelial cells form the molecular

17

scaffold for synapse formation between egg-laying MNs and their vulval epithelial cell (VC) targets. The

removal of the guidepost cell or the molecular signals produced by this cell prevents the MN from forming

synapses with its VC target and leads it to form ectopic synapses on body wall muscle cells that normally

do not receive synaptic input from this MN class (Shen & Bargmann, 2003; Shen et al, 2004). Meanwhile,

in the mouse cerebellum, mossy fibers synapse preferentially with granule and Golgi cells, but not with

other cerebellar cell types, especially Purkinje cells (Palay and Chan-Palay, 1974). Nevertheless, in

weaver or reeler mutant mice, which have severe cerebellar defects including absence of granule cells

and heterotopic formation of Purkinje cells, some mossy fibers form aberrant synapses on Purkinje cells

(Sotelo, 1975a; Mariani et al, 1977). This result suggests that far from being rigidly fixed, synaptic

connections may be influenced by the local cellular context.

In contrast to the deterministic models, a theory of promiscuous synaptic specificity has also been

proposed in which neurons are incompletely specified and have relatively indeterminate connections.

These are then refined through activity-dependent mechanisms and removal of inappropriate

connections, ultimately leading to stereotyped and specific wiring of neural circuits (Sanes & Yamagata,

2009). This concept has arguably been best-studied in the developing vertebrate retinotopic map, in

which the selective, activity-dependent removal of inappropriate synapses between RGCs and their

targets in the optic tectum/superior colliculus and dorsal lateral geniculate nucleus leads to highly precise

patterns of connectivity (Katz & Shatz, 1996; Sanes & Yamagata, 2009).

Regardless, it is highly likely that most neural networks are formed through parallel processes of

genetic predetermination and variable, context-dependent wiring. The coalescence of these processes

could account for two essential properties of a properly functioning nervous system: the need for neural

circuits to be assembled with high accuracy while also enabling the organism to adapt to its own

experiences and context. In the following section I will briefly describe some molecular mechanisms

underlying synaptic specificity in neural circuits.

Molecular cues guiding synaptic specificity

When an axon reaches its target destination, how does it select its specific synaptic partner

among many other potential candidates? One prominent mechanism that is the logical extension from

18

Sperry’s chemoaffinity hypothesis is the “lock-and-key” model, in which the presynaptic axon and

postsynaptic target cell both express matching recognition signals. Intuitively, this model suggests that

the molecular recognition between synaptic partners induces synapse formation by facilitating attractive

interactions. On the other hand, recognition molecules might also function to inhibit synapse formation

with inappropriate targets, in particular by inducing a repulsive signal.

Several families of cell-surface receptors have been implicated in the direct matching of synaptic

partners (Sanes & Yamagata, 2009). In the Drosophila neuromuscular system, the leucine-rich repeat

(LRR) proteins Capricious (Caps) and Tartan are expressed in specific presynaptic neurons and their

postsynaptic targets. Loss-of-function and overexpression studies indicate that Caps and Tartan are

required for proper target selection and synapse formation through homophilic interactions (Shishido et al,

1998; Kurusu et al, 2008). Similarly, in the mouse retina, different subpopulations of INs and RGCs

express different variants of immunoglobulin-superfamily (IgSF) molecules (e.g. Sidekick-1, Sidekick-2,

Dscam, and DscamL), each of which engages in homophilic interactions. Cells expressing each of these

IgSF proteins arborize in specific laminae of the inner plexiform layer of the retina, with pre- and

postsynaptic cells expressing the same gene arborizing and forming synapses in the same sublamina

(Yamagata & Sanes, 2008). These examples and others suggest that homophilic recognition and

attraction is a conserved mechanism for generating precise patterns of connectivity.

As mentioned earlier, repulsive signals arising from inappropriate partners can also shape

synaptic specificity. Another example from the Drosophila larval neuromuscular system illustrates this

concept well: During normal development, muscle 13, but not the adjacent muscle 12, secretes Wnt4

protein. In animals lacking Wnt4 or its receptor Drizzled-2, MN axons that normally innervate muscle 12

now ectopically synapse on muscle 13. Conversely, ectopic expression of Wnt4 in muscle 12 inhibits

those MNs from now innervating muscle 12. Thus, Wnt4 is a repellent signal for MNs that leads to the

inhibition of synapse formation on inappropriate targets (Inaki et al, 2007).

In many cases, the formation of a synaptic contact between pre- and postsynaptic partners does

not mean that the synapse is fully functional or even permanent. Synapse assembly often requires the

differentiation and maturation of pre- and postsynaptic specializations for complete functionality, as

demonstrated by neurexin-neuroligin complex in mammalian cells. Here, postsynaptic cells express

19

neuroligins, which is sufficient to trigger presynaptic specializations, including increase in number of

synaptic vesicles per terminal. The interaction between neurexin and neuroligin also elicits differentiation

of the postsynaptic cell, including promoting the formation of dendritic spines and recruitment of

postsynaptic scaffolding proteins and receptors to synaptic sites (Scheiffele et al, 2000; Dean et al, 2003;

Chih et al, 2005; Shen & Scheiffele, 2010). Thus, synaptic organizers like the neurexin-neuroligin

adhesion complex help to encode connection specificity by ensuring that only correct synaptic

connections are functionally expressed and strengthened.

Finally, it is a well-established phenomenon that developing neurons receive many more inputs

than are seen in the adult nervous system. The elimination of supernumerary inputs is an important

mechanism for generating specific patterns of connectivity since it involves the removal of inappropriate

connections and the strengthening of proper innervations. In many cases, synaptic pruning is an activity-

dependent program. For example, in the Drosophila neuromuscular system, immature neuromuscular

junctions (NMJ) receive inputs from multiple MN axons. Eventually, all but one axon will be eliminated,

resulting in a mono-innervated, mature NMJ. Genetic studies have shown that this process is activity-

dependent process since (1) active axons always outcompete inactive axons, (2) blocking synaptic

activity throughout the NMJ prevents synapse elimination, and (3) excessive activity accelerates pruning

(Lichtman & Colman, 2000; Buffelli et al, 2003). Synapse elimination is also widely used in vertebrate

circuit remodeling, such as in the cortex and retina (Katz & Shatz, 1996; Tran et al, 2009).

Importantly, none of the molecular mechanisms described above are mutually exclusive; rather, it

is likely that multiple, redundant strategies are used by the same neuron at different development stages

to assemble and remodel its specific synaptic connections. Investigation of the developmental programs

establishing synaptic specificity in different neuronal cell types will undoubtedly advance our

understanding of why and how certain neuronal subtypes wire up in functionally distinct neural circuits.

In vitro recapitulation of synaptic specificity Despite considerable progress in understanding how simple neural circuits are assembled, our

ability to fully understand the mechanisms underlying formation of mammalian neural circuits is limited by

both the higher level of complexity of the mammalian brain and the current lack of experimentally

20

accessible models for probing synaptic specificity in a more systematic manner. The use of cultured

neurons would circumvent this problem, but whether isolated cells can recapitulate in vivo patterns of

synaptic connectivity is an unresolved issue. In this section, I will summarize some experiments that have

examined whether neurons can accurately select their appropriate synaptic partners in vitro.

Although performed in a simpler invertebrate system, studies in the marine mollusk Aplysia

californica helped set precedent for specific synapse formation in dissociated culture. In this preparation,

isolated cholinergic INs were capable of making synapses with its appropriate synaptic partners while

avoiding synapses with nontarget neurons (Camardo et al, 1983). Subsequent studies using dissociated

mammalian neurons have produced more mixed results. In hippocampal neurons, pre- and postsynaptic

components are matched according to function. That is, glutamate receptors and other excitatory

synapse-associated molecules cluster opposite glutamate-releasing terminal but not opposite GABA-

releasing terminals, and vice versa (Craig et al, 1994). To test if presynaptic inputs are required for the

proper clustering of postsynaptic specializations, hippocampal neurons were isolated from mouse brain

and grown in microisland culture. Surprisingly, these neurons showed significant mismatch of pre- and

postsynaptic components, with a large proportion of GABA receptor clusters localized opposite

glutamatergic terminals, suggesting promiscuous synaptogenesis in this system (Rao et al, 2000).

Another study of cultured hippocampal cells demonstrated the opposite result. To investigate the

specific formation of synapses between dentate gyrus (DG) and CA3 mossy fiber neurons, Williams et al

cultured DG, CA3 and non-target CA1 neurons together in microisland culture (Williams et al 2011). Their

results using fluorescent visualization of synaptic contacts as well as paired electrophysiological

recordings show that cultured DG neurons synapse on CA3 neurons significantly higher than chance, and

at higher frequency than on other DG neurons and CA1 neurons. In this system, synaptic specificity is

established early and is driven by selective synapse formation onto correct targets and not by synaptic

elimination from incorrect targets. Furthermore, this work implicates a role for the cell-adhesion molecule

cadherin-9 in synapse formation, as cadherin-9 is specifically expressed in DG and CA3 neurons and is

required for the formation and differentiation of this particular circuit. While the authors note that some

inappropriate synaptic connections are made in vitro, the overall results of this study suggest not only can

21

neurons retain their synaptic specificity in dissociated culture, but also that this culture system can be

used to systematically investigate the role of novel recognition factors in establishing specific connectivity.

Mouse cerebellar neurons have also been reported to form selective synaptic connections under

in vitro conditions (Ito & Takeichi, 2009). In the cerebellum, granule cell (GC) dendrites preferentially

synapse with mossy fibers, but not climbing fibers. Importantly, upon synapse formation with their mossy

fiber partners, GC dendrites form characteristic claw-like morphology and localize synapses exclusively to

their distal ends. To test if these features are conserved in culture, Ito & Takeichi cultured dissociated

GCs with pontine nuclei explants, where mossy fibers originate. As controls, they also cocultured GCs

with explants of inferior olivary nuclei, the major source of climbing fibers, or hippocampus, which are

never encountered by GCs in vivo. Using live cell imaging and immunostaining for pre- and postsynaptic

markers, they found that GCs preferentially formed synapses with pontine axons compared to the other

axonal types. Remarkably, not only were some cultured GC neurons able to recapitulate the claw-like

morphology observed in vivo, but pontine axons were able to specifically induce distal end-localization of

synapses in GC dendrites, but not in hippocampal neurons.

These results alongside the work described above from Williams et al (2011), suggest that

neurons can correctly recognize their targets for synaptogenesis even in vitro. Importantly, as discussed

in an earlier section, while synaptogenesis can occur between non-specific partners both in vivo and in

vitro, such synapses may not be as fully differentiated or functional as synapses formed between correct

pairs of neurons. Thus far, I have presented a few examples of pioneering studies in various model

organisms that has provided the foundational basis for our current understanding of nervous system

development, in particular the generation of neuronal diversity and the role of synaptic specificity in

assembly of functionally distinct neural circuits. While I have highlighted many diverse systems in both

invertebrates and vertebrates, one prominent system was deliberately omitted despite being the

benchmark for many of these discoveries and models. In the following section, I will discuss the

generation of neuronal diversity in the vertebrate spinal cord, focusing in particular on the construction of

a specific spinal microcircuit modulating the monosynaptic stretch reflex.

22

C. Spinal cord neurogenesis

The spinal cord is the command center for movement, one of the most essential animal

behaviors. Circuits largely residing in the dorsal spinal cord are responsible for the relay and processing

of cutaneous sensory information to higher brain centers, including nociception, thermosensation,

pruriception, and mechanosensation. Conversely, ventral spinal circuits are involved in integrating

descending inputs from higher brain centers as well as proprioceptive sensory information to produce

coordinated locomotor behaviors (Goulding, 2009). The establishment of mature spinal circuits begins

with the specification of distinct classes of neurons from progenitor cells located at defined positions in

the developing neural tube (Jessell, 2000).

Rostrocaudal spinal cord patterning

During vertebrate embryonic development, the initially flat plate of neuroepithelial cells

invaginates and folds into a closed tubular structure. Differential signaling along the rostrocaudal axis of

the neural tube subsequently divides the CNS into forebrain, midbrain, hindbrain and spinal cord

structures (Fig 1.2) (Lumsden & Krumlauf, 1996; Jessell, 2000). Recent evidence suggests that neural

progenitors of the caudal (posterior) nervous system, which gives rise to the spinal cord, have distinct

developmental origins from progenitors forming the rostral (anterior) brain (Diez del Corral et al, 2003;

Wilson et al, 2009; Gouti et al, 2014). Gradients of Fgf8 and Wnt3a signaling produced from the caudal

primitive streak of the elongating embryo induce a population of neuromesodermal progenitors (NMP)

that can give rise to paraxial mesoderm, which give rise to somites of the early embryo, or spinal cord

cells. Thus, as NMPs migrate out from the caudal region, they choose to adopt one of two fates:

presomitic mesoderm or preneural tube (PNT), with each of these intermediate fates being defined by a

distinct transcriptional identity. PNT cells entering the spinal cord are exposed to increasing

concentrations of retinoic acid (RA) emanating from the developing somites, which represses Fgf8 and

Wnt3a signaling, leading to upregulation of TFs associated with neural progenitor identity, including Pax6

and Irx3 (Gouti et al, 2014; Diez del Corral et al, 2003). The expression of these TFs precipitates a switch

of PNT cells into bona fide spinal neuron progenitors (Gouti et al, 2014; Gouti et al, 2015).

23

Complementary to this more recent model of how posterior progenitor identity is specified is the

classical paradigm for the development of spinal cord identity (Jessell, 2000). Namely, opposing gradients

of RA and FGF signals oriented rostral-to-caudal and vice versa, respectively, control the expression of

the evolutionarily conserved Hox gene family, which encode homedomain TFs that function to confer

regional identity to embryonic tissue (Papalopulu & Kintner, 1996; Muhr et al, 1999; Lamb & Harland,

1995; Cox & Hemmati-Brivanlou, 1995; Dasen & Jessell, 2009). In vertebrates, the 39 Hox genes are

organized in four clusters (Hoxa-d), which are expressed according to their position along the

chromosome such that genes at the 3’ end are expressed in more rostral areas and those at the 5’ end

are expressed more caudally (Durston et al, 2011; Kmita & Duboule, 2003). Progressive RA/FGF

patterning along the rostrocaudal axis induces differential Hox gene expression profiles, resulting in both

the differentiation of posterior spinal cord identity from more anterior fates, as well as segmentation of the

spinal cord into cervical, brachial, thoracic and lumbar regions (Dasen et al, 2003; Dasen & Jessell,

2009). It is thought that graded FGF signals initiate expression of caudal Hoxc6-9 genes controlling

brachial and thoracic spinal identity, while RA is required to induce repress anterior neural fates and

induce cervical Hoxc5 genes (Liu et al, 2001).

The role of RA signaling in establishing rostrocaudal identity in the spinal cord has been

particularly well studied. In the developing embryo, the paraxial mesoderm expresses retinaldehyde

dehydrogenase 2 (Raldh2), the enzyme necessary for RA synthesis (Fig 1.3) (Maden, 2002).

Consequently, Raldh2 mutant animals lose caudal hindbrain and cervical spinal cord structures with

expanded rostral hindbrain and relatively normal forebrain and rostral midbrain (Neiderreither et al, 2000).

As discussed above, opposing gradients of FGF and RA are required for rostrocaudal patterning,

suggesting that these two signaling systems repress each other’s expression (Maden, 2002). In caudal

neural explants, ectopic activation of RA signaling components, including constitutive activation of the

retinoid receptors RAR/RXR leads to further downregulation of Fgf8 expression and increased neuronal

differentiation. Conversely, in Raldh2 mutant mice and vitamin A-deficient quail embryos, both of which

are unable to synthesize RA, Fgf8 expression is strengthened and neuronal differentiation is inhibited,

resulting in fewer neurons generated in the spinal cord (Maden et al, 1996; Diez del Corral, 2003;

24

Molotkova et al, 2005). Thus, RA is required to repress FGF expression for proper establishment of the

cervical spinal cord, as well as to induce neural differentiation throughout the spinal cord.

In addition to RA and FGFs, Wnts and TGFβ family members are also important for the

caudalization of neural tissue. Wnt3a mutant animals exhibit expansion of anterior brain structures at the

expense of caudal tissue and ectopic expression of Wnt3a leads cells to adopt caudal neuron identities

(Kiecker & Niehrs, 2001; Liu et al, 1999; Nördstrom et al, 2002). Likewise, growth differentiation factor 11

(Gdf11), a TGFβ family member protein, is required in conjunction with high levels of FGF to ensure

proper Hox gene expression (e.g. Hoxc10) at the most caudal regions of the spinal cord (Liu et al, 2001).

Dorsoventral spinal cord patterning

Coincident with the transformation of PNT cells to spinal cord neuron progenitors is their

newfound competence to respond to graded BMP and Shh signals produced from the dorsal and ventral

poles of the spinal cord, respectively. Pioneering studies have firmly established the importance of graded

BMP and Shh signaling in dorsoventral patterning of spinal progenitor domains (Fig 1.4) (Jessell, 2000;

Briscoe & Ericson, 2001). These two signals establish anti-parallel gradients that control the expression of

HD and bHLH TFs, the combinatorial expression of which will induce and maintain eleven progenitor

domains positioned along the dorsoventral axis of the spinal cord from which the basic cell types of the

spinal cord will be generated. Cells exposed to BMP signals early on generate neural crest cells (NCC),

which produce SNs of the dorsal root ganglia, among other structures, while later exposure to BMP

signals generates six dorsal progenitor domains (dP1-dP6) that produce spinal dorsal INs (dI1-dI6). (Lai

et al, 2016). At the ventral pole of the neural tube, differential exposure to Shh specifies four IN progenitor

domains (p0-p3); the MN progenitor domain (pMN); and floor plate cells, which are non-neuronal cells

that take over production of Shh (Briscoe et al, 2000; Alaynick et al, 2011). While BMP and Shh constitute

the primary drivers of dorsal and ventral fates, respectively, signaling by other TGFβ family members,

Wnts and RA have also been shown to contribute to dorsoventral patterning. In the following sections, I

will discuss the molecular mechanisms underlying BMP/Shh-mediated patterning, including models of

how these signals specify distinct neuronal populations in the dorsal and ventral spinal cord.

25

TGFβ signaling pathway

BMPs are members of the TGFβ protein superfamily of highly conserved, secreted extracellular

ligands. In addition to BMPs, the TGFβ family includes TGFβs, activins, inhibins, and Gdfs (Liu &

Niswander, 2005). Signaling by this family of proteins is critical for normal embryonic development as well

as maintenance of adult tissues, and involve the control of processes as fundamental as cellular growth,

differentiation, and apoptosis (Kingsley, 1994; Hogan, 1996). Within a subfamily, these ligands bind to

different type I or type II transmembrane serine/threonine kinase receptors (Wang et al, 2014; Liu &

Niswander, 2005). In the canonical TGFβ signaling pathway, the binding of different TGFβ ligands

induces the formation of heterotetrameric complexes comprised of two homo- or heterodimers of type I

and type II receptors. The constitutively active type II receptor phosphorylates the type I receptor, leading

to the phosphorylation and specific activation of different members of the Smad family of proteins. In

particular, ligand binding first activates the receptor-regulated Smads (R-Smads), including Smad1/5/8 for

BMP receptors and Smad2/3 for other TGFß signals. The R-Smads subsequently associate with the

common-mediator Smad (co-Smad) Smad4 to form a multimeric complex that translocates to the nucleus

where it interacts with transcriptional coactivators and corepressors to regulate gene expression (Liu &

Niswander, 2005). Importantly, TGFβ signaling is modulated in a variety of ways, including through the

expression of extracellular antagonists such as follistatin, noggin and chordin, that block the interaction of

the ligand with the receptor. The expression of these antagonists is regulated by TGFβ signaling itself in a

feedback loop that refines the duration and strength of the signal. Inhibitory Smads (I-Smads) such as

Smad6 and Smad7 compete with Smad1/5/8 for binding of the activated type I receptor and thus are also

involved in feedback inhibition of this signaling pathway (Wang et al, 2014; Liu & Niswander, 2005).

Development of the dorsal spinal cord

Classical embryological studies have demonstrated that constitutive signaling by the TGFβ family

member BMP4 functions to repress neural differentiation in ectodermal cells (Munoz-Sanjuan &

Brivanlou, 2002). Studies in the frog Xenopus laevis later revealed that inhibition of BMP4 signaling by

factors emanating from the “organizer” region of the embryo was sufficient to induce neural differentiation

of ectodermal cells. Indeed, gain- and loss-of-function studies showed that these factors – follistatin,

26

noggin, and chordin – each act to bind and inactivate BMP4 signaling in order to promote neural

differentiation and repress ectodermal differentiation (Spemann & Mangold, 1924; Hemmati-Brivanlou &

Melton, 1992; Hemmati-Brivanlou & Melton, 1994; Hemmati-Brivanlou et al, 1994; Zimmerman et al, ,

1996).

In addition to its role in neural induction, BMP signaling provides positional information in the

dorsal neural tube, including in the spinal cord (Lee & Jessell, 1999). In mice, the six early-born dorsal

progenitor domains are spatially organized such that dI1 INs are born closest to the developing roof plate,

with the dP2 domain situated just ventral to the dP1, and so on. Genetic ablation of the roof plate, one of

the primary reservoirs for BMP signals in the dorsal spinal cord along with the epidermal ectoderm,

results in (1) failure to properly pattern dorsal progenitor domains, (2) absence of dI1 and dI2 INs

normally generated from those domains, and (3) shift of dI3 INs to more dorsal regions (Millonig et al,

2000; Lee et al, 2000). Based on their timing and pattern of expression in the roof plate, BMP2, BMP4

and BMP7 proteins are specifically implicated in this process, as well as several other TGFβ family

members (Gdf7, activin, and dorsalin) (Liem et al, 1995; Lee et al, 1998).

Experiments in chick and mouse embryos provide abundant evidence that TGFβ signaling,

especially BMP signals, is required and sufficient to induce dorsal fates in the spinal cord. In chick spinal

cord explant cultures, the addition of BMP4/7 or ActivinA is sufficient to induce dI1 and dI2 IN identity

(Liem et al, 1997). Furthermore, overexpression of constitutively active BMP receptors in mice induces

dorsal IN gene expression programs and prevents the generation of intermediate and ventral spinal

neuron types (Panchison et al, 2001; Timmer et al, 2002). Conversely, BMP receptor mutants exhibit loss

of dI1 INs, with positional shift of dI2 and dI3 classes into more dorsal domains (Chesnutt et al, 2004;

Wine-Lee et al, 2004; Tozer et al, 2013). Accordingly, the generation of dI1 and dI2 INs is also inhibited

by treatment with the TGFβ antagonists noggin and follistatin (Liem et al, 1997; Lee et al, 1998).

The current model for dorsal neuron specification is that a diffusible gradient of BMP signaling

emanating from the epidermal ectoderm and roof plate progressively specifies dorsal spinal neuron fates.

Specifically, exposure of spinal progenitors to different levels of BMP signaling promotes the progressive

generation of dI1-3 spinal INs, such that cells exposed to high concentrations of BMP acquire dI1 IN fate

while lower BMP concentrations induce dI3 IN identity (Liem et al, 1997; Timmer et al, 2002; Tozer et al,

27

2013). Indeed, the specific dose of BMP signaling appears to be critical for establishing the proper size of

dorsal IN domains. Strong activation of BMP signaling in the chick dorsal spinal cord causes a decrease

in the number of dI2 INs, while weaker activation of BMP signaling causes ventral expansion of the dP2

progenitor domain (Timmer et al, 2002). While the formation of the more intermediate dI4-6 spinal

interneuronal populations is thought to be independent of BMP signaling since these domains are largely

intact in BMP signaling mutants, some evidence indicates that intermediate dorsal IN types are also

sensitive to changes in BMP concentrations. Zebrafish BMP mutants have an increase of Lhx1/5-

expressing dI4 INs, but the injection of the BMP antagonist chordin in these mutants causes a decrease

in dI4 INs, suggesting that intermediate spinal cell types such as dI4 INs may also be generated by an

intermediate level of BMP signaling (Lee et al, 2000; Gross et al, 2002; Nguyen et al, 2000).

In addition to spatial patterning, the duration of exposure to BMP signaling in dorsal progenitors

also influences the formation of distinct dorsal spinal neuron subtypes, suggesting that dorsal progenitors

are also temporally specified in response to BMP signals (Tozer et al, 2013). As mentioned earlier, at

early stages of embryonic development neural progenitors respond to BMP signaling by adopting NCC

fate while later BMP exposure promotes the generation of dorsal spinal INs (Liem et al, 1997). More

recent studies in mice and chick have shown that dorsal progenitors are sequentially fated to acquire dI3,

dI2, and finally dI1 identity through exposure to increasing levels of BMP signaling. Furthermore, when

BMP signaling is blocked early, progenitors fated to become dI1 become dI3 INs instead, but late

blockade results in a switch from dI1 to dI2 identity (Tozer et al, 2013).

Other TGFβ family members, especially Activin and Gdf7, are also involved in dorsal spinal

patterning (Timmer et al, 2005; Zechner et al, 2007). Ex vivo treatment of neural explants with Activin

ligands results in the repression of ventral genes and induction of dI1 and dI2 IN markers (Liem et al,

1995; Liem et al, 1997; Pituello et al, 1995). Subsequent studies using in ovo electroporation of a

constitutively active form of the Activin receptor ActR-1B demonstrated that Activin signaling was

sufficient to promote the differentiation of dI3 INs (Timmer et al, 2002). Likewise, Gdf7 induces

differentiation of dI1 INs specifically, while Gdf7 mouse mutants lose the dorsal-most subclass of dI1 INs

(dI1A) (Lee et al, 1998). Importantly, these studies show that the effects of Activin and Gdf7

28

manipulations is distinct from BMP-mediated patterning, suggesting that TGFβ family members have non-

redundant functions in the differentiation of dorsal IN cell types.

Non-TGFβ factors are also involved in the specification of dI1-3 INs, specifically the Wnt family of

signaling molecules. Like BMP and other TGFβ family members, Wnt1 and Wnt3a are thought to be

secreted in a gradient from the roof plate, acting as morphogens to control dorsal IN specification

(Muroyama et al, 2002; Hirabayashi et al, 2004). In particular, Wnt1/3a signals are proposed to interact

with BMPs to control the expression of Msx1/2 and Olig3, TFs that are required for the specification of

dI1-3 INs (Müller et al, 2005; Zechner et al, 2007). Mice lacking both Wnt1 and Wnt3a exhibit significant

reduction in the development of dI1 and dI2 INs with increase in dI3 INs, while overexpression of Wnt3a

protein induces dI1 and dI2 INs in chick neural explants (Muroyama et al, 2002; Hirabayashi et al, 2004).

In addition to its role in patterning, Wnt signaling has also been shown to be important for the proliferation

of dorsal progenitors since misexpression of Wnts results in massive enlargement of the dorsal spinal

cord while removal of Wnt signaling causes significant reductions in the size of dorsal progenitor domains

(Dickinson et al, 1994; Megason & McMahon, 2002; Chesnutt et al, 2004; Xie et al, 2011). Current

evidence suggests that canonical Wnt signaling regulates patterning and proliferation independently in

the dorsal spinal cord, and that these processes are separable and differentially dependent on distinct

Wnt target genes (Bonner et al, 2009).

The binding of TGFβ and Wnt ligands with their cognate receptors leads to the regulation of HD

TFs, in particular Pax3, Pax7, Msx1, and Msx2, which initially pattern the developing dorsal spinal cord.

For example, Msx1/2 genes are thought be directly regulated by BMP signaling, and are themselves

expressed in a gradient-like pattern dorsoventrally, providing additional evidence of the morphogenetic

function of BMP signaling Lee et al, 1999; Caspary & Anderson, 2003). Subsequently, BMP signaling

helps to set the expression boundaries of several proneural bHLH TFs, the combinatorial expression of

which define the six dorsal progenitor domains and are required for their differentiation into mature INs.

Atoh1 is expressed in the progenitor cells adjacent to the roof plate (dI1); Ngn1 and Ngn2 are expressed

in dI2 progenitors; and Ascl1 is expressed in dI3-dI5 precursors, while expression of another bHLH factor,

Ptf1a, is restricted to dP4 progenitors (Pierani et al, 1999; Kriks et al, 2005; Ma et al, 1997; Gowan et al,

2001; Helms et al, 2005; Muller et al, 2005; Glasgow et al, 2005; Alaynick et al, 2011; Lai et al, 2016). In

29

addition to bHLH factor regulation, BMP signaling also regulates the expression of HD TFs in postmitotic

cells, especially Lbx1, the expression of which divides dorsal spinal INs into two classes: Class A (dI1-3)

INs, which do not express Lbx1 and are generated in response to BMP signaling versus Class B (dI4-6)

INs, which express Lbx1 and are not BMP-responsive. Accordingly, misexpression of Lbx1 blocks the

differentiation of Class A neurons, while Lbx1 mutant mice exhibit defects in class B neuron specification,

as well as aberrant dorsal spinal circuitry (Müller et al, 2002; Gross et al, 2002).

Cross-repression between the bHLH factors further refines the borders of adjacent progenitor

domains: in Atoh1 mouse mutants, dI1 INs are lost while Ngn1 expression is expanded dorsally and

excess dI2 INs are produced (Gowan et al, 2001). Moreover, in Ptf1a-deficient mice, dI4 INs are absent

but dI5 INs are generated in excess due to unopposed Ascl1 activity (Glasgow et al, 2005; Mizuguchi et

al, 2006; Wildner et al, 2006). As mentioned earlier, Lbx1 is involved in the generation of dI4-6 INs, partly

through its regulation of the TF Pax2. Pax2, however, is only expressed in dI4 and dI6 INs and not dI5

cells since it is required for GABAergic inhibitory neuronal fate. Recent studies show that Tlx3, which is

expressed in excitatory dI5 INs, represses Lbx1, resulting in specific inhibition of Pax2 in dI5 cells. Thus,

Lbx1 and Tlx3 function as opposing switches specifying excitatory versus inhibitory neuronal fate in these

dorsal neuron populations (Cheng et al, 2004; Cheng et al, 2005).

The end result of these molecular specification programs is the generation of six non-overlapping

dorsal progenitor domains, from which six distinct types of dorsal INs originate. These dorsal IN

populations differ widely in terms of their TF expression programs, use of excitatory or inhibitory

neurotransmitters and neuropeptides, settling positions in the dorsal and intermediate spinal cord, axon

trajectories, and choice of synaptic partners (Alaynick et al, 2011; Lai et al, 2016). Later, I will discuss the

development of a specific class of dorsal cells, the Ptf1a-derived dI4 INs, as an access point to

understanding its function in somatosensory circuits. In the following section, I will describe how different

ventral neuron cell types are generated in response to another secreted morphogen, sonic hedgehog.

Sonic hedgehog signaling

In the ventral spinal cord, the BMP-equivalent signal is Shh, which is produced in a gradient by

the notochord then by floor plate cells and is critical for the formation of the five classes of ventral spinal

30

progenitors: p0, p1, p2, pMN, p3 (Briscoe & Ericson, 2001; Ribes & Briscoe, 2009). In vitro and in vivo

experiments have shown that Shh is both necessary and sufficient to induce the formation of ventral

neuron cell types (Roelink et al, 1995; Marti et al, 1995; Ericson et al, 1996; Chiang et al, 1996; Briscoe et

al, 2001). Like the BMPs, Shh acts in a concentration-dependent manner such that cells generated at the

most ventral regions of the neural tube (e.g. floor plate cells and p3 progenitors) require higher

concentrations of Shh for their differentiation compared to cells generated at more dorsal positions (e.g.

p0 and p1 progenitors) (Ericson et al, 1997; Briscoe et al, 2000). Indeed, in chick neural explants, two-to-

threefold increases in the concentration of recombinant Shh and/or increasing the duration of exposure to

Shh signaling progressively switches the identity of cells towards more ventral fates (Ericson et al, 1997;

Briscoe et al, 2000; Dessaud et al, 2007).

The intracellular transduction of Shh signaling is mediated by two transmembrane proteins,

Patched1 (Ptc1) and Smoothened (Smo). Shh binding to Ptc1 receptor releases the inhibition of Smo,

leading to the regulation of Gli TFs, the activators Gli1 and Gli2 and the repressor Gli3, which together

control the expression of Shh target genes (Matise & Joyner, 1999; Jacob & Briscoe, 2003; Vokes et al,

2007; Vokes et al, 2008). Activated Smo initiates downstream processes that switch the balance from

predominantly repressive Gli activity to transcriptionally active Gli, resulting in the expression of HD

factors controlling ventral progenitor identity (Briscoe & Ericson, 2001). Indeed, Smo and Gli activity are

both required and sufficient for the response of ventral neural cells to graded Shh signals (Wijgerde et al,

2002; Zhang et al, 2001; Bai et al, 2004; Lei et al, 2004; Dessaud et al, 2008).

Ventral spinal cord patterning

In the classical view of ventral spinal patterning, HD TFs induced by the Shh-Ptc1-Smo-Gli

signaling axis can be divided into two classes based on their regulation by Shh: Class I genes are

expressed more dorsally and are repressed by high levels of Shh signaling, while Class II genes are

expressed more ventrally and are activated by high levels of Shh signaling. As such, the combinatorial

expression of these HD TFs defines each of the ventral progenitor domains and also acts to specify the

identity of the neuronal cell types produced by each domain (Briscoe et al, 2000; Briscoe et al, 1999).

Cross-repressive interactions between neighboring Class I and Class II proteins delineate the boundaries

31

between neighboring progenitor domains and provide unequivocal assignment of ventral neuronal identity

(Ericson et al, 1997; Briscoe et al, 2000; Chen et al, 2011). For example, Dbx2 (Class I) and Nkx6.1

(Class II) genes, which are expressed in adjacent progenitor domains, are engaged in cross-repressive

interactions to control the ventral boundary of the p1 progenitor domain. In Nkx6.1 mouse mutants, Dbx2

expression extends ventrally, leading to selective loss of MNs and V2 INs, which are normally generated

from progenitor domains expressing Nkx6.1; and increased generation of Dbx2-derived V1 INs due to the

ventral expansion of the p1 progenitor domain (Briscoe et al, 2000; Sander et al, 2000).

Shh-independent signaling pathways have also been implicated in acquisition of distinct ventral

neuronal cell fates. More recent studies have shown that retinoid and FGF signals are also crucial to

establishing proper ventral spinal patterning. Indeed, the different timing of Shh, RA, and FGF signals

during spinal cord development contributes to the control of Class I and II HD protein expression, with

early FGF signals repressing both Class I and II HD proteins, while later RA signaling counteracts Fgf8

expression to switch on high-level expression of Class I HD proteins, which is counteracted by focal,

repressive Shh signals to induce Class II HD protein expression (Novitch et al, 2003).

Furthermore, a specific role for RA signaling in ventral spinal patterning has been described for

the generation of the most dorsal ventral spinal domains. Retinoids secreted from the paraxial mesoderm

can function in a Shh-independent manner to specify the p0 and p1 progenitor domains (Pierani et al,

1999). Evidence for the role of RA in ventral patterning comes from mouse embryological studies

examining the development of neurons arising from Dbx1/2-expressing progenitor domains, p0 and p1,

which are the most dorsal domains in the ventral neural tube and which generate V0 and V1 INs. These

studies show that while Shh is sufficient to induce the p0 and p1 progenitor domains, Shh signaling is not

required for the differentiation of V0 and V1 INs. Accordingly, the use of vitamin A-deficient quail

embryos, which cannot synthesize RA, or inhibition of RA signaling in chick neural tube by

overexpression of dominant-negative RA receptors shows that RA is also required for the specification of

V0 and V1 INs. Still, some V0 INs can be generated in animals completely lacking Shh signaling (i.e.,

Smoothened knockouts), suggesting that specification of the p0 domain relies on other signaling

pathways, including RA (Wijgerde et al, 2002; Pierani et al, 1999; Wilson et al, 2004). It is clear, though,

that the generation of V1 INs, meanwhile, requires the confluence of RA and Shh signaling.

32

In addition to RA, BMP and FGF signals are also involved in ventral neuron specification.

Zebrafish embryos mutant for various BMP factors exhibit dorsally-expanded ventral progenitor domains

(Barth et al, 1999; Nguyen et al, 2000). Additionally, neural progenitor cells exposed to increasing

concentrations of BMP in the context of fixed Shh levels acquire increasingly dorsal neuronal fates,

suggesting that BMPs function to limit the graded activity of Shh in the ventral spinal cord (Liem et al,

2000). FGF signals, on the other hand, have been shown to repress both Class I and Class II TFs when

expressed at high levels. Furthermore, FGF can interact with RA and Shh signaling in a context-

dependent manner to regulate the generation of the most ventral neuron types, V3 INs and MNs,

suggesting that FGF signals might also be involved in ventral spinal cord patterning (Novitch et al, 2003).

The convergence of these signaling pathways, especially Shh-mediated patterning, divides the

ventral neural tube into five sharply delineated progenitor domains marked by distinct TF codes. As in the

dorsal spinal cord, each of these ventral progenitor domains generates discrete classes of postmitotic

neurons with distinct molecular and functional characteristics. Next, I will discuss the specific

development of spinal MNs as the prototype for understanding how spinal neuron subtype diversity is

generated and subsequently co-opted in the construction of functionally distinct spinal circuits.

Motor neuron subtype differentiation

During ventral spinal cord development, the progenitor domain giving rise to MNs expresses the

HD proteins Pax6 and Nkx6.1, as well as the bHLH factor Olig2. Newly postmitotic neurons from this

domain express a distinct set of HD factors, including Hb9, Isl1/2, and Lhx3, the combinatorial expression

of which endow them with MN-specific identity (Briscoe et al, 2000; Novitch et al, 2001; Tanabe et al,

1998; Tsuchida et al, 1994, Dasen, 2009). Although spinal MNs as a class share a core set of features,

especially cholinergic neurotransmission and projection of axons outside the CNS to innervate muscles,

they can also be subdivided into distinct subclasses based on differences in gene expression, position

within the spinal cord, and patterns of connectivity (Catela et al, 2015; Dasen, 2009).

The acquisition of MN subtype identity begins at the earliest stages of spinal cord development.

To ensure the precise control of the hundreds of anatomically and functionally distinct muscle groups

distributed throughout the body, different sets of MNs are specified at cervical, brachial, thoracic and

33

lumbar regions of the developing spinal cord. In particular, these discrete types of MNs are

topographically organized into motor columns within the spinal cord according to their projection patterns

(Fig 1.5). For example, at cervical levels of the spinal cord, MNs expressing the LIM-domain TF Lhx3 can

be found in the medial motor column (MMC), which innervates body axial muscles, or phrenic motor

column (PMC), which specifically innervates the diaphragm. At brachial and lumbar levels of the spinal

cord, MMC MNs are generated as well as MNs innervating limb muscles. These limb-innervating MNs

reside in the lateral medial column (LMC), downregulate Lhx3 expression, and are uniquely marked by

FoxP1 TF expression (Dasen et al, 2008).

The generation of different sets of MNs at different rostrocaudal levels is controlled by Hox

transcriptional networks (Dasen et al, 2003; Dasen et al, 2005). As mentioned in the spinal cord

patterning section, signaling gradients of RA, FGFs and Wnts are integrated to established Hox gene

expression domains, the boundaries of which are sharpened by cross-repressive interactions between

pairs of Hox genes (Dasen et al, 2005; Wu et al, 2008; Jung et al, 2010). The ability of Hox TFs to control

MN subtype differentiation depends also on the activity of other TFs acting in parallel with the Hox genes.

In the absence of FoxP1, a putative cofactor for Hox transcriptional activity, Hox-dependent steps of MN

columnar differentiation are lost, resulting in reversion of MN identities back to an ancestral MN subtype

identities (Dasen et al, 2008; Rousso et al, 2008). Interestingly, some Hox activities in MNs are

unaffected in FoxP1 mutant animals, indicating that other factors control these specific functions. Indeed,

Pbx proteins were recently shown to be required for the specification and peripheral connectivity of Hox-

dependent MN subtypes, as well as for the columnar organization of spinal MNs (Hanley et al, 2016).

Hox-dependent pathways also function to segregate LMC MNs into distinct subgroups based on

their axonal trajectories and innervation of specific limb muscles (Dasen, 2009; Catela et al, 2015). First,

LMC MNs are subdivided into lateral (LMCl) and medial divisions (LMCm) based on their innervation of

dorsal and medial muscles, respectively (Kania et al, 2000; Kania & Jessell, 2003; Bonanomi & Pfaff,

2010). Hox programs establish the expression of LIM homeodomain proteins Lhx1 and Isl1 which act to

segregate these LMC subtypes (Tsuchida et al, 1994; Kania & Jessell, 2003; Alaynick et al, 2011). In

particular, the specification of the LMCl subtype depends on the retinoid signaling produced by LMCm

neurons, which selectively express the Raldh2 enzyme for RA synthesis (Sockanthan & Jessell, 1998).

34

Unlike the Raldh2 expressed in the paraxial mesoderm, which is present throughout the rostrocaudal axis

of the spinal cord, the specific expression of Raldh2 in lateral LMC neurons is controlled by FoxP1-

dependent Hox activity. Accordingly, FoxP1 mutants exhibit significant reduction in Raldh2 levels

specifically in LMC MNs (Dasen et al, 2008; Rousso et al 2008). During MN genesis, later-born LMCl MN

progenitors differentiate and migrate away from the VZ past earlier-born medial LMC MNs and in doing so

receive RA signals from this Raldh2-expressing population. The transduction of RA signals in lateral LMC

neurons leads to the induction of Lhx1 TF expression specifying LMCl fate and restricting Isl1 expression

to LMCm MNs (Kania & Jessell, 2003; Sockanathan & Jessell, 1998). The expression of these LIM HD

proteins in LMC neurons appears to regulate genetic programs controlling expression of guidance

molecules and receptors such as the Eph/ephrin, semaphorin/plexin, and GDNF/ret pathways (Kania &

Jessell, 2003; Huber et al, 2005; Kramer et al, 2006a; Bonanomi & Pfaff, 2010).

The second level of LMC MN diversification controlled by Hox genetic programs is the

organization of LMC neurons into ~50 distinct motor pools innervating select muscle groups in the limb. In

particular, each motor pool occupies a characteristic spatial position in the ventral spinal cord that reflects

MN-muscle topography: motor pools innervating the proximal muscles in each limb are positioned more

rostrally within the LMC whereas pools innervating distal muscles are located at more caudal levels

(Hollyday & Jacobson, 1990; Landmesser, 1978b). The clustering of LMC neurons into motor pools and

the establishment of pool-specific muscle connectivity is directly controlled by Hox TF networks regulating

the expression of motor-pool specific TFs (Dasen et al, 2005; De Marco Garcia & Jessell, 2008; Dasen,

2009). Furthermore, as has been shown for the establishment of MN segmental and columnar identity,

cross-repressive interactions between Hox paralogs function to define and maintain specific motor pool

identity (Dasen et al, 2005; Dasen & Jessell, 2009).

While cell-intrinsic programs such as Hox and LIM-domain TF regulatory networks have a major

role in the subtype differentiation of MNs, other mechanisms are also required to create the immense

diversity of MN subtypes evident in the mature spinal cord. Examples include the role of non-canonical

Wnt signaling in the choice between MMC and hypaxial motor column (HMC) fate in the thoracic spinal

cord, as well as the requirement of limb muscle-derived neurotrophic signals for specification of certain

motor pools (Agalliu et al, 2009; Haase et al, 2002; Lin et al, 1998). Furthermore, as in other regions of

35

the CNS, Notch signaling is suggested to be involved in subtype differentiation in the spinal cord, and in

MNs specifically (Yeo & Chitnis, 2007; Skaggs et al, 2011). Indeed, in the p2 progenitor domain of the

ventral spinal cord, which generates V2 INs, mutations of Notch signaling components indicate that Notch

signaling is required for acquisition of V2b over V2a cell fate (Del Barrio et al, 2007; Peng et al, 2007;

Rocha et al, 2009; Ramos et al, 2010 Marklund et al, 2010). In the pMN domain, Notch signaling had

been proposed to control MN subtype identity by controlling the timing of neurogenesis (Crawford &

Roelink, 2007; Sabharwal et al, 2011). More recently, Notch signaling has been shown to be required for

the specification of MMC MN identity; blocking Notch signaling either genetically or pharmacologically

results in the specification of HMC identity under rostralizing conditions and medial LMC identity in

caudalizing conditions (Tan et al, 2016). Together, these data suggest that Notch signaling is involved in

the control of MN subtype identity and the generation of neuronal subtype diversity in the spinal cord.

Over the past several decades, we have learned a great deal about the cell-intrinsic and extrinsic

programs controlling spinal cord patterning along the rostrocaudal and dorsoventral axes; induction of the

basic spinal neuronal cell types; and, in the case of MNs, specification of molecularly and functionally

diverse subtypes. Whether spinal IN domains are similarly diversified through comparable mechanisms is

not known. In the final section, I will propose to study the development of spinal IN subtype diversity by

examining the specification of two distinct classes of spinal INs mediating the monosynaptic reflex circuit.

D. Monosynaptic stretch reflex circuits

In its most basic configuration, the monosynaptic stretch reflex circuit consists of a MN, a pSN,

and a muscle target in the periphery (Sherrington, 1906; Eccles et al, 1957). Group Ia pSNs have two

projections: an efferent axon that synapses peripherally onto muscle spindles to detect stretch, and an

afferent axon that synapses in the ventral spinal cord onto MNs innervating the same muscle. The central

projections of pSNs onto homonymous MNs in the spinal cord are established with remarkable specificity,

even in the absence of activity, suggesting that these circuits are genetically hardwired (Eccles et al,

1957; Mears & Frank, 1997; Frank, 1990; Mendelson & Frank, 1991). In recent years, a growing body of

work has uncovered some of the underlying mechanisms for establishing the precise connectivity of pSNs

36

onto MNs, including the role of the guidance molecules semaphorin/plexin, coordinated expression of

motor pool-specific TFs such as Pea3, and position-dependent programs involving FoxP1 (Pecho-

Vrieseling et al, 2009; Vrieseling & Arber, 2006; Dasen et al, 2008; Surmeli et al, 2011).

Spinal cord INs regulate the output of the monosynaptic reflex circuit by facilitating or impeding

the excitatory signals sent by pSNs onto MNs. In my thesis, I will focus on the role of inhibitory INs in

modulating this particular sensory-motor circuit: dorsally-derived GABApre INs that act to filter sensory

excitatory transmission onto MNs; and ventrally-derived V1 inhibitory INs which synapse directly onto

MNs to modulate the duration and degree of MN excitability, including a distinct group of V1 INs

mediating recurrent inhibition of MNs to provide feedback inhibition (Renshaw, 1941).

Proprioceptive sensory neurons

Dorsalizing BMP signals early during spinal cord development promote the formation of neural

crest cells, which delaminate from the developing neural tube and migrate away to generate the dorsal

root ganglion (DRG). Extrinsic signals emanating from the adjacent somites and spinal cord and intrinsic

TF programs induces the differentiation of multiple types of SNs in the DRG, including nociceptive,

mechanoreceptive, and proprioceptive subtypes (Marmigère & Ernfors, 2007; Lallemend & Ernfors,

2012). Proprioceptive sensory neurons are generated in the first wave of sensory neurogenesis, which

gives rise to large, myelinated DRG neurons that include pSNs marked by the expression of TrkC, the

receptor for neurotrophin-3 (NT-3); as well as mechanoreceptors expressing TrkB, the receptor for brain-

derived neurotrophic factor (BDNF) or Ret, the receptor for glial-derived neurotrophic factor (GDNF). The

second and third waves of sensory neurogenesis, meanwhile, produce non- or thinly myelinated

nociceptors, which express TrkA, the receptor for nerve growth factor (NGF) (Liu & Ma, 2011).

The proneural activity of the TFs Ngn1 and Ngn2 is required for the formation of DRG neurons,

with early specification of proprioceptors being chiefly dependent on Ngn2 (Ma et al, 1999). Newly born

SNs turn off these neurogenic factors and express the homeobox TFs, Pou4f2 (also known as Brn3a) and

Isl1, which are required for their terminal differentiation and to suppress non-DRG cell fates, including

dorsal spinal IN and other mesodermal tissues (Lanier et al, 2009; Sun et al, 2008; Liu & Ma, 2011). Early

during sensory neurogenesis, some neurons fated to become TrkB mechanoreceptors also express TrkC

37

receptor. The consolidation of proprioceptive identity requires expression of the runt-domain TF Runx3 in

pSNs as well as retrograde NT-3 signals derived from their muscle target in the periphery.

In particular, Runx3 expression appears to be required to repress the TrkB lineage, either by

inhibiting TrkB receptor expression itself or by suppressing mechanoreceptor-specific TFs such as Shox2.

Furthermore, Runx3 mutant animals fail to express other proprioceptive-specific molecular markers, such

as the ETS TF Er81 (Etv1) and the calcium-binding protein parvalbumin, suggesting that Runx3 is

required for full differentiation of pSNs identity (Levanon et al, 2002; Inoue et al, 2002; Inoue et al, 2007;

Nakamura et al, 2008). Er81 in particular is required for the formation of central connections between

pSNs and MNs in the ventral spinal cord as loss of Er81 results in the failure of proprioceptive afferents to

project past the intermediate zone and into the ventral spinal cord (Arber et al, 2000). Importantly, Runx3

is not sufficient to drive pSN identity, since overexpression of Runx3 alone results in only modest

increase in TrkC neurons compared to the dramatic reduction of TrkB/Ret neurons (Kramer et al, 2006).

NT-3 signaling is also essential for pSN identity and function since NT-3 is required for the

survival of pSNs and for the proper targeting of their central projections (Ernfors et al, 1994). In mice

mutant for both the proapoptotic gene Bax and NT-3, pSNs survive to the first postnatal week, but due to

loss of Er81 expression, peripheral axons of pSNs and the associated muscle spindles do not develop

properly while central projections of the pSNs terminate inappropriately in the intermediate spinal cord

and do not send collaterals into the ventral spinal cord, resulting in the failure to form normal

monosynaptic stretch reflex circuits, a phenotype similar to what is seen in Er81-/- mice (Patel et al, 2003;

Arber et al, 2000).

GABApre interneuron development

Inhibitory INs formed in the dorsal spinal cord gate the duration and magnitude of excitatory

inputs coming from SNs, including pSNs (Frank & Fuortes, 1957; Eccles et al, 1961). A subset of these

dorsally-derived INs have been proposed to mediate presynaptic inhibition of primary sensory afferents

onto MNs in the ventral spinal cord, which functions to indirectly dampen motor output (Betley et al,

2009). GABApre INs form axo-axonic synapses with sensory afferent terminals in the ventral spinal cord,

38

and in doing so control sensory gain as a means to suppress motor oscillations and ensure smooth

movement during goal-directed reaching behaviors (Fink et al, 2014).

GABApre INs can be distinguished from other inhibitory INs by their dual expression of the

GABA-synthesizing enzymes, glutamic acid decarboxylase-67 (Gad67) and glutamic acid decarboxylase-

65 (Gad65), which are expressed in the cytosol and synaptic vesicles of GABAergic neurons, respectively

(Betley et al, 2009; Hughes et al, 2005). Interestingly, the molecular and morphological maturation of

GABApre synapses requires secreted BDNF selectively produced from pSN terminals (Betley et al,

2009). Recent studies propose that the specific connectivity of GABApre INs with pSNs is directed by

molecular recognition signals between the IN and pSN. Expression of the IgSF protein Contactin5/NB2

and its co-receptor Caspr4 by the SN is complemented by CHL1/NrCAM ligands expressed on the

GABApre IN to mediate formation of GABApre synaptic boutons onto pSN terminals (Ashrafi et al, 2014).

This program appears to be activity-dependent, as glutamate release from pSN terminals is required to

activate BDNF release for the regulation of Gad65 expression in GABApre terminals (Mende et al, 2016).

Despite our increasing understanding of the molecular underpinnings of GABApre IN synaptic

specificity within the context of the monosynaptic stretch reflex circuit, the developmental programs

differentiating GABApre INs from other dorsally-derived inhibitory INs are largely unknown. In the dorsal

spinal cord, GABAergic and glycinergic inhibitory INs are generated from the Ptf1a-expressing dP4

progenitor domain, which produces both early-born (embryonic day e10-11.5) dI4 INs and later-born

(e12-14.5) dILA INs (Glasgow et al, 2005). Early-born dI4 INs migrate from the dP4 progenitor domain

and settle nearby in the intermediate zone of the spinal cord, with a subset projecting axons to the ventral

spinal cord where they synapse onto pSN afferent terminals (Glasgow et al, 2005; Betley et al, 2009).

Meanwhile, dILA INs migrate dorsally to the most superficial layers of the spinal cord where they intermix

with dP5 domain-derived excitatory dILB neurons. Here, dILA neurons presumably provide inputs onto

other dorsal spinal INs and cutaneous sensory afferents that terminate in these zones (Wildner et al,

2006; Betley et al, 2009; Wildner et al, 2013). How dI4 and dILA inhibitory IN subtypes are differentially

specified from the same progenitor domain during dorsal spinal cord development is not well understood.

Furthermore, although GABApre INs are likely a subset of early-born dI4 INs based on their positioning in

39

the intermediate spinal cord and their preferential synaptic connectivity with pSNs, how GABApre INs are

distinguished from other dI4-derived neurons is also unknown.

Developmental studies relying largely on mouse mutants have demonstrated that the

differentiation of both dI4 and dILA INs depends on a trimeric TF complex containing Ptf1a, RBPJ, and an

E-protein, which is essential for acquisition of GABAergic identity and repression of glutamatergic

programs in the dorsal spinal cord, as well as in retina, hindbrain and cerebellum (Hori et al, 2008; Henke

et al, 2009; Meredith et al, 2009; Borromeo et al, 2014; Glasgow et al, 2005; Hoshino et al, 2005; Pascual

et al, 2007; Fujitani et al, 2006; Nakhai et al, 2007). While Ptf1a expression is dependent on Ascl1,

suggesting that Ptf1a is downstream of Ascl1, the precise relationship between these bHLH factors is not

fully understood in the mammalian dorsal spinal cord (Mazurier et al, 2014; Mizuguchi et al 2006).

Recent studies comparing the phenotypes of Ascl1 and Ptf1a mutant mice indicate that Ascl1 and

Ptf1a may act in opposition to specify excitatory versus inhibitory dorsal spinal neuronal subtypes: Ascl1

mutant exhibit dramatic loss of excitatory dI3, dI5 and dILB IN subtypes, while Ptf1a-/- lose all inhibitory

INs (i.e., Pax2- and Lhx1/5- expressing dI4 and dILA IN subtypes) in the dorsal spinal cord. Previous

studies had also shown that Ascl1 is only required for the generation of late-born dILA neurons and not

early-born dI4 neurons, which express low or no Ascl1. Accordingly, Ascl1 null mutants lose dILA INs as

well as excitatory dI3, dI5, and dILB IN populations in the spinal cord while exhibiting a modest gain in

early-born dI4 INs (Helms et al, 2005; Mizuguchi et al, 2006; Wildner et al, 2006). Indeed, in Ascl1-/- mice

at e18.5, there is a 50% reduction in the number of dorsal spinal GABAergic, Gad67-expressing INs

compared to wildtype controls, a phenotype that is most pronounced in the superficial dorsal horn.

Furthermore, while ~1/3 of dorsal spinal inhibitory neurons also use glycine as a neurotransmitter, these

GlyT2-expressing neurons, which reside mostly in the deep dorsal horn, are relatively well-preserved in

Ascl1-/- animals, providing secondary support to the finding that loss of Ascl1 selectively targets dILA INs

(Wildner et al, 2013). Interestingly, this study did not find any significant changes in the number of

excitatory (Tlx3-expressing) INs in the spinal cord of Ascl1-/- mice, contrary to previous reports (Helms et

al, 2005; Mizuguchi et al, 2006; Chang et al, 2013). Regardless, these studies together provide strong

evidence that while both subsets of dP4-derived INs require Ptf1a for their development, dILA, but not dI4

INs are also dependent Ascl1 for their specification.

40

This insight has been used to identify candidate genetic markers for distinguishing dI4 and dILA

INs (Fig 1.6) (Wildner et al, 2013). In particular, genes downregulated in both Ascl1-/- and Ptf1a-/- mice

are postulated to mark Ascl1-dependent dILA inhibitory INs, while genes downregulated in Ptf1a-/- mice

only should label Ascl1-independent dI4 INs. Microarray expression analysis showed that thirty genes are

significantly downregulated in Ptf1a-/- mice, with 21/30 of these genes also reduced in Ascl1-/- mice.

Importantly, four of these genes, pDyn, Kcnip2, RORβ, and Tfap2b, are expressed in non-overlapping,

layer-like domains in the dorsal horn, with pDyn expressed most superficially, followed by Kcnip2, RORβ,

and Tfap2b ventrally, suggesting that these might represent distinct dI4-derived subtypes beyond simple

division of dI4 and dILA classes. Reassuringly, these results have been corroborated by expression

profiling of superficial compared to deep dorsal horn neurons (J. Kaltschmidt, unpublished).

Most significantly, genes encoding for prodynorphin (pDyn), an opioid polypeptide hormone, and

Kcnip2, a protein that interacts with voltage-gated potassium channels, are restricted to the superficial

dorsal horn and highly downregulated in Ascl1-/- and Ptf1a-/- mice. Two additional genes, Npy, which

encodes for neuropeptide Y, and RORβ, encoding for the retinoic acid related orphan receptor were also

significant downregulated in Ascl1-/- mice, and to a greater extent, Ptf1a-/- animals, suggesting that they

label both early and late-born subsets of dI4 INs. Since their expression is not completely abrogated in

Ascl-/- or Ptf1a-/- mice, it is likely that some excitatory INs also express these genes. Interestingly, Ptf1a-

derived, GABAergic INs destined to target cutaneous sensory afferents in the dorsal horn express NPY

on their synaptic boutons, while GABApre INs targeting pSNs do not express this neuropeptide (Betley et

al, 2009). Conversely, neurons expressing the TF TFAP2b were completely absent in Ptf1a-/- mice but

not affected in Ascl1-/- mice, suggesting that this marker selectively labels at least a subset of early-born

dI4 INs. Furthermore, TFAP2b-expressing neurons reside almost exclusively in the deep dorsal horn

making this gene an intriguing candidate for genetically accessing GABApre INs specifically (Wildner et

al, 2013; Levine et al, 2014; J. Kaltschmidt, unpublished).

Once specified, GABApre INs migrate a short distance away from their progenitor domain and

settle in the deep dorsal horn (Glasgow et al, 2005; Betley et al, 2009). Although their cell bodies persist

in the dorsal spinal cord, they project long-range axons into the ventral horn to synapse onto pSN afferent

terminals located on MN cell bodies and proximal dendrites. Indeed, in spite of the remarkably close

41

proximity of sensory afferent terminals and MNs, pSN terminals are the exclusive synaptic target of

GABApre INs in the ventral spinal cord, with approximated sensory:motor preference of ~2000:1. As

mentioned before, when pSNs are specifically removed from the circuit using genetic tools, GABApre INs

fail to undergo proper synaptic differentiation and withdraw their axons from the ventral spinal cord rather

than form ectopic synapses with other spinal neurons, including MNs (Betley et al, 2009). Thus, GABApre

INs derived from Ptf1a-expressing progenitors provide a distinct mode of inhibition to sensory-motor

circuits through highly specific restriction of presynaptic sensory neurotransmitter release onto MNs.

V1 interneuron subtype diversity

While GABApre INs are an important source of inhibitory signals shaping sensory-motor circuits,

motor output is further sculpted by the activity of other spinal INs providing more conventional,

postsynaptic inhibition of MN excitability (Goulding, 2009; Kiehn, 2016). Ventrally-derived V1 INs

constitute just one-third of all inhibitory cells in the ventral spinal cord but provide ~80% of postsynaptic

inhibitory inputs onto MN cell bodies and proximal dendrites, with the remainder arising from ventral

commissural INs as well as certain dorsal neuronal populations (Betley et al, 2009; Kiehn, 2006; Sapir et

al, 2004; Lanuza et al, 2004; Jankowska, 1992; Tripodi et al, 2011; Wilson et al, 2010). However,

although all V1 INs share a common progenitor and some unifying characteristics, including ipsilateral

projections in the ventral spinal cord and expression of inhibitory neurotransmitters (GABA and/or

glycine), the V1 class is highly heterogeneous, comprising a mixture of dozens of molecularly distinct IN

subtypes (Sapir et al, 2004; Saueressig et al, 1999; Alvarez et al, 2005; Siembab et al, 2010; Benito-

Gonzalez & Alvarez, 2012; Francius et al, 2013; Bikoff et al, 2016; Gabitto et al, 2016). Whether all or

only select V1 IN subtypes provide monosynaptic inhibitory inputs onto MNs is not known. Furthermore,

while the synaptic preferences of some V1 subtypes are known, the overall organizational logic of V1

inhibitory circuits remains obscure. Finally, assuming that molecularly distinct V1 IN subtypes also acquire

specialized synaptic connectivity in spinal circuits, the developmental programs underlying the

diversification of V1 IN subtype identity and function are also largely undetermined.

All V1 INs arise from the p1 progenitor domain in the ventral spinal cord that is delineated by

combinatorial expression of the HD TFs Pax2, Dbx2, and Nxk6.2. Shortly after p1 progenitors exit cell

42

cycle, they express the TF Engrailed-1 (En1) (Burrill et al, 1997; Matise and Joyner, 1997; Ericson et al,

1997; Sauressig et al, 1999; Vallstedt et al, 2001). While En1 is a cell-type-specific marker for V1 INs in

the spinal cord, it is not required for the specification of V1 INs, acting instead as a determinant for V1 IN

axonal guidance and synaptogenesis (Saueressig et al, 1999; Sapir et al, 2004). Recent studies have

shown that the TF Prdm12, a member of the Prdm family of epigenetic zinc-finger transcriptional

regulators, is specifically expressed in p1 progenitors, suggesting that it might be a novel regulator of V1

IN specification (Kinameri et al, 2008; Zannino et al, 2014; Thélie et al, 2015). Indeed, loss- and gain-of-

function studies using frog, chick, and mice showed that Prdm12 is required for En1 expression, and is

sufficient to induce ectopic V1 INs in the spinal cords of frog and chick (Thélie et al, 2015). Studies of

Prdm12 have also revealed a novel functional role for V1 INs in locomotor circuits, as knockdown of

Prdm12 in zebrafish results in impaired escape response and inability to coordinate swimming

movements (Goulding 2009; Higashijima et al, 2004; Li et al, 2004; Zannino et al, 2014). While we cannot

directly compare swimming fish and walking mammals, in the mouse, genetic ablation of En1-expressing

neurons results in elongation of the locomotor step cycle, suggesting that in mammalian circuits, V1 INs

are important for controlling the speed of locomotion (Gosgnach et al, 2006). Nevertheless, these studies

confirm that V1 INs contribute to the central pattern generators in the ventral spinal cord that control

locomotion (Goulding, 2009; Grillner & Jessell, 2009; Kiehn, 2016).

Despite their common origins from Prdm12-expressing progenitors and En1-expressing early

postmitotic cells, mammalian V1 INs are highly heterogeneous and the subtypes responsible for these

behavioral phenotypes have not been isolated. Recent studies from the Jessell lab has shown that the V1

IN class can be subdivided into ~50 molecularly distinct subtypes based on combinatorial expression of

19 TFs significantly enriched in the V1 population compared to dorsally-derived dI4 INs (Fig 1.7) (Bikoff et

al, 2016; Gabitto et al, 2016). These transcriptionally-defined V1 subsets exhibit stereotyped spatial

positioning in the ventral horn reminiscent of the topographical organization of MN pools (Surmeli et al,

2011). Indeed, analysis of two contrasting V1 IN subtypes with distinct dorsoventral settling positions

indicates that the positional segregation of distinct V1 subtypes influences the source and degree of

sensory inputs onto these cell types and potentially also their patterns of motor connectivity, suggesting

43

that V1 IN diversity may have arisen from the need to construct specific microcircuits for controlling

specific muscles (Bikoff et al, 2016).

Renshaw cell development and function

While Bikoff and colleagues have examined the physiological and connectivity differences of a

few molecularly-defined V1 subtypes, V1 IN functional diversity has not yet been extensively catalogued.

However, classic studies using electrophysiology have shown that at least two V1 IN subtypes have very

specialized functions. Although both provide monosynaptic inhibitory inputs to MNs, RCs mediate

recurrent inhibition of MNs while Ia inhibitory INs (IaIN) are involved in reciprocal inhibition of antagonistic

MNs. More specifically, RCs receive excitatory inputs from α-MN axon collaterals and provide inhibitory

inputs to the same MNs and their synergists (Renshaw, 1941; Renshaw, 1946; Eccles et al, 1954; Eccles

et al, 1961). IaINs, meanwhile, receive monosynaptic inputs from group Ia pSN afferents and mediate

reciprocal inhibition of MNs of antagonistic muscles to ensure proper flexor-extensor alternation (Eccles

et al, 1956; Feldman & Orlovsky, 1975; Zhang et al, 2014). The functional connectivity differences

between RCs and IaINs make them ideal cell types for studying the diversification of V1 IN subtype

identity, yet this analysis is complicated by the recent discovery that IaINs are derived from both V1 and

V2b cells (Zhang et al, 2014; Britz et al, 2015). Therefore, in my thesis, I focused on the specification and

synaptic specificity of RCs as an entry point for understanding how V1 IN diversity is established and

manifested in spinal motor circuits.

Although they have been studied since the early 1940s, the precise function of RCs in locomotor

circuits remains the subject of intense speculation (Windhorst, 1996; Alvarez & Fyffe, 2007). The original

hypothesis that RCs act in a simple feedback circuit to prevent excessive MN excitability has largely been

dismissed, in part because RCs also synapse on non-MNs in the spinal cord, including other RCs and

IaINs, as well as receive modulatory inputs from many different cell types, including descending systems

and other RCs (Eccles et al, 1954; Alvarez & Fyffe, 2007; Bhumbra et al, 2014). Currently, the most

appealing hypothesis is that the RC-MN recurrent inhibitory circuit acts as a variable gain regulator to

modulate motor outputs over a wide range of contractile forces, with RCs being largely inhibited during

strong contractions and activated during weak ones (Hultborn et al, 1979; Hultborn & Pierrot-Deseilligny,

44

1979a). Notably, RCs have primarily been studied in the lumbar, or hindlimb, spinal cord; while recurrent

inhibition is thought to act on MNs targeting most limb, neck and diaphragm muscles, a thorough

accounting of RC number and function in cervical, brachial and thoracic spinal cord has not yet been

provided (Hahne et al, 1988; Brink & Suzuki, 1987, Lipski et al, 1985; Saywell et al, 2013). Indeed,

recurrent inhibitory circuits are absent for some MN pools, including those innervating digits (hands and

feet) and the muscles of mastication (Ilert & Kümmel, 1999; Shigenaga et al, 1989). The significance of

this difference is currently unknown.

Renshaw cells, which make up ~10% of the adult V1 population, are distinguished from other V1

IN subtypes based on a number of properties, including gene expression profile, morphology, spatial

position, and physiology (Alvarez et al, 2005; Zhang et al, 2014; Alvarez et al, 2013). Among En1-

expressing V1 INs, RCs uniquely express high levels of the calcium-binding protein, calbindin (Cb), which

is abundant in their axons and dendrites (Carr et al, 1998; Alvarez et al, 1999; Geiman et al, 2000).

Although Cb function in RCs has not been demonstrated, it is proposed to be involved in facilitating fast

synaptic transmission (Blatow et al, 2003; Alvarez et al, 2013). Importantly, RCs are not the only Cb-

expressing cells in the spinal cord. Non-RC neurons include dorsally-derived IN cell types as well as

immature V1 INs that transiently express one or more calcium buffering proteins (i.e., parvalbumin and

calretinin) (Geiman et al, 2000; Mentis et al, 2006; Alvarez et al, 2005; Alvarez & Fyffe, 2007).

Nevertheless, there is an overall decrease in the number of non-RC Cb-expressing INs during postnatal

spinal development (Zhang et al, 1990; Smith et al, 2015; Alvarez et al, 2005).

In the adult, Cb-expressing RCs are located in the most ventral regions of laminae VII and IX of

the ventral spinal cord (“Renshaw cell area”), where they come in close contact with MN cell bodies, their

axon collaterals, and their exiting axons at the ventral root (Sapir et al, 2004; Alvarez et al, 2005). During

early embryonic development, RCs take a characteristic path towards this final settling position, migrating

away from the p1 progenitor domain shortly after becoming postmitotic and traveling a circumferential

route navigated initially by their ventrally oriented neurites. These neurites surround the motor pools

located at the lateral edge of the embryonic ventral horn, eventually stopping at the ventral root exit to

wait for their cell bodies to “catch up.” Indeed, at the time when RCs are born, there are few other cells in

the ventral horn than MNs, which are born in overlapping periods with RCs (Benito-Gonzalez & Alvarez,

45

2012; Stam et al, 2012; Wu et al, 2006; Novitch et al, 2001). Furthermore, later-born V1 IN subtypes do

not travel the same route as RCs and settle dorsomedially to MNs (Alvarez et al, 2013). These

observations suggest that from their early genesis, RCs and MNs experience a unique and close

relationship not afforded to other V1 INs, potentially resulting in preferential targeting of RCs to MNs and

vice versa to form the specialized recurrent inhibitory circuit.

In the postnatal spinal cord, RCs are the sole IN target of MN axon collaterals (MNs themselves

also receive auto-synaptic input from recurrent motor axon collaterals (Alvarez et al, 2013; Lagerback et

al, 1981). Immunohistochemical analyses detecting vesicular acetylcholine transporter (VAChT) as a

proxy for cholinergic MN inputs show that each RC inputs from 75 MNs, with an average of 6 cholinergic

contacts onto each RC (Alvarez et al, 1999). A recent study using paired electrophysiological recordings

between single MNs and RCs also concluded that there are ~6 MN contacts per RCs, but that only ~4

MNs converge onto individual RCs (Moore et al, 2015). While there may be any number of reasons for

these discrepancies, it’s clear that there is a large convergence of motor inputs onto a relatively small

population of RCs, suggesting that individual RCs likely play an outsized role in the control of motor

output. Furthermore, in addition to their high density on RC somas and dendrites, VAChT terminals on

RCs are also relatively larger in size than VAChT terminals on other ventral INs, presumably correlating

with the strength of synaptic connectivity between MNs and RCs (Alvarez et al, 1999).

Interestingly, while MN axons in the periphery exclusively release acetylcholine (ACh) at NMJs,

recent studies suggest that in mouse neonates, MN collaterals in the spinal cord may co-release ACh and

the excitatory amino acids (EAA) glutamate and/or aspartate onto RCs (Mentis et al, 2005; Nishimaru et

al, 2005; Richards et al, 2014). In particular, application of cholinergic blockers (muscarinic and nicotinic)

was unable to completely block excitatory postsynaptic potentials (EPSP) in RCs elicited from antidromic

activation of ventral roots. The remaining 20-30% of the EPSP was abolished by application of glutamate

receptor (NMDA and AMPA) antagonists (Mentis et al, 2005; Nishimaru et al, 2005). Whether MN-RC

synapses express glutamatergic synaptic components (e.g. vesicular glutamate transporters, VGLUT) is

still under investigation. Interestingly, a recent study has proposed that MNs release EAAs, such as

aspartate, that do not rely on VGLUTs for their transport into synaptic vesicles, indicating MN synapses

on RCs may be complex than previously thought (Curtis et al, 1960; Richards et al, 2014).

46

The formation of RC-MN recurrent circuitry begins with MN synaptogenesis on RCs at e13 in the

mouse. Following their migration into the lateral ventral horn and axon extension into the ventral

funiculus, RCs begin innervating MNs (e15), with frequency of synapse formation increasing into

postnatal development (Alvarez et al, 2013). Interestingly, RC differentiation and synaptogenesis does

not require MN inputs or synaptic activity. RCs are found in normal numbers and settling positions both in

choline acetyltransferase (ChAT) knockout mice and when synaptic activity is globally disrupted in the

spinal cord by expression of tetanus toxin light chain fragment (TeNT), which acts to prevent synaptic

vesicle exocytosis (Myers et al, 2005; Stam et al, 2012; Zhang et al, 2008).

It is also important to note that while MN collaterals are considered the classic input to RCs, RCs

receive other modulatory inputs, including from excitatory group 1a pSN afferents, excitatory V3 INs,

other inhibitory INs, and descending inputs (Alvarez et al, 2013; Geiman et al, 2000; Geiman et al, 2002;

Gonzalez-Forero et al, 2005; Mentis et al, 2005; Mentis et al, 2006; Nishimaru et al, 2010; Siembab et al,

2016). Interestingly, RCs receive VGLUT1-expressing pSN synapses beginning at late embryonic stages,

which proliferate from P0 to P15 before being weakened and deselected in favor of VAChT-expressing

synapses from MN collaterals (Siembab et al, 2016; Mentis et al, 2005; Mentis et al, 2006). These studies

suggest that the RC-MN recurrent circuitry in the adult is likely the end result of a prolonged process of

connectivity refinement involving transient sensory inputs, and presumably others.

In addition to receiving non-MN inputs, RCs are also known to synapse directly on non-MN cell

types in the ventral spinal cord. These include IaINs, ventral spinocerebellar tract neurons, as well as

other RCs (Hultborn et al, 1979; Windhorst, 1990; Jankowska, 1992; Alvarez & Fyffe, 2007). It is

hypothesized that RCs modulate the recurrent inhibitory circuits produced by other RCs in order to

provide mutual inhibition for the refinement of motor behaviors (Ryall, 1970). RCs also famously function

to inhibit IaINs, controlling their activation in order to regulate the co-contraction of antagonist muscles

that are themselves inhibited by IaINs (Jankowska, 1992; Alvarez et al, 2013). Overall, the diversity of RC

synaptic inputs as well as neuronal targets likely reflects the central role that they inhabit in the

construction and elaboration of spinal circuits underlying coordinated and adaptable motor behaviors.

Nonetheless, in the classic MN recurrent inhibitory circuit, RC innervation of α-MNs leads to long

duration inhibitory postsynaptic potentials (IPSP) in the MN (Alvarez et al, 2013). While adult RCs are

47

thought to be predominantly glycinergic, some evidence suggests that at least some RCs also use GABA

to prolong their inhibitory synaptic action on MNs since GABA activation results in slower decay of

synaptic currents compared to glycine-mediated inhibition (Renshaw, 1946; Eccles et al, 1954; Ryall,

1970; Cullheim & Kellerth, 1981; Schneider & Fyffe, 1992). Accordingly, recurrent IPSPs in MNs can be

abolished by application of strychnine and bicuculline, which are glycinergic and GABAergic blockers,

respectively (Schneider & Fyffe, 1992; Alvarez et al, 2013). While some inhibitory INs can release both

neurotransmitters, it is currently unknown if individual RCs are capable of releasing both GABA and

glycine, or if these represent distinct subpopulations of RCs (Jonas et al, 1998; Schneider & Fyffe, 1992).

Previously, it was thought that RCs provide relatively minor inhibitory effects on MN excitability

(Windhorst et al, 1978; Windhorst, 1996; Maltenfort et al, 2004; Bhumbra et al, 2014). For example, RC

contacts on MN dendrites are distal compared to those of IaINs, suggesting that they provide relatively

weaker input for modulating MN activity (Curtis & Eccles, 1959; Burke et al, 1971; Fyffe, 1991). Moreover,

in earlier studies, RC activation was shown to evoke only minimal effects on MN membrane potential

(Van Keulen, 1981; Hamm et al, 1987). However, a recent study using paired recordings of RCs and MNs

indicates that RCs may be strategically positioned to produce maximal inhibitory shunting effects on MN

activity, and that evoking just a single action potential from one RC is sufficient to silence MN spike firing

(Bhumbra et al, 2014). This study thus provides some of the strongest support for the effectiveness of

RC-mediated recurrent inhibition, suggesting that RCs wield significant influence over MN output.

In addition to their unique circuitry with MNs as well as other spinal cellular components, RCs

also differ from other V1-derived neurons by their morphology, neurotransmitter receptor clustering, and

physiology. Compared to other ventral INs, RCs have smaller soma size, multipolar or fusiform

morphology, and dendritic arbors that are sparsely branched with relatively short projections (Fyffe, 1990;

Geiman et al, 2000; Bui et al, 2003; Alvarez & Fyffe, 2007). While other V1 INs primarily project rostrally,

RCs extend bifurcating axons 1-2 segments in both directions to synapse onto the soma and proximal

dendrites of MNs in those adjacent segments (Fyffe, 1991; Saueressig et al, 1999; Sapir et al, 2004).

Interestingly, while most RC projections in the ventral funiculus travel ~3 mm from the soma, some

projections have been observed up to 12 mm away (Van Keulen 1979; Lagerbäck & Kellerth, 1985a;

48

Jankowska & Smith, 1973). It is proposed that the short-range projections primarily provide inhibition of

MNs, while the longer-range projections may act on other RC targets (Jankowska & Smith, 1973).

RCs can also be distinguished by their distinct clustering of neurotransmitter receptors and

synaptic components. In particular, cholinergic contacts from MNs onto RCs are primarily nicotinic (i.e.,

blocked by dihydro-ß-erythroidine but not atropine) and as such they express α2 and α4 nicotinic

acetylcholine receptor subunits (encoded by Chnra2 and Chnra4 genes, respectively) (Curtis & Ryall,

1964; Dourado & Sargent, 2002). While the α4 subunit is expressed by other spinal INs, α2 expression is

restricted to RCs and has recently been used to genetically access this cell type in the spinal cord (Ishii et

al, 2005; Perry et al, 2015). The selective expression of the α2 subunit in RCs is thought to underlie the

cells’ high and rapid sensitivity to acetylcholine (Alvarez et al, 2005). In addition to specific acetycholine

receptors, all RCs also uniquely express distinctively large and complex postsynaptic gephyrin clusters, a

scaffolding protein that functions to anchor inhibitory neurotransmitter receptors at the postsynaptic

membrane (e.g. glycine receptors, GlyRs) (Alvarez et al, 1997; Gonzalez-Forero et al, 2005). These so-

called postsynaptic densities (PSD) on RCs are on average 10-100X larger than typical inhibitory PSDs

on other neuronal cell types, (Alvarez et al, 1997; Alvarez & Fyffe, 2007).

Finally, compared to other neuronal cell types in the spinal cord, adult RCs are capable of firing

high-frequency bursts of discharges, with frequencies up to 1000 Hz in the cat, in response to single

ventral root stimulation, as well as from single action potentials from single MNs, although with shorter

burst duration (Eccles et al, 1954; Van Keulen, 1981; Eccles et al, 1961; Walmsley & Tracey, 1981;

Hamm et al, 1987). In mouse neonates, the range of firing of RCs is 83-122 Hz at resting membrane

potential (Mentis et al, 2005; Mentis et al, 2006). Whether embryonic RCs in the mouse can produce

similar responses is not known, although it is suggested that RC burst firing may be elicited in embryonic

spinal cords at hyperpolarized (<80 mV) membrane potentials (F. Alvarez, personal communication).

A recent examination of the physiological profiles of molecularly distinct V1 IN subtypes

demonstrated that non-overlapping subpopulations of RCs and those expressing FoxP2 and Pou6f2 TFs

could be distinguished by their passive and active membrane properties. Using whole-cell current-clamp

electrophysiological recordings, they found that RCs exhibited large, low-threshold depolarizations and

short after-hyperpolarizations, resulting in strong spike bursting phenotype. In contrast, FoxP2-expressing

49

INs exhibited no low-threshold depolarizations, a large and fast-rising afterhyperpolarization, and no initial

spike bursting; whereas Pou6f2-expressing segregated into a lateral subset characterized by bursting and

transient low-threshold depolarizations, and a medial subset without bursting and weak low-threshold

depolarizations (Bikoff et al, 2016). These data together suggest that RCs can be distinguished from

other identified V1 INs based on their specialized physiological properties. Although it is widely accepted

that RC burst firing is generated by long-duration EPSPs from motor axons, the functional consequence

of burst firing is currently unknown (Eccles et al, 1961; Walmsley & Tracey, 1981).

Despite the abundance of information about RC-specific morphological and physiological

properties, developmental programs controlling RC differentiation from other V1 IN subtypes are largely

unknown. Recent studies, however, indicate that RCs are born early and progress through a different

transcriptional program than other V1 INs (Fig 1.8) (Stam et al, 2012; Benito-Gonzalez & Alvarez, 2012).

Specifically, RCs are the earliest-born among V1 INs, exiting cell cycle and differentiating between e9.5

and e10.5, while other V1 subtypes, including IaINs, are born during the second wave of neurogenesis

between e10.5 and e12.5. Temporal differentiation of distinct spinal IN subtypes has been observed for

other progenitor domains, including early-born versus late-born dI4/dILA and dI5/dILB, dI6 and V2 INs,

suggesting that temporal patterning may be a widespread mechanism for spinal IN subtype diversification

(Glasgow et al, 2005; Stam et al, 2012; Tripodi & Arber, 2012). Nevertheless, the molecular mechanisms

underlying temporal genesis of different V1 subtypes are unknown. Interestingly, Notch signaling was

investigated as a possible molecular mechanism for V1 progenitor diversification domain, in particular as

a driver of temporal specification (Stam et al, 2012; Marklund et al, 2010; Ramos et al, 2010). However, a

role for Notch signaling was largely abandoned based on lack of effect using presenilin and Jagged1

knockout animals to globally disrupt the Notch pathway (Stam et al, 2012).

Perhaps as a result of their early neurogenesis, RCs are generated by a different transcriptional

code compared to other V1 INs. Based on gene expression profiling of V1 INs at e12.5, a time-point when

multiple V1 subtypes are being generated and acquiring specialized properties, V1 INs can be sorted into

two broad subsets: a Cb-expressing RC population that expresses the TFs Onecut1 (Oc1), Onecut2

(Oc2) and MafB; and a larger, non-overlapping population that expresses the TF FoxP2. Furthermore, the

TF Foxd3, which is initially expressed broadly in all newly born V1 INs, is quickly downregulated by all

50

other subtypes except RCs, which maintain high levels of Foxd3. Further analysis reveals that all early-

born V1 INs express either Oc2 or a combination of Oc1 and Oc2 even before En1 suggesting that the

Onecut TFs are the earliest RC-specific TFs to be expressed. Thus, V1 progenitors fated to adopt RC

identity activate Oc1/2, En1, Foxd3, MafB and Cb in sequence; while non-RCs initially activate and

downregulate Foxd3 while upregulating FoxP2 expression (Stam et al, 2012).

Mouse mutant analyses confirm the importance of this TF expression program for RC

specification. In Oc1/2 double knockout animals, RC differentiation is initiated to a limited extent, as

evident by the successful development and migration of some MafB- and Cb-expressing neurons in the

RC area at e10.5. However, by e12.5, there is severe reduction of MafB- and Cb-expressing cells,

suggesting that in the absence of Oc1/2, the RC differentiation program cannot be maintained, leading to

delayed but considerable depletion of RCs from the ventral spinal cord. Conditional Foxd3 deletion in V1

progenitors recapitulates some of the effects of deleting Oc1 and Oc2, suggesting that Foxd3 may act in

parallel with these factors to program RC-specific identity. Foxd3 mutants fail to induce the expression of

RC-specific markers MafB and Cb, and thus the formation of the recurrent inhibitory circuit, indicating that

Foxd3 is required for initiation of RC differentiation. Furthermore, Foxd3 deletion in postmitotic V1 INs

also results in marked reduction of MafB and Cb-expressing cells in the RC area, suggesting that Foxd3

is necessary to both induce and maintain the RC program. Finally, deletion of MafB, which is expressed

late during RC development, indicates that while MafB is not necessary for RC differentiation and the

formation of RC-MN recurrent circuitry, its expression is required for the maintenance of Cb expression in

RCs as well their survival (Stam et al, 2012).

Altogether, these results indicate that RCs are differentially specified from the V1 progenitor

domain based on a temporally-regulated transcriptional program. However, the molecular mechanisms

underlying RC temporal differentiation are still unknown. Furthermore, whether RCs are generated from

multipotent V1 progenitors or from a restricted set of progenitors remains to be determined. A yet

unresolved question is whether there is spatial patterning of V1 progenitors, as has been demonstrated

for V0 INs, which produces functionally distinct subtypes from genetically distinct ventral and dorsal

subdomains (Pierani et al, 2001; Moran-Rivard et al, 2001; Griener et al, 2015). By understanding the

programs underlying V1 IN diversification, specifically the generation of a well-defined, highly specialized

51

subtype such as the RC, we can begin to unravel the logic behind spinal IN subtype diversification as it

relates to the construction of spinal sensorimotor circuits controlling the immense range of motor

behaviors that complex organisms are capable of generating.

While much of our understanding of spinal cord diversity derives from investigations of MN

subtype diversification, it is likely that most, if not all, spinal IN classes will also be revealed to be highly

heterogeneous, not only in terms of molecular identity but also functionality (Bikoff et al, 2016; Bjornfors &

El Manira, 2016; Griener et al, 2015; Del Barrio et al, 2013). The lack of specific genetic access to the

majority of these IN cell types and the relatively small size of many of these populations makes the

pursuit of these questions additionally challenging. For example, RCs comprise 1-10% of all V1 INs

(Alvarez et al, 2005; Zhang et al, 2014). In the final section of this introduction, I will propose the use of

embryonic stem cell-derived neurons for studying questions of IN development and circuitry.

E. Approach

Over the past few years, major advances have been reported in the directed differentiation of

pluripotent stem cells to diverse neuronal cell types throughout the neuraxis, offering researchers a

powerful tool for studying molecular mechanisms underlying neuronal development (Wichterle et al, 2002;

Salero et al, 2007; Gaspard et al, 2009; Maroof et al, 2013; Nicholas et al, 2013; Gouti et al, 2014, etc).

Using appropriate combinations of developmentally relevant patterning signals, mouse and human

embryonic stem cells, as well as induced pluripotent stem cells, can be induced to differentiate into neural

cells with high efficiency in a process that closely recapitulates in vivo neural development both

molecularly and temporally (Gaspard et al, 2010; Petros et al, 2011). Moreover, recent attempts have

focused on defining the subtype identity of in vitro-generated neurons, including development of protocols

to generate specific populations of cortical pyramidal neurons, cortical INs, and spinal MNs (Gaspard et

al, 2008; Ideguchi et al, 2010; Chen et al, 2013; Maroof et al, 2013; Tyson et al, 2015; Peljto et al, 2010;

Nedelec et al, 2012; Machado et al, 2014).

Directed differentiation of ESCs into spinal MNs was first described almost fifteen years ago (Fig

1.9) (Wichterle et al, 2002). In the presence of RA and a Shh-related smoothened agonist (SAG), mouse

52

and human ESCs can be induced to differentiate first into spinal cord MN progenitors then into bona fide

spinal MNs through a pathway that mirrors in vivo MN development. Using a mouse ESC line in which the

MN-specific gene Hb9 drives high expression of green fluorescent protein (Hb9::GFP), MNs can be

visually identified as well as specifically isolated using fluorescently activated cell sorting (FACS)

techniques. Under RA and SAG conditions, ~30-50% of all cells acquire MN identity, including expression

of MN progenitor and postmitotic genes in the appropriate temporal sequence, MN-like morphological

features, and electrophysiological properties (Wichterle et al, 2002; Li et al, 2005; Miles et al, 2004).

Remarkably, ES-MNs transplanted into the developing chick spinal cord are able to migrate appropriately

into the ventral horn, project axons out of the spinal cord into the periphery, and form recognizable

cholinergic synapses with target muscles (Wichterle et al 2002; Soundararajan et al, 2006; Wichterle et

al, 2009; Peljto et al, 2010; Bryson et al, 2014). These pioneering studies provided convincing evidence

that neurons with molecular and functional identity of spinal MNs could be efficiently generated in vitro,

providing direct access to limitless amount of MNs to study their developmental specification as well as

their role in amyotrophic lateral sclerosis and spinal muscular atrophy, neurodegenerative disorders

affecting MNs, potentially with direct clinical applications (Peljto et al, 2010; Chen et al, 2012; Machado et

al, 2014; Tan et al, 2016; Ho et al, 2016; Sanses et al, 2016; Simon et al, 2016).

ES-MNs have also been shown to recapitulate aspects of in vivo MN subtype diversity (Peljto et

al, 2010). While ES-MNs induced with RA and SAG exhibit rostral cervical MN identity, as evident by their

Hox gene expression profile and expression of MMC marker Lhx3 but absence of LMC marker FoxP1

expression, retinoid-independent differentiation conditions relying on SAG and endogenously secreted

Wnt and FGF signals can induce MNs with brachial and thoracic identities, including the generation of

FoxP1-expressing limb-innnervating subtypes. Notably, transplanted ESC-derived FoxP1-expressing

MNs migrate into LMC territory in the ventral horn of the chick spinal cord, preferentially project axons into

the limb, and acquire motor pool-specific markers (Peljto et al, 2010). Altogether, these results indicate

that ES-MNs not only acquire generic MN features, but they may be induced to exhibit subtype-specific

characteristics depending on their response to extrinsic signals.

Certainly, the ability to efficiently differentiate ESCs into defined neuronal cell types represents a

unique opportunity for studying mechanisms of neuronal specification, subtype diversification, and

53

potentially, functional connectivity. ESC-derived MNs in particular have been shown to be a convenient

system for experimentally dissecting complex questions concerning MN development, including the

instructive role of Notch signaling during selection of MMC versus HMC identity, the establishment and

maintenance of MN-specific gene expression programs, and the contribution of extrinsic (e.g., Wnt and

FGF) and intrinsic factors (e.g. Hox gene expression) to MN subtype diversification (Tan et al, 2016;

Rhee et al, 2016; Mazzoni et al, 2013b).

Importantly, whether spinal neurons such as V1 and dI4 INs can be generated with similar

efficiency to MNs had not been shown. Previous studies have reported limited success in generating

spinal INs with characteristics of inhibitory INs, including Pax2 and GABA enzyme expression, though the

neurons generated in these attempts could not definitively labeled as belonging to either V1 or dI4 IN

classes due to lack of cell-type specific markers (Gottlieb & Huettner, 1999; Murashov et al, 2004; Kim et

al, 2009; Najafi et al, 2009). Furthermore, these studies failed to provide convincing molecular or

functional analyses to establish that in vitro-generated inhibitory INs recapitulated in vivo developmental

programs. In this thesis, I will present abundant evidence that ESCs can be efficiently differentiated into

V1 and dI4 INs that are indistinguishable from their in vivo counterparts both molecularly and functionally.

I will then focus on using ES-V1 and ES-dI4 INs to identify molecular mechanisms underlying their

subtype diversification, as well as to establish a robust in vitro co-culture system to study their subtype-

specific connectivity with MNs and pSNs.

54

Figure 1.1

Figure 1.1 Distinct origins of inhibitory inputs modulating the monosynaptic stretch reflex Local inhibitory INs impinge on the monosynaptic stretch reflex circuit in a synapse-specific manner. GABApre INs are a subset of dorsally-derived dI4 INs (red) providing presynaptic inhibition of MNs by forming axo-axonic synapses on primary sensory afferent terminals. In contrast, ventral V1 INs (green) have been shown to provide direct postsynaptic inhibition of MNs (adapted from Betley et al, 2009).

55

Figure 1.2

Figure 1.2 Morphogen signaling during neural tube development During embryogenesis, different brain regions (FB, forebrain; MB, midbrain; HB, hindbrain; SC, spinal cord) are formed at the intersection of the morphogenetic signaling molecules Shh, Wnt, Fgf8, and RA. Spinal MNs are patterned by RA from the paraxial mesoderm somites and Shh from the ventral floor plate (FP) and notochord (NC). (adapted from Aguila et al., 2012; Alloydi & Hedlund, 2014).

56

Figure 1.3

Figure 1.3 Expression of RA-synthesizing enzyme Raldh2 in developing vertebrate spinal cord (A) Raldh2 expression in chick neural tube at the 10-13 somite stage (A, anterior; P, posterior). Posterior expression of Raldh2 is limited by Fgf8 signals (not shown) (Diez del Corral & Storey, 2004). (B) (Left) Raldh2 expression in Hamburger-Hamilton (HH) Stage 27 chick brachial spinal cord. Raldh2 is expressed by LMC MNs and roof plate. (Right) Raldh2 expression is present in the roof plate but not MNs in the thoracic (T) spinal cord (Sockanathan & Jessell, 1998).

57

Figure 1.4

Figure 1.4 Dorsoventral patterning of the developing spinal cord After specification in the proliferative ventricular zone, postmitotic neurons migrate to the outer mantle layer of the spinal cord. The floor plate and notochord are sources of Shh, while the roof plate secretes BMP and Wnt molecules. Progressively more dorsal progenitor domains are exposed to a decreasing concentration of Shh, while more ventral progenitor domains experience lower concentrations of BMPs and Wnts. These secreted factors act in opposing gradients to pattern the spinal cord by acting on prepattern and proneural genes in different dorsoventral territories. The boundaries between progenitor domains are defined and sharpened by cross-repressive interactions between pairs of HD and bHLH genes. The combinatorial code of these factors specifies different progenitor domains (Dp1–Dp6, Vp0–Vp3 and pMN), which give rise to distinct neuronal cell types (dorsal interneurons dI1–dI6, ventral interneurons V0–V3 and motor neurons, respectively) (Gómez-Skarmeta et al, 2003).

58

Figure 1.5

Figure 1.5 Hox expression patterns in the spinal cord underlie MN subtype diversity In vertebrates, 39 Hox genes are distributed across 4 clusters, with each gene expressed in discrete rostrocaudal domains. In the spinal cord, expression of Hox4-11 genes align with MN columnar and pool types (PMC, phrenic motor column; LMC, lateral motor column; HMC, hypaxial motor column; PGC, preganglionic motor column; MMC, medial motor column). Peripheral targets of each motor column are shown. LMC MNs further diversify in ~50 motor pools targeting limb muscles at brachial and lumbar levels (Philippidou & Dasen, 2013).

59

Figure 1.6

Figure 1.6 Molecular diversity of dorsal inhibitory interneurons Microarray expression profiling of e12.5 mouse spinal cord reveals genes downregulated in in Ascl1-/- and/or Ptf1a-/- mutants. (A) In situ hybridization results for four downregulated genes with non-overlapping expression in the dorsal horn, suggesting that these comprise molecularly distinct subpopulations of inhibitory interneurons, including dI4 and dILA INs derived from the Ptf1a-expressing spinal progenitor domain. (B) Schematic representation of layer-specific distribution of inhibitory subpopulations (right) and their dependence on Ascl1 (left) (Wildner et al, 2013).

60

Figure 1.7

Figure 1.7 V1 interneuron subtype diversity (A) Expression of 19 TFs enriched in V1 INs compared to dI4 INs in p0 L3-L5 lumbar spinal segments. Anti-FoxP2 (blue), MafA (green), Pou6f2 (yellow) and Sp8 (red) antibodies label 64.2% ± 0.6% of V1 INs. (B) Spatial distribution of 7 different V1 IN subsets in p0 lumbar spinal cord. (C) V1 INs segregate into four discrete clades defined by mutually exclusive expression of FoxP2, MafA, Pou6f2, and Sp8 (<1% overlap). Clades are further subdivided by distinct TFs (black). Dotted line represents additional V1 cell types. The number of V1 cell types is indicated in parentheses (Bikoff et al, 2016).

61

Figure 1.8

Figure 1.8 Renshaw cell neurogenesis (A) Birthdating of V1 interneuron subtypes in the lumbar spinal cord. Sections through the mouse cervical spinal cord at E12.5 showing the time course of BrdU incorporation into V1 INs. Cb-expressing RCs in the ventrolateral quadrant of the spinal cord are labeled with BrdU pulses at E9.5, whereas few are labeled at E10.5. Conversely, many non-RC V1 subtypes incorporate BrdU at E10.5 and E11.0. (B) Model showing the temporal generation of RCs and other V1 interneuron subtypes. The development of RCs is dependent on maintenance of high expression of Foxd3 TF, which is initially broadly expressed in postmitotic V1 INs. Additionally, selective activation of the Onecut transcription factors OC1 and OC2, as well as MafB, during the first wave of V1 IN neurogenesis is essential for the RC differentiation program. Conversely, non-RC V1 INs express low levels of Foxd3, with a large subset expressing the TF FoxP2 (Stam et al, 2012).

62

Figure 1.9

Figure 1.9 Directed differentiation of spinal motor neurons from ESCs (A) Timeline of ES cell differentiation. Temporal profile of expression of the key developmental markers is shown above the axis with outline of the differentiation protocol is below the axis. (B) Typical shape of ES cell colonies (arrows) prior to trypsinization. (C) Embryoid bodies at 2 days of differentiation. (D, E) Expression of motor neuron progenitor marker Olig2 and post-mitotic motor neuron markers Hb9 and Isl1/2 in immunostained sections of EBs. Abbreviations: PE, primitive ectoderm; NP, neural plate; pMN, progenitor motor neuron; GDNF, glial cell line–derived neurotrophic factor (Wichterle & Peljto, 2008). (F) Transplant of Hb9::GFP MNs in HH Stage 15-17 chick spinal cord. Transverse sections through stage 27 spinal cord at lumbar levels after grafting MN-enriched EBs. The pathway of axons is detected by neurofilament (NF) expression (Wichterle et al, 2002).

63

Chapter 2: Directed differentiation of spinal inhibitory interneurons from stem cells Introduction

Spinal neuronal development depends on the intersection of multiple axes of extrinsic signaling

molecules, chief among them being RA produced by the paraxial mesoderm and gradients of BMP and

Shh emanating from the roof and floor plates of the spinal cord, respectively (Jessell et al, 2000).

Although differentiation of spinal MNs from mouse ESCs is well established, whether different spinal

inhibitory IN cell types can be also be efficiently generated from mESCs has not previously been

demonstrated (Wichterle et al, 2002). Using prior knowledge of spinal patterning cues and developmental

timing, as well as experience gained from directed differentiation of spinal MNs, I first tested conditions for

efficient differentiation of V1 and dI4 inhibitory INs from mESCs.

While V1 INs are generated from a ventral spinal domain marked by expression of the TF En1,

dI4 INs are produced from a dorsal progenitor domain expressing the TF Ptf1a. Based on their spatial

origins, we predicted that V1 INs could be generated by modulating the concentration and/or timing of

Shh signals during differentiation. Since V1 INs are produced from a spinal progenitor domain that lies

dorsal to the MN progenitor domain and thus farther away from the Shh source at the floor plate of the

neural tube, we anticipated that V1 INs would require less Shh signals for their specification compared to

MNs (Briscoe & Ericson, 2001; Wichterle et al, 2002). Conversely, dI4 IN differentiation should depend on

dorsalizing signals, especially BMPs and Wnts (Lee & Jessell, 1999; Caspary & Anderson, 2003).

Whether patterning and specification of intermediate spinal domains such as the dP4 progenitor domain

giving rise to dI4 INs depends on BMP and Wnt signaling has not been definitively established.

Evidence against a role for BMP or Wnt signaling in dI4 IN generation comes from BMP and Wnt

mouse mutants which do not exhibit overt changes in the establishment of the dP4 domain or dI4 IN

production (Lee et al, 2000; Nguyen et al, 2002; Muroyama et al, 2002; Timmer et al, 2005). An

alternative hypothesis is that generation of dI4 INs during spinal cord development requires inhibition of

dorsalizing signals (Müller et al, 2002; Gross et al, 2002). Thus, to determine the role of dorsalizing

signals on dI4 IN generation, I differentiated dI4 INs using different concentrations and timings of TGFß

and Wnt agonists as well as antagonists.

64

To determine the mechanisms controlling intermediate spinal cord patterning, I developed

reporter ESC lines for dI4 and V1 IN populations systematically probed effects of diverse concentrations

of ventralizing and dorsalizing agonists and antagonists. In addition to optimizing conditions for efficient

differentiation of V1 and dI4 INs, I examined the molecular profile of ESC-derived V1 and dI4 INs using

immunocytochemical and global gene expression profiling approaches at several key developmental

stages. Finally, to test if ES-V1 and ES-dI4 INs also exhibit known functional properties of their in vivo

spinal counterparts, I reintroduced these neurons into embryonic spinal cord using transplantation into

developing chick embryo, an established proxy for testing in vivo functionality of in vitro-derived

mammalian cells (Wichterle et al, 2009).

Results Derivation of V1 and dI4 lineage reporter stem cell lines V1 INs arise from a ventral spinal domain that transiently expresses the TF En1 while dI4 INs are

produced from the Ptf1a-expressing dP4 progenitor domain in the dorsal spinal cord (Fig 2.1A). I derived

ESC lines from En1::cre or Ptf1a::cre mice crossed to mouse strains in which fluorescent proteins are

expressed upon Cre-mediated excision of floxed stop sequence (Fig 2.1B) (Kimmel et al, 2000;

Kawaguchi et al, 2002; Madisen et al, 2010; Srinivas et al, 2001; Buffelli et al, 2003). Thus, all cells

generated from En1 or Ptf1a lineages are permanently fluorescently labeled and can be isolated using

fluorescence-activated cell sorting (FACS). En1::cre x ROSA::tdTomato line, which expresses the red

fluorescent protein tdTomato under control of the ROSA locus, will be referred to as En1-tdTomato, while

a green fluorescent reporter line will be referred to as En1-GFP. Similarly, Ptf1a::cre x ROSA::tdTomato

will be referred to as Ptf1a-tdTomato. I also derived lines from Ptf1a::cre x Thy1::YFP crosses, which are

referred to as Ptf1a-Thy1YFP – these produced mosaic expression of fluorescent reporter expression,

labeling ~20-30% of successfully differentiated cells (Betley et al, 2009, data not shown).

65

Optimized differentiation of V1 interneurons

Using the new ESC reporter lines, I first optimized protocols for directed differentiation of ESCs to

V1 INs (ES-V1). Treatment of nascent embryoid bodies (EBs) on Day 2 of differentiation with low

concentration (0.5-5 nM) of smoothened agonist (SAG), a downstream effector of Shh signaling, in the

presence of high RA (1µM) yields 31.6% En1-tdTomato fluorescent positive (FP) V1 INs by Day 8 of

differentiation, with similar results for the En1-GFP line (Fig 2.2A,B). By comparison, MN differentiation

requires 100X more SAG, suggesting that EBs respond to graded variations in Shh signaling to produce

different ventral neuron cell types, as in vivo (Fig 2.2C) (Briscoe & Ericson, 2001). Although the

percentage of FP cells generated was greatest after Day 8 of differentiation, FP cells were produced

beginning on Day 5 (1.3%) and slowly increased through Days 6 and 7 (6.4 and 8.0%, respectively) until

reaching peak generation on Day 9 (36.1%) (Fig 2.2B). Thus, En1-lineage cells are generated in a narrow

time window from Days 5-8 of in vitro differentiation under conditions of high RA and low SAG. En1-FP

cells could be visualized using fluorescence microscopy and efficiently isolated by enzymatic dissociation

of EBs followed by FACS. Shortly after onset of reporter expression, FP cells adopt neuron-like

morphologies in EBs and dissociated cultures, including long projections from cell bodies, diverse

dendritic arbors, and growth cone protrusions (Fig 2.2A, data not shown). Immunostaining for En1 protein

in differentiating EBs showed significant overlap between En1 immunoreactivity and En1-tdTomato

reporter early during differentiation (~75%), but since En1 is only transiently expressed in early

postmitotic V1 INs, most reporter cells quickly downregulate protein expression (Fig 2.2D).

Immunocytochemical characterization of ES-V1 INs I next used immunocytochemistry (ICC) to test if En1-derived FP cells acquire molecular

characteristics of spinal V1 inhibitory INs and recapitulate steps of normal V1 development (Fig 2.3). First,

to confirm that RA and low SAG treatment is sufficient to caudalize progenitors to spinal neuron identity,

Hox gene expression was examined in Day 8 EBs. While expression of anterior Hox TFs such as Hoxa2

was largely absent, there was significant expression of caudal Hox genes, including Hoxc4-6, as well as

some expression of Hoxc8 and Hoxc9 (Fig 2.4A, data not shown). Thus, ESCs differentiated under

RA/low SAG conditions generate neurons of mostly cervical (Hoxc4-6) and brachial (Hoxc6,8) spinal cord

66

identity (Wichterle et al, 2002; Peljto et al, 2010; Philippidou & Dasen, 2013). Hox expression was also

examined in e12.5 cervical and brachial spinal cords of En1::cre x Tau.lsl.mGFP.IRES.nLacZ mice, in

which V1 INs are marked by expression of nuclear LacZ (Hippenmeyer et al, 2005; Bikoff et al, 2016).

Comparison of En1-lineage cells in the spinal cord and in EBs confirms that ESC-derived cells adopt

mixed cervical and brachial spinal cord identity and suggests that V1 INs in vivo and in vitro express

different combinations of Hox TFs, as shown for MNs (data not shown) (Dasen, 2009; Peljto et al, 2010;

Mazzoni et al, 2013b).

Having established that En1-FP cells in vitro acquire spinal identity, I next performed ICC on early

(Day 4-6) and late (Day 8) EBs to determine if they can differentiate into p1 progenitors and acquire V1

IN-specific molecular identity. Spinal p1 progenitors giving rise to V1 INs express the ventral HD TFs

Pax6, Dbx2, Irx3 and Nkx6.2, while excluding Dbx1, Nkx6.1, and Nkx2.2 (Briscoe & Ericson, 2001).

Accordingly, many cells in early EBs (Day 5) express high levels of Pax6 and Nkx6.2, while some cells

express Dbx1, Nkx6.1 and Irx3, but not Nkx2.2 (Fig 2.4A, data not shown). This TF expression profile

indicates that EBs exposed to RA and low SAG produce mixed populations of ventral progenitors giving

rise to V0-V2 INs and MNs, but not V3 INs (Briscoe et al, 2000; Wichterle et al, 2002). While Dbx2

expression could not be examined using ICC, the increased number of cells expressing Nkx6.2 relative to

Dbx1 and Nkx6.1 suggests RA/low SAG conditions preferentially produce V1 progenitors (Sander et al,

2000; Alaynick et al, 2011). In addition to these HD TFs, EBs on Days 5-6 also expressed the bHLH TFs

Ngn1/2, which are expressed in p0-p2 and pMN progenitors, while few cells expressed Olig2, a bHLH TF

required for MN generation, providing further evidence that RA/low SAG conditions bias the formation of

intermediate ventral neuron types (Novitch et al, 2001).

By Days 5-6, progenitor markers are downregulated as FP cells begin to appear, suggesting that

En1-expressing cells are starting to be born at this time point. Immunostaining for En1 protein establishes

its transient expression pattern, recapitulating in vivo observations (75.2% on Day 5 versus 13.0% on Day

6) (Fig 2.4A and Fig 2.2D). Foxd3, another TF that is essential for V1 IN specification, is also expressed

broadly and transiently in early postmitotic V1 INs (58.3% on Day 5). By the time V1 IN differentiation

plateaus on Day 8, virtually all FP cells express the neuronal-specific markers Tuj1 and NeuN, as well as

Pax2 and Lhx1/5, TFs involved in specifying inhibitory cell identity (Fig 2.4A, data not shown) (Burrill et al,

67

1997; Batista et al, 2008; Pillai et al, 2007; Sapir et al, 2004). Importantly, En1-tdTomato cells do not co-

express TFs marking En1-lineage cells in the midbrain (e.g., Lmx1b-expressing dopaminergic neurons),

consistent with lack of expression of forebrain/midbrain marker Otx2; nor do they express HD TFs

labeling other ventral spinal neuronal classes (i.e., Evx1/2, Chx10, Lhx3, and Isl1/2 for V0, V2, and MNs,

respectively) (Fig 2.4B, data not shown) (Arenas et al, 2015; Alaynick et al, 2011). Altogether, these

results not only indicate that En1-FP cells in vitro recapitulate V1 IN molecular development in vivo, but

also that reporter expression is restricted to En1-derived spinal INs in EBs.

Finally, following dissociation of EBs and an additional 1-2 weeks of culture in defined neuronal

media, ESC-derived V1 INs express markers suggestive of functional inhibitory neuron maturation,

including proteins necessary for inhibitory GABA and glycine neurotransmission (Gad65/67, VGAT,

GlyT2) and synapse components (synapsin; synaptic vesicle glycoprotein, SV2a) (Fig 2.4C, data not

shown) (Sapir et al, 2004; Alvarez et al, 2005). Interestingly, there was a switch in GABA versus glycine

neurotransmitter expression, with upregulation of GlyT2 in cells cultured on maturation-promoting

astrocytes compared to cells cultured on extracellular matrix (laminin/fibronectin) only (9.2 versus 41.8%),

a change that was concomitant with an increase in neurite outgrowth and branching (Fig 2.4D) (Clarke &

Barres, 2013).

Optimized differentiation of dI4 INs

Having differentiated ESCs into neurons with V1 IN-like molecular and morphological features, I

next focused on optimizing conditions for efficient differentiation of dI4 INs from ESCs. Whereas Shh

signaling is considered paramount for specification of ventral neuron types such as V1 INs and MNs, the

specific molecular signals required to generate intermediate dorsal spinal neuron types such as dI4 INs

are not as well established. Using the Ptf1a::cre x ROSA::tdTomato ESC line, I first treated EBs on Day 2

of differentiation with high RA only (1 µM) to establish a baseline differentiation efficiency (Fig 2.5A,B).

Interestingly, Ptf1a-FP cells emerged one day after En1-FP cells, with no FP cells on Day 5 but 1.3% FP

cells on Day 6. By comparison, 1.3 and 6.4% of cells in V1 IN differentiation are FP by Days 5 and 6,

respectively (Fig 2.4A). Moreover, during MN differentiation from ESCs, the MN-specific Hb9::GFP

reporter first appears on late Day 4/early Day 5 (Wichterle et al, 2002). In the developing spinal cord, dI4

68

INs are born starting at e10.5, while V1 INs are born at e9.5 and MNs at e8.5, suggesting that

specification of spinal INs from ESCs temporally matches in vivo cells. (Glasgow et al, 2005; Wu et al,

2006; Alvarez et al, 2013). Production of dI4 INs steadily increased until Days 8 and 9, with up to 9.6% of

all cells expressing the reporter (Fig 2.5C). Ptf1a-FP cells generated from Ptf1a-Thy1YFP ESCs

developed along a similar timeline, but these neurons could be visualized only after ICC detecting for

GFP and, as noted earlier, exhibited mosaic reporter expression (Fig 2.5B). Finally, <10% of reporter cells

co-expressed Ptf1a protein at any timepoint, consistent with transient expression of this progenitor marker

during the transition to postmitotic neuronal fate (Fig 2.6A).

To increase the generation of Ptf1a-FP cells, EBs were treated with RA in combination with

different TGFß and Wnt signaling agonists and antagonists at varying concentrations and times (Fig

2.5D,E). Surprisingly, the majority of the added factors did not significantly increase the yield of FP cells

above RA only baseline when added on Day 2 of differentiation. These included BMP4, Wnt3a, Gdf7 and

TGFß2; as well as the BMP signaling-specific antagonists dorsomorphin, LDN-191389, and Noggin; or

the tankyrase inhibitor XAV939, which blocks Wnt signaling. Gdf11 had a small but significant effect on

Ptf1a-FP cell differentiation, despite having a more established role in rostrocaudal patterning of the

spinal cord (Liu et al, 2001; Liu, 2006). Yet, the most pronounced effect was due to another TGFß family

member, ActivinA, which binds to different receptors than BMPs and activates distinct R-Smads

(Smad2/3 versus Smad1/5/8) for downstream signaling from the receptors (Liu & Niswander, 2005).

RA+ActivinA (25 ng/mL) treatment on Day 2 generated 39.3% FP cells by Day 8 of differentiation, almost

4X higher than RA only conditions (Fig 2.5F). Using RA+ActivinA, FP cells were produced starting a day

earlier than when treated with RA only, with a steady increase until Day 9, indicating that ActivinA not only

increases the generation of Ptf1a-derived cells, but also accelerates their development (Fig 2.6A).

Accordingly, there were more Ptf1a-immunoreactive cells at Day 5 in ActivinA -treated EBs compared to

RA only (Fig 2.6B).

Interestingly, treatment of EBs with activin receptor-specific inhibitors SB-431542 and follistatin

had opposing effects: while SB-431542 caused a small increase in FP cell generation (10.4%), follistatin

decreased the differentiation efficiency (4.2%), suggesting that proper modulation of TGFß signaling

might be required for the production of Ptf1a-derived neurons (Fig 2.5E) (Villapol et al, 2013). To examine

69

this effect in more detail, EBs were treated with both RA+ActivinA and either SB-431542 or follistatin.

Both inhibitors significantly depressed the effect of ActivinA on FP cell production (27.8 and 15.3%,

compared to 43.4% without any inhibitors), suggesting that they both function to inhibit ActivinA signaling

to affect the generation of Ptf1a-derived cells (Fig 2.6C).

Finally, to fully assess the requirement of ActivinA signaling for efficient differentiation of Ptf1a-FP

cells from ESCs, EBs were treated with different concentrations of ActivinA on Day 2, or with a singular

concentration (25 ng/uL) on different days of the differentiation protocol. Ptf1a-FP cells were produced at

highest efficiency in the 10-50 ng/uL range, with the effect tapering off at 100 ng/uL (Fig 2.6D).

Furthermore, we identified a narrow time window for addition of ActivinA to affect Ptf1a-tdTom cell

differentiation: between Days 2 and 4, ActivinA treatment produced similar efficiencies (43.0-52.6% FP

cells), while treatment on Day 5 or later produced significantly smaller effects (25.9% or less) (Fig 2.6E).

Altogether, these results indicate that efficient Ptf1a-FP cell differentiation is dependent on both the

concentration and timing of ActivinA signals.

Immunocytochemical characterization of ES-dI4 INs As with V1 IN differentiation from ESCs, EBs differentiated with RA+ActivinA exhibited Hox gene

expression profiles consistent with formation of spinal neurons of mixed cervical and brachial identity,

including Hoxa2 expression and upregulation of caudal Hox5 and Hox6 clusters (Dasen, 2009;

Philippidou & Dasen, 2013). Interestingly, while Hox TFs such as Hoxa5, Hoxa7, Hoxc6, Hoxc8 and

Hoxc9 are expressed in other cells in the EB, Ptf1a-tdTomato cells do not co-express any Hox factors,

suggesting that dI4 IN development might be independent of Hox regulation (data not shown) (Liu et al,

2001). While Hox TFs are expressed in dorsal spinal domains, detailed analyses of their expression in

different dorsal lineages, including Ptf1a-derived cells, have not been carried out (Dasen et al, 2005).

Next, to examine if spinal progenitors in RA+ActivinA-treated EBs are appropriately dorsalized

and express dP4 domain-specific markers, I examined the expression of HD and bHLH TFs previously

shown to be expressed in the dorsal and intermediate spinal cord. During dorsal spinal cord development,

the dP4 progenitor domain expresses the HD TFs Pax3/6/7, Irx3, Msx1, Gsh1/2, and Ascl1, as well as the

bHLH TFs Ascl1 and Ptf1a (Alaynick et al, 2011; Lai et al, 2016). Following 2-3 days of RA+Activin

70

differentiation, when relatively few FP cells are generated, Ptf1a-tdTomato EBs contain many cells

expressing Pax3/6/7, as well as Ascl1 and Ptf1a (Fig 2.7A, data not shown). Furthermore, while some

Dbx1 and Nkx6.1-expressing cells are generated in RA only differentiation conditions, these are mostly

absent in RA+ActivinA-treated EBs, indicating that ActivinA sufficiently dorsalizes spinal progenitors to

exclude intermediate and ventral fates (data not shown).

As Ptf1a-FP cells appear beginning Days 5 and 6 and continuing until Days 8 and 9, progenitor

markers are downregulated while postmitotic TF are increasingly expressed in reporter cells. These

include the HD TF Lbx1, which is transiently expressed in spinal dI4-dI6 INs and is required together with

Ptf1a to specify dI4/dILA over dI5/dILB IN identity (Schubert et al, 2001; Gross et al, 2002; Müller et al,

2002). In EBs, Lbx1 is expressed broadly, encompassing most Ptf1a-FP cells and non-FP cells, indicating

that RA+ActivinA differentiation conditions are conducive to generation of Lbx1-expressing dI4-dI6 INs.

Lbx1 expression in vitro is also transient, as evidenced by the fact that 43.3% of Ptf1a-FP cells express

Lbx1 on Day 6 while only 35.3% of FP cells express it on Day 8 (Fig 2.7A, data not shown). In addition to

Lbx1, the TFs Pax2 and Lhx1/5 are also co-expressed by the majority of FP cells (71.5 and 78.5%,

respectively) (Fig 2.7A). Pax2 and Lhx1/5 expression is frequently used to identify GABAergic dI4 (and

dI6) INs in the spinal cord since their expression is required for maintaining inhibitory neurotransmitter

and neuropeptide expression, respectively (Glasgow et al, 2005; Bröhl et al, 2008; Pillai et al, 2007).

By Day 8 of differentiation, virtually all cells in the EB express the neuronal specific marker NeuN.

Ptf1a-tdTomato EBs dissociated on Day 8 and cultured for an additional 1-2 weeks either on

laminin/fibronectin-coated surface or on primary mouse cortical astrocytes also express neuronal-specific

Tuj1 and the synapse marker synapsin, as well as inhibitory neurotransmitters Gad65, Gad67, and GlyT2

(Fig 2.7B, data not shown). Unlike V1 IN differentiation, there was no apparent switch from GABAergic to

glycinergic neurotransmitter expression, since ~70% of FP cells expressed Gad67 (versus ~20% GlyT2-

expressing cells) whether cultured on laminin/fibronectin or astrocytes (data not shown).

Since RA+ActivinA treatment also induces non-dI4 dorsal neurons, as indicated by strong

expression Lbx1 in non-FP cells, I also examined expression of Isl1/2 and Pou4f1, which are TFs

expressed in dI1-3 and dI5 populations, but not dI4 INs in the spinal cord (Lai et al, 2016). While Isl1/2-

expressing cells are induced in these conditions and some cells in the EBs expressed Pou4f1, these were

71

largely mutually exclusive with reporter expression. Ptf1a-tdTomato cells also lacked expression of

VGLUT2, the vesicular glutamate transporter 2, which is utilized in excitatory neurons, suggesting that

they are exclusively inhibitory. Interestingly, ES-dI4 INs receive many VGLUT2 inputs, as evidenced by

the immunoreactive puncta on their dendrites and cell bodies (Fig 2.7C). Together, these data indicate

that ES-dI4 INs differentiated using RA+ActivinA recapitulate key steps of dI4 IN molecular development

in vivo, including expression of dI4 IN-specific markers and exclusion of other dorsal IN markers.

Effects of ActivinA on dorsal spinal patterning

While we were persuaded that Ptf1a-FP cells acquire dI4 IN identity, we were also interested in

further investigating the cell types generated by activating the ActivinA signaling pathway, especially

since its role has been poorly studied compared to other TGFß members (Lee & Jessell, 1999). Indeed,

an earlier study reported that ActivinA affected dI3 IN specification without effecting dI4 INs (Timmer et al,

2002). Using ICC to detect for TFs expressed by different dorsal IN cell types, I compared effects of RA

only, RA+BMP4 and RA+ActivinA treatments on dorsal patterning of Ptf1a-tdTomato EBs:

Although RA treatment alone is not sufficient to generate Olig3 and Lhx2/9-expressing dI1 INs,

some FoxP2-expressing dI2 INs and Isl1/2-expressing dI3 INs are generated (Fig 2.8A). However, the

majority of cells express Ptf1a or Pou4f1/Lmx1b TFs, indicating that the RA only differentiation conditions

generates some dorsal but mostly intermediate spinal neuron cell types, as previously reported (Wichterle

et al, 2002). Meanwhile, the addition of RA and BMP4 on Day 2 of differentiation was sufficient to repress

dI4 IN generation, as evidenced by the relative absence of Ptf1a-tdTomato cells as well as Ptf1a-

immunoreactive cells on Day 8 (Figs 2.5E and 2.8B). Furthermore, BMP4 signaling generates Olig3- and

Lhx2/9-expressing dI1 INs, as well as more FoxP2-expressing dI2 INs compared to RA only

differentiation, consistent with its well-established role in specifying the most dorsal IN cell types (Fig

2.8B) (Caspary & Anderson, 2003; Tozer et al, 2013). Using this relatively high concentration and early

timing of BMP4 addition (25 ng/uL added on Day 2), few dI3-5 INs are generated, with the few Pou4f1-

expressing cells presumably generated from dP1 or dP2 domains and not dP5 (Alaynick et al, 2011).

Finally, treatment of Day 2 EBs with ActivinA, another TGFß family member, generated many Ptf1a-

immunoreactive cells on Day 6 and consequently many reporter cells by Day 8 of differentiation. Unlike

72

with BMP4, few dI1 INs were generated, while a small number of dI2, dI3 and dI5 were produced from

these conditions (Fig 2.8C). Thus, the combination of RA+ActivinA primarily induces intermediate neuron

types such as dI4 INs, while BMP4 signaling preferentially generates more dorsal spinal types.

As mentioned earlier, the majority of FP cells generated using RA+ActivinA expressed Pax2 and

Lhx1/5 TFs, which are required for inhibitory neuron specification (Fig 2.7A). Nevertheless, a small

subpopulation did not express either of these markers, especially when compared to RA only-treated

cultures (Fig 2.9A,E,F; data not shown). Interestingly, ActivinA treatment on Day 2 also leads to an

increase in the number of Ptf1a-tdTomato cells co-expressing Lmx1b, a TF uniquely marking excitatory

dI5/dILA INs (Fig 2.9B,E,F). Importantly, Pax2 and Lmx1b expression never coincide in the spinal cord or

in vitro, indicating that the increase of Lmx1b-expressing FP cells does not occur within the Pax2-

expressing population (Fig 2.9A-D, data not shown). Together, these data suggest that RA+ActivinA

produces a significant population of dI4 INs that do not express Pax2 or Lhx1/5, and may ectopically

express dI5 markers (See discussion).

In addition to changes in the proportion of Pax2/Lmx1b-expressing FP cells, ActivinA treatment

also leads to the generation of non-neuronal glia expressing the marker glial fibrillary acidic protein

(GFAP) (Hol et al, 2015; Bushong et al, 2004; Freeman, 2010). Indeed, while virtually all Ptf1a-tdTomato

cells produced in RA only conditions express the neuronal-specific marker NeuN, ActivinA-treated EBs

generate ~5% GFAP-expressing FP cells on Day 8 of differentiation, which increases to ~20-25% in

dissociated culture after one week (Fig 2.9A,B,E,F and data not shown). Thus, ActivinA signaling is not

only sufficient to generate Ptf1a-expressing dI4 INs, but also potently induces astrocyte formation from

the dP4 domain. While less characterized than the ventral spinal domains giving rise to glia, the formation

of astrocytes from dorsal progenitor domains has been well-documented (Hochstim et al, 2008; Pringle et

al, 1998; Rao et al, 1998; Gregori et al, 2002; Tsai et al, 2012). The specific dorsal domains from which

astrocytes arise are not known, but a recent study indicates that Ascl1-lineage cells produce dorsally-

restricted spinal cord astrocytes (Vue et al, 2014). This finding is also consistent with observations from

late embryonic dorsal spinal cord, in which some Ptf1a::cre x ROSA::tdTomato lineage-traced cells adopt

astrocytic “bushy” or “spongiform” morphologies with thick primary processes and dense networks of fine

processes (J. Kaltschmidt, unpublished).

73

Faced with these complex phenotypes, we were interested in identifying conditions that would

yield more homogeneous populations of dI4 INs. As reported earlier, addition of ActivinA early during

ESC differentiation (Days 2-4) is required for the most efficient generation of Ptf1a-FP cells, while

introducing ActivinA later (Days 5-6) produces less FP cells (25.4% compared to 39.0% when added on

Day 2) (Fig 2.6E). I tested if adding ActivinA on Day 5 of differentiation would result in enriched

generation of Pax2-expressing FP cells. However, while less FP cells co-expressed Lmx1b when ActivinA

is added later, the proportion of Pax2+ FP cells was mostly unchanged (Fig 2.9C, data not shown).

Since Ptf1a-tdTomato cells lacking Pax2 expression could conceivably represent glial precursors,

I hypothesized that inhibiting Notch signaling could bias the formation of earlier-born neuronal cell types

over later-born glial cells (Novitch et al, 2001; Lu et al, 2002; Zhou & Anderson, 2002; Tan et al, 2016).

Previous studies have shown that the small molecule gamma-secretase inhibitor DAPT can be used to

effectively abolish Notch signaling (Geling et al, 2002; Tan et al, 2016). DAPT treatment in differentiating

Ptf1a-tdTomato EBs leads to a small but significant increase in FP cells generated on Day 8 when added

on Days 5 or 6, indicating that Notch inhibition in dI4 INs promotes neuronal differentiation at the expense

of glia formation, as shown for other cell types (Fig 2.9G) (Artavanis-Tsakonas et al, 1999). Indeed, when

DAPT is added on Day 5 with ActivinA (RA is added on Day 2 for proper neuralization and caudalization),

>90% of FP cells now express Pax2 and NeuN while GFAP expression is minimized in Day 8 EBs as well

as Day 15-22 dissociated cultures (Fig 2.9D, data not shown). Accordingly, when ActivinA and DAPT are

both added on Day 5, there is no significant change in FP cells generated compared to ActivinA only,

suggesting that while DAPT increases dI4 IN differentiation from ESCs, its effect is counterbalanced by

the loss of non-neuronal glial cells (Fig 2.9H). Importantly, earlier addition of ActivinA (Days 2,3 or 4)

combined with late DAPT on Day 5 does not recapitulate this effect, nor does the addition of other

dorsalizing signals (e.g. BMP4, Gdf7, XAV939) to ActivinA (data not shown). Thus, while ActivinA is

required and sufficient to specify dI4 INs from ESCs, early activation of ActivinA signaling leads to

generation of heterogeneous populations of neurons and glia from the Ptf1a spinal domain while later

activation combined with Notch-mediated inhibition of gliogenesis leads to more homogeneous

differentiation of Pax2-expressing dI4 INs.

74

RNA-seq gene expression profiling While ICC studies of ES-V1 and dI4 INs were useful for confirming the molecular identity of the in

vitro-generated cells, such analyses are limited by availability and specificity of antibodies for ICC, as well

as incomplete knowledge of the molecular factors involved in their specification. Therefore, to examine

their gene expression profiles more systematically, I dissociated early and late En1-tdTomato and Ptf1a-

tdTomato EBs, used FACS to purify FP cells, and performed expression profiling using RNA-sequencing

(Wang et al, 2009). Here, dI4 INs were differentiated using RA only to avoid glial contamination.

Comparison of V1 and dI4 INs allows for identification of genes specifically enriched in each of

these classes while filtering out genes generic to spinal neurons, particularly spinal inhibitory INs. First, to

identify genes required for early V1 or dI4 IN specification, Day 5 ES-V1 INs were compared to D6 ES-dI4

INs (Fig 2.10A). Although En1 is not on early enough to label p1 progenitors, we reasoned that Day 5 ES-

V1 INs were immature enough in their development to provide insight into the factors required for V1 IN

specification. Furthermore, to confirm these results, I also compared non-FP cells from Day 5 En1-

tdTomato EBs with D6 ES-dI4 INs since many of non-FP cells likely represent p1 progenitors (Fig 2.4A,

data not shown). I focused on TF expression since many studies have shown that neuronal identity is

largely controlled by regulatory networks of TFs (Lee & Pfaff, 2001; Shirasaki & Pfaff, 2002).

Not surprisingly, En1 is the most enriched gene in early ES-V1 INs, followed by Six6, Foxd3, and

Bhlhe22 (Fig 2.10A). The HD TF Six6 (Optx2) has been shown to be important for regulation of retinal

progenitor cell proliferation, where it acts downstream of Pax6 function, as well as pituitary/hypothalamus

development (Tréteault et al, 2008; Jean et al, 1999). However, a role for Six6 in spinal cord development

has not yet been demonstrated. Meanwhile, a recent study of the Olig-related protein Bhlhe22 (Bhlhb5)

showed that expression of this TF in the spinal cord specifically promotes the formation of dI6, V1 and

V2a spinal IN progenitors by acting downstream of RA signaling and Pax6 function, as well as by

regulating Notch signaling to promote neurogenesis (Skaggs et al, 2011). Furthermore, Bhlhe22 is one of

19 TFs recently shown to be differentially expressed in V1 INs and used to subdivide this spinal neuron

class into ~50 distinct subtypes (Bikoff et al, 2016). Finally, the zinc-finger PRDM family member Prdm12

is also in the list of top 10 TFs enriched in ESC-derived V1 INs. Prdm12, which is expressed in p1

progenitors specifically, has recently been shown to be required and sufficient to generate En1-

75

expressing V1 INs in the chick and mouse spinal cord (Kinameri et al, 2008; Thélie et al, 2015; Zannino et

al, 2014). Thus, gene expression profiling of early ES-V1 INs reveals known progenitor and early

postmitotic markers of V1 INs (e.g., En1, Foxd3, Bhlhe22, Prdm12), as well as potentially novel regulators

of V1 IN identity (e.g. Six6) (Fig 2.10A).

To trace the molecular development of ES-V1 INs specifically, I then performed comparisons of

early (Day 5) and late (Day 8) En1-tdTomato cells with ESCs (Fig 2.10B). Based on genes differentially

expressed between Day 8 ES-V1 and dI4 INs, hierarchical clustering of these three groups revealed that

the gene expression profile of Day 5 ES-V1 INs more closely resembles Day 8 V1 INs than ESCs,

although Day 5 ES-V1 INs appeared to be transitioning between the two states. For example, while En1

expression is virtually absent in mESCs, it is highly induced at V1 INs on both Days 5 and 8 (0.056, 5.45,

5.63 log2 fold-change, or log2FC, respectively, p<0.001). Furthermore, the p1 progenitor marker Prdm12

is similarly low in ESCs, expressed at high levels in Day 5 FP cells, but is eventually downregulated by

Day 8 (0.054, 6.34, 4.68 log2FC, respectively, p<0.001). (Fig 2.10C). These data indicate that directed

differentiation of ESCs to V1 INs largely follows the molecular development of V1 INs in vitro, with

appropriate regulation of progenitor and postmitotic gene expression.

Similar analyses of TFs upregulated in ES-dI4 INs on Day 6 of differentiation reveals that Gsx1,

Msx3, Prdm13, Pax3, Lbx1, and Ptf1a TFs are highly upregulated in dP4 progenitors, consistent with in

vivo analyses (Fig 2.10A) (Kriks et al, 2005; Müller et al, 2005; Borromeo et al, 2014; Seto et al, 2014;

Glasgow et al, 2005; Gross et al, 2002; Müller et al, 2002; Liu et al, 2014; Chang et al, 2013). TFAP2a/b

and Pou4f2 (Brn3b) TFs are also highly enriched: these have been shown to be induced in e12.5 Ptf1a-

derived cells from mouse spinal cord, with TFAP2a/b acting downstream of Ptf1a to specify inhibitory

amacrine cells in the retina, and Pou4f2 likely having a role in specification of superficial dorsal inhibitory

INs (Wildner et al, 2013; Jin et al, 2015; Zou et al, 2012). Interestingly, Zic1 is involved in specification of

Pax2-expressing inhibitory INs from Ptf1a progenitors in the mouse cerebellum, as well as in the

proliferation of dorsal spinal progenitors; whether it has a specific role in dP4 domain patterning has not

yet been demonstrated (Aruga et al, 2002a; Aruga et al, 2002b; Pascual et al, 2007; Seto et al, 2014).

Thus, similarly to analysis of ES-V1 INs, both known and novel TFs expressed in dP4 progenitors are

revealed by the RNA-seq profiling of ES-dI4 INs, suggesting that not only do ES-dI4 INs recapitulate in

76

vivo developmental programs, but expression profiling of in vitro-derived neurons may uncover new

candidate factors for probing development of dI4 INs specifically.

V1 and dI4 INs share spinal inhibitory neuron identity, manifested by common expression of TFs

(Pax2, Lhx1/5) required to maintain inhibitory neuron status and the downstream genes they regulate

(Gad1 and Gad2, encoding for Gad65 and Gad67, respectively), as well as many other generic spinal

neuronal genes (Fig 2.10D). Nevertheless, V1 and dI4 INs in the spinal cord have highly distinct

functional properties, including settling in different spinal laminae and distinct synaptic target selections

(Betley et al, 2009; Bikoff et al, 2016). Accordingly, many genes are differentially expressed between Day

8 ES-V1 and dI4 INs, with 361 genes enriched versus 212 genes downregulated in ES-V1 compared to

dI4 INs (log2FC >2.0, padj <0.05). For example, genes enriched in Day 8 ES-V1 INs include En1, Foxd3,

FoxP2, and Calb1, while ES-dI4 INs express Tfap2a, Tfap2b, Lbx1, and Npy, among others (Fig 2.10D).

These data indicate that directed differentiation of ESCs towards V1 or dI4 INs produces two highly

distinct IN cell types, on par with their in vivo counterparts (Bikoff et al, 2016).

To assess the similarity of ESC-derived INs with spinal neurons, I compared RNA-seq gene

expression profiling data from Day 8 ES-V1 and dI4 INs with microarray-based profiling data generated

for V1 and dI4 INs from e12.5, p0 and p6 mouse spinal cord (Bikoff et al, 2016). In addition, I performed

differential expression analysis using RNA-seq data from ESC-derived MNs and ESCs to determine

whether ES-V1 and dI4 INs were more similar to ESCs or other ESC-derived neurons than to their IN

counterparts in vivo (Wichterle et al, 2002; Rhee et al, 2016; M. Closser, unpublished; Stadler et al,

2011). Day 8 ES-V1 INs, Day 6 ES-MNs and Day 0 ESCs were compared against Day 8 ES-dI4 INs to

determine their relative FPKM levels for each transcript (expressed as log2FC). These differential

expression values were then cross-referenced to the most highly differentially expressed genes between

V1 versus dI4 INs from spinal cord at e12.5, p0, and p6 (FC > 3.0, p-value <0.02) (Bikoff et al, 2016).

Hierarchical clustering ESCs, ES-V1 INs, ES-MNs, and spinal V1 INs at e12.5 and p0 indicates that ESC-

derived V1 INs share similar transcriptional profile with spinal V1 INs, especially p0 V1 INs, than with

other ES cell types (Fig 2.10E). Interestingly, ESCs and ES-MNs were more similar to e12.5 spinal V1 INs

than ES-V1 INs or p0 spinal V1 INs; the significance of this result is unclear.

77

Transplant of ESC-derived interneurons into developing neural tube In addition to assessing the molecular profile of ESC-derived spinal INs, we also tested them for

functional properties exhibited by these cell types in the spinal cord. Once born, developing V1 INs

migrate away from the p1 progenitor domain and settle ventrally in the ipsilateral spinal cord, occupying

laminae VII and IX by p20. Following migration, V1 INs project axons ipsilaterally into the ventral horn and

rostrocaudally one to two segments away (Alvarez et al, 2005; Sapir et al, 2004). In contrast, Ptf1a-

derived dI4 INs migrate into superficial laminae I and II or the intermediate zone of the dorsal horn of the

spinal cord, with some dI4 INs projecting axons into the ventral spinal cord to innervate sensory afferent

terminals (Glasgow et al, 2005; Betley et al, 2009). To test the migratory and axon guidance properties of

ESC-derived V1 and dI4 INs, I transplanted early (Day 5) En1-tdTomato and Day 6 Ptf1a-tdTomato EBs

containing mixed populations of FP and non-FP cells into the central canal of the developing chick neural

tube at Hamburger-Hamilton Stage 16 (HH16), a time when endogenous V1 and dI4 spinal INs are born

(Fig 2.11A) (Hollyday & Hamburger, 1977). Four days after transplantation, at HH30, I analyzed the

distribution of ESC-derived V1 vs dI4 INs along the dorsoventral axis. In vitro-generated V1 INs migrated

into appropriate ventral spinal domains and project axons locally and along designated V1 IN trajectories

within the ventral funiculus (Fig 2.11B,C). In contrast, transplanted ES-dI4 INs migrate dorsally and send

axons into both dorsal and ventral compartments of the spinal cord (Fig 2.11B,C).

Therefore, based on the results presented, V1 and dI4 inhibitory spinal INs can be efficiently

generated from ESCs using developmentally-relevant extrinsic signals; recapitulate appropriate steps of

in vivo molecular development; and, most remarkably, acquire class-specific functional properties, in

particular cell migration and axon guidance.

Discussion Generation of spinal inhibitory interneurons from stem cells

While spinal INs are important components of sensorimotor circuits, efforts to study the

specification, subtype diversity and function of distinct IN classes have been impeded by the relatively

small numbers of these cells in the spinal cord, as well as difficulty in isolating these populations from

78

primary tissue. Embryonic stem cell-derived neurons have the potential to circumvent these obstacles,

but while the directed differentiation of ESCs to spinal MNs is well-established, whether similar

approaches can lead to efficient differentiation of spinal INs has not been convincingly demonstrated

(Wichterle et al, 2002). Indeed, previous attempts to differentiate spinal IN cell types from mouse and

human ESCs or induced pluripotent stem cells (iPSC) have met with variable success with generally low

efficiencies. Furthermore, in these studies, identification of different spinal IN cell types relied on detection

of TFs and other markers that are expressed only transiently (e.g. En1, Ptf1a, Chx10, etc) and/or are not

specific to single neuronal classes (e.g. Isl1/2, Lhx3, Pax2, Pou4f1) (Iyer et a, 2016; Xu et al, 2015;

Brown et al, 2014; Meinhardt et al, 2014; Murashov et al, 2005; Maury et al, 2015).

While several of studies have reported differentiation of excitatory V2a and V3 spinal INs from

ESCs, the production of either ventral or dorsal spinal inhibitory IN cell types has not been systematically

investigated (Iyer et al, 2016; Xu et al, 2015; Brown et al, 2014). Differentiation of pluripotent stem cells

into non-spinal inhibitory INs has also previously been reported, including cortical and cerebellar INs, but

these protocols generally produce IN progenitors that require additional transcriptional programming or

reintroduction into the cortical environment via transplantation to acquire cell type-specific identity

(Watanabe et al, 2005; Li et al, 2009; Maroof et al, 2010, 2013; Goulburn et al, 2011; Danjo et al, 2011;

Au et al, 2013; Nicholas et al, 2013; Maroof et al, 2013; Salero & Hatten, 2007).

Thus, the advantages of the current study are two-fold: First, I derived lineage-reporter stem cell

lines to indelibly and specifically mark successfully differentiated cells, enabling us to track the molecular

and functional development of these cell types, as well as to isolate them using FACS for cell-specific

analyses (e.g. gene expression profiling or synapse formation assays). Second, I have developed

optimized protocols for directed differentiation of ESCs into two specific spinal inhibitory IN cell types, V1

and dI4 INs, using developmentally-appropriate extrinsic signals. Indeed, these results highlight how the

understanding of normal CNS development can lead to rational strategies for generating predefined

classes of neurons from pluripotent stem cells, in particular through systematic modifications of the type,

concentration and timing of extrinsic patterning factors. In this study, we show here that V1 and dI4 INs

can indeed be efficiently differentiated through systematic optimization of extrinsic factors, with rates

79

comparable to the best MN differentiation from mESCs. Whether other spinal neuron cell types can also

be generated using the same minimal approach remains to be determined.

Shh-mediated generation of ventral neuron cell types from ESCs

The combination of high RA and low (5 nM SAG) is able to efficiently generate ventral spinal

neurons with molecular and functional properties of V1 INs. The range of SAG concentrations producing

V1 INs is relatively narrow (0.5 to 5 nM), with virtually no FP cells generated with 0 or 10 nM SAG. At

these low concentrations of SAG, other ventral spinal neuron cell types are generated, although at much

lower numbers (<5% of all cells), including Evx1/2-expressing V0 INs, Chx10-expressing V2a INs, and

Isl1/Hb9-expressing spinal MNs. Based on ICC and RNA-seq results, Lhx3-expressing V2 INs are likely

the second most common cell type generated. Previous studies have shown that Lhx3-expressing V2

INs, especially excitatory Chx10+ V2a IN subtype, are relatively enriched at 50 nM SAG, while spinal

MNs are generated most efficiently at 500 nM to 1 µM range (Wichterle et al, 2002; Chen et al, 2011;

Maury et al, 2015). These in vitro differentiation results reflect the dorsoventral patterning of the spinal

progenitor domains giving rise to these different IN cell types, which itself is a response to the Shh

gradient from the floor plate of the developing neural tube, providing further confirmation that we can

harness knowledge of in vivo patterning processes to rationally generate different neuron cell types from

ESCs (Jessell, 2000; Briscoe & Ericson, 2001).

TGFß signaling and dorsal interneuron specification

Although it has long been known that roof-plate derived signals, especially TGFß family members

like BMPs, are important for the specification of the most dorsal spinal neuron types (dI1-dI3 INs), the

precise role of roof-plate signaling in the specification of more intermediate cell types such as dI4-dI6 INs

is less clear (Lee & Jessell, 1999; Helms & Johnson, 2003). Here, I used ESC-derived dorsal spinal

progenitors to show that BMP4 signaling is indeed sufficient to induce Lhx2/9- and FoxP2-expressing cell

fates, corresponding to dI1-dI2 INs, respectively, at the expense of Isl1/2-expressing dI3 INs or Lbx1-

dependent dI4-6 INs, consistent with previous studies (Millonig et al, 2000; Lee et al, 2000; Timmer et al,

2002). Conversely, ActivinA, another TGFß family member, preferentially induces the formation of Ptf1a-

80

expressing dI4 INs, with a small increase in dI3 INs and Pou4f1/Lmx1b-expressing dI5 INs. This result

significantly diverges from earlier findings using in ovo chick electroporation of constitutively active

ActivinA-specific receptor constructs (caALK4), which showed that while Activin signaling is not involved

in spinal patterning, overexpression of caALK4 is sufficient to specifically generate excess dI3 INs by

promoting their precocious differentiation without affecting any other dorsal IN class (Timmer et al, 2005).

The finding that ActivinA signaling acts in a graded fashion to induce formation of Isl1-expressing dI3 INs

in chick neural explants without changing other IN cell types was also reported by Liem et al (1997),

although this study did not look at the specification of intermediate dorsal IN types such as dI4 INs. Our

results show that although ActivinA induces more Isl1/2-expressing dI3 INs than in RA only treated

cultures, the biggest effect was observed in the production of Ptf1a-derived dI4 INs.

One hypothesis for this difference is that high RA signaling during ESC differentiation biases the

formation of intermediate dorsal progenitors, which are then acted on by ActivinA ligands to promote dI4

IN cell fates over other cell types (Wichterle et al, 2002). Since increasing concentrations of ActivinA

produces more Ptf1a-expressing cells in EBs, ActivinA likely regulates Ptf1a expression in neural

progenitors to activate dI4 IN specification (data not shown). Interestingly, ActivinA signaling can induce

formation of mesendodermal tissue from ESCs, which can be further differentiated into Ptf1a-expressing

pancreatic lineages, providing further evidence that ActivinA can activate Ptf1a-related programs in the

appropriate cellular context (Kubo et al, 2004; Delaspre et al, 2013). Since high RA signaling is sufficient

to neuralize developing tissues with the majority of cells expressing either NeuN or GFAP, Ptf1a-

tdTomato FP cells in our system are unlikely to represent pancreatic progenitor cells.

An alternative hypothesis is that electroporation of caALK4 is not sufficient to generate a graded

response that would affect the formation of distinct spinal domains. Although Timmer et al (2005)

modified the levels of caALK4 expression by changing construct concentrations as well as through use of

a weaker promoter driving caALK4 expression, they reported that these changes only either enhanced

the generation of dI3 INs or had no effect. However, the range of concentrations used in the chick

electroporation experiments may either be insufficient to activate more intermediate dorsal spinal neural

fates such as dI4 INs, or conversely, function to repress dI4 IN fates while specifying dI3 INs. Thus, in the

81

ESC-based system, the systematic titration of extrinsic signals might better replicate concentration

gradients in vivo, revealing novel actions of ActivinA on dorsal spinal patterning and neurogenesis.

Cross-repression of glutamatergic versus GABAergic dorsal interneuron fates

During RA+ActivinA-induced differentiation of ESCs into dI4 INs, a larger percentage of all FP

cells co-expressed the glutamatergic dI5/dILB marker Lmx1b (13.1% compared to 6.7% in RA only

cultures). The balanced formation of excitatory and inhibitory neurons in the dorsal spinal cord depends

on the complementary expression of TFs specifying glutamatergic fates (e.g. Tlx1/3, Lmx1b) and those

specifying GABA/glycinergic fates (e.g. Pax2, Lhx1/5) (Qian et al, 2002; Ding et al, 2004; Cheng et al,

2004; Burrill et al, 1997; Batista & Lewis, 2008; Pillai et al, 2007). By inhibiting Lbx1, Tlx3 functions to

repress the default GABAergic fate defined by Pax2 expression in order to determine glutamatergic fate

(Cheng et al, 2004; Cheng et al, 2005). Thus, Ptf1a-derived cells, which express high levels of Pax2,

should not express Tlx1/3 or Lmx1b. As such, spinal neurons in EBs as well as embryonic spinal cord

never co-express Pax2 or Lmx1b. What, then, are Lmx1b-expressing FP cells, and how do they arise?

One potential explanation for the ectopic expression of Lmx1b in Ptf1a-FP cells in Day 8 EBs is

that RA+ActivinA induces bipotent progenitors capable of generating either dI4 or dI5 IN cell types. In the

spinal cord, Ascl1-expressing progenitors can give rise to either dILA (Pax2-expressing) or dILB (Lmx1b-

expressing) (Mizuguchi et al, 2006). As such, although ActivinA-driven Ptf1a expression is sufficient to

activate the Pax2-specific transcriptional program in the majority of Ascl1-lineage cells fated to become

inhibitory neurons, a subset of these might be incompletely specified and thus unable to activate Pax2

expression and/or repress glutamatergic Lmx1b expression. Intriguingly, Notch signaling may be involved

in this secondary process since inhibition of Notch signaling using DAPT suppresses the formation of

Lmx1b-expressing Ptf1a-FP cells while also increasing the number of FP cells co-expressing Pax2. Some

evidence for this comes from recent studies showing that Notch is required for the cell fate switch

between glutamatergic dILB and GABAergic dILA INs in the spinal cord (Mizuguchi et al, 2006).

Notch signaling and the fate of Ptf1a-derived progenitors

During dI4 IN development in vivo, Ptf1a acts in a trimeric complex with an E-protein and the TF

82

RBPJ to specify GABAergic neuronal identity. While RBPJ normally acts to transduce activated Notch

signaling to affect neurogenesis, RBPJ function in this complex is thought to be independent of Notch

(Beres et al, 2006; Hori et al, 2008; Borromeo et al, 2014). Nevertheless, a role for Notch signaling has

been suggested for the specific formation of late-born inhibitory dILA INs; in particular, low Notch levels in

conjunction with Ascl1 predispose the formation of inhibitory dILA over excitatory dILB INs (Mizuguchi et

al, 2006). Our results indicate that inhibition of Notch signaling through using the gamma-secretase

inhibitor DAPT is sufficient to promote neural differentiation. Whether the extranumerary Ptf1a-FP cells

formed under DAPT-treated conditions preferentially adopt early-born dI4 versus late-born dILA IN

identity remains to be determined. Finally, DAPT-mediated inhibition of Notch signaling also leads to

enrichment of Pax2-expressing Ptf1a-FP cells in RA+ActivinA conditions. As noted earlier, Notch

signaling is not required for generation of dI4 INs, but its role in controlling cell fate specification within the

dP4 domain requires further investigation.

Besides its effect on neurogenesis, Notch signaling can impinge on gliogenesis by regulating the

timing of neurogenesis of bipotent progenitors (Nye et al, 1994; Wang et al, 1998). Since Notch functions

to preserve a pool of neural stem cells by preventing their differentiation, inhibition of Notch would be

predicted to deplete this progenitor pool, resulting in the formation of earlier-born neurons at the expense

of glia formation. Notch signaling has also been shown to be directly involved in gliogenesis (Wang &

Barres, 2000; Morrison et al, 2000; Gaiano et al, 2000; Furukawa et al, 2000). In the pMN domain of the

spinal cord, Olig2-expressing neural progenitors first produce MNs, followed by oligodendrocytes and

astrocytes (Masahira et al, 2006; Novitch et al, 2001; Zhou & Anderson et al, 2002; Zhou et al, 2001).

Here, Notch signaling acts to maintain Olig2-expressing progenitors in a proliferative state; inhibition of

Notch using DAPT leads to depletion of Olig2+ progenitors and enhanced neurogenesis (Tan et al, 2016).

Similarly, DAPT-mediated inhibition of Notch signaling during dI4 IN differentiation in vitro results in

enhanced neurogenesis and reduced astrocyte generation; whether Notch directly controls gliogenic fate

through prevention of progenitor differentiation remains to be determined.

83

Transcriptional profiles of ESC-derived V1 and dI4 interneurons

In this study, we show that ES-V1 and dI4 INs express molecular markers specific to these spinal

domains, including progenitor and postmitotic TFs, as revealed through ICC and genome-wide RNA-seq

expression profiling. While these studies confirm existing data about the molecular identity of these cells,

the RNA-seq datasets also provide a unique opportunity to assess factors that might be involved in the

specification of these distinct neuronal cell types since ESC-derived INs give us unique access to some of

the earliest developmental stages. Indeed, our screen of ES-V1 INs identified Prdm12 as a novel V1 IN-

specific progenitor regulator, which has been recently supported by in vivo genetic manipulations (Thélie

et al, 2015; Zannino et al, 2014). Since Day 5 En1-tdTomato EBs contain p1 progenitors as well as early

postmitotic V1 INs, the ESC-based system can provide insights into V1 development not readily available

with current genetic tools.

Similarly, expression profiling of ES-dI4 INs reveals that they express high levels of the TFs

TFAP2a and TFAP2b compared to ESCs, ES-V1 INs, and ES-MNs, suggesting that these factors are

specific to dI4 IN development. Microarray profiling of Ptf1a-derived cells from developing spinal cord

have also shown that these cells express high levels of TFAP2b, but these studies were performed

relatively late, evidenced by lack of expression of most progenitor markers, including Ptf1a (Wildner et al,

2013; Bikoff et al, 2016). Thus, while TFAP2b may function as a marker of a subset of early-born dI4 INs,

as suggested by recent studies, the expression pattern of these TFs in ES-dI4 INs indicates that TFAP2a

and/or TFAP2b may function more generally to regulate the development of Ptf1a-lineage cells (Wildner

et al, 2013; Levine et al, 2014; J. Kaltschmidt, unpublished). This hypothesis is consistent with the role of

these TFs in the developing retina, where they have been shown to regulate amacrine cell development,

an inhibitory IN cell type that also originates from a Ptf1a-expressing domain (Jin et al, 2015).

Using the RNA-seq datasets to uncover novel factors enriched in V1 or dI4 INs will also aid in the

identification of genes essential for their specification. In particular, the generation of different cell types

from ESCs can be accomplished through forced expression of developmentally-regulated factors,

generally TFs, in a process that can be as or more efficient than directed differentiation using extrinsic

patterning signals (Kyba et al, 2002; Andersson et al, 2006; Panman et al, 2011; Martinat et al, 2006;

Takahashi & Yamanaka, 2006). For example, recent studies show that spinal MNs can be generated

84

with ~80% efficiency through joint induction of three TFs, Ngn2, Isl1/2, and Lhx3 (Mazzoni et al, 2013a).

Further investigation of the molecular programs establishing V1 and dI4 IN-specific identity may identify

transcriptional modules that can efficiently program generic V1 and dI4 INs from ESCs, even specific

desired subtypes.

Conclusion

Altogether, these studies mark the first demonstration of efficient differentiation of specific spinal

inhibitory INs from pluripotent stem cells. Our results demonstrate that the cells express appropriate

molecular markers associated with the lineages of V1 and dI4 INs and that differentiated cells exhibit cell-

type specific migratory patterns to populate relevant laminae in the developing spinal cord upon

transplantation in vivo. ESC-derived INs can be used to study the roles of extrinsic patterning signals and

intrinsic signals underlying specification of different spinal IN cell types during embryonic development,

including molecular programs underlying their subtype diversification. Such studies may provide novel

insights into how molecularly distinct neuronal cell types acquire cell type-specific function during spinal

circuit formation. Additionally, spinal circuits involving different classes of inhibitory INs, including V1-

derived Renshaw cells, have been implicated in neurological diseases such as amyotrophic lateral

sclerosis (ALS). Richer understanding of developmental programs underlying spinal IN specification will

produce more efficient approaches for differentiating stem cells into clinically-relevant cell types for

modeling disease, drug discovery, and cellular replacement therapy.

85

Figure 2.1

Figure 2.1 Derivation of lineage reporter ESC lines for V1 and dI4 interneuron differentiations A. Immunostaining of mouse spinal cord tissue reveals transient expression of En1 and Ptf1a protein in postmitotic V1 INs and dP4 progenitors, respectively. B. En1::cre and Ptf1a::cre mice crossed to ROSA-lsl-tdTomato or GFP lines (Kimmel et al, 2000; Kawaguchi et al, 2002) enables permanent fluorescent labeling of En1- and Ptf1a-lineage cells. Scale bars = 100 µm.

86

Figure 2.2

87

Figure 2.2 Directed differentiation of V1 inhibitory spinal interneurons from mouse ESCs (A) Strategy for differentiating ESCs to V1 INs using 1 µM RA for neuralization and posteriorization and 5 nM SAG for ventralization of neural progenitors in Day 2 nascent embryoid bodies. Using En1::cre crossed to ROSA-lsl-tdTomato or ROSA-lsl-EYFP (En1-tdTomato and En1-GFP, respectively), En1 lineage cells can be identified by fluorescent reporter expression and isolated using FACS. On Day 8 of differentiation, 28.4% and 32.4% of total cells in the EB are GFP and tdTomato positive. Depicted are Day 8 EBs grown in suspension from both tdTomato and GFP lines with native reporter expression (unstained, left), fixed and cryosectioned EBs with native reporter expression (middle) and dissociated FP cells after 1 week culture on laminin/fibronectin substrate (right). Scale bars = 50 µm. (B) Quantification of fluorescent reporter positive (FP) cells on Days 5-10 of differentiation for both En1-GFP and En1-tdTomato reporter lines using flow cytometry. For all FP cell quantifications using flow cytometry, at least 3 independent differentiations were performed, with values shown as mean ± standard error of the mean (SEM). (C) Quantification of FP cells differentiated with 1 µM RA and different concentrations of SAG (0 to 500 nM). Note that ES-MNs are most efficiently differentiated using 100X SAG concentration (500nM). (D) Transient expression of En1 protein in En1-tdTomato fluorescent reporter cell line. On Day 5 of differentiation, ~75% of FP cells co-express En1, decreasing to ~10% by Day 8 of differentiation.

88

Figure 2.3

Figure 2.3 Molecular developement of V1 interneurons in vivo and in vitro In the embryonic mouse spinal cord, V1 interneurons are born from e9.5 to e12.5 in two waves, the first generating early-born subtypes such as RCs and the second wave generating non-RC subtypes such as FoxP2-expressing V1 INs, with more FoxP2-expressing cells generated compared to RCs (Stam et al, 2012; Benito-Gonzalez & Alvarez, 2012). On the left and bottom axes are the stages of V1 IN development and corresponding days of V1 IN development in vitro, respectively (days in vitro in parentheses). Progenitors from the p1 spinal domain express different homeodomain (HD) and basic-helix-loop-helix (bHLH) TF proteins during ventral spinal patterning, including Pax6, Dbx2, Nkx6.2, Ngn1/2. Recently, the zinc-finger PR-domain containing protein Prdm12 was shown to be specifically expressed in p1 progenitors in the spinal cord. As V1 IN development proceeds, progenitors exit cell cycle and differentiate, upregulating postmitotic TFs such Foxd3, En1, Pax2, Lhx1/5. A subset of V1 INs that later become RCs also selectively activate OC2 expression, and maintain Foxd3 expression while other V1 INs downregulate this protein. Since V1 INs comprise an inhibitory neuron class, mature V1 INs express proteins indicative of inhibitory neurotransmission, including Gad67, which is required for GABA synthesis, and the glycine transporter (GlyT2), which is involved in glycinergic neurotransmission. V1 INs can be grouped into molecularly distinct subpopulations, including Calbindin-expressing RCs, which also express the TFs MafA and MafB in addition to the aforementioned OC1/2. Non-RC V1 INs express other molecular markers including FoxP2, Sp8, and Parvalbumin.

89

Figure 2.4

90

Figure 2.4 ESC-derived V1 interneurons recapitulate in vivo molecular developmental programs (A) Differentiation of ESCs using RA and low SAG (5 nM) results in cells with spinal Hox expression profile (e.g. HoxC6). Notably, forebrain Hox expression is absent (e.g. HoxA2). Furthermore, Day 4-6 EBs exhibit HD TF profiles consistent with ventral spinal progenitor identity, including Pax6 and Nkx6.2, as well as Irx3, Ngn1/2, Dbx1 (not shown). As progenitors become postmitotic, they express TFs indicative of V1 IN identity, including En1, Foxd3, Pax2, and Lhx1/5. (Right) Quantification of En1-tdTomato FP cells co-expressing early (En1, Foxd3) and late (Pax2, Lhx1/5) postmitotic TFs. (B) En1-tdtomato FP cells do not ectopically express TFs expressed in non-spinal En1 lineages (Lmx1b) or non-V1 spinal neurons (Evx1/2, Chx10, Lhx3, Isl1/2). (C) Dissociated En1-tdTomato FP cells express neuronal-specific markers (NeuN) and presynaptic proteins (Synapsin), as well as markers suggestive of inhibitory neuron maturation (Gad67, GlyT2). (D) Dissociated En1-tdTomato FP cells cultured on laminin/fibronectin-coated glass coverslips have simpler neuronal morphology than cells cultured on monolayer of primary cortical astrocytes. Furthermore, V1 INs cultured on astrocytes switch from Gad67 to GlyT2 expression, suggesting functional maturation. (ANOVA *p<0.05). Scale bars = 50 µm EBs, 20 µm dissociated cells.

91

Figure 2.5

Figure 2.5 Directed differentiation of ESCs to dI4 spinal interneurons (A) Strategy for differentiating dI4 INs using 1 µM RA for neuralization and caudalization. Treatment of ESCs with RA alone induces expression of dorsal and intermediate spinal HD and bHLH TFs (Wichterle et al, 2002). (B) ESC lines were derived from Ptf1a::cre mice crossed to ROSA-LSL-tdTomato or Thy1::YFP reporter mice for fluorescent labeling of Ptf1a-derived cells. RA treatment alone yields ~10% Ptf1a-tdTomato FP cells, with less in Ptf1a-Thy1YFP EBs due to mosaic expression. Scale bars = 50 µm. (C) Quantification of FP cells from Ptf1a-tdTomato line over Days 5-9 of differentiation with RA only. (D) Table listing dorsalizing factors and concentrations used to identify signals required for efficient differentiation of dI4 INs from ESCs. Blue text denotes agonists; green text denotes antagonists of the pathway. (E) Quantification of differentiation efficiency on Day 8 using dorsalizing signals. Dark gray bars are BMP-associated agonists and antagonists. Light gray bars are Wnt-associated agonists and antagonists. Red bars are TGFß-associated agonists and antagonists. ActivinA is the most effective agent for differentiation of Ptf1a-tdTomato FP cells from ESCs, yielding ~40% FP cells on Day 8 of differentiation, compared to <10% in control, RA-only treated EBs. (ANOVA, *p<0.05, **p<0.01, ***p<0.001). (F) Quantification of FP cells from Ptf1a-tdTomato line over Days 5-9 of differentiation with RA and 25 ng/mL ActivinA added on Day 2 of differentiation.

92

Figure 2.6

Figure 2.6 ActivinA induces Ptf1a expression in differentiating ESCs (A) Ptf1a protein is transiently expressed in differentiating Ptf1a-tdTomato EBs. Scale bars = 50 µm. (B) ActivinA treatment (25 ng/mL on Day 2) accelerates production of Ptf1a-immunoreactive cells compared to RA-only treated EBs. (C) TGFß antagonists (SB-431542 and Follistatin) affect generation of Ptf1a-tdTomato FP cells by antagonizing ActivinA signaling, especially the Activin-specific inhibitor Follistatin. Shown are quantifications of dissociated Day 8 EBs. (D) Concentration-dependent effect of ActivinA signaling on production of Ptf1a-tdTomato FP cells on Day 8 of differentiation. (E) Timing-dependent effect of ActivinA (25 ng/mL) signaling on production of Ptf1a-tdTomato FP cells on Day 8 of differentiation. ActivinA added on Days 2-4 produce >40% FP cells, with reduced efficiency when added on Day 5 (25.9%) or beyond.

93

Figure 2.7

Figure 2.7 Molecular development of in vitro-generated dI4 interneurons (A) ESC differentiation using RA+ActivinA leads to the induction of dorsal spinal patterning genes, including HD and bHLH TFs (Pax6, Pax7, Ascl1 and Ptf1a) involved in generating the dP4 spinal progenitor domain (Days 4-6 early EBs). As dorsal progenitors become postmitotic, they express markers of dI4 INs in vivo, including Lbx1, Pax2, and Lhx1/5 (Day 8 late EBs). (B) Ptf1a-tdTomato FP cells express the neuron-specific markers NeuN and synapsin, suggestive of neuronal maturation and synapse formation. FP cells also express proteins required for inhibitory neurotransmission, including Gad65, Gad7, and GlyT2. (C) Ptf1a-tdTomato FP cells don’t express TFs specific to other dorsal spinal IN cell types, or VGLUT2, which is used for excitatory neurotransmission. Scale bars = 50 µm.

94

Figure 2.8

Figure 2.8 Comparison of different TGFß family members on dorsal spinal patterning (A) Although RA treatment alone is not sufficient to generate Olig3 and Lhx2/9-expressing dI1 INs, some FoxP2-expressing dI2 INs and Isl1/2-expressing dI3 INs are generated. However, the majority of cells express Ptf1a or Pou4f1/Lmx1b TFs, indicating that the RA only differentiation conditions generates some dorsal but mostly intermediate spinal neuron cell types, as previously reported. (B) Addition of RA and BMP4 (25 ng/µL) on Day 2 of differentiation is sufficient to repress dI4 IN generation, as evidenced by the relative absence of Ptf1a-tdTomato cells as well as Ptf1a-immunoreactive cells on Day 8. BMP4 signaling generates Olig3- and Lhx2/9-expressing dI1 INs, as well as more FoxP2-expressing dI2 INs compared to RA only differentiation. Few dI3-5 INs are generated, with the few Pou4f1-expressing cells presumably generated from dP1 or dP2 domains and not dP5. (C) Treatment of Day 2 EBs with RA and ActivinA (25 ng/mL) produces many Ptf1a-immunoreactive cells on Day 6 and consequently many reporter cells by Day 8 of differentiation. Unlike with BMP4, few dI1 INs were generated, while a small number of dI2, dI3 and dI5 were produced from these conditions. Scale bars = 50 µm.

95

Figure 2.9

Figure 2.9 ActivinA treatment generates distinct cell types, including glia (A, B) Comparison of cell types formed in Day 8 Ptf1a-tdTomato EBs using RA only or with addition of ActivinA (25 ng/mL) added on Day 2. RA+ActivinA produces a significant population of dI4 INs that do not express Pax2 or Lhx1/5 or ectopically express the dI5 marker, Lmx1b. ActivinA signaling is sufficient to generate Ptf1a-expressing dI4 INs and induces GFAP+ glia formation from the dP4 domain. (C) Adding ActivinA on Day 5 of differentiation results in less FP cells generated overall, but ectopic expression of Lmx1b is abolished. However, many Pax2 negative FP cells remain. (D) DAPT added to differentiating Ptf1a-tdTomato EBs on Day 5 with ActivinA results in >90% of FP cells expressing Pax2 and NeuN while GFAP expression is minimized in Day 8 EBs. (E) Quantification of Pax2 or Lmx1b-expressing Ptf1a-tdTomato FP cells in Day 8 RA only compared to RA+ActivinA (added Day 2). (F) Quantification of FP cells co-expressing Lmx1b, Pax2, as well as cells negative for NeuN in RA only or RA+ActivinA (added Day 2) conditions, normalized to differentiation efficiency (9.6 vs 39.0%, respectively). (G) DAPT treatment in differentiating Ptf1a-tdTomato EBs leads to a small but significant increase in FP cells generated on Day 8 when added on Days 5 or 6. (H) Quantification of FP cells generated with ActivinA and/or DAPT treatment on Day 2 vs Day 5 of differentiation. Scale bars = 50 µm.

96

Figure 2.10

97

Figure 2.10 RNA-seq gene expression profiling of ESC-derived spinal interneurons (A) Most highly enriched genes (log2 fold-change, L2FC) in FACS purified ESC-derived V1 progenitors (Day 5) compared to dI4 progenitors (Day 6) (p-adj <0.01). En1 and Ptf1a are highly enriched transcripts in V1 and dI4 INs, respectively. (B) Genes up- and down-regulated (red and green, respectively) in ESCs, Day 5 and Day 8 ESC-derived V1 INs compared to dI4 INs. Hierarchical clustering analysis suggests that Day 5 V1 INs are in transition from ESCs to more mature Day 8 V1 INs. (C) Comparison of gene expression of ESCs, Day 5 and Day 8 V1 INs centered on the most enriched genes in Day 5 V1 INs. (D) Comparison of gene expression of ESC-derived V1 and dI4 INs on Day 8 (mean L2FC of FPKM values). (E) Comparison of ESCs, ESC-derived MNs, ESC-derived V1 INs (RNA-seq) and En1-derived V1 INs dissociated from e12.5 and p0 spinal cord (microarray, Bikoff et al, 2016). Plotted are L2FC values relative to ESC-derived dI4 INs (ESCs, ESC-MNs or ESC-V1 INs) or spinal Ptf1a-derived dI4 INs. ESC-V1 INs cluster closest to spinal p0 and e12.5 V1 INs.

98

Figure 2.11

Figure 2.11 Distinct migration and axonal projections of transplanted ES-V1 and dI4 interneurons (A) Schematic of transplantation paradigm into developing chick neural tube. Early En1-GFP (Day 5) or Ptf1a-tdTomato (Day 6) EBs were grafted into lesioned Hamburger-Hamilton (HH) Stage 16 chick embryonic spinal cord and examined 4 days later for differences in their migratory patterns and/or axonal projections. (B) ES-V1 INs migrate and project axons in the ventral horn, while ES-dI4 INs migrate dorsally and project axons in the dorsal white matter (n ≥4). On the right, quantification of En1-GFP or Ptf1a-tdTomato cell migration into distinct regions of the spinal cord (6 equivalent dorsoventral bins). (C) Additional examples of the axonal trajectories of transplanted ES-dI4 and V1 INs. Scale bars = 100 µm.

99

Chapter 3. Renshaw cell subtype specification Introduction

During neural circuit formation, neurons acquire distinct subtype identities that determine their

ability to recognize their appropriate synaptic partners. V1 INs in the spinal cord exhibit extensive

molecular heterogeneity, the consequence of which is division of the parental population into more than

fifty molecularly distinct subtypes based on combinatorial expression of 19 different TFs (Francius et al,

2013; Bikoff et al, 2016; Gabitto et al, 2016). However, developmental mechanisms underlying V1

progenitor domain diversification are largely unknown, partly due to lack of genetic tools to specifically

access V1 progenitors in vivo. Thus, the ability to generate V1 progenitors from ESCs provides a unique

opportunity to study V1 subtype diversification in a simplified and experimentally accessible system.

Previously, I established the feasibility of using ESC-derived V1 INs to model and study

mechanisms of V1 subtype diversity and connectivity. ES-V1 INs acquire molecular characteristics of

their in vivo counterparts and migrate and project axons into proper spinal domains following

transplantation into the developing neural tube. Remarkably, in vitro-generated V1 INs also exhibit

significant molecular diversity with similar enrichment of TFs defining distinct V1 IN subpopulations in

vivo. While the majority of V1 IN subtypes in the spinal cord have not yet been characterized, some ES-

V1 INs acquire molecular features of Renshaw cells (RC), one of the few well-studied spinal neuron cell

types known for their role in mediating recurrent inhibition of MNs (Renshaw, 1946; Eccles et al, 1954;

Sapir et al, 2004; Alvarez et al, 2005; Alvarez & Fyffe, 2007; Alvarez et al, 2013). Although all V1-derived

neurons express the TF En1, RCs are distinguished by their exclusive co-expression of En1 and the

calcium-binding protein, calbindin (Cb). Nevertheless, how RCs are uniquely specified from other V1

populations is unclear, although recent studies have shown that RCs are born earlier than other V1 INs

and transit through distinct transcriptional states (Stam et al, 2012; Benito-Gonzalez & Alvarez, 2012).

Notch signaling has been shown to be important for controlling timing of progenitor cell

differentiation in other CNS cell types (Artavanis-Tsakonas et al, 1999). High Notch activity maintains

neural progenitors by upregulating TFs that repress the expression of proneural basic-helix-loop-helix

(bHLH) genes. When Notch activity is low, proneural bHLH TFs are de-repressed to initiate progenitor cell

exit from cell cycle and neuronal differentiation (Yoon et al, 2004; Yoon & Gaiano, 2005). Inhibition of

100

Notch signaling early during neurogenesis results in precocious neuronal differentiation and enrichment of

early-born neuronal subtypes at the expense of later-born neurons. Thus, Notch may similarly function to

control timing of V1 IN generation, contributing to V1 subtype diversification. In addition to having a

permissive role in regulating timing of neurogenesis, Notch has also been shown to have an instructive

role in specifying alternative cell fates, indicating that Notch signaling may promote V1 subtype diversity

by selecting and/or repressing specific V1 subtype identities (Jadhav et al, 2006a; Jadhav et al, 2006b;

Yaron et al, 2006; Livesey & Cepko, 2010; Mizeracka et al, 2013).

Previous studies have used global Notch manipulations to conclude that Notch signaling is not

involved in V1 IN development (Stam et al, 2012; Marklund et al, 2010; Ramos et al, 2010). However,

these analyses are likely to be confounded by non-cell autonomous effects of disrupting Notch signaling

in all cell types, as well as by the lack of temporal specificity in global knockout models. Yet, accurate

assignment of the role of Notch in V1 subtype specification requires the ability to specifically manipulate

Notch signaling in V1 progenitors, a difficult task given the current lack of genetic tools to access the p1

progenitor domain. Here, I used an ESC-based system to show that manipulations of the Notch signaling

pathway in in vitro-generated p1 progenitors impinges on V1 neurogenesis and subtype differentiation

and acts to significantly enhance the generation of RCs.

In addition to Notch signaling, I also examined the role of extrinsic signals for specifying V1

subtype identity. During embryonic development, RA produced from the paraxial mesoderm is required to

neuralize and caudalize cell fates towards spinal neuron identity, as well as to pattern ventral progenitor

domains giving rise to MNs and different classes of INs, including V1 INs (Jessell, 2000; Pierani et al,

2001; Novitch et al, 2003). Later, RA is specifically produced by Raldh2-expressing LMCm MNs in limb-

innervating regions of the spinal cord in order to specify LMCl subtype identity, suggesting that RA has a

regionally restricted role in the specification of MN subtype identity (Sockanathan et al, 2003). Previous

studies suggest that calbindin-expressing RCs may be more abundant at brachial, or limb-innervating

regions, of the developing spinal cord, suggesting a spatial component to RC specification, as well as a

potential role for MN-derived signals to control IN subtype identity to match motor circuit demands

(Francius et al, 2013). To test this hypothesis, I used ESC-derived V1 INs to determine if RA signaling is

required for RC generation compared to other V1 subtypes. By studying how RA and/or Notch signaling

101

informs the specific generation of RCs, we can begin to infer the logic and mechanism of the extensive

subtype diversity within the V1 IN class.

Results

Molecular heterogeneity of ESC-derived V1 interneurons

Using RNA-seq gene expression profiling and ICC approaches, I found that ESC-derived V1 INs

exhibit significant subtype diversity at the molecular level. RNA-sequencing was performed on ES-V1 and

dI4 INs on Day 8 of differentiation to identify genes enriched in V1 compared to dI4 INs, with log2FC

cutoff >2 (Fig 2.10). The FPKM fold-change enrichment from this profiling approach was compared to

microarray-based fold-enrichment of gene expression in e12.5, p0, or p6 V1 INs from mouse spinal cord.

I focused on the expression of the select 19 TFs enriched in V1 over dI4 INs in the spinal cord, used to

divide V1 INs into molecularly and spatially distinct subpopulations (Bikoff et al, 2016; Gabitto et al, 2016).

Remarkably, ESC-derived V1 INs show ≥3 fold-change enrichment (FPKM) in 15 of the 19 TFs

(Fig 3.1). As apparent in our earlier comparison of ESC-derived neurons with spinal cord neurons, ES-V1

INs more closely resemble p0 and p6 than e12.5 stage V1 INs in their expression of these 19 TFs (Fig

2.10E, Fig 3.1). MafA, MafB and Prox1 are expressed, but not enriched, in ESC-derived V1 INs

compared to ES-dI4 INs (FPKM range: 1.98-62.57; MafA: 31.16, MafB: 12.83, Prox1: 12.38). These TFs

are also either not enriched in the spinal cord dataset (MafB and Prox1) or missing from the microarray

(MafA), but were included because of their localized expression in the ventral spinal cord. Nr4a2 is also

not differentially expressed in ES-V1 INs or in the spinal cord at e12.5 or p0, but is moderately enriched at

p6 (Fig 3.1). Altogether, these expression profiling results indicate that ESC-derived V1 INs adopt the

molecular signature of in vivo V1 IN subtype diversity.

Immunostaining for several of these markers in EBs confirmed their expression in subsets of cells

(Fig 3.2A). Under RA/low SAG conditions, a small percentage of cells expressed Pou6f2 or Sp8 TFs

(<2% of FP cells), while Nr5a2 and FoxP1-expressing subsets were significantly larger (7.8 and 31.0% of

FP cells, respectively) (Fig 3.2B). By comparison, Pou6f2, Sp8, Nr5a2, and FoxP1-expressing cells mark

13, 13, 10, and 25% of all V1-derived INs in the lumbar spinal cord at p0 and p6 (Bikoff et al, 2016).

102

Nonetheless, although Sp8-expressing V1 INs were studied for their sensory-motor connectivity and the

physiological properties of the Pou6f2 subtype were examined by Bikoff et al (2016), the precise function

of these as well as most of the ~50 molecularly-defined V1 IN subtypes is currently unknown. Indeed,

while it is proposed that molecularly distinct V1 IN subtypes have specialized functional roles in spinal

motor circuits, how different types of V1 INs impinge on MN excitability remains to be determined.

Among the few well-studied IN cell types in the spinal cord are Renshaw cells (RC), which are

derived from the V1 domain and uniquely provide recurrent inhibition to MNs (Sapir et al, 2004). To

determine if RCs are produced during ESC to V1 IN differentiation under RA/low SAG conditions, I

performed ICC to detect calbindin (Cb) expression. While some non-RCs within the V1 IN class may

transiently express low levels of Cb, especially nascent V1 INs, Cb is most highly induced in RCs and its

expression persists into the adult (Geiman et al, 2000; Mentis et al, 2006; Alvarez et al, 2005; Alvarez &

Fyffe, 2007). Remarkably, during ESC differentiation to V1 INs, 16.5% of FP cells acquire Cb expression

in Day 8 EBs (Fig 3.2C). Conversely, >90% of Cb-expressing cells co-express the En1-tdTomato or En1-

GFP reporter on Day 8 (data not shown). For comparison, ESC-derived V1 INs also differentiate into a

larger subset marked by the expression of the TF FoxP2 (33.4% of all FP cells). FoxP2-expressing V1

INs have been shown to be both molecularly and functionally distinct from RCs in vivo, thus providing a

useful counterpart for subsequent analyses of V1 subtype-specific development and function. Indeed,

FoxP2 and Cb expression are non-overlapping in Day 8 EBs (Fig 3.2C). In the spinal cord, RCs comprise

~10% of all lumbar V1 INs at any time, while FoxP2 encompasses ~34% in p0 lumbar spinal cord (down

from ~50% in e12.5 animals), indicating that our in vitro system approximates the prevalence of these

subtypes in the spinal cord (Morikawa et al, 2009; Stam et al, 2012; Bikoff et al, 2016).

Out of these 6 molecular markers, 4 demarcate non-overlapping subpopulations of V1 INs –

Pou6f2, Sp8, FoxP2 and Cb. Together, these four factors constitute ~69% of all V1 INs at p0 and 51% of

ESC-derived V1 INs on Day 8 of differentiation using RA/low SAG conditions (Bikoff et al, 2016; data not

shown). Intriguingly, the magnitude of expression of some of these markers changes with adjustments in

timing and/or conditions of in vitro differentiation. For example, Sp8-expressing cells were many times

more prevalent on Day 12 of differentiation compared to Day 8 (0.23 vs 5.62% of FP cells), suggesting

103

that their development is delayed compared to other V1 subtypes (Fig 3.2D). By comparison, Cb and

FoxP2 expression in En1-FP cells is stable from Days 8 to 14 of in vitro differentiation (data not shown).

Interestingly, we observed a small, yet significant increase in Pou6f2 and Nr5a2-expressing cells

at 0.5 nM concentration SAG compared to 5 nM (Pou6f2: 3.0 vs 0.76%; Nr5a2: 15.8 vs 7.8%, p<0.001),

while FoxP2-expressing cells decreased as SAG concentrations increased from 0.5 to 5 to 10 nM (45.0

vs 33.3 vs 3.9%). Meanwhile, at 10 nM concentration of SAG, Cb-expressing cells constitute 33.4% of all

FP cells generated, a 31% increase from standard RA/low SAG differentiation conditions (Fig 3.2E, data

not shown). However, at 10 nM, V1 IN differentiation is reduced to 6.1% from 31.6% (Fig 2.1C, Fig 3.2E).

These results suggest that Shh signaling may not only be required for the proper patterning of the p1

progenitor domain to generate the correct number of V1 INs, but may function in dorsoventral spatial

patterning of the p1 domain to generate molecularly distinct subtypes (See discussion).

Characterization of ESC-derived Renshaw cells

In order to confirm that Cb-expressing V1 INs acquire other molecular properties of RCs, I used

ICC to detect for TFs known to be important for RC-specific development in the spinal cord, including

Onecut1 (OC1), OC2, and MafB (Fig 3.3A) (Stam et al, 2012). At e12.5 in mouse brachial spinal cord,

about 10% of En1-expressing cells express OC2, while 26% of Cb-expressing V1 INs co-localize OC2

(data not shown). Similarly, in p0 spinal cord, ~17% of V1 INs express OC2, while 10% express OC1 and

15-25% express MafB (Bikoff et al, 2016; Benito-Gonzalez & Alvarez, 2012). While the expression of

OC1/2 in Cb-expressing V1 INs in the postnatal spinal cord has not yet been addressed, a recent study

reported that MafB is expressed in 100% of RCs residing in the ventrolateral spinal cord (Benito-

Gonzalez & Alvarez, 2012; Alvarez & Fyffe, 2007). In addition to these other factors, the TF MafA is

expressed in 5% of V1 INs at p0, with 47% of Cb-expressing V1-derived cells expressing MafA. By

comparison, 62% of RCs express MafA at e13.5, suggesting a developmental component to MafA

expression in RCs. While the role of MafA in V1 INs has not yet been determined, MafA expression in

combination with MafB and OC2 represents one of four distinct clades (besides Pou6f2, FoxP2 and Sp8)

of V1 INs within which RCs reside (Bikoff et al, 2016).

104

ESC-derived RCs on Day 8 also express these markers, albeit at lower levels than observed in

vivo. While approximately 10% of V1 INs express MafA, MafB, or OC2, only ~2% of Cb-expressing V1

INs do (Fig 3.3B). Potentially, these TFs may be more dynamically expressed in in vitro-generated RCs,

and are downregulated by Day 8 of differentiation. Indeed, assessment of these markers on Day 6

indicates that they are all more highly induced during early differentiation (data not shown). Relatedly,

although MafB, OC1 and OC2 are initially required for the generation and maintenance of RC identity, at

later stages of development these TFs, along with MafA, might specify distinct RC subtypes, with each TF

constituting only a small portion of Cb-expressing V1 INs. Evidence for this last hypothesis arises from

previous reports about the morphological heterogeneity among RCs in the spinal cord (Fyffe, 1990).

In addition to molecular profiling of ESC-derived RCs, I also performed functional testing of this

V1 IN subtype using the transplantation paradigm for chick neural tube. During mouse development,

subsets of V1 INs settle in distinct locations in the ventral spinal cord (Alvarez et al, 2005; Bikoff et al,

2016). While FoxP2-expressing V1 INs are broadly distributed, RCs cluster in the ventrolateral aspect of

the ventral horn, adjacent to MN axons (Stam et al, 2012; Benito-Gonzalez et al, 2012, Bikoff et al, 2016).

To test the migratory properties and spatial positioning of ESC-derived RCs, I transplanted Day 5 En1-

GFP EBs into HH16 chick neural tube and assessed the subtype distribution along the mediolateral and

dorsoventral axes 3 days later in the HH30 chick spinal cord. Since Cb expression is virtually absent in

the chick spinal cord at this stage, any apparent Cb signal originates from transplanted mouse derived-

cells (data not shown). Surprisingly, while R-interneurons, the avian equivalent of RCs, are dorsomedial

to MN pools in the chick spinal cord, ESC-derived RCs migrate ventrolaterally, occupying a restricted

region in the ventral horn embedded with MN cell bodies and neurites (Wenner & O’Donovan, 1999; Xu et

al, 2005). Conversely, FoxP2-expressing V1 INs have wider distribution along both the mediolateral and

dorsoventral axes in the ventral horn (Fig 3.3C). While chick transplantation does not address if the

migratory properties of V1 INs are intrinsically determined or established by extrinsic cues, mouse-

derived V1 IN subtypes nonetheless exhibit spatial properties reminiscent of their counterparts in vivo.

Inhibition of Notch signaling enhances Renshaw cell differentiation

How RCs are differentially specified from other V1 INs is not well established, yet understanding

105

the molecular mechanisms underlying RC differentiation will provide insights into the logic of V1 subtype

diversification, as well as potential routes to more efficient generation of functionally important and

clinically-relevant cell types such as RCs (Raynor & Shefner, 1994; Li, et al, 2000; Mazzocchio & Rossi,

2010; Casas et al, 2013; Wootz et al, 2013; Ramirez-Jarquin et al, 2014). Recent studies show that RCs

are born early during V1 neurogenesis in vivo, suggesting that temporal patterning of V1 progenitors

might be important for specification of V1 subtype identity. While RCs are born between e9.5-10.5,

FoxP2-expressing V1 INs are born later, between e11.5-12.5 (Stam et al, 2012; Benito-Gonzalez &

Alvarez, 2012). By performing BrdU birthdating, we determined that ESC-derived RCs are also born early

compared to FoxP2-expressing INs, with peak generation on Day 4 of in vitro differentiation compared to

Days 6-7 for the FoxP2 subset (Fig 3.4).

Though it has previously been proposed that a temporally regulated transcriptional program

controls the differentiation of RCs, the instructive signals that produce this program have not been

identified (Stam et al, 2012). Given its well-known role in generating cellular diversity in the CNS, we

hypothesized that the Notch signaling pathway might be involved in this transformation. In particular,

Notch signaling has been shown to be important for controlling the timing of neurogenesis of different

neuronal cell types, with high Notch activity maintaining neural progenitor status while low Notch activity

promotes progenitor cell exit from cell cycle and the onset of neuronal differentiation (Heitzler et al, 1996;

Artavanis-Tskakonas et al, 1999; Huang et al, 2014). Yet, while Notch signaling components are

expressed in the p1 progenitor domain, prior studies have largely excluded a role for Notch in V1 IN

development due to overall lack of effect in global knockout models (Ramos et al, 2010; Skaggs et al,

2011; Stam et al, 2012). However, these analyses are likely to be confounded by non-cell autonomous

effects of disrupting Notch signaling in all cell types and by lack of temporal specificity.

Using pharmacological and genetic approaches, I manipulated Notch signaling during early V1

neurogenesis in vitro to assess if shifts in Notch activity affects the generation of RCs in particular. The

interaction between the Notch receptor and its ligand leads to gamma secretase-mediated cleavage and

release of the intracellular domain of the Notch receptor (NICD) into the cytoplasm. NICD translocates to

the nucleus, where it interacts with the Notch effectors RBP-J and Mastermind-like-1 (Maml1) to recruit

transcriptional co-activators to Notch target genes such as Hes5 (Fig 3.5A). Overexpression of the

106

dominant-negative form of Maml1 (DnMaml1) can be used to inhibit Notch signaling while overexpression

of NICD can be used to activate Notch. Using ESC lines carrying either DnMaml1 or NICD under the

control of a doxycycline-inducible promoter I differentiated ESCs to V1 INs using RA/low SAG and added

doxycycline on Day 4 of differentiation when there are many V1 progenitors (Tan et al, 2016). DnMaml1

expression was induced in the majority of cells and caused a 4-fold reduction of the expression of the

Notch target gene Hes5 by quantitative reverse transcription polymerase chain reaction (qRT-PCR).

NICD was similarly efficient, resulting in a 2-fold increase of Hes5. Alternatively, numerous studies have

indicated that the small molecule gamma-secretase inhibitor DAPT can be used to efficiently inhibit Notch

signaling without the requirement of specific genetic tools (Hori et al, 2013; Geling et al, 2002; Kessaris et

al, 2001; Tan et al, 2016). Indeed, DAPT treatment on Day 4 of differentiation causes a 3.6-fold decrease

in Hes5 expression, similar to DnMaml1 induction (Fig 3.5B).

DAPT-mediated Notch inhibition results in overall increase in V1 INs generated when added on

Days 4-6 of differentiation, when p1 progenitors marked by Pax6, Nkx6.2, and Ki67 are abundant (~45%

FP cells in DAPT-treated vs 34.9% in control EBs) (Fig 3.5C, data not shown). Since RCs are born earlier

than other V1 INs, DAPT-mediated Notch inhibition should not only increase V1 neurogenesis, but also

enhance the differentiation of early-born subtypes such as RCs. To test this hypothesis, I performed ICC

for Cb and FoxP2, finding that DAPT-mediated Notch inhibition results in 72% increase in Cb-expressing

V1 INs in Day 8 EBs (26.3 vs 15.3% in control EBs when added on Day 4, p<0.001), with an even more

striking decrease in FoxP2-expressing cells (3.0 vs 29.4%, p<0.001) (Fig 3.5D,E). Thus, DAPT treatment

during a critical period of V1 neurogenesis functions to increase RC genesis at the expense of later-born

subtypes, including a broad population of FoxP2-expressing V1 INs.

To examine the mechanism of DAPT action on V1 IN differentiation, I performed RNA-seq gene

expression profiling of FACS-purified En1-tdTomato from Day 8 EBs treated with DAPT on Day 4. As

expected, many Notch target genes were repressed in DAPT-treated cells compared to untreated

controls, including Hes5, Nrarp, and Notch1/3 receptors (Fig 3.6A) (Lamar et al, 2001; Artavanis-Tsakonis

et al, 1999). Important, hierarchical clustering of DAPT-treated ES-V1 INs, DAPT-treated ES-MNs and

control ES-V1 INs revealed that DAPT-treated V1 INs were more closely related to non-treated V1 INs,

107

indicating that DAPT treatment doesn’t convert distinct spinal progenitors to a uniform cell fate, but acts to

modify V1 IN molecular identity specifically (data not shown).

DAPT and control V1 INs were then compared against ES-dI4 INs from Day 8 to determine if

DAPT-treated cells had different molecular signature from control V1 INs. Using log2FC cut-off >2 and

padj <0.05, 119 genes V1-enriched genes were upregulated using DAPT treatment, while 201 genes

were downregulated. Upregulated genes included Cb (2.56 log2FC increase), as well as several markers

expressed by early postmitotic V1 INs, such as Foxd3, Hmx3, Hmx2, and Otp (Fig 3.6B and Fig 2.10A-C).

Conversely, transcripts downregulated by DAPT treatment included FoxP2 (-5.52 log2FC) (Fig 3.6B).

Since the majority of genes affected by DAPT treatment have not been well-studied in V1 INs, I

subsequently focused my analysis on the regulation of the 19 TFs shown to be enriched in V1 INs in vivo

(Fig 3.1, Bikoff et al, 2016). Amazingly, only TFs associated with the MafA clade in the V1 subtype

analysis, which includes RCs, are upregulated by DAPT treatment (red bars: Pou6f2, OC1, OC2, Cb,

MafB, and MafA), while all other TFs were repressed (blue bars) (Fig 3.6C). Together, these results

indicate that Notch inhibition strongly promotes the formation of the earliest-born V1 IN subtypes, while

preventing the formation of the majority of other V1 INs. Whether Notch inhibition enhances RC

generation by affecting the temporal differentiation of p1 progenitors and/or by directly specifying RC

identity over other V1 IN subtypes remains to be determined.

As demonstrated earlier, ES-RCs co-express MafA, MafB and OC2 TFs at surprisingly low levels

(Fig 3.3B). To test if DAPT treatment has an effect on these expression levels, I performed ICC on Day 8

EBs to detect for these markers, finding that Notch inhibition on Day 4 leads to large increase in overall

expression of these TFs in EBs, as well as in En1-tdTomato FP cells specifically (Fig 3.7A). Remarkably,

in long-term cultures of dissociated En1-tdTomato FP cells (Day 21), ~50% of FP cells expressed Cb and

OC2 in DAPT-treated conditions, while only 14% of cells acquired RC molecular identity in untreated cells

(Fig 3.7B). Thus, Notch inhibition increases the expression of RC-specific TFs compared to normal

RA/low SAG conditions, with even a short pulse of DAPT (Days 4 and 6) being sufficient for sustained

expression of one of these TFs (OC2) in the long-term, providing strong evidence that Notch signaling

has a role in the development of RC-specific identity in differentiating V1 INs.

108

Sustained retinoid signaling is required for specific generation of Renshaw cells

While Notch signaling likely has a role in the generation of RCs from the p1 progenitor domain,

Notch inhibition does not lead to exclusive production of RCs. DAPT treatment generates 30% Cb-

expressing cells in Day 8 EBs, which increases to 53% by Day 21 of differentiation (Fig 3.5E and 3.7B).

During V1 IN specification in vivo and in vitro, the extrinsic signaling molecules RA and SAG are jointly

required for the patterning of the p1 progenitor domain giving rise to V1 INs (Fig 3.8A,B; Pierani et al,

2001). Interestingly, removal of RA specifically on Days 3 and 4 of ESC differentiation results in only 3.1

and 17.0% FP cells generated on Day 8, compared to 31.7% under normal differentiation conditions with

RA/SAG, or when RA is removed on Days 5 or 6 (33.5 and 31.3% FP cells, respectively) (Fig 3.8B).

Conversely, specific removal of SAG on either Days 4 or 6 of differentiation does not grossly affect V1

neurogenesis (Fig 3.8C). Thus, early RA signaling is specifically and critically required for the generation

of V1 INs from ESCs, but expendable during later stages of differentiation.

Based on observations that Cb-expressing cells are more abundant at brachial regions of the

spinal cord, where there is secondary source of RA signals coming from Raldh2-expressing, limb-

innervating MNs, I investigated if RA signaling might also be required for specific generation of RCs,

beyond its primary role in establishing the p1 progenitor domain (Francius et al, 2013; data not shown).

To test this hypothesis, RA was removed from differentiating EBs on Day 5 and ICC was performed to

detect for Cb and FoxP2-expressing cells. Day 5 was chosen for RA manipulations since removal of RA

at this time point does not affect total V1 neurogenesis, but p1 progenitors or early postmitotic V1 INs at

this stage may still be influenced by extrinsic factors to acquire different subtype fates (Fig 3.8B)

Removal of RA on Day 5 led to significant decrease in Cb-expressing En1-tdTomato cells (2.0%

compared to 22.5% in control). This effect was specifically due to RA removal, as replacement of RA only

on Day 5 produced the normal cohort of Cb-expressing FP cells (20.4%), while replacement of SAG only

did not (1.1%) (Fig 3.8D). Remarkably, removal of RA specifically affected the formation of Cb-expressing

En1-tdTomato cells, as FoxP2-expressing cells were generated in normal quantities (26.7 vs 29.2% in

control) (Fig 3.8E). Since RA has been shown to directly regulate calbindin expression, I also checked if

other RC markers are changed in EBs where RA is removed early (Wang et al, 1995; Matsumoto et al,

1998). MafA expression was also downregulated in FP cells after removal of RA (0.42 vs 10.45% in

109

control EBs), as well as specifically in Cb-expressing V1 INs (0.20 vs 3.13% in control) (Fig 3.8F). In

contrast, MafB expression was unchanged between the two conditions (Fig 3.8G). Thus, RA appears to

be required for the specific generation of MafA-expressing RCs in vitro. Overall, these results suggest

that once the p1 progenitor domain is established, RA signaling may additionally be involved in the

specification of distinct V1 IN subtypes, especially RCs.

As mentioned earlier, RCs may be differentially distributed along the rostrocaudal axis of the

spinal cord, with enhanced numbers of RCs in limb-innervating regions. In brachial and lumbar spinal

cord, lateral LMC MNs express the RA-synthesizing enzyme Raldh2, which has been shown to influence

the differentiation of later-born medial LMC MNs (Sockanathan et al, 2003). To test if Raldh2-expressing

MNs could rescue the loss of Cb-expressing RCs, I performed a series of co-culture experiments with

En1-tdTomato ES-V1 INs and spinal MNs derived from Hb9::GFP or inducible HoxC8 (iHoxC8-V5) ESC

lines (Wichterle et al, 2001; Peljto et al, 2010; Mazzoni et al, 2011; Mazzoni et al, 2013b; Tan et al, 2016).

Under RA/high SAG conditions, ESCs differentiate to MNs with rostral cervical spinal cord identity, while

doxycycline-mediated induction of the Hox TF HoxC8 induces MNs to adopt brachial MN properties,

including expression of LMC TF FoxP1 (Fig 3.9A).

MN and V1 IN differentiations were performed independently, followed by mixture of early En1-

tdTomato EBs (Day 5) with either progenitor or postmitotic MN EBs (Fig 3.9B). At this step, RA was

specifically removed in some cultures to determine if MNs could produce RA to maintain Cb-expressing

RCs. Three days after mixture, ICC was performed on EBs to detect for Cb or FoxP2-expressing V1 INs

(Fig 3.9C). Neither pMNs or postmitotic MNs produced under normal RA/SAG conditions were sufficient

to rescue the loss of RCs after removal of RA on Day 5 (Fig 3.9D). Conversely, FoxP1- and Raldh2-

expressing MNs generated by Hoxc8 induction did increase the generation of Cb-expressing FP cells in a

dose-dependent manner, with no dox, 1 µgl/mL dox, and 3 µg/mL dox producing 18.3, 22.4, and 23.2%

FP cells, respectively, compared to 8.3% without HoxC8-expressing MNs and 20.7% under normal

RA/SAG conditions (Fig 3.10A,B; data not shown). Interestingly, differentiation HoxC8-expressing MNs in

media lacking vitamin A, a requisite precursor for Raldh2-mediated RA synthesis, results in abrogation of

this effect (Fig 3.10B) (Maden et al, 2002; Maden, 2007). Furthermore, cell-cell contact between V1 INs

and MNs was not required to produce this effect, as V1 INs cultured in conditioned media produced by

110

HoxC8-expressing MNs were competent to generate Cb-expressing cells (data not shown). Thus,

Raldh2-expressing MNs are capable of recovering the generation of Cb-expressing RCs when

exogenous RA is removed during early stages of V1 IN subtype diversification. These RA-mediated

effects appear to be specific to RC generation, since FoxP2-expressing FP cells did not appear to be

affected by these manipulations. (data not shown).

While these results suggest that RA may potentially be required for RC specification, I also tested

if additional RA can promote the formation of Cb-expressing cells over V1 IN subtypes. Importantly, late

addition of high RA (2 µM or 5 µM) on Day 5 of differentiation did not significant change the percentage of

En1-tdTomato cells in Day 8 EBs compared to control (data not shown). Although there was a slight

increase in Cb-expressing FP cells at higher concentrations of RA (25.3 and 30.3% compared to 20.7% in

standard 1 µM differentiation), the change was not significant, suggesting that RA is likely not sufficient

for specification of RC identity from uncommitted p1 progenitors (Fig 3.10C).

Confluence of Notch and RA signaling to specify Renshaw cells

Retinoids have been shown to be involved in regulating Notch signaling in neural progenitors of

the spinal cord (Paschaki et al, 2012; Ryu et al, 2015). To test if RA and Notch signaling pathways act

together to specify ES-RCs, I dissociated early ES-V1 IN EBs (Day 5) before co-culturing them with

Hb9::GFP MNs or HoxC8-inducible MNs. Dissociation of neural progenitors has been shown to disrupt

Notch signaling (Shimojo et al, 2008). Thus, using this assay, I could determine if manipulations of RA

signaling are potentiated by disruptions of Notch signaling, and vice versa. Interestingly, FoxP2-

expressing V1 INs are significantly reduced from baseline even after EBs are allowed several days to

reform aggregates and re-establish cell-cell interactions (Fig 3.10D). Co-culture of ES-V1 INs and MNs

further decreases the proportion of FoxP2-expressing V1 INs, suggesting that MN presence prevents

maximal restoration of V1-V1 interactions, blocking FoxP2+ V1 IN cell fate. Furthermore, in this context,

RA signaling remains dispensable for FoxP2+ V1 IN specification as HoxC8 induction in MNs does not

improve the generation of FoxP2-expressing FP cells (Fig 3.10D). In contrast, while Cb-expressing RCs

are drastically reduced in conditions devoid of retinoid signals, co-culture of dissociated ES-V1 INs with

ES-MNs is sufficient to rescue Cb cells (Fig 3.10E). Thus, prevention of efficient cell-cell interactions

111

between V1 INs by MNs leads to sustained suppression of Notch to promote RC generation. This effect is

enhanced when Raldh2-expressing MNs are introduced, suggesting that synergistic interactions between

RA and Notch signaling pathways support RC specification. These results were confirmed by additional

experiments using DAPT, rather than physical dissociation, to inhibit Notch signaling (data not shown).

Discussion Specification of spinal V1 INs from mouse ESCs recapitulates in vivo V1 IN development,

including progression through distinct progenitor stages marked by unique expression of TFs specific to

V1 INs (see Chapter 2). Stem cell-derived V1 INs thus provide a unique opportunity to investigate

molecular mechanisms leading to transformation of apparently homogeneous p1 progenitors into dozens

of molecularly, and potentially, functionally distinct subtypes. Indeed, ES-V1 INs provide unprecedented

experimental access to p1 progenitors in order to systematically probe the role of extrinsic signals and

intrinsic factors underlying V1 IN subtype diversification.

We show here that ESC-derived V1 INs show enrichment for TFs used to subdivide V1 INs in the

spinal cord, including 4 TFs (FoxP2, MafA, Pou6f2, and Sp8) demarcating non-overlapping clades of V1

IN cell types (Bikoff et al, 2016; Gabitto et al, 2016). Together, ES-V1 INs expressing FoxP2, MafA,

Pou6f2 and Sp8 factors constitute 56% of all En1-tdTomato FP cells on Day 8 of differentiation,

compared to 64% at p0 in the mouse spinal cord, demonstrating not only that in vitro-V1 INs recapitulate

key aspects of V1 IN subtype diversity in vivo, but also that ES-V1 INs may potentially be more diverse

than their spinal counterparts. Indeed, V1 IN subtype diversity has only been examined in the lumbar

spinal cord while ES-V1 INs likely acquire cervical and brachial segmental identity based on their Hox

expression profiles (Fig 2.4A, see Chapter 2). Recent observations suggest that Hox TF codes control the

specification of different V1 IN subtypes at thoracic versus brachial levels of the spinal cord, suggesting

not only that V1 IN and MN subtype diversification may be enacted by similar Hox transcriptional

regulatory programs but distinct subtype repertoires may be generated at different segmental levels (L.

Sweeney, personal communication).

Most impressively, ES-V1 INs acquire complex pattern of gene expression characteristic of two

distinct V1 subtypes previously characterized in vivo, Renshaw cells and FoxP2-expressing neurons. RCs

112

comprise a relatively small population of V1 INs with many shared molecular and functional features,

including their expression of Cb and a developmentally-regulated sequence of Foxd3, OC1/2, MafA, and

MafB TFs; distinct settling position in the ventrolateral horn; and role in recurrent inhibition of MNs (Stam

et al, 2013; Alvarez et al, 2005; Alvarez & Fyffe, 2007; Alvarez et al, 2013). Conversely, FoxP2-

expressing V1 INs constitute a much broader and heterogeneous subpopulation of V1 INs, as

demonstrated by their wide dispersion in the ventral spinal cord and their molecular heterogeneity (data

not shown; Siembab et al, 2010; Benito-Gonzalez & Alvarez, 2013; Zhang et al, 2014). At least some

FoxP2-expressing cells constitute IaINs, which are V1- and V2b-derived spinal INs that mediate

reciprocal inhibition of MNs (Benito-Gonzalez & Alvarez, 2012; Zhang et al, 2014; Britz et al, 2015). Cb-

and FoxP2-expressing V1 INs in vitro recapitulate these differences, most notably reproducing

appropriate migratory patterns in the ventral spinal cord upon engraftment of ES-V1 INs into the

developing chick neural tube. These results indicate that ES-V1 INs, especially RCs, acquire molecular

programs enabling them to respond appropriately to extrinsic guidance cues to navigate the

cytoarchitecture of the developing spinal cord (Wichterle et al, 2002; Peljto et al, 2010; Shen & Scheiffele,

2010). Whether these molecular programs are genetically hardwired or require exposure to the native

embryonic environment is not known. Interestingly, transplantation of ES-V1 INs later during their

development (Day 8) results in significantly fewer cells migrating from the graft site into the ventral horn

compared to transplants with Day 5 or 6 EBs, although these cells still project axons appropriately into the

ventral horn, suggesting that early, but not late, ES-V1 INs are competent to respond to local signals for

their proper migration (data not shown).

Establishment of spinal interneuron subtype diversity – a role for Notch signaling

To understand molecular mechanisms underlying V1 IN subtype diversification, we focused on

the differentiation of RCs from other V1 INs. In general, it is currently unknown if all V1 INs are generated

from a common progenitor that is competent to produce different V1 IN subtypes in response to different

spatial or temporal cues. Alternatively, multiple distinct subpopulations of p1 progenitors may be specified

to produce only certain V1 IN subtypes. While lineage-based clonal analyses and

heterotopic/heterochronic transplantation studies of V1 INs have not yet been performed to answer these

113

questions, largely due to lack of specific genetic access to p1 progenitors, some insights have been

obtained regarding specification of RCs from other V1 IN cell types. In particular, recent studies have

shown that not only are RCs in vivo born earlier than other V1 INs, but RC fate is determined by a

temporal TF program that is required for the establishment and maintenance of RC-specific identity (Stam

et al, 2012; Benito-Gonzalez & Alvarez, 2012). Thus, although V1 INs appear genetically and

morphologically homogeneous early during their development, their subtype-specific cell fates may be

already specified during early postmitotic stages, if not earlier.

After establishing that ES-RCs are also born earlier than other V1 INs, we tested if manipulation

of Notch signaling could bias the formation of early-born RCs over later-born subtypes such as FoxP2-

expressing V1 INs. Indeed, not only did inhibition of Notch signaling using DAPT significantly increase the

generation of Cb-expressing V1 INs, but Notch inhibition also virtually eliminated FoxP2-expressing cells.

Interestingly, in vivo, different Notch signaling mutants generate normal numbers of Cb-expressing RCs

and FoxP2-expressing V1 INs (Stam et al, 2012; Marklund et al, 2012; Ramos et al, 2010). However,

these global knockout models lack the temporal and spatial precision of the ES-based system for

manipulating Notch signaling in V1 progenitors. Furthermore, several of the Notch mutants exhibit cardiac

and other systemic defects affecting embryo fitness and survival, thereby complicating analyses of V1 IN-

specific changes (M. Goulding, personal communication). ES-derived V1 INs thus provide a simpler,

more experimentally accessible system to interpret effects of Notch on V1 subtype differentiation.

Nevertheless, several important questions remain about the mechanism used by Notch signaling

to specify distinct V1 IN subtypes. First, the mechanism by which Notch signaling controls p1 progenitor

domain patterning to specify different V1 subtype fates is currently unknown. High Notch activity

maintains neural progenitors while low Notch activity promotes progenitor cell exit from cell cycle and

neuronal differentiation. Therefore, Notch may act permissively to regulate the time at which neural

progenitor cells differentiate and respond to appropriate specification signals (Louvi & Artavanis-

Tsakonas, 2006). In the retina, Notch primarily plays a permissive role in regulation of cell fate by

controlling the timing of progenitor differentiation into postmitotic neurons. Inhibition of Notch signaling

early during retinal neurogenesis results in precocious differentiation and enrichment of early-born retinal

subtypes at the expense of later-born neurons (Jadhav et al, 2006; Yaron et al, 2006; Livesey & Cepko,

114

2001). Thus, one possibility is that Notch also functions as a permissive signal in V1 progenitors to control

the timing of V1 subtype generation. Thus, progenitors differentiating early may become competent to

acquire RC identity, while those differentiating later acquire other subtype identities.

Alternatively, Notch signaling may act instructively by regulating binary cell fate choice, in which

cells receiving Notch signals acquire one specific cell type identity while cells lacking Notch acquire a

separate and distinct identity (Satou et al, 2012; Hori et al, 2013). As such, Notch may act instructively to

specify subtype identity in V1 INs, including directly selecting or repressing RC specific identity. In the

spinal cord, cells receiving Notch signals acquire V2b IN identity while cells deprived of Notch acquire

V2a IN fate (Yang et al, 2006; Peng et al, 2007; Kimura et al, 2008; Misra et al, 2014; Okigawa et al,

2014; Zou et al, 2015). Furthermore, instructive Notch signaling has recently been shown to contribute to

the specification of MMC versus HMC spinal motor column identity, as well as lateral versus medial

divisional identity of limb-innervating LMC MNs (Yang et al, 2006; Tan et al, 2016). Distinguishing

between the permissive versus instructive roles of Notch signaling in the V1 spinal domain will be

important for understanding how RC identity is specified, resulting in more efficient and homogenous

methods of producing RCs; as well as for understanding the lineage relationships between different V1

subtypes, providing important information about how the p1 domain is patterned to generate such

tremendous cellular diversity.

Second, the identity of other V1 INs generated during DAPT-mediated Notch inhibition is not

known. During ESC-to-V1 IN differentiation with DAPT treatment on Day 4, V1 neurogenesis increases

overall, but the increase in Cb-expressing RC generation does not match the magnitude of the

corresponding loss of FoxP2-expressing V1 INs. Therefore, what is the identity of V1 INs lacking both Cb

and FoxP2 expression? RNA-seq expression profiling of DAPT-treated cells suggests that TFs marking

non-RC-lineage cells are generally downregulated compared to control, while TFs associated with the RC

developmental program are markedly increased. Immunostaining for MafA, MafB, and OC1/2 confirm that

these TFs are not only upregulated in Cb-expressing V1 INs, but in other V1 INs as well. Thus, one

hypothesis is that DAPT-mediated Notch inhibition represses FoxP2+ V1 IN lineages while promoting

lineages defined by MafA, MafB, or OC1/2 expression, which include, but may not be restricted to, RCs.

Indeed, in the embryonic and postnatal spinal cord, MafA, MafB, and OC2 also mark non-RC populations.

115

MafA, for example, is expressed in half of all RCs but also in a subset of V1 INs positioned more dorsally

to the classic RC area (Bikoff et al, 2016). An alternative hypothesis is that p1 progenitors are only

competent to produce early-born RCs following DAPT treatment, but their molecular maturation is

delayed such that only a fraction of MafA, MafB, or OC1/2-expressing V1 INs are competent to express

Cb on Day 8 of differentiation. Indeed, by 21 days in vitro, many more V1 INs express both Cb and OC2

(~50%) compared to Day 8 (26.3%), indicating that while Notch inhibition may promote RC neurogenesis,

Notch signaling may be important for other aspects of RC development

Retinoid signaling and Renshaw cell generation

Although Notch signaling clearly plays an important role in V1 IN subtype differentiation, it is

unlikely to be the only regulator of this process. Indeed, Notch inhibition does not lead to the

transformation of all V1 INs to RC fate. We therefore examined whether additional extrinsic cues could

influence V1 IN subtype development (Lu et al, 2015). Manipulation of SAG concentration in ESC-derived

p1 progenitors changes the composition of V1 IN subtypes generated, especially FoxP2 and Cb-

expressing V1 INs, suggesting that dorsoventral spatial patterning may have a role in V1 IN subtype

diversification. In addition, we explored the role of rostrocaudal patterning cues in V1 IN specification.

Different V1 IN subtypes may predominate at distinct rostrocaudal levels of the developing spinal cord, as

suggested by a recent study exploring the molecular diversity of ventral spinal neuron types, including V1

INs (Francius et al, 2013). In this study, it is suggested that Cb-expressing V1 INs are more prominent at

brachial, or limb-innervating, levels of the spinal cord compared to cervical or thoracic segments, which

do not contain limb-innervating MNs. Concomitant with this expression pattern are several additional

pieces of evidence suggesting a unique developmental relationship between RC and MNs. First, RCs are

the earliest-born of V1 INs, with their neurogenesis largely overlapping with MNs (RCs: e9.5-e10.5; MNs:

e8.5-e10.5) (Stam et al, 2012; Benito-Gonzalez & Alvarez, 2012; Rowitch et al, 2002). Once born, RCs

migrate circumferentially from the progenitor domain to settle in a region of the ventral horn dense with

MN axons and cell bodies (Alvarez et al, 2005; Alvarez et al, 2013). Second, decades of

electrophysiological and anatomical studies have indicated that RCs exhibit a unique functional

116

relationship with MNs by being the exclusive provider of recurrent inhibition onto MNs (Renshaw, 1946;

Eccles et al, 1954; Alvarez & Fyffe, 2007).

At brachial levels of the spinal cord, medial LMC MNs produce Raldh2, which account for the

increase in retinoid signaling at limb levels of the spinal cord (Wang et al, 1996; Zhao et al, 1996; Rossant

et al, 1991; McCaffery & Dräger, 1994; Solomin et al, 1998; Sockanathan & Jessell, 1998). During spinal

MN development, LMCm MNs are generated earlier than lateral division LMC MNs. Early-born LMCm

MNs expressing Raldh2 produce retinoids that not only function to increase overall MN generation in the

brachial spinal cord, but also specifically influence the development of later-born LMCl neurons migrating

through LMCm territory to reach their final position (Sockanathan & Jessell, 1998). These studies not only

provide precedence for RA signaling having a spatially-restricted role in the postmitotic specification of

spinal neuron subtype identity, but also establish LMCm MNs as an additional source of RA during spinal

cord development.

Here, we demonstrate that the role of RA signaling in RC specification can be divided into two

stages. First, retinoid signaling, in conjunction with Shh, is required for the generation of p1 progenitors

from ESCs, confirming in vivo findings. Importantly, the paraxial mesoderm abutting the developing spinal

cord is the most likely source of this early RA signal (Pierani et al, 1999; Wilson et al, 2004). In the

second stage, sustained RA signaling is required for the specific generation of Cb-expressing RCs.

Indeed, removal of RA during intermediate stages of ES-V1 development, after initial patterning of the p1

progenitors is complete, results in suppression of ES-RC development while FoxP2-expressing cells are

generated in normal numbers.

However, whether RA is required for the establishment and/or maintenance of RC cell fate is not

known. One hypothesis is that sustained RA signaling is required for the full expression of the RC-specific

transcriptional program involving MafB, OC1, OC2, and likely MafA (Stam et al, 2012; Bikoff et al, 2016).

Removal of RA signaling might then pause the progression of this program. Indeed, removal of RA on

late Day 4 and early Day 5 of differentiation is sufficient to eliminate Cb (and MafA) expression, while later

removal does not affect the yield of RCs. Although it is unclear from these experiments if RA signaling is

required for RC specification in progenitor or postmitotic cells, these data nonetheless suggest that RA is

likely required during a narrow time window for the complete specification of RC fate. Additionally, some

117

evidence suggests that Cb expression can be rescued if RA is re-supplied later during in vitro

differentiation, suggesting that lack of RA pauses the RC developmental program, but which can be

restarted upon exposure to the right cues (data not shown).

Interestingly, co-culture of ES-V1 INs devoid of RA signaling with Raldh2-expressing ES-MNs is

also sufficient to rescue the loss of Cb-expressing V1 INs. Experiments using conditioned media from

MNs, or media lacking vitamin A (a requisite precursor for retinoid synthesis) suggest that these effects

are due to secreted retinoids from limb-innervating ES-MNs. However, while these studies demonstrate

that HoxC8-expressing MNs are competent to synthesize RA to rescue RC-like cells, whether these

additional RA signals are derived from MNs or paraxial mesoderm cannot be determined from these

experiments. Indeed, whether MN-derived RA signals are necessary for RC-specific development

requires genetic tools to specifically ablate LMC MNs in vivo. To confirm the role of RA in RC specification

in vivo, experiments to block or activate RA signaling in vivo should also be performed in order to

determine if RA is required and sufficient to generate RCs at different rostrocaudal levels of the

developing spinal cord (See General Discussion).

Role of output neurons on recruitment of distinct interneuron subtypes

Although sustained RA signaling appears to be required for efficient generation of Cb-expressing

V1 INs, whether MNs are the source of this signal in vivo is currently unclear. Why might MNs produce

RA? As mentioned before, RA secreted by Raldh2-expressing LMCm MNs influences the specification of

LMCl MNs. MN-derived RA is also involved in increasing overall MN generation in the brachial spinal cord

(Sockanathan & Jessell, 1998). Whether MN-derived RA can affect the differentiation of other spinal

neuron cell types is not known. Interestingly, MNs lacking recurrent inhibition, such as in the digits of the

hand and foot, express the limb-specific marker FoxP2, but these MNs lack Raldh2 expression (Illert &

Kummel, 1999; A. Mendelsohn & T. Jessell, unpublished).

In the spinal cord, MNs are the sole output neuron. One hypothesis is that MNs actively control

the wiring of motor circuits, including recruiting distinct IN inputs, potentially by producing paracrine

signals required for their developmental specification. Recent studies have shown that excitatory

projection neurons, which are the output neurons of the cortex, provide laminar information to developing

118

inhibitory INs, influencing their migration and positioning in the cortex (Lodato et al, 2011). Furthermore,

in vivo lineage reprogramming of callosal projection neurons into corticofugal projection neurons leads to

differential recruitment of inhibitory inputs from parvalbumin-expressing INs, suggesting that cortical

output neurons can directly control the recruitment of afferent inputs (Ye et al, 2015). However,

mechanisms used by projection neurons to influence cortical IN subtype lamination and synaptic

recruitment are not known. Indeed, whether output neurons such as cortical projection neurons or spinal

motor neurons have a role in developmental specification of their circuit partners, especially different IN

subtypes, has not been shown. Nevertheless, based on several lines of evidence, we suggest that MNs

might provide an instructive signal for RC differentiation, thus having a unique and novel role in V1 IN

subtype diversification.

119

Figure 3.1

Figure 3.1 Enrichment of V1 interneuron subtype-specific transcription factors TFs enriched ESC-derived V1 INs (Day 8) compared to e12.5, p0 and p6 spinal V1 INs, relative to dI4 INs. Right column, FPKM values for ESC-derived V1 INs from RNA-seq profiling.

120

Figure 3.2

121

Figure 3.2 Subtype diversity of ESC-derived V1 interneurons (A) Subsets of En1-tdTomato FP cells express TFs delineating molecularly distinct V1 subtypes in the embryonic and early postnatal spinal cord. (B) Quantification of FP cells expressing Pou6f2, Sp8, Nr5a2 and FoxP1 TFs. (C) Non-overlapping subpopulations of En1-tdTomato FP cells express Calbindin and FoxP2 (left). Quantification of Cb- and FoxP2-expressing subsets (right). (D) Increased production of Sp8-expressing V1 INs in vitro over time. (E) SAG concentration influences the generation of different V1 IN subsets in vitro, indicating a potential role for Shh-mediate intradomain patterning. Scale bars = 50 µm.

122

Figure 3.3

Figure 3.3 Calbindin-expressing V1 interneurons in vitro acquire Renshaw cells properties (A) Molecular differentiation of RCs compared to other V1 INs. (B) A subset of Cb-expressing En1-tdTomato FP cells co-express MafA, MafB, and Onecut2 TFs, which comprise a transcriptional program considered to be crucial for RC development in vivo. Scale bars = 50 µm. (C) Transplanted Cb-expressing V1 INs in the developing chick spinal cord migrate ventrolaterally compared to FoxP2-expressing subset. Scale bars = 100 µm.

123

Figure 3.4

Figure 3.4 BrdU birthdating of ESC-derived V1 interneurons Based on BrdU pulse-labeling of differentiating EBs, Calbindin-expressing FP cells are born earlier (Day 3-5) compared to FoxP2-expressing subset (Day 4-6). Scale bars = 50 µm.

124

Figure 3.5

Figure 3.5 Notch inhibition promotes the formation of Calbindin-expressing V1 interneurons (A) Schematic of Notch signaling. (B) QPCR results show that downregulation of Notch signaling using inducible DN-Maml1 ESC line or pharmacological DAPT treatment in differentiating V1 INs results in similarly decreased relative expression of Hes5, a downstream target of Notch signaling; while activation of Notch signaling using inducible NICD ESC line causes upregulation of Hes5 expression. (C) Notch inhibition with DAPT treatment on Days 4 to 6 results in small increase in generation of En1-tdTomato FP cells, suggesting that low Notch enhances V1 neurogenesis. (D) DAPT-mediated Notch inhibition on Day 4 of differentiation results in dramatic loss of FoxP2-expressing FP cells in Day 8 EBs. (E) Quantification of DAPT effect on generation of Cb-expressing versus FoxP2-expressing En1-tdTomato FP cells in Day 8 EBs. The greatest effect is observed when DAPT is added on Day 4 or 5 of differentiation, with 72% increase in Cb-expressing cells and even more striking decrease in FoxP2-expressing cells. (n>3, ANOVA *p<0.05, **p<0.01, ***p<0.001).

125

Figure 3.6

126

Figure 3.6 RNA-seq expression profiling of DAPT-treated V1 interneurons (A) DAPT-mediated Notch inhibition of Day 4 differentiating V1 INs results in downregulation of most Notch target genes. (B) Top up- and downregulated genes in Day 8 ESC-derived V1 INs treated with DAPT on Day 4 compared to control (L2FC ≥2, padj <0.01). (C) Based on 19 TFs used to define V1 IN subtype diversity, DAPT treatment leads to upregulation of transcripts associated with RC identity (red bars), while the majority non-RC subtype markers are downregulated (blue bars). Dashed line represents 1.5 L2FC cut-off for significance.

127

Figure 3.7

Figure 3.7 DAPT treatment upregulates MafA, MafB and OC2 expression (A) Notch inhibition on Day 4 leads to large increase in overall expression of these TFs in EBs, as well as in En1-tdTomato FP cells specifically. Scale bars = 50 µm. (B) In long-term cultures of dissociated En1-tdTomato FP cells (Day 21), ~50% of FP cells expressed Cb and OC2 in DAPT-treated conditions, while only 14% of untreated cells acquired RC molecular identity.

128

Figure 3.8

Figure 3.8 Sustained retinoid signaling is required for Renshaw cell specification (A) V1 neurogenesis in vitro requires RA. (B) Temporal requirement of RA signaling in V1 neurogenesis. Premature removal of RA on Days 3 and 4 leads to diminished generation of En1-tdTomato FP cells during ESC-to-V1 IN differentiation. Removal of RA later, on Days 5 and 6, produces similar numbers of FP cells as EBs in which RA/SAG are maintained throughout differentiation. (C) Removal of SAG on Days 4 or 6 does not affect V1 IN differentiation. (D) Cb-expressing V1 INs are specifically reduced when RA is removed on early Day 5 of differentiation. Replacement of RA, but not SAG, is sufficient to promote the development of Cb-expressing cells. (E) FoxP2-expressing V1 INs are not affected by early removal of RA. (F) MafA-expressing V1 INs are also diminished by early RA removal, especially MafA/Cb-co-expressing V1 INs. (G) MafB expression is not affected by early RA removal. Scale bars = 50 µm.

129

Figure 3.9

Figure 3.9 Motor neuron-V1 interneuron co-cultures (A) Hb9::GFP and iHoxC8-V5 MNs differentiated using RA and high SAG (500 nM) exhibit similar differentiation efficiency (Hb9 immunostaining), but iHoxC8-V5 MNs express high levels of FoxP1, indicating that HoxC8 induction produces limb-innervating MN subtypes. (B) Day 5 En1-tdTomato EBs were mixed with Day 6 Hb9::GFP EBs for 3 days then fixed and cryosectioned. Shown are three examples of V1 IN-MN EB interactions in co-cultures. (C) Immunocytochemical detection of Cb-expressing V1 INs in mixed co-cultures. (D) Mixture of V1 INs with either Hb9::GFP pMN progenitors (Day 4) or postmitotic MNs (Day 6) is not sufficient to loss of RCs from early RA removal. Scale bars = 50 µm.

130

Figure 3.10

Figure 3.10 Raldh2-expressing motor neurons rescue Renshaw cell loss (A) Induction of HoxC8-V5 expression with dox treatment in differentiating MNs leads to increased FoxP1 and Raldh2 expression in a dose-dependent manner. Scale bars = 50 µm. (B) Compared to Hb9::GFP MNs, which have rostrocervical identity, mixture of V1 INs with Raldh2-expressing iHoxC8-V5 MNs is sufficient to rescue loss of RCs from early removal of RA. This effect is partially dependent on RA synthesis, since co-culture of V1 INs and MNs in vitamin A-deficient media leads to reduced formation of Cb-expressing V1 INs. (C) Increased RA (5 µM) promotes, but does not significantly increase, generation of Cb-expressing cells compared to control (1 µM RA) or 2 µM RA. (D) To test if Notch and RA signaling pathways interact to specify RCs, ES-V1 IN EBs (Day 5) were dissociated prior to co-culture with Hb9::GFP or HoxC8-induced MNs in order to disrupt Notch signaling. FoxP2-expressing V1 INs are significantly reduced from control even after EBs are allowed several days to reform aggregates and re-establish cell-cell interactions. Co-culture of ES-V1 INs and MNs further decreases the proportion of FoxP2-expressing V1 INs. RA signaling is dispensable for FoxP2+ V1 IN specification as HoxC8 induction in MNs does not improve the generation of FoxP2-expressing FP cells. (E) Co-culture of dissociated ES-V1 INs with ES-MNs is sufficient to rescue Cb cells even after early removal of RA, suggesting that prevention of efficient cell-cell interactions between V1 INs by MNs leads to sustained suppression of Notch to promote RC generation. This effect is enhanced when Raldh2-expressing MNs are introduced, suggesting synergism between Notch and RA signaling.

131

Chapter 4. Synaptic specificity of ESC-derived Renshaw cells Introduction

The establishment of precise and highly specific synaptic connectivity is required for proper brain

functioning. Synaptic specificity requires developing neurons to project axons to correct target areas,

recognize their appropriate synaptic partners among many other potential partners, and build functional

synapses with the target neuron (Shen & Scheiffele, 2010). Although significant progress has been made

in understanding mechanisms of axon guidance and synapse assembly, relatively less is known about

the molecular development of synaptic target recognition. Furthermore, much of our current

understanding of synaptic specificity comes from simple model organisms such as C. elegans and

Drosophila. Whether similar mechanisms and molecules are utilized in the mammalian central nervous

system is still unclear. Indeed, our understanding of the molecular basis of synaptic specificity in more

complex organisms could be significantly advanced by our ability to model it in a simplified system that is

similarly amenable to experimentation.

Embryonic stem cell (ESC)-derived neurons represent such a system. Mouse ESCs can be

directed to differentiate into distinct neuronal cell types in a process that recapitulates normal embryonic

development both molecularly and functionally (Petros et al, 2011; Wichterle et al, 2002). Spinal motor

circuits are an especially ideal system to study the development of synaptic specificity since many

synaptic partners of MNs have been identified, including multiple classes of spinal interneurons (IN)

providing excitatory or inhibitory inputs onto MNs to modulate MN activity (Goulding, 2009). Historically,

the best characterized premotor IN is the Renshaw cell (RC). While RCs, which belong to the ventral V1

IN class, have other synaptic partners in the spinal cord, they are best known for their unique role

providing recurrent inhibition of MNs, in which RCs inhibit MNs while receiving excitatory inputs from the

same MNs via axon collaterals to form a negative feedback loop (Alvarez & Fyffe, 2007). Given our ability

to generate MNs and INs from ESCs, including V1 INs with RC-like features, we asked if we could use

the in vitro system to test their synaptic selectivity in co-culture with MNs.

Here we use an ESC-based system to establish a novel assay for testing synaptic connectivity of

molecularly-defined V1 subtypes with MNs. Using synaptic marker expression analysis, transsynaptic

rabies virus tracing, and optogenetics-based electrophysiological approaches, we confirmed that stem

132

cell-derived MNs and RCs preferentially engage in circuits suggestive of recurrent inhibitory connections.

Importantly, comparative analysis of other ESC-derived V1 IN subtypes as well as more dorsally-

positioned dI4 spinal INs demonstrates the selectivity of RC-MN circuitry. Thus, synaptic specificity of a

molecularly-defined spinal cell type can be recapitulated in a simple in vitro co-culture system.

Results Co-culture of stem cell-derived spinal interneurons and motor neurons Having shown that ESC-derived RCs acquire functional features of RCs in vivo, including

migration into spinal regions populated by MN axons following transplantation into spinal cord, we asked

if ESC-derived RCs also exhibit synaptic preference for MNs. We first developed a co-culture assay of

ES-MNs and ES-INs, utilizing primary cortical astrocytes as permissive substrate for synaptic maturation

and for long-term survival (Albuquerque et al, 2009; Takazawa et al, 2012; Johnson et al, 2007; Meshul et

al, 1987; Slezak & Pfrieger, 2003; Clarke & Barres, 2013; Chung et al, 2015). MNs were dissociated from

EBs at Day 6 of differentiation, the peak of MN genesis in vitro, and purified using fluorescent or magnetic

activated cell sorting (FACS or MACS), based on the ESC reporter line used (Hb9::GFP, Hb9::RFP or

Hb9::Cd14-GFP) (Wichterle et al, 2002; Bryson et al, 2014; Tan et al, 2016; Rhee et al, 2016).

Meanwhile, ES-V1 or dI4 INs were dissociated from EBs on Day 8, FACS purified based on tdTomato or

GFP reporter expression, and mixed with MNs in varying densities.

In co-culture, ESC-derived V1 and dI4 INs exhibited qualitatively different types of contacts with

MNs. ES-V1 INs preferentially formed close cellular contacts with MNs even in sparse cultures (25% of

V1 interactions vs 3% of dI4). Interestingly, while ES-V1 IN axons were generally tightly entwined around

MN cell bodies and proximal dendrites, ES-dI4 IN axons encircled MN cell bodies and traveled loosely

around the MNs (30 vs 6% of interactions compared to V1 INs). Furthermore, a greater proportion of ES-

dI4 INs failed to interact with MNs at all, even at cell densities similar to V1 INs (6 vs 0%) (Fig 4.1A,B).

These initial cellular interaction analyses imply that ES-V1 INs as a population maintain a physically

closer relationship to MNs in vitro. In order to further define the interactions of specific subtypes of ES-V1

133

INs, in particular ESC-derived RCs, we developed ESC-based synaptic mapping tools to probe the

functional synaptic connectivity of ESC-derived spinal INs and MNs in co-culture in vitro.

Monosynaptic rabies virus tracing reveals subtype-specific inputs onto MNs Renshaw cells and IaINs are well-known for providing inhibitory inputs onto MNs to provide

recurrent and reciprocal inhibition, respectively, but whether all V1 IN subtypes synaptically target MNs is

not known (Fig 4.2A). Immunocytochemistry (ICC) for GABAergic synapse markers reveals that Cb-

expressing RCs, as well as other non-Cb+ En1-tdTomato FP cells, form Gad67 and synapsin-

immunoreactive synaptic boutons on MN cell bodies and proximal dendrites in vitro, suggesting that

different V1 INs provide GABAergic inputs onto MNs (data not shown). In order to systematically assess

the subtype identity of V1 INs synapsing onto MNs, I adapted monosynaptic rabies virus (RABV) tracing

for retrograde tracing of premotor V1 INs. The use of recombinant RABV with its glycoprotein (G)

replaced with GFP restricts viral spread to initially infected neurons unless a copy of G protein is also

supplied to the cell, while also allowing RABV-infected cells to be easily identified through GFP

expression. The SADΔG-GFP virus is also pseudotyped with an avian envelope protein (EnvA) so that it

can only infect cells carrying the cognate TVA receptor (Wickersham et al, 2007a,b; Callaway, 2008;

Osakada, 2011). Using an ESC line carrying the MN-specific reporter Hb9::GFP, I stably transfected a

transgene containing both TVA and G protein (TVA/G) so that while initial SADΔG-GFP RABV is

restricted to MNs, viral spread is permitted to first-order presynaptic neurons (Fig 4.2B). Co-culture of

TVA/G transgenic MNs with V1 INs enables unambiguous identification of monosynaptically-connected

V1 INs by their joint expression of tdTomato fluorescent reporter and SADΔG-GFP (Fig 4.2C).

En1-tdTomato V1 INs and TVA/G-expressing MNs were co-cultured for one week prior to addition

of SADΔG-GFP or SADΔG-dsRed, an RFP variant of RABV. Three to seven days after RABV addition,

V1 INs and MNs were analyzed for their expression of SADΔG-GFP and SADΔG-dsRed, respectively, to

determine their infection rates (Fig 4.2B). Seven days was chosen as the endpoint because of evidence

of cellular toxicity by that stage, including atypical neuronal morphology, blebbing of processes, and

excessive cellular debris (data not shown), similar to in vivo (Schnell et al, 2010; Callaway and Luo, 2015;

Ghanem & Conzelman, 2015; Reardon et al, 2015). In addition to fluorescent reporter expression, I also

134

confirmed rabies infection in MNs and V1 INs by immunostaining for Rabies-N nucleoprotein (Papaneri et

al, 2012). Motor neurons were successfully infected after 24-48 hours of RABV addition at high

efficiencies (1º infection: 91% using SADΔG-dsRed and 97% using Rabies-N), with infection increasing

over several days (Fig 4.2D,E, data not shown). Conversely, V1 INs did not begin to express SADΔG-

GFP until three days after infection, similar to in vivo reports (Jovanovic et al, 2010; Coulon et al, 2011).

At 4 days post-infection (2º infection), 9.4% of V1 INs express GFP, increasing to 14.3% by Day 7 (Fig

4.2F). Moreover, on Day 4 post-infection, the transsynaptic rate was similar at both low or high ratios of

V1 INs to MNs (2:1 vs 5:1), as well as at different cell densities, indicating that the retrograde transfer of

RABV is likely not determined by cellular composition, including proximity of V1 INs and MNs, of the co-

culture (Fig 4.2G).

Using this assay, we were next interested in assigning subtype identity to premotor V1 INs. I

performed ICC to detect for cells triple-labeled by SADΔG-GFP, En1-tdTomato and either Cb or FoxP2 to

determine if Cb-expressing RCs were more likely to synapse onto MNs than FoxP2-expressing subtypes

(Fig 4.3A). There was a small, but not statistically significant, increase in the percentage of rabies-

infected V1 INs expressing Cb over FoxP2 (16.6 vs 10.6%) (Fig 4.3B). However, FoxP2-express V1 INs

are significantly enriched in long-term cultures (without DAPT) compared to Cb-expressing cells (47.3 vs

25.7% of all En1-tdTomato cells) (Fig 4.3C). Therefore, we normalized the subtype-specific RABV-based

synaptic connectivity to compute a “connectivity index” or “C.I.,” which takes into account IN prevalence in

the culture in determining the likelihood of their MN connectivity. As such, a C.I. of 1.0 indicates that the

cell type is equally likely to be monosynaptically connected to MNs as not. Using this index, we

determined that Cb-expressing RCs had a C.I. of 1.66, compared to 1.05 for FoxP2-expressing V1 INs,

indicating that RCs are more likely to provide direct synaptic inputs onto MNs. Intriguingly, the C.I. of

ESC-derived V1 INs lacking both Cb and FoxP2 expression was found to be less than 1.0, suggesting

that there may exist subpopulations of V1 INs that actively avoid synapsing with MN targets (Fig 4.3D).

Thus, based on the RABV retrograde tracing, Cb-expressing V1 INs synapse onto MNs at higher than

expected rates based on their numbers in culture. However, many FoxP2-expressing FP cells also

provided inputs onto MNs, suggesting that MNs receive mixed inhibitory inputs from different V1 INs.

135

Differential VAChT contacts on V1 interneuron subtypes

In the mouse spinal cord, RCs are the exclusive synaptic target of MN collaterals (Alvarez et al,

2013). Here, we used the co-culture system to determine if ES-MNs exhibit similar synaptic specificity for

RCs, or if they promiscuously provide inputs to other cell types in vitro. (Fig 4.4A). Since dI4 INs reside far

from the ventral funiculus where MN axon collaterals project, we anticipated that few dI4 INs should

receive VAChT+ MN inputs. Therefore, we first compared MN synaptic inputs on ES-dI4 versus V1 INs in

order to establish a baseline of in vitro MN synaptogenesis. En1-tdTomato V1 INs or Ptf1a-tdTomato dI4

INs were co-cultured with Hb9::GFP MNs at varying cellular densities for 2 weeks, then fixed and

immunostained for VAChT, which marks cholinergic synapses (Fig 4.4B,C). Remarkably, while 46.8% of

all V1 INs exhibited dense VAChT+ synaptic puncta on their cell bodies and proximal dendrites, only

11.5% of dI4 INs did so, making V1 INs more than 4X more likely to receive MN inputs than dI4 INs (Fig

4.4C,D). Why some dI4 INs receive cholinergic inputs is not known (See Discussion and Chapter 5).

Often, dI4 INs did not co-localize VAChT synapses even when they were found nearby VAChT-

expressing cells (Fig 4.4C).

Next, I examined the subtype identity of V1 INs receiving MN collateral inputs. Here, RCs were

identified by joint expression of En-tdTomato reporter, Calbindin and Onecut2 (Fig 4.5A). Renshaw cells

defined by Cb+OC2 expression comprised 56.7% of all VAChT-innervated V1 INs, while cells expressing

Cb-only, OC2-only or neither of those markers made up 1.8, 34.5, and 7.2% of the remaining V1 INs (Fig

4.5B). In addition, by comparing RCs and non-RC V1 INs, I found that 75.5% of RCs received VAChT+

MN inputs compared to 20.6% of non-RC V1 INs (Fig 4.5C). Closer analysis of the non-RC V1 subtypes

receiving cholinergic inputs revealed that while 35.4% of OC2-only expressing FP cells had VAChT

staining, just 11.8% of Cb-only FP cells and 13.7% of FP cells without Cb or OC2 did (Fig 4.5D). Finally,

taking into account the prevalence of these different V1 subtypes in culture, I calculated a connectivity

index for RCs versus non-RC V1 INs: while RCs had a C.I. of 1.97 non-RCs had C.I. of 0.62, indicating

that RCs are highly likely to receive MN inputs while non-RCs had significantly lower than chance

likelihood of doing so (i.e., C.I. = 1.0) (Fig 4.5E). Thus, while anomalous connections between MNs and

unknown, non-RC targets do occur, analysis of differential VAChT immunoreactivity reveals that RCs

136

receive significantly more MN inputs than other V1 IN subtypes, indicating that this simple assay can be

used to uncover the synaptic connectivity preferences of molecularly-defined neuronal cell types.

Physiological signature of ESC-derived Renshaw cells

While immunocytochemical observation of synaptic contacts and monosynaptic RABV tracing are

suggestive of synaptic connections between V1 INs and MNs, validation of functional synapses between

these cell types requires electrophysiological analysis. Thus, we performed whole-cell patch-clamp

electrophysiology recordings on En1-tdTomato FP cells. Long-term cultures of V1 INs were established,

then recorded to determine if they develop subtype-specific physiological properties, including passive

and active membrane responses (Fig 4.6A). The subtype identity of intracellularly-filled (e.g. Neurobiotin)

recorded cells was post hoc identified with Cb and OC2 immunoreactivity (Fig 4.6B). Importantly, we

differentiated ESCs to V1 INs using DAPT-mediated Notch inhibition to enhance the formation of RCs,

recording from 35 RCs and 30 non-RCs overall.

The passive membrane properties of RC-V1 INs versus non-RCs were assessed following

injections of steps of negative and positive currents (Fig 4.6C,D). Based on the slope of the linear current-

to-voltage relationship, RCs have increased input resistance compared to other V1 IN subtypes (421.7

MΩ ± 30.1 vs 264.0 MΩ ± 16.1, p<0.0001) (Fig 4.6D,G). Accordingly, RC soma area is significantly

smaller than other V1 INs (259.5 µm2 ± 13.3 vs 423.2 µm2 ± 31.2; p<0.0001) (Fig 4.6E,F). Moreover, RCs

have significantly decreased rheobase and time constant, and their resting membrane potential and

threshold potential are similar (Fig 4.6H-K). Altogether, these results indicate that ESC-derived RCs have

increased membrane excitability compared to non-RC V1 INs.

In the mammalian spinal cord, RCs exhibit high frequency burst firing (up to 1000 Hz) in response

to antidromic stimulation of ventral roots or evoked potentials in single MNs (Eccles et al, 1954; Van

Keulen, 1981; Eccles et al, 1961; Walmsley & Tracey, 1981; Hamm et al, 1987). We performed current-

clamp recordings in ESC-derived V1 INs to test if in vitro-generated RCs could be distinguished from

other V1 INs by their active membrane properties, including their ability to reproduce burst firing of action

potentials (AP). Surprisingly, there was no significant difference in the firing response of RCs compared to

non-RC V1 INs (Fig 4.7A,B). Furthermore, while some V1 INs produced initial bursts of APs followed by

137

tonic firing, the maximum firing frequency for both RCs and non-RCs was similar (~15 Hz) (Fig 4.7A,C).

Thus, at this stage in their development, ESC-derived RCs be distinguished from other V1 INs based on

their passive, but not active, firing properties.

Optogenetic approach to studying MN-RC synaptic specificity in vitro After assessing RC-specific functional properties, we used optogenetics-mediated MN stimulation

coupled with whole-cell patch-clamp recordings of V1 INs to probe MN-RC functional synaptic

connectivity in the in vitro co-culture system. En1-tdTomato FP cells were co-cultured with Hb9::GFP MNs

expressing the light-sensitive ion channel channelrhodopsin-2 (ChR2) to enable optical stimulation of

MNs specifically (Nagel et al, 2003; Boyden et al, 2005; Bryson et al, 2014) (Fig 4.8A). Optogenetic

stimulation of MNs using a brief pulse of light produced single APs, which were then able to elicit APs in

synaptically-connected V1 INs, including molecularly-identified RCs (Fig 4.8B,C). RC responses were

completely abolished using a combination of the cholinergic blockers mecamylamine and atropine (Fig

4.8D). Reinforcing our prior analyses using VAChT immunoreactivity, RCs were more likely to generate a

depolarization compared to other V1 INs in response to MN optogenetic activation (86.4% ± 9.43 vs

22.2% ± 22.2, p<0.05) (Fig 4.8E).

To confirm the monosynaptic response of RCs to MN photoactivation, we examined the response

onset variability, or jitter, of the RC response over multiple trials at different frequencies (Fig 4.9A).

Previous studies have shown monosynaptic responses exhibit small latency jitters (i.e., the difference

between minimum and maximum latency of the response onset) as stimulation frequency increases,

while disynaptic or polysynaptic responses show increased variability (Shneider et al, 2009). Overall, the

latency from the MN AP to the onset of the RC response was ~4 milliseconds, suggestive of a

monosynaptic response (Fig 4.9B). At 0.1 and 1 Hz stimulation frequencies, the variability of the RC

response was minimal (coefficient of variation = 0.066 ms ± 0.0087 and 0.06 ms ± 0.0046, respectively),

confirming that the RC response is likely monosynaptic (Fig 4.9C). At 10 Hz, jitter increased a small

amount (0.091 ms ± 0.0065), but this difference was not significant and in line with prior in vivo

observations (data not shown) (Shneider et al, 2009).

138

Discussion

In the postnatal mouse lumbar spinal cord, ~50 candidate V1 IN subtypes have been predicted

based on combinatorial expression of 19 TFs as well as their restricted settling positions in the ventral

horn (Bikoff et al, 2016; Gabitto et al, 2016). While functional properties of most of these newly defined V1

subtypes have not been characterized, V1-derived Renshaw cells have long been known to provide

postsynaptic inhibition onto α-motor neurons (Renshaw, 1941; Renshaw, 1946; Eccles et al, 1954; Eccles

et al, 1961). Conversely, RCs are the only spinal IN cell type to receive MN axon collaterals (Alvarez et al,

2013). How RCs are uniquely specified to form recurrent inhibitory connections with MNs is not generally

known. Besides RCs, only IaINs, which are derived from both V1 and V2b progenitors, have been shown

to provide direct inhibitory inputs onto MNs to mediate reciprocal inhibition for flexor-extensor alternation

(Eccles et al, 1956; Feldman & Orlovsky, 1975; Zhang et al, 2014; Britz et al, 2015). Whether all V1-

derived neurons provide monosynaptic inhibitory inputs onto spinal MNs is also not currently known.

Renshaw cell-motor neuron synaptic specificity in vitro

Using ESC-derived V1 INs and MNs, I developed a simple co-culture assay for studying the

synaptic connectivity of V1 IN subtypes, especially RCs. Previously, we demonstrated that ES-RCs

acquire molecular signature of their in vivo counterparts as well as their unique spatial coordinates when

transplanted in the embryonic spinal cord (See Chapter 3). Using whole-cell patch-clamp

electrophysiological recordings, we show that ES-RCs also adopt distinct membrane properties compared

to other V1 INs, including significantly decreased soma area, increased input resistance, decreased

rheobase and increased time constant under steady resting membrane and threshold potential. Our

results are compatible with in vivo observations showing that Cb-expressing RCs are smaller in size than

other INs, with larger input resistance compared to IaINs and MNs (Bui et al, 2003; Mentis et al, 2006;

Alvarez & Fyffe, 2007). Together, these data suggest that RCs may be more excitable than other V1 INs,

although the significance of this is not currently known. Interestingly, ES-RCs did not fire high frequency

burst discharges in response to current injection, a trademark of RC firing in vivo, although a spike

doublet is observed in some molecularly defined RCs, which has been reported to be a characteristic

response of early postnatal RCs to ventral root stimulation in vivo (Lamotte d’Incamps & Ascher, 2008;

139

Perry et al, 2015). Whether ES-RCs can be coaxed to fire burst discharges of APs is not known, although

some evidence suggests that hyperpolarization of early postnatal mouse RCs in vivo is required to induce

them to fire burst discharges up to 80 Hz, suggesting that burst firing may be a developmentally-regulated

phenomenon (F. Alvarez & G. Mentis, personal communication).

ES-RCs not only acquire distinct membrane properties compared to other V1 INs in vitro, but they

also exhibit distinct patterns of synaptic connectivity as well. Based on monosynaptic RABV tracing, ES-

RCs provide monosynaptic inputs on MNs at higher frequency than expected, although non-RC V1 INs,

including FoxP2-expressing subtype, also synapse onto MNs in vitro. Conversely, using VAChT-

immunoreactivity analysis and by recording V1 IN subtype response to optogenetic stimulation of MNs,

we show that MNs provide significantly more monosynaptic, cholinergic inputs onto ES-RCs than other

non-RCs. These data demonstrate that in vitro-generated RCs and MNs recapitulate spinal patterns of

synaptic connectivity without having ever been exposed to developmental signals in the spinal cord.

Mechanisms of Renshaw cell-motor neuron synaptic specificity

Indeed, the development of a robust in vitro co-culture system paves the way for studies

investigating molecular mechanisms underlying RC-MN synaptic specificity. Using ESC-derived neurons,

we show that synaptic specificity between RCs and MNs is largely retained in vitro, suggesting that

signals mediating RC-MN interactions can be sufficiently induced even in a simple co-culture system

largely devoid of spatial cues and cellular interactions present in the spinal cord. These data indicate that

RC synaptic connections may be hardwired from initial specification based on their lineage and birth

order. RCs are firstborn among V1 INs, encountering MN axons early during their development as they

migrate through MNs to reach their final settling position near the ventral root exit of MN axons (Stam et

al, 2012; Benito-Gonzalez & Alvarez, 2012; Alvarez & Fyffe, 2007; Alvarez et al, 2013). During early

motor circuit formation, RCs are the only other cell type in the ventral horn, their migratory and settling

patterns potentially predisposing them to form synaptic connections with MNs. Correlation between timing

of neurogenesis and synaptic connectivity has also been described in other spinal circuits, as well as in

the hippocampus, cerebellum, and olfactory bulb (Arber, 2012; McLean & Fetcho, 2009; Tripodi et al,

140

2011; Deguchi et al, 2011; Espinosa & Luo, 2008; Imamura, 2011). Thus, RC-MN synaptic specificity in

the spinal cord may have arisen from a process of precisely timed neurogenesis.

One important question is subtype-specific birth order is translated into functional specialization of

different neuronal cell types? Stam et al, (2012) showed that by virtue of being the earliest-born V1 IN cell

type, RCs acquire a specialized TF program distinguishing them from other V1 INs during. Whether these

RC-specific TFs programs are involved in RC-specific synaptic connectivity has not been shown, but it is

tempting to consider that these TF programs coordinate RC-MN synapse formation via regulation of

genes underlying RC responsiveness to MNs. Indeed, during development, many aspects of synapse

formation, differentiation and function rely on communication between presynaptic axons and their

postsynaptic targets. Anterograde and/or retrograde signaling, mediated by secreted and/or membrane

bound factors, may be involved in the specific formation of recurrent inhibitory connections between RCs

and α-MNs. (Sanes & Yamagata, 2009). While signals mediating RC-MN connectivity have not yet been

identified, expression profiling of DAPT-treated ES-V1 INs, which are enriched in RCs, may yield

candidate cell surface molecules and secreted signals for further investigation.

In addition to temporal factors determining RC-MN connectivity, spatial factors have the potential

to be involved in RC-MN circuitry in vivo. Shortly after their specification, RCs migrate into a region of the

ventral horn densely packed with MNs. Once they reach their final position in this region, RCs and MNs

might form recurrent synaptic connectivity based on Peter’s rule, which posits that synaptic contacts

occur whenever dendrites and axons happen to be in apposition (Peters and Feldman, 1976;

Stepanyants et al, 2002). However, our results using ESC-derived neurons argue against this conclusion.

ES-RCs and MNs, after all, are never exposed to the spinal cord environment and are not constrained by

spatial boundaries, yet RCs still receive the large proportion of MN cholinergic inputs. Currently, we have

no evidence that RCs closer to MNs than other V1 INs. However, even in this circumstance, MN axons

travel widely in the in vitro setting, making it unlikely that they would be limited to only synapsing on cells

closest to them (data not shown). In addition, recent evidence shows that there is no obvious correlation

between the strength of recurrent inhibitory connections between MNs and RCs and the relative distance

between them (Moore et al, 2015). Finally, the increased monosynaptic connectivity of RCs to MNs

141

revealed by RABV tracing, especially considering their relatively small numbers in culture, suggests that

RCs in vitro recognize MNs as an appropriate synaptic partner even without spatially-derived cues.

Motor neuron synaptic connectivity of non-Renshaw cell V1 interneurons in vitro

In addition to being used as a tool for identification of signals mediating RC-MN synaptic

specificity, our system can also be used to determine the degree and significance of synaptic connectivity

of other V1 IN subtypes with MNs. While RCs and IaINs have been shown to provide postsynaptic

inhibition of MNs, the synaptic connectivity of other V1-derived cell types is mostly unknown. Interestingly,

En1 mouse mutants in vivo show dramatic reduction of Cb+ contacts on MNs, suggesting that En1

regulates the number of RC to MN synaptic contacts. Conversely, loss of En1 does not change MN-

VAChT inputs onto MNs, indicating normal MN to RC synaptic contacts (Sapir et al, 2004). Although En1

expression has been shown to be required for synapse formation of RCs onto MNs specifically, all V1-

derived INs express En1 during their early development and En1 is required for their proper axonal

pathfinding and fasciculation (Saueressig et al, 1999; Matise & Joyner, 1997). Therefore, En1 might

regulate synapse formation of all V1 INs onto MNs, indicating that postsynaptic inhibition of MNs is a trait

shared by all V1-derived cells.

Nevertheless, our RABV transsynaptic tracing data indicate that RCs are more likely to provide

monosynaptic inputs onto MNs based on their relative numbers in vitro. Conversely, V1 INs not

expressing either Cb or FoxP2 are significantly less likely to be synaptically connected to MNs. Therefore,

although other V1 INs may provide monosynaptic inputs to MNs, RCs might be responsible for the bulk of

postsynaptic inhibitory inputs onto MNs, possibly because of their spatial proximity to MNs or

developmental convergence. Interestingly, recent analyses of another V1 IN subtype expressing the TF

Sp8 demonstrates that although Sp8-expressing V1 INs provide monosynaptic inputs on MNs, they do so

significantly less frequently than molecularly defined RCs (~1:5 ratio) (Bikoff et al, 2016; J. Bikoff & T.

Jessell, personal communication). In addition, contrary to previous reports of RCs providing only a minor

contribution to MN firing, more recent investigations using in vivo paired recordings of reciprocally

connected RC-MN pairs suggest that RCs provide strong and numerous inhibitory inputs onto MNs

(Bhumbra et al, 2014). Therefore, while all V1-derived cell types may be competent to form functional

142

synapses onto MNs, RCs may be specialized to provide increased and/or stronger inhibitory inputs onto

MNs by virtue of their recurrent inhibitory connections. Finally, to further assess functionality of V1

synaptic connections onto MNs, I have also generated En1-tdTomato reporter lines carrying ChR2-YFP

(data not shown).

Notably, ES-MNs also provide cholinergic inputs onto non-RC V1 INs. First, while optogenetic

stimulation of MN-dI4 IN cultures has not yet been performed, 11.5% of Ptf1a-tdTomato FP cells received

VAChT+ inputs. One possibility is that these VAChT synapses do not originate from MNs, but instead

from other cholinergic cell types in the culture residual from FAC sorting for INs and MNs (Barber et al,

1984; Phelps et al, 1984; Huang et al, 2000). Cholinergic Pitx2-expressing V0c INs, which provide C

boutons on MNs, are the most likely candidate, since differentiation conditions for V1 and dI4 INs

generate some Evx1/2-expressing V0 INs (Zagoraiou et al, 2009). Alternatively, VAChT-immunoreactivity

on dI4 INs might represent transient physical contacts and not functional synapses, the consequence of

MN axons traversing the area in search of their proper synaptic partners. In general, VAChT+ contacts on

dI4 INs were less dense than on V1 INs, although co-culture of all three spinal neuron types would

provide more accurate comparison (data not shown).

Analysis of V1 IN subtypes shows that cells lacking expression of Cb and/or OC2 represented

43.5% of all VAChT-innervated V1 INs, while 22.2% of these cells received monosynaptic inputs from

optogenetically-activated MNs. Interestingly, V1 INs expressing only OC2 not only received the largest

share of MN-VAChT inputs to non-RCs (34.5%), but a significant fraction of this subtype received

cholinergic inputs (~33%), whereas very few V1 INs expressing Cb only or neither Cb/OC2 received

VAChT-immunoreactive inputs. The identity of non-RC OC2-expressing cells is not known. In the

postnatal lumbar spinal cord, 17% of V1 INs express OC2; while these are most densely distributed in the

RC area, there is a significant population of OC2-expressing V1 INs dorsal to this area (Bikoff et al,

2016). Although MN collaterals have been shown to form synaptic connections with RCs in vivo, whether

some OC2-expressing INs also recruit MN collaterals has not been directly examined. In the spinal cord,

the likelihood of cellular interactions is limited not only by cell-type specific programs, but also by spatial

constraints. Using an in vitro co-culture system, we strip the cells of these restraints, potentially revealing

their intrinsic connectivity preferences.

143

Furthermore, the identity of Cb only-expressing V1 INs is unclear. In the spinal cord, 7-8% of the

neurons in the RC area at p19 weakly express Cb and do not express large clusters of gephyrin,

suggesting that these cells might represent a functionally distinct class of INs (Sapir et al, 2004).

Furthermore, large V1-derived Cb-expressing cells are also present near the central canal of the

postnatal spinal cord, far from the RC area, indicating that while >90% of Cb-expressing V1 INs are RCs,

a small minority may have distinct functional properties (Floyd & Ladle, 2015). Finally, V1 INs lacking both

Cb and OC2 expression express little VAChT and largely do not respond to MN inputs. Although the

identity of these cell types is not known, we can take advantage of expression profiling data of ES-V1 INs,

including from DAPT-treated cells, to identify candidate molecular markers for these cell types.

Finally, an alternative explanation for MN cholinergic inputs onto non-RC V1 INs is that although

these experiments provide strong indication of synaptic specificity between ES-RCs and MNs even in

vitro, aberrant connections may also arise between different V1 IN subtypes and MNs, either due to the

lack of additional spinal-derived cues to regulate synaptogenesis, or because ESC-derived V1 INs have

not yet acquired mature neuronal identity. Indeed, while adult RCs receive few primary sensory afferents

compared to IaINs, these inputs are more abundant on early postnatal RCs, suggesting that RCs undergo

synaptic remodeling during their development (Mentis et al, 2006; Siembab et al, 2010). One possibility is

that other synaptic inputs on non-RCs in vivo, such as from sensory afferents or other INs, function to

deselect and weaken early MN inputs. Alternatively, non-RC V1 INs in the spinal cord may be shielded

from MN inputs based on their relative distance from MN axons, while MN axons in the in vitro setting

have liberty to form synapses on cell types they normally do not encounter. Nevertheless, our findings

that MNs still selectively form synapses with RCs suggests that RC-MN synaptic specificity is largely

retained even in dissociated culture.

144

Figure 4.1

Figure 4.1 Differential interactions of ESC-derived interneurons with motor neurons In co-culture, ESC-derived V1 and dI4 INs exhibited qualitatively different types of contacts with Hb9::GFP MNs. ES-V1 INs preferentially form close cellular contacts with MNs even in sparse cultures (25% of V1 interactions vs 3% of dI4). While ES-V1 IN axons tightly entwine around MN cell bodies and proximal dendrites, ES-dI4 IN axons encircle MN cell bodies and travel loosely around the MNs (30 vs 6% of interactions compared to V1 INs). Furthermore, a greater proportion of ES-dI4 INs fail to interact with MNs at all, even at cell densities similar to V1 INs (6 vs 0%).

145

Figure 4.2

Figure 4.2 Monosynaptic rabies virus tracing of V1 interneuron-motor neuron connectivity (A) Schematic depicting potential synaptic connectivity of RC and non-RC V1 INs with MNs in the ventral spinal cord. (B) (Top) TVA-2A-Rabies G protein construct delivered into Hb9::GFP ESCs for MN-specific initial viral infection and retrograde monosynaptic transfer; (Bottom) Timeline of V1 IN-MN co-culture for monosynaptic RABV tracing. (C) Depiction of RABV tracing, with large, GFP-expressing MN surrounded by En1-tdTomato V1 INs (red), including SAD∆G-GFP-expressing V1 INs, which are monosynaptically connected to MNs (yellow). (D, E) Efficiency of initial RABV infection of MNs, using SAD∆G-dsRed virus variant and immunostaining for Rabies-N nucleoprotein (n=3). (F) Quantification of timing required for efficient secondary infection of V1 INs directly synapsing onto rabies-infected MNs (n=3 each time point). (G) Comparison of efficiency of secondary infection of V1 INs based on ratio of INs to MNs (n=3). (Student’s t-test, *p<0.05, **p<0.01). Scale bars = 50 µm.

146

Figure 4.3

Figure 4.3 RABV tracing reveals V1 interneuron subtype-specific connectivity with motor neurons (A) After viral tracing, immunocytochemistry detecting Calbindin and FoxP2 was used to reveal the subtype identity of premotor V1 INs. Scale bar = 50 µM. (B) There is a modest, but not statistically significant, increase in Cb-expressing premotor V1 INs compared to FoxP2+ V1 IN subtype. (C) There are ~2X as many FoxP2-expressing V1 INs generated as Cb-expressing V1 INs. Student’s t-test (**p<0.01). (D) Taking into account V1 IN subtype prevalence in culture to calculate a connectivity index (C.I.), Cb-expressing V1 INs are significantly more likely to provide monosynaptic inputs onto MNs than other V1 IN subtypes, especially cells not expressing either Cb or FoxP2. Red dotted line marks C.I. of 1, signifying 50% likelihood of synapsing with MNs. (n≥3). ANOVA (*p<0.05, **p<0.01).

147

Figure 4.4

Figure 4.4 Differential VAChT-immunoreactive inputs on V1 and dI4 inhibitory interneurons in vitro (A) Schematic depicting potential MN cholinergic inputs onto RC and non-RC V1 INs. (B) Timeline of V1 IN-MN co-culture for monosynaptic RABV tracing. (C) Lack of VAChT-immunoreactive synaptic inputs on Ptf1a-tdTomato FP cell body or proximal neurites despite proximity to VAChT+ cells in culture. (D) Quantification of VAChT-immunoreactive inputs on ESC-derived V1 versus dI4 spinal INs (n=3; Student’s t-test, *p<0.05).

148

Figure 4.5

149

Figure 4.5 Stem cell-derived Renshaw cells preferentially receive VAChT+ cholinergic inputs (A) Immunostaining of Day 22 dissociated and FACS purified En1-tdTomato FP cells for RC markers Calbindin/Onecut2 and VAChT for identification of MN-cholinergic inputs. (B) V1 IN subtype-specific recruitment of VAChT-immunoreactive inputs (n=3, ANOVA, **p<0.01). (C) ~80% of molecularly-defined RCs receive VAChT cholinergic inputs compared to only 20% of non-RC V1 INs (n=3, ANOVA, ***p<0.001). (D) Additional partition of V1 INs into four molecularly distinct groups indicates that RCs and OC2 only-expressing V1 INs are highly likely to receive VAChT inputs compared to Cb only-expressing V1 INs or cells not expressing with Cb or OC2 (n=3, ANOVA, ***p<0.001). (E) Calculation of connectivity index for RCs versus non-RCs (Student’s t-test, *p<0.05).

150

Figure 4.6

Figure 4.6 ESC-derived Renshaw cells exhibit distinctive passive membrane properties (A) Timeline of V1 IN-MN co-culture for electrophysiological recordings. (B) Post hoc immunostaining of Neurobiotin-filled En1-tdTomato recorded cell for subtype identification. Renshaw cells were strictly identified as En1-tdTomato FP cells co-expressing Calbindin and OC2. Scale bar = 50 µM. (C) Superimposed membrane responses (upper traces) following current injection (lower traces) in RC and non-RC V1 INs in vitro. (D) Current/voltage relationships for RCs versus non-RCs. Based on the slope of the linear current-to-voltage relationship, RCs have increased input resistance compared to other V1 IN subtypes (421.7 MΩ ± 30.1 vs 264.0 MΩ ± 16.1, Student’s t-test, p<0.0001). (E) Scatterplot depicting relationship between soma size and input resistance for RC vs non-RC V1 INs. (F) Passive membrane properties of ESC-derived RC and non-RC V1 INs. Compared to non-RC V1 INs (blue bars), RCs (red bars) have decreased soma size, increased input resistance, decreased rheobase, and increased time constant without significant differences in resting membrane potential or threshold potential, indicating that ES-RCs may be hyperexcitable compared to non-RCs (35 RCs vs 30 non-RCs, Student’s t-test, *p<0.05, ***p<0.001).

151

Figure 4.7

Figure 4.7 Active membrane properties of ESC-derived Renshaw cells Current-clamp recordings were performed in ESC-derived V1 INs to test if in vitro-generated RCs could be distinguished from other V1 INs by their active membrane properties, including their ability to reproduce burst firing of action potentials (AP). (A) Depicted are examples of repetitive firing (top) and single AP (bottom), elicited in ES-RCs cultured for 2 weeks on astrocyte monolayer. The asterisk denotes spike doublet firing. (B) There was no significant difference in the firing response of RCs compared to non-RC V1 INs. (C) While some ES-V1 INs produced initial bursts of APs followed by tonic firing, the maximum firing frequency for both RCs and non-RCs was similar (~15 Hz).

152

Figure 4.8

Figure 4.8 Selective motor neuron cholinergic inputs onto ESC-derived Renshaw cells (A) ESC line expressing Hb9::CD14-IRES-GFP for MN identification and MACS sorting, and CAG::ChR2-YFP for optogenetic stimulation (Bryson et al, 2014). Depicted are dissociated MNs differentiated from this line, immunostained for MN-specific markers choline acetyltransferase (ChAT) and Hb9. (B, C) Optogenetic stimulation of MNs using brief pulse of light (green line, 25 ms) produced single APs, which were able to elicit APs in synaptically-connected V1 INs, including molecularly-identified RCs, as revealed by current-clamp electrophysiological recordings. (D) RC responses were completely abolished using a combination of the cholinergic blockers mecamylamine and atropine. (E) ES-RCs are significantly more likely to depolarize in response to MN photoactivation compared to non-RC V1 INs (n=3, Student’s t-test, *p<0.05).

153

Figure 4.9

Figure 4.9 Monosynaptic motor neuron connections onto Renshaw cells in vitro (A) Response onset variability, or jitter, of the RC response over multiple trials at different frequencies (0.1 and 1 Hz). (B) The latency from the MN AP to the onset of the RC response was ~4 sec, suggestive of a monosynaptic response. (B) At 0.1 and 1 Hz stimulation frequencies, the variability of the RC response was minimal (coefficient of variation = 0.066 ms ± 0.0087 and 0.06 ms ± 0.0046, respectively), confirming that the RC response is likely monosynaptic.

154

Chapter 5: Specification of dI4 interneuron subtypes: in search of the GABApre Introduction

All dI4 INs emerge from a common Ptf1a-expressing spinal dp4 progenitor domain, including an

early-born dI4 population that settles in the intermediate zone of the spinal cord and projects to the

ventral horn to synapse onto proprioceptive sensory neurons (pSN), and a late-born dILA population that

migrates into the superficial dorsal horn and provides inhibition to cutaneous sensory afferents there

(Glasgow et al, 2005; Betley et al, 2009; Wildner et al, 2013). The precise combination and timing of

molecular cues specifying distinct Ptf1a-derived subtypes in the spinal cord has not been explored. Here,

I used a stem cell-based system to dissect potential pathways involved in the generation of dI4 versus

dILA inhibitory INs. The goal of these studies is to determine if GABApre INs, a subset of early-born dI4

INs, can be generated from ESCs in order to model their synaptic specificity with pSNs.

The efficient differentiation of ESCs to dI4 INs requires RA signaling to neutralize and caudalize

progenitors to spinal cord identity, as well as ActivinA signals to dorsalize neuronal fates and activate the

Ptf1a transcriptional program. To determine the subtype identity of dI4 INs generated under these

conditions, I analyzed expression profiling data from dI4 INs at early and late developmental time points

to identify candidate factors distinguishing dI4 and dILA IN populations (Wildner et al, 2013; Meredith et

al, 2013; Bikoff et al, 2016). Thereafter, I used ESC-derived dP4 progenitors to screen different TGFß

family member ligands in order to identify the molecular code enriching for early-born dI4 IN population

containing GABApre INs (Lee & Jessell, 1999; Mizuguchi et al, 2006; Hori et al, 2008)

To test the functionality of molecularly-defined ESC-derived dI4 IN subtypes, I transplanted

nascent Ptf1a-tdTomato FP cells into the developing chick neural tube and examined their migration and

axonal projections into appropriate spinal laminae. Furthermore, to assess their synaptic connectivity with

pSNs over MNs in vitro, I first developed co-culture assays using ES-dI4 INs and ES-MNs for determining

their synaptic connectivity using immunocytochemical and monosynaptic rabies virus tracing approaches.

I also optimized methods for generating pSNs through direct transcriptional programming of ESCs (Yang

et al, 2011; Mazzoni et al, 2013a; Marmigère & Ernfors, 2007; Dykes et al, 2011). Efficient generation of

pSNs from ESCs will provide an invaluable resource for studying the formation of GABApre IN synaptic

specificity in the developing spinal cord.

155

Results

Molecular differentiation of stem cell-derived dI4 IN subtypes

In the dP4 spinal domain, progenitors giving rise to early-born dI4 INs exhibit low or no

expression of the bHLH factor Ascl1, while progenitors of later-born dILA INs express high levels of Ascl1

(Fig 5.1A). High Ascl1 expression is also seen in progenitor domains immediately dorsal (dP3) and

ventral (dP5) to the Ptf1a spinal domain (Helms et al, 2005; Mizuguchi et al, 2006; Wildner et al, 2006;

Helms & Johnson, 2003). To determine if ESC-derived dP4 progenitors also express different levels of

Ascl1, I performed ICC to detect for Ptf1a and Ascl1 co-expression in Day 6 EBs differentiated with RA

only (Fig 5.1B). Equal cohorts of Ptf1a-expressing cells expressed high, low and no Ascl1, suggesting

that Ptf1a-expressing progenitors at this stage are likely competent to produce either dI4 IN subtype (Fig

5.1C). Nevertheless, without an Ascl1-lineage tracer in the Ptf1a-tdTomato or Thy1YFP ESC lines, it is

not currently possible to follow the fate of progenitors expressing no, low or high levels of Ascl1.

Lacking other molecular markers for distinguishing early-born, intermediate zone dI4 INs and

later-born, superficial horn dILA INs, I next used expression profiling from embryonic and early postnatal

Ptf1a-derived neurons in the spinal cord to screen for candidate genes uniquely expressed in these two

populations (Wildner et al, 2013; Bikoff et al, 2016; Sunkin et al, 2013). While few genes were found to be

both enriched in dI4 INs and have restricted expression in the deep dorsal horn, the TF TFAP2b was

identified in multiple screens as being highly expressed in dI4 INs compared to V1 INs at different stages

of development, as well as in Ascl1-independent dorsal spinal cord neurons (data not shown) (Bikoff et al,

2016; Wildner et al, 2013). Indeed, analysis of TFAP2b expression in the spinal cord of Ptf1a::cre x

ROSA::tdTomato lineage reporter mice reveals restricted expression of TFAP2b in the intermediate spinal

cord near the central canal early (e12.5), as well as in the most ventral subpopulation of Ptf1a-tdTomato

FP cells at later stages (e12.5, e18.5) (Fig 5.2A,C). Although TFAP2b emerged as the strongest

candidate for molecular identification of early-born dI4 INs, which include GABApre INs, not all Ptf1a-

tdTomato cells in the deep dorsal horn express TFAP2b (~8.7% at late embryonic stages) and not all

TFAP2b cells are Ptf1a-derived (data not shown).

Although our goal is to optimize differentiation of GABApre INs, we were also interested in the

developmental pathways involved in generation of non-GABApre cell fates. Indeed, increased

156

understanding of dILA IN specification may provide insights into mechanisms of dP4 progenitor

diversification. While relatively few dI4 IN-enriched genes have restricted expression in the deep dorsal

horn, many have broad expression in the superficial dorsal horn (e.g. Zic1, Skor1, Skor2, Sall3) (data not

shown) (Wildner et al, 2013; Bikoff et al, 2016; Sunkin et al, 2013). Likewise, the TF Ebf1 (Early B-cell

factor 1) is also enriched in superficial laminae I and II of the dorsal horn, where dILA neurons migrate

after they are born (Fig 5.2B) (Wildner et al, 2013; Mizuguchi et al, 2006; Lai et al, 2016). Ebf1 is

expressed in Ptf1a-tdTomato FP cells at e18.5, as well as in non-FP cells in the dorsal horn, suggesting

that it is likely not a specific marker for dILA INs (Fig 5.2C). Indeed, excitatory Lmx1b-expressing INs in

laminae I and II have also been to express Ebf1 (Ding et al, 2004). However, the high density of Ebf1-

expressing cells in superficial layers of the dorsal horn where dILA neurons settle prompted further

investigation of its role in dI4 IN subtype diversification.

Under RA only differentiation conditions, RNA-seq expression profiling data reveals that ES-dI4

INs are also enriched for TFAP2b and Ebf1 TFs compared to V1 INs (log2FC 8.86 and 2.31, respectively)

(data not shown). Yet, on Day 8 of differentiation, 23.3% of Ptf1a-tdTomato FP cells co-express TFAP2b,

while only 4.5% of FP cells express Ebf1 (Fig 5.2D). Previously, we have shown that efficient

differentiation of Ptf1a-derived neurons in the spinal cord likely relies on the confluence of RA and TGFß

signaling from the roof plate and overlying ectoderm. Compared to RA only differentiation, TGFß family

agonists and antagonists (SB-431542, ActivinA, and BMP4) slightly decreased the production of TFAP2b-

expressing FP cells (20.6, 18.9 and 9.9%, respectively) (data not shown). Conversely, while SB-431542

and BMP4 treatment resulted in a small decrease in Ebf1-expressing FP cells (1.7 and 2.5%,

respectively), ActivinA strongly promoted the formation of Ebf1-expressing FP cells (22.6%, p<0.01) (Fig

5.2D). Since overall dI4 neurogenesis increases with ActivinA treatment (9.6 vs 39.3%), the number of

cells expressing Ebf1 is dramatically upregulated in RA+ActivinA conditions (data not shown). Thus, the

dorsalizing effects of ActivinA induces the formation of Ebf1-expressing dI4 INs, while decreasing the

proportion of TFAP2b-expressing FP cells. As in vivo, TFAP2b and Ebf1 expression never co-localize in

the in EBs, indicating that they mark discrete populations of Ptf1a-derived cells (Fig 5.2C,D).

Interestingly, transplant of early (Day 6) Ptf1a-tdTomato EBs differentiated with RA only into the

developing chick neural tube also dramatically increases the generation of Ebf1-expressing Ptf1a-

157

tdTomato cells (Fig 5.2E). At this stage during normal chick spinal development (HH Stage 30 or E6-7),

Ebf1 expression is largely restricted to the ventral spinal cord, with a small, dense population of Ebf1-

expressing cells in the superficial dorsal horn adjacent to the overlying ectoderm (data not shown). In the

engrafted chick neural tube, however, many FP cells migrate from the central graft site into the superficial

dorsal horn where they express high levels of Ebf1. Moreover, transplanted Ebf1-expressing cells migrate

more dorsally than TFAP2b-expressing cells, including those co-expressing the Ptf1a-tdTomato reporter

(Fig 5.3E). These data suggest that in vitro-generated dP4 progenitors acquire subtype-specific

characteristics upon exposure to spatial patterning cues present in the developing spinal cord.

Synaptic connectivity of ESC-derived dI4 INs in vitro

As described previously, transplantation of ESC-derived neurons is a useful method for assessing

functional properties of molecularly-defined subtypes, including their migration and axonal projections.

Yet, efforts to study the circuit integration of engrafted ESC-derived neurons, including their synaptic

specificity, have proved technically challenging. Therefore, we established an in vitro co-culture system to

study the synaptic connectivity of ES-derived dI4 INs with different spinal neuron cell types. In the spinal

cord, dI4-derived GABApre INs preferentially synapse on pSN afferent terminals while eschewing

synaptic connections with MNs in the vicinity (Betley et al, 2009). As described in Chapter 4, dI4 INs

exhibit different cellular interactions with MNs compared to V1 INs, including a circling behavior in which

axons traveled loosely around MN cell bodies, as well as decreased interaction events in total (Fig4.1A,B)

However, in many instances, GABAergic synapse markers, including Gad67, Gad65, and VGAT co-

localized with synapse components (synapsin, SV2b) at sites of Ptf1a-tdTomato axonal contacts with MN

cell bodies and dendrites, suggesting that dI4 INs in vitro may form physical synapses with ES-MNs.

To test the functionality of these synapses, I used a similar approach described previously for

ESC-derived V1 INs and MNs in which I adapted monosynaptic rabies virus (RABV) tracing for

determining the synaptic connectivity of ES-dI4 INs with ES-MNs (Wickersham et al, 2007a,b; Callaway,

2008; Osakada, 2011). Surprisingly, in vitro-generated dI4 INs form many monosynaptic connections with

MNs, particularly when compared to V1 INs, which are a bona fide premotor IN (Sapir et al, 2004;

Saueressig et al, 1999; Alvarez et al, 2005). Four days post-RABV infection, 13.7% of dI4 INs express

158

SAD∆G-GFP, suggesting they are monosynaptically connected to MNs, compared to 9.4% of V1 INs

(p<0.05). This difference is even more striking three days later, when 24.4% of dI4 INs compared to

14.3% of V1 INs appear to provide monosynaptic inputs onto MNs (p<0.05) (Fig 5.4). Thus, based on

RABV-mediated transsynaptic tracing, in vitro-generated dI4 INs not only form synaptic contacts directly

on MNs, but they do so at a higher rate than V1 INs.

Notably, in our RABV co-culture studies, dI4 INs were generated using RA only without ActivinA

treatment, which produces ~28.3% TFAP2b-expressing dI4 INs and 18.0% Ebf1+ dI4 INs on Day 19 of in

vitro culture. A significant proportion of Ptf1a-tdTomato FP cells providing monosynaptic inputs onto MNs

co-expressed TFAP2b (35.5% on Day 4 post-RABV infection), while relatively few cells expressed Ebf1

(2.5%). Calculation of connectivity index for TFAP2b versus Ebf1-expressing cells yielded 1.27 and 0.13,

respectively, indicating that TFAP2b-expressing cells were slightly more likely than chance to provide

synaptic inputs onto MNs, while Ebf1-expressing cells are relatively unlikely to do so (data not shown).

Programming proprioceptive sensory neurons from stem cells

Although reasons why dI4 IN form prolific synapses with MNs in vitro are still unclear (see

Discussion), we wanted to test if dI4 INs could form appropriate synaptic connections with sensory targets

in vitro. Initial studies using heterogeneous SN populations dissected from embryonic mouse spinal

dorsal root ganglion (DRG) provided evidence that axons emanating from ES-dI4 IN EBs preferentially

interacted with SNs in DRG explants compared to ES-MN EBs (data not shown). However, these

experiments provided limited insights into cell type-specific patterns of synaptic connectivity. Therefore,

we optimized a method for transcriptional programming of pSNs directly from ESCs. Although directed

differentiation of ESCs using extrinsic patterning signals has proven highly effective at generating certain

neuronal cell types, other classes of cells have proven more challenging to produce (Petros et al, 2011;

Sandoe & Eggan, 2013; Tsunemoto et al, 2015). Earlier attempts to differentiate select SN cell types from

pluripotent stem cells have generally led to enrichment of cutaneous SNs, with poor yield of pSNs

(Wainger et al, 2015; Maury et al, 2015; Blanchard et al, 2015). Alternatively, many recent studies have

shown that understanding the transcriptional regulatory network underlying specification of different cell

fates allows for the direct programming of cell identity through forced expression of key TFs (Kyba et al,

159

2002; Andersson et al, 2006; Panman et al, 2011; Martinat et al, 2006; Vierbuchen et al, 2010; Takahashi

& Yamanaka, 2006). Indeed, spinal MNs can be rapidly and efficiently programmed from ESCs by forced

expression of 3 TFs – Ngn2, Isl1/2, and Lhx3, with efficiencies far exceeding normal RA/SAG

differentiation (Mazzoni et al, 2013a).

By adapting the same system for spinal MN programming from ESCs, we induced expression of

TFs involved in pSN-specific development in vivo. All SN precursors transit through transcriptional states

defined by Ngn2, Isl1/2, Pou4f1 (Brn3a) and FoxS1 expression (Lanier et al, 2009; Sun et al, 2008; Liu &

Ma, 2011; Montelius et al, 2007). Subsequently, other TFs differentially expressed among different

classes of sensory neurons function to activate or repress the expression of sensory modality-defining

features, including TrkA-C receptors for different neurotrophic factors and neuropeptide expression

(Marmigère & Ernfors, 2007; Lallemand & Ernfors, 2012). Indeed, the TF Runx3 has been shown to be

essential for the development of pSNs specifically by establishing or maintaining expression of pSN-

specific factors, including TrkC, the receptor for neurotrophin-3 (NT-3); the calcium-binding protein

parvalbumin (Pv); and the ETS TF Er81 (Etv1) (Liu & Ma, 2011).

Using transgenic lines expressing different combinations of these TFs under control of a

tetracycline-responsive element (TRE), I induced TF expression by addition of doxycycline (dox) and

examined co-expression of the V5-epitope tag on Pou4f1 and Isl1/2 two days after induction (Fig 5.5A).

To control the spinal identity of the programmed cells, high RA was added with dox to ensure proper

caudalization and expression of spinal Hox genes (Mazzoni et al, 2013a). V5 induction was robust and

homogenous for all three combinations (Fig 5.5B). Unsurprisingly, ESCs and EBs expressing Isl1/2

transgene (Pou4f1-Isl1/2-Ngn2, or PIN) induced both V5 and high levels of Isl1/2. By comparison, while

Pou4f1-Ngn2-FoxS1 (PNF) overexpression in ESCs was able to induce Isl1/2 expression, overexpression

of Pou4f1-Ngn2-Runx3 (PNR) could not (Fig 5.5A,B). Nevertheless, PNF was not sufficient to induce

Runx3 expression, which is required for pSN molecular differentiation, and adopted large and flat

morphologies when dissociated and cultured for several days, suggesting that this combination of TFs

does not form neurons (Fig 5.5C, data not shown). Conversely, PIN-differentiated EBs produce cells co-

expressing Isl1/2 and Runx3 (13.7%), which are largely overlapping with cells co-expressing Pou4f1 and

Runx3 (15.1%), and these acquired NeuN expression and neuronal morphologies (Fig 5.5C). Thus, the

160

combinatorial expression of Ngn2, Isl1/2, and Pou4f1 TFs in ESCs is able to directly program ESCs to

Runx3-expressing pSN precursors.

During their development, SNs including pSNs require neurotrophic support for their appropriate

specification. Neurotrophin-3, which is produced from their target muscle spindles in the periphery,

promotes survival of pSNs, as well as the proper targeting of their central afferent projections (Ernfors et

al, 1994). To test if ESC-derived pSNs also require NT-3 signaling specifically for their specification, I

induced expression of PIN then cultured the programmed neurons in media containing NT-3 and/or nerve

growth factor (NGF), which promotes the survival of TrkA-expressing nociceptive SNs (Liu & Ma, 2011).

NT-3, but not NGF, significantly promoted the development of Runx3-expressing cells on Day 6 when

added on Days 2-4 of differentiation (Fig 5.5D, data not shown). For example, NT-3 addition on Day 2

increased Runx3/V5-expressing cells from 8.0 to 23.1% (p<0.001) (Fig 5.5D).

To confirm the proprioceptor identity of programmed Runx3-expressing cells, I performed ICC on

neurons dissociated from EBs on Day 6 and cultured for 4 days on extracellular matrix (laminin and

fibronectin) in neuronal media containing NT-3. Cultured neurons acquired cellular morphology

reminiscent of the pseudo-unipolar morphology of SNs from DRG in vivo, including a large, bulbous cell

body and single short processes extending from the soma, which then bifurcate into two axons. In vivo,

one of those axons would terminate in the spinal cord, the other projecting to peripheral muscle targets

(Fig 5.6A) (Marmigère & Ernfors, 2007). Furthermore, they expressed molecular markers of pSNs in vivo,

including Runx3, Neurofilament-Heavy (NF-H), and TrkC (upper panel); as well as Pv (lower panel) (Fig

5.6B) (Fornaro et al, 2008; Marmigère & Ernfors, 2007). Indeed, quantification of Pv expression indicates

that Runx3-expressing neurons are highly enriched for Pv after 4 days in NT-3-enriched culture, with

62.5% of Runx3 cells co-expressing Pv on Day 10, compared to 17.5% in EBs on Day 6 (Fig 5.6C).

Altogether, these findings suggest that direct programming of pSN cell fate is possible through

efficient induction of generic sensory-lineage TFs (Ngn2, Isl1/2 and Pou4f1), followed by culture in NT-3

enriched media to promote specific development of TrkC-expressing proprioceptors. Direct programming

of ESCs to sensory neurons can establish distinct DRG pseudo-uniplar morphologies, and induce pSN-

specific molecular developmental programs.

161

Discussion

Ptf1a-derived GABAergic interneuron subtypes in the dorsal spinal cord In the early developing dorsal spinal cord, six distinct IN domains are established by the specific

expression of bHLH TFs (Atoh1, Ngn1/2, Ascl1, and Ptf1a) in their progenitor cells. As these progenitor

cells differentiate, their distinct cellular identities are consolidated by their expression of different

combinations of HD TFs and their timing of neurogenesis (Lai et al, 2016). In particular, the Ptf1a-

expressing spinal domain generates dI4 INs, which constitute a Pax2- and Lhx1/5-expressing GABAergic

inhibitory IN population that can be subdivided into at least two subpopulations: early-born dI4 and later-

born dILA INs (Glasgow et a, 2005; Betley et al, 2009).

Intriguingly, molecular markers purported to distinguish early and late-born dI4 INs mark only

subsets of cells within each population (Wildner et al, 2013; Betley et al, 2009). For example, GABApre

interneurons, which provide presynaptic inhibition of primary pSN terminals in the ventral horn, are

distinguished by their joint expression of the GABAergic enzymes Gad67, which has subcellular

localization in the cytosol, and Gad65, which is primarily bound to synaptic vesicles (Monyer & Markram,

2004; Soghomonian & Martin, 1998; Hughes et al, 2005; Betley et al, 2009). Yet, only 54.5% Ptf1a-

derived cells in the deep dorsal horn co-express Gad67 and Gad65, suggesting that only a subset of

early-born dI4 neurons function as GABApre INs (Betley et al, 2009). Meanwhile, other Ptf1a-derived

cells express Gad67 and/or glycine transporter, GlyT2, including dILA INs providing presynaptic inhibition

to cutaneous sensory afferents in the superficial dorsal horn (Zeilhofer et al, 2006; Betley et al, 2009).

In addition to differential expression of inhibitory neurotransmitters, other factors are only

expressed by subsets of dI4 or dILA INs. For example, only some early-born dI4 INs in the intermediate

zone express the TFAP2b (<10% of Ptf1a-lineage cells). In addition, recent studies show that only a

subset of Ptf1a-lineage cells in the deep dorsal horn expresses the TF Satb2; these inhibitory neurons

are distinct from TFAP2b-expressing cells from the Ptf1a-lineage, although both are suggested to

mediate the transformation of sensory information into motor output (Levine et al, 2014; Hilde et al, 2016).

Conversely, among genes highly enriched in the superficial dorsal horn, the TF Ebf1 is expressed in only

a fraction of Ptf1a-lineage reporter cells in laminae I and II, where dILA neurons settle, as are the

neuropeptides prodynorphin and neuropeptide Y (data not shown; Wildner et al, 2013; Betley et al, 2009).

162

Thus, while a systematic analysis of dI4 IN subtype diversity has not yet been completed, dI4 INs may in

fact constitute a similarly molecularly and functionally diverse class of inhibitory INs in the dorsal spinal

cord, as has recently been shown for V1 INs in the ventral horn, allowing for development of complex and

sophisticated spinal circuits for sensorimotor control.

Subtype diversity of Ptf1a-derived cells in the central nervous system

Evidence for more extensive diversification of Ptf1a-derived neurons comes from lineage-tracing

studies in the mammalian hindbrain and cerebellum. In the hindbrain, the Ptf1a domain produces diverse

types of neurons based on spatial position (Fuyiyama et al, 2009; Hori & Hoshino, 2012; Kohl et al, 2015).

Dorsal hindbrain dB1 neurons express the TFs Ptf1a, Lbx1, and the LIM-HD TFs Lhx1 and Lhx5, similar

to spinal dI4 INs (Glasgow et al, 2005; Hori et al, 2008; Meredith et al, 2009; Storm et al, 2009). Rostral

dB1 INs contribute to the DCN and project inhibitory axons to the inferior colliculi, while more caudally-

positioned dB1 neurons generate the ION, which provides excitatory inputs to Purkinje cerebellar cells.

Some dB1 INs also help to generate the vestibular nuclei (Yamada et al, 2007; Renier et al, 2010). Cell

fates within each of these nuclei are further diversified: for example, in the DCN, Ptf1a+ progenitors

produce GABAergic Golgi and stellate cells, as well as glycinergic cartwheel cells (Fuyiyama et al, 2009).

Lineage-tracing of dB1 INs confirms that they project axons in discrete tracts to form synapses on

multiple target sites, including Purkinje cells, midbrain vestibular nuclei, auditory nuclei, and the medulla.

Meanwhile, in the mouse cerebellum, all GABAergic neurons are produced from a Ptf1a-

expressing domain in the cerebellar ventricular zone (VZ), including Purkinje, Golgi, Lugaro, basket, and

stellate cells of the cerebellar cortex; as well as deep cerebellar nuclei projection INs (Hoshino et al,

2005; Sudarov et al, 2011). Several TFs have been reported to be differentially expressed between

early-born Purkinje cells and later-born neurons in the VZ. For example, Purkinje cells are GABAergic

projection neurons and express Corl2/Skor2 TF while all other Ptf1a-derived cell types in the cerebellum

are Pax2-expressing INs (Minaki et al, 2008; Maricich et al, 1999). In addition, the LIM-HD TFs Lhx1/5

and their cofactor Ldb1 have been shown to be involved in the specification of Purkinje cells specifically

(Zhao et al, 2007). Interestingly, in Ascl1-null cerebellum, Pax2-positive neurons, but not Purkinje cells,

are reduced, while Purkinje cell production is specifically affected in Ngn1-/- mice, suggesting that

163

different bHLH factors function alongside Ptf1a during specification of distinct cerebellar neuron types,

similar to what has been shown in the dorsal spinal cord (Lundell et al, 2009; Sudarov et al, 2011).

Birthdating studies in the cerebellum have also revealed that distinct cerebellar neuronal cell

types are produced in stereotyped birth order in an inside-out manner, with deep cerebellar nuclei

neurons produced first, followed by Purkinje cells, then Golgi and Lugaro cells, and finally basket and

stellar cells (Sultan, 2002; Leto et al, 2006; Sudarov et al, 2011). Thus, temporal patterning may be

involved in the specification of different cell types from a common Ptf1a progenitor domain in the

cerebellar VZ. Heterotopic and heterochronic transplantation studies have revealed that while early Ptf1a-

expressing progenitors differentiate into all types of GABAergic neurons, including Purkinje cells,

progenitors taken from the postnatal cerebellum generate only Pax2-expressing INs, suggesting that

cerebellar GABAergic progenitors have distinct temporal identities (Jankovski et al, 1996; Carletti et al,

2002). Recent studies suggest that the temporal identity transition of Ptf1a progenitors is controlled by the

bHLH TFs Olig2 and Gsx1, which are expressed in dorsally-located Purkinje cell progenitors and ventral

Pax2-expressing IN progenitors, respectively (Seto et al, 2014). Alternatively, more recent transplantation

studies suggest that VZ Pax2+ precursors in particular may not be restricted in their potential to produce

different cell types over time but rather acquire mature identities in response to extrinsic cues from their

environment (Leto et al, 2006; Leto et al, 2009).

Altogether, studies in the dorsal spinal cord, hindbrain and cerebellum suggest that Ptf1a-

expressing progenitor domains likely generate multiple types on GABAergic neurons based on their

spatial and temporal patterning during development. While our studies have focused on the differentiation

of the GABApre IN providing presynaptic inhibition of pSNs in the monosynaptic stretch reflex,

understanding molecular mechanisms of Ptf1a domain diversification overall will likely yield insights into

GABApre molecular and functional specialization, including their synaptic specificity with pSNs.

Molecular expression profiling of dI4 and dILA IN subtypes

While early-born dI4 and late-born dILA INs both require Ptf1a for their development, dILA, but

not dI4, INs are also dependent Ascl1 for their specification (Mazurier et al, 2014; Mizuguchi et al 2006;

Glasgow et al, 2005). This insight has been used to identify candidate genetic markers for distinguishing

164

dI4 and dILA INs. Four of these genes, pDyn, Kcnip2, RORβ, and Tfap2b, are expressed in non-

overlapping, layer-like domains in the dorsal horn, with pDyn expressed most superficially, followed by

Kcnip2, RORβ, and Tfap2b ventrally, suggesting that these might represent distinct dI4-derived subtypes

beyond simple division of dI4 and dILA classes (Wildner et al, 2013). In particular, cells expressing the TF

TFAP2b were completely absent in Ptf1a-/- mice but not affected in Ascl1-/- mice, suggesting that this

marker selectively labels at least a subset of early-born dI4 INs. Furthermore, TFAP2b-expressing

neurons reside almost exclusively in the deep dorsal horn in laminae V, making this gene an intriguing

candidate for genetically accessing GABApre INs specifically (Wildner et al, 2013; Levine et al, 2014; J.

Kaltschmidt, unpublished).

TFAP2b belongs to the Activating Enhancer Binding Protein 2 family with 4 other members,

including TFAP2a. Both TFAP2a and 2b have been shown to be involved in eye, ear, neural tube, kidney,

and limb development, with recent studies showing that these TFs act downstream of Ptf1a in the murine

retina to control specification of inhibitory amacrine cells (Eckert et al, 2005; Hilger-Eversheim et al, 2000;

Bassett et al, 2012; Jin et al, 2015). Whether TFAP2b also has a similar role in the specification of Ptf1a-

derived spinal inhibitory IN types remains to be determined. Interestingly, TFAP2b-expressing cells in the

deep spinal cord have been suggested to act as motor synergy encoders, neurons in the spinal cord

serving as a central node for coordination of corticospinal and sensory pathways. In this study, the

majority of TFAP2b-expressing neurons are shown to express Gad65 and/or Gad67, indicating that they

may constitute GABApre INs, while a minority express glutamatergic neurotransmitters, including

VGLUT2. However, in their proposed role as motor synergy encoders, TFAP2b neurons necessarily

synapse with MNs, which is experimentally demonstrated by rabies virus-mediated transsynaptic tracing

to identify cells making monosynaptic connections onto MNs (Levine et al, 2014). This is in stark contrast

with the proposed wiring diagram of GABApre INs, which selectively synapse onto pSN afferent terminals

in the spinal cord. Importantly, TFAP2b-expressing cells comprise <10% of all Ptf1a-lineage cells in the

lumbar dorsal spinal cord at p0, indicating that either only a very small population of deep dorsal horn

Ptf1a-expressing cells provides all presynaptic inhibition to pSN afferents, or TFAP2b-expressing cells

constitute only a subset of cells with this function. Therefore, while TFAP2b-expressing INs are the most

165

promising candidates for segregating GABApre INs from other Ptf1a-derived cells, whether they actually

fulfill this role is unknown.

While not as highly enriched as other candidate genes for dI4 IN subtype diversification, the TF

Ebf1 is highly expressed in a distinct subset of dI4 INs occupying the superficial-most dorsal horn (Hu et

al, 2012; John et al, 2012). Previously, Ebf1 has been shown to be important for cell cycle exit and

neuronal differentiation in the developing spinal cord (Garcia-Dominguez et al, 2003; Garel et al, 1999;

Green & Vetter, 2011). Furthermore, Ebf1 is suggested to act downstream of Lmx1b to control the

differentiation and migration of excitatory laminae I-II neurons and the ingrowth of cutaneous sensory

afferents into the dorsal horn; whether it acts in a similar manner in the inhibitory Ptf1a-lineage remains to

be determined (Ding et al, 2004). Finally, in the hindbrain, Ebf1 is involved in the proper migration of facial

branchiomotor neurons; while Ebf1-/- mutants exhibit abnormal thalamocortical projections during basal

ganglia development (Garel et al, 2000; Garel et al, 2002). Indeed, whether Ebf1 is involved in the

specification of dILA subtype-specific identity and/or dorsal migration to laminae I-II from the Ptf1a

progenitor domain is not currently known.

Extrinsic patterning signals involved in dI4 IN subtype diversification

Interestingly, the differentiation of Ebf1-expressing cells in the Ptf1a lineage depends on

exposure to TGFß signals, in particular ActivinA. Here, we show that ESCs differentiated with RA only

produce only very small populations of cells co-expressing Ptf1a and Ebf1, while addition of ActivinA

results in 2-fold increase of this population. Furthermore, transplantation of early EBs treated only with RA

results in dramatic migration of Ptf1a-tdTomato FP cells from the graft site into the superficial dorsal horn,

suggesting that early Ptf1a progenitors are competent to respond to local instructive signals from the

developing spinal cord. This phenotype is reminiscent of heterotopic/heterochronic transplantation studies

of embryonic and postnatal cerebellar progenitors showing that cerebellar Ptf1a-derived cells adopt their

mature neuronal subtype identities in response to environmental cues (Leto et al, 2006; Leto et al, 2009;

Leto & Rossi 2012). Thus, GABApre INs and other Ptf1a-derived cell types in the spinal cord might

acquire subtype-specific molecular identities based on their location in the developing spinal cord. Indeed,

expression of Dynorphin, a neuropeptide used by many GABAergic dorsal horn neurons, including some

166

dILA INs, is normally very low in in RA only-treated EBs (Kardon et al, 2014; Sardella et al, 2011; Wildner

et al, 2013; Polgar et al, 2013). Transplant of these EBs into the developing chick neural tube, however, is

sufficient to induce expression of Dynorphin in a subset of these cells, suggesting that extrinsic regulation

of dI4 IN subtype identity is likely not specific to Ebf1-expressing INs (data not shown).

Synaptic specificity of dI4 INs in vivo and in vitro

In the spinal cord, axo-axonic synapses between GABApre INs and pSN afferent terminals act to

selectively filter excitatory sensory inputs onto MNs, providing presynaptic inhibition to control sensory-

motor drive (Rudomin et al, 1999; Hughes et al, 2005; Betley et al, 2009). Recent studies suggest that the

synaptic targeting of GABApre INs is restricted to sensory terminals in the ventral spinal cord – in

circumstances when pSNs are genetically removed from the circuit, GABApre INs withdraw from the

ventral horn rather than make ectopic synapses onto MNs (Betley et al, 2009). Using RABV retrograde

tracing adapted for in vitro co-culture, we show here that ES-dI4 INs, which comprise a mixed population

of Ptf1a-derived cells, readily form monosynaptic synapses with spinal ES-MNs. Indeed, by this assay,

dI4 INs are more likely to be synaptically connected to MNs compared to spinal V1 INs, are significant

subset of which are known to provide postsynaptic inhibition of MNs.

Several hypotheses may account for the lack of synaptic specificity of dI4 INs in vitro. First,

although a significant subset of Ptf1a-tdTomato FP cells express TFAP2b, a TF enriched in the

population of deep dorsal horn dI4 INs from which GABApre INs likely arise, whether these cells adopt

GABApre molecular and functional identity in vitro is not known. A recent study suggests that inhibitory

TFAP2b-expressing cells in lamina V of the spinal cord may serve as a group of motor synergy encoders

receiving direct inputs from motor cortex and sensory pathways to provide monosynaptic inputs onto

spinal MNs to influence motor output (Levine et al, 2014). Furthermore, ESCs differentiated to dI4 INs

also generate some Ebf1-expressing FP cells that are likely a subset of late-born dILA neurons providing

synaptic inhibition to cutaneous sensory afferents in the dorsal cord (Glasgow et al, 2005). Since these

never encounter MNs in vivo, whether dILA INs also exhibit stringent synaptic specificity for SN afferent

terminals and avoid MNs is not known.

167

Interestingly, it was also reported that during the first postnatal month in the mouse, GABApre INs

form “transient” Gad67+ synapses with MNs which never mature into functional synapses due to failure of

presynaptic inhibition (which requires sensory-derived BDNF signals, etc) and lack of post-synaptic GABA

receptors. (Betley et al, 2009). Whether all Gad67-immunoreactive boutons between dI4 INs constitute

functional synapses, and whether RABV transfer requires fully mature synapses for efficient transfer are

open questions to be determined. Indeed, ES-dI4 INs may require further development to acquire the

ability to accurately discern SN versus MN synaptic targets. Expression profiling of ES-dI4 INs indicates

that they likely share molecular identity with embryonic to early postnatal spinal dI4 INs; whether relatively

immature cells have equal potential to form selective synaptic connections is not known (data not shown).

Before molecular description of stringent synaptic specificity of GABApre and pSNs, electron

microscopy studies suggested that presynaptic inhibitory INs formed triadic synaptic junctions between

SNs and MNs in the spinal cord. Indeed, presynaptic terminals, or P boutons, were shown to contact both

the primary sensory afferent terminal and the MN cell membrane in the ventral horn, although it is argued

that the physical contact with the MN does not necessarily constitute a synapse (Gray, 1962; Conradi,

1969; Conradi & Skoglund, 1969). Interestingly, P boutons were also found on group Ia pSN afferent

terminals contacting unidentified ventral INs (Maxwell et al, 1990; Walmsley et al, 1995; Walmsley et al,

1987). Overall, these ultrastructural studies suggest that physical contacts between presynaptic inhibitory

INs (i.e. GABApre INs) may not be surprising; whether they constitute functional synapses is less clear.

Importantly, ES-dI4 INs are generated in an environment entirely devoid of both SN and MNs, as

well as most other spinal cell types, signals and interactions. Although it has been shown that SN-derived

signals are required for functional maturation of GABApre INs, whether other cellular interactions also

have a role in GABApre differentiation has not been determined. Alternatively, GABApre axonal

projections and synaptic connectivity may be entirely cell-intrinsically determined through a hardwired

genetic program that coordinates the expression of molecular matching cues with their preferred synaptic

partners, as recently suggested by Ashrafi et al (2014). Nonetheless, molecular programs differentiating

GABApre INs from other spinal neuron cell types are not yet known, making assessment of this possibility

challenging with our current tools.

168

Finally, a promising approach for testing the synaptic specificity of in vitro-generated dI4 INs is to

force the cells to choose between SN and MN synaptic partners. Our preliminary results suggest that

Ptf1a-tdTomato neurons preferentially project axons towards DRG explants compared to EBs containing

ES-MNs. However, whether this is due to repulsive signals emanating other ventral spinal cells or

attractive cues from DRGs is not easily determined using mixed populations of neurons in explants and

EBs. Therefore, we are developing co-cultures with ES-pSNs generated from direct transcriptional

programming using SN lineage-relevant TFs and pSN-specific neurotrophic factors. Co-culture studies

with dI4 INs, pSNs and MNs will provide increased insight into whether presynaptic inhibitory circuits can

be recapitulated in vitro. This assay will also be invaluable for testing the synaptic connectivity of different

dI4 IN subtypes differentiated from ESCs.

169

Figure 5.1

Figure 5.1 Ascl1-dependent and independent dI4 interneuron subpopulations (A) Ptf1a-expressing dP4 progenitors in e10.5 mouse dorsal spinal cord express low or no Ascl1 compared to surrounding dP3 or dP5 spinal domains. Scale bars = 100 µm. (B) Day 6 EBs differentiated with RA only showing that Ptf1a-expressing cells express variable levels of Ascl1, although generally lower than non-Ptf1a-expressing cells in the EB. (C) Quantification of Ascl1 (no, low, high) protein expression in Ptf1a-expressing cells in Day 6 EBs by immunostaining. Scale bars = 50 µm.

170

Figure 5.2

Figure 5.2 Molecular and spatially distinct subsets of dI4 interneurons in vivo and in vitro (A) In situ hybridization of TFAP2b and Ebf1 transcripts in p4 mouse spinal cord shows their distinct localization in the dorsal horn (Allen Brain Atlas). (B) Spinal cord sections from e12.5 and e18.5 Ptf1a::cre x ROSA-LSL-tdTomato lineage reporter mice immunostained for Ebf1 and TFAP2b proteins. E18.5: right hemisegment of dorsal horn only. Scale bars = 100 µm. (C) Immunostaining of Day 8 Ptf1a-tdTomato EBs shows that TFAP2b and Ebf1 are non-overlapping populations, and that ActivinA treatment significantly enhances generation of Ebf1-expressing FP cells. Quantification of TFAP2b and Ebf1-expressing Ptf1a-tdTomato FP cells on the right (n=3, ANOVA, *p<0.05, **p<0.01). Scale bars = 50 µm.

171

Figure 5.3

Figure 5.3 Transplanted dI4 interneuron subsets migrate into distinct dorsal horn laminae Day 6 Ptf1a-tdTomato EBs differentiated with RA only were engrafted in HH Stage 16 chick neural tube and examined 4 days later for cell migration and axonal projections. During RA only differentiation in vitro, few Ebf1-expressing dI4 interneurons are generated. However, transplanted Ptf1a-tdTomato FP cells highly upregulate Ebf1 expression, and these cells migrate into superficial dorsal horn laminae, while TFAP2b-expressing FP cells largely remain in the deep dorsal horn (n=3). Scale bars = 100 µm.

172

Figure 5.4

Figure 5.4 Monosynaptic RABV tracing reveals premotor connections of dI4 interneurons Ptf1a-tdTomato FP cells were purified and co-cultured with Hb9::GFP MNs expressing TVA-2A-Rabies G for monosynaptic RABV tracing for one week prior to addition of SAD∆G-GFP RABV. On day 4 or 7 post-infection, co-cultures were fixed and Ptf1a-tdTomato FP cells were examined for their synaptic connectivity with MNs. Compared to parallel cultures with ESC-derived V1 INs, dI4 INs formed significantly more monosynaptic connections with MNs, as determined by the percentage of V1 or dI4 INs expressing SAD∆G-GFP (n=3, ANOVA, *p<0.05, **p<0.01). Scale bars = 15 µm.

173

Figure 5.5

Figure 5.5 Transcriptional programming of proprioceptive sensory neurons from ESCs (A) Transgene constructs for generating inducible pSN cell lines. (B) Initial testing of dox-induced clones Pou4f1-Ngn2-FoxS1 (PNF), Pou4f1-Ngn2-Runx3 (PNR), and Pou4f1-Isl1/2-Ngn2 (PIN) ESC cell lines (8-10 clones tested for each line). Immunostaining for epitope tag V5 and Isl1/2 shows that PIN and PNF lines robustly induce Isl1/2 expression. (C) Further testing of PNF and PIN lines shows that PIN induces Runx3-expressing Isl1/2 and Pou4f1 cells, suggestive of proprioceptor identity. (D) Treatment of dox-induced EBs with NT-3 (10 ng/mL) increases generation of Runx3/V5+ cells on Day 6 of differentiation. (n=3, ANOVA, ***p<0.001). Scale bars = 50 µm.

174

Figure 5.6

Figure 5.6 Morphology and molecular maturation of induced proprioceptive sensory neurons (A) PIN-induced pSNs cultured until Day 10 on laminin/fibronectin substrate acquire pseudo-unipolar morphology of DRG sensory neurons in vivo, with single short processes extending from the soma, which then bifurcate into two axons. (B) Induced pSNs express other proprioceptor-specific molecular markers, including TrkC, the receptor for NT-3 and parvalbumin (Pv). Sensory neurons, including pSNs, also express Neurofilament heavy-unit (NF-H) in vivo. (C) Quantification of PIN-V5 cells co-expressing Pv on Days 6 and 10 of culture, indicating that maturing pSNs acquire additional Pv expression.

175

Chapter 6: General discussion and future directions Summary

The precise wiring of neural circuits requires that developing neurons acquire distinct subtype

identities that determine their ability to recognize their appropriate synaptic partners. How apparently

uniform neural progenitors are transformed into distinct cell types with specialized identity, connectivity

and function within neural circuits is a mostly unresolved issue in the field of developmental neurobiology.

For my dissertation studies, I focused on the development of a relatively simple sensorimotor circuit in the

mammalian spinal cord, investigating the developmental specification, subtype diversification and

synaptic connectivity of V1 and dI4 inhibitory INs providing essential regulation of the spinal

monosynaptic stretch reflex circuit. In the spinal cord, V1-derived Renshaw cells are known to provide

postsynaptic inhibition to MNs, while dI4-derived GABApre INs provide presynaptic inhibition to pSNs

innervating MNs. I began my studies examining the specification of these subtypes and their synaptic

specificity with two major questions in mind: First, given the molecular heterogeneity of both V1 and dI4

spinal neuron classes, is there a role for cell-intrinsic programs in directing distinct IN subtypes towards

their preferred synaptic partners, or is matching between pre and postsynaptic cells principally

determined by cell non-autonomous signals and interactions? Second, are all or only certain subsets of

V1 and dI4 INs capable of integrating into this spinal circuit? To begin to address these questions, and to

circumvent the cellular complexity of the spinal cord, I developed novel stem cell-based tools that allowed

for simple modeling of V1 and dI4 IN specification and synaptic connectivity, including the ability to

systematically manipulate the signals, timing, and cellular interactions potentially involved in these

processes. I divided my results into four major parts:

In Chapter 2, I showed that mouse ESCs could be efficiently differentiated into cells with

molecular and functional identity of spinal V1 and dI4 INs using developmentally-relevant extrinsic

patterning signals. In particular, V1 INs were most efficiently generated using low level Shh signaling, as

predicted from models of ventral spinal patterning (Briscoe & Ericson, 2001). Conversely, while the TGFß

family member BMP has been best studied for its role in dorsal spinal patterning, my results show that

another TGFß ligand, ActivinA, most efficiently generates cells with dI4 IN identity (Helms & Johnson,

2003). ESC-derived V1 and dI4 INs acquire gene expression and functional properties of their in vivo

176

counterparts, including appropriate migration and axonal projections after transplant into the developing

spinal cord. Thus, distinct spinal IN cell types can be efficiently produced from stem cells, providing the

opportunity to study the development and function of these cells in a more mechanistic manner.

In Chapter 3, I took advantage of our ability to produce p1 progenitors from ESCs to study the

developmental specification of distinct V1 IN subtypes, particular Renshaw cells mediating recurrent

inhibition of MNs. Recent studies have shown that V1 INs comprise a highly heterogeneous class of

neurons, yet how distinct V1 IN subtypes such as Renshaw cells are uniquely specified is very poorly

understood (Bikoff et al, 2016). ESC-derived V1 INs express many of the molecular factors used to

subdivide the V1 IN class in vivo and can also be segregated into non-overlapping subpopulations. Using

ESC-derived p1 progenitors, I showed that the evolutionarily conserved Notch signaling pathway is likely

involved in the differentiation of RCs from other V1-derived neurons, either by controlling the timing of V1

subtype neurogenesis or by providing an instructive cue to specify RC fate over other potential V1 IN

identities (Cepko, 2014). In addition, I found that retinoic acid signaling may be specifically involved in RC

development. In the absence of RA signals, Cb-expressing RCs are not formed, while other V1 IN cell

types are generated in normal numbers. While MN-derived RA has previously been shown to be

important for specification of select MN subtypes, a role for RA signaling in IN cell fate specification has

not yet been described, raising the possibility that MNs may be directly involved in the construction of

motor circuits. These studies suggest that we might be able to improve on the yield of RCs from ESCs by

taking advantage of their unique regulation by Notch and RA signals compared to other V1 IN subtypes.

In Chapter 4, I showed that ESC-derived V1 IN subtypes have differential synaptic connectivity

with MNs. Monosynaptic RABV tracing studies revealed that while MNs receive direct synaptic inputs

from both Cb-expressing RCs and non-RC FoxP2-expressing V1 INs, RCs are significantly more likely

than other V1 INs to form synapses onto MNs. Accordingly, studies using VAChT-immunoreactivity and

whole-cell patch-clamp electrophysiological recordings of ESC-derived RCs in conjunction with

optogenetic stimulation of MNs indicated that that RCs were also significantly more likely to receive

cholinergic inputs from MNs than other V1 IN subtypes. Thus, our data show that RCs exhibit synaptic

specificity for MNs, and vice versa, even in an in vitro setting largely devoid of spinal contextual cues.

177

Finally, in Chapter 5, I showed that ESC-derived dI4 INs can also be subdivided into molecularly

distinct subtypes, including those expressing the TFs TFAP2b and Ebf1, which migrate into the superficial

and deep dorsal horn, respectively, when transplanted into the developing chick neural tube. TFAP2b-

expressing cells likely belong to the early-born dI4 IN class in the deep dorsal horn, which has been

shown to give rise to GABApre INs mediating presynaptic inhibition of pSN afferent terminals in the spinal

cord (Glasgow et al, 2005; Betley et al, 2009; Wildner et al, 2013). Interestingly, ActivinA treatment

significantly induces the formation of Ebf1-expressing cells, which are a subset of late-born dILA INs in

the superficial dorsal horn, suggesting that extrinsic signaling factors may be involved in patterning of the

dP4 progenitor domain to give rise to different subtypes. Surprisingly, monosynaptic RABV tracing

reveals that dI4 INs form synapses on MNs at higher frequency than V1 INs. Whether dI4 IN-MN synaptic

contacts are a result of incomplete specification of ESC-derived dI4 INs or a manifestation of purported

GABApre, MN, and SN triadic synapses is not known (Gray, 1962; Conradi, 1969; Conradi & Skoglund,

1969). To further examine the synaptic specificity of dI4 INs, I also optimized generation of pSNs from

ESCs through direct transcriptional programming using SN-lineage TFs, as well as exposure to pSN-

specific neurotrophic factors. Establishment of a co-culture assay of GABApre INs with MNs and SNs

might reveal novel insights into the target specificity of these neurons.

Altogether, my results indicate that many aspects of V1 and dI4 IN development are recapitulated

in vitro, including their subtype-specific molecular and functional diversity. ESC-derived neurons can be

used to probe molecular mechanisms underlying subtype diversification of V1 and dI4 INs, including

establishing the roles of Notch and RA signaling in specifying RC-specific subtype identity, as well as the

role for ActivinA in inducing Ebf1-expressing dILA INs. Finally, we can use ESC-derived V1 and dI4 INs to

model the synaptic connectivity of these neuronal subtypes with MNs and SNs, including prospectively

identifying signals required to control their specific synaptogenesis during motor circuit formation.

In vitro modeling of spinal interneuron subtype specification The mammalian spinal cord is a complex neuronal network consisting of diverse cell types

generated in precise spatial and temporal patterns throughout development. However, the molecular and

genetic mechanisms directing relatively uniform populations of spinal progenitors into diverse neuronal

178

subtypes remains a significant challenge. The advent of pluripotent stem cell technology allows for the

generation of diverse neuronal populations in vitro, facilitating the analysis of neuronal development on a

cellular and molecular level previously unattainable. In my thesis, I showed that ESC-derived V1 INs can

be used to pinpoint novel roles for Notch and RA signaling in RC specification from p1 progenitors.

Furthermore, I showed that different dI4 IN subtypes can be generated in vitro through application of

extrinsic ActivinA signals. In the next section, I will discuss ongoing and future studies to further probe the

development of V1 and dI4 IN subtypes:

Cell fate determination of p1 progenitors

How is cellular diversity generated in the p1 progenitor domain? Several models for cell fate

determination of V1 IN cell types can be considered. According to one intrinsic model of cell fate

determination, p1 progenitors are multipotent and pass through an invariant series of competence states,

during each of which progenitors are competent to produce a subset of V1 subtypes. However, each p1

progenitor may not make every type of cell that it is competent to produce, potentially as a result of

intervening signaling pathways, stochastic mechanisms and/or local environmental signals. For example,

in the retina, high Notch activity maintains cells in the undifferentiated progenitor state without changing

their temporal identity, while acting later on in postmitotic retinal cells to promote one cell fate over

another (e.g. photoreceptor versus another cell type). As such, p1 progenitors differentiating at different

times produce distinct types of V1 INs, but Notch signaling can influence when the progenitor is allowed

to differentiate, effectively changing the cell fate choice without affecting the overall normal progression of

p1 progenitor temporal states. Furthermore, Notch signaling may act later in postmitotic V1 INs to control

selection of distinct subtype identities (Jadhav et al, 2006a; Jadhav et al, 2006b; Cepko, 2014).

Alternatively, p1 progenitors might dynamically interpret local extrinsic signals such as RA to choose one

distinct V1 subtype fate (e.g. Renshaw cell) over another.

On the other hand, distinct p1 progenitors may be established at early stages of V1 IN

development, each of which then goes through its own intrinsic program to produce only specific types of

progeny. In this case, Notch signaling might be involved in the establishment of distinct p1 progenitors

and/or regulation of alternative cell fates of the postmitotic daughter cells produced from different p1

179

progenitors. Conversely, distinct p1 progenitors might be established at different dorsoventral and

rostrocaudal positions in the spinal cord due to the differential intensities of local morphogenetic signals

such as Shh and RA (Briscoe & Ericson, 2001; Sockanathan et al, 2003). Finally, the most extreme

extrinsic model of cell fate determination suggests that p1 progenitors are equivalent at all times and

competent to produce all V1 IN subtypes, with extrinsic cues inducing the different cell fates in these

progeny. As such, Notch and/or RA signaling might be involved in this final step of cell fate selection (Oh

et al, 2007; Swaroop et al, 2010; Cepko, 2014). Ultimately, it is likely that a combination of intrinsic and

extrinsic mechanisms interact to influence the wide array of V1 IN subtypes observed in the ventral horn

of the mature spinal cord. Future studies will rely on heterochronic and heterotopic transplantation of p1

progenitors in developing spinal cord, as well as clonal analyses in vivo and in vitro to disentangle the

likely complex regulatory mechanisms for generating V1 IN cellular diversity.

Mechanism of Notch signaling on Renshaw cell specification

Recent studies have demonstrated that distinct types of V1 INs are generated at different

developmental stages, with RCs born first and FoxP2-expressing V1 INs born later (Stam et al, 2012;

Benito-Gonzalez & Alvarez, 2012). Whether other V1 subtypes are generally produced in a conserved

temporal sequence is unclear. In particular, BrdU incorporation studies can be used to determine the

timing of neurogenesis for distinct V1 IN subtypes. While we know that Notch signaling is involved in RC

development in vitro, the mechanism(s) used by Notch to control RC specification from other V1 INs is

unclear. On the one hand, Notch signaling may act as permissive signal in p1 progenitors to control the

timing of V1 subtype generation, as proposed by the intrinsic model of V1 cell fate determination. In

particular, progenitors differentiating early may be competent to acquire RC identity, while those

differentiating later acquire other subtype identities. Notch signaling would interfere here to regulate the

timing of p1 progenitor cell cycle exit to bias the formation of different V1 subtype identities. Alternatively,

Notch signaling may act to directly instruct subtype identity of V1 INs, including repressing RC specific

identity since DAPT-mediated Notch inhibition increases RC generation (Fig 6.1) (Cepko, 2014).

In ongoing studies, I am using DAPT-mediated Notch inhibition and genetic Notch activation

using inducible NICD ESC lines (see Chapter 3) to test these different possibilities. In particular, using

180

BrdU birthdating of DAPT-treated p1 progenitors, we can test if Notch inhibition changes the timing of V1

neurogenesis to accelerate V1 differentiation, resulting in the production of early-born V1 subtypes at the

expense of later-born subtypes. Conversely, to determine if Notch directly specifies or inhibits RC identity,

Notch inhibition of late p1 progenitors which are past the proposed early temporal competence window for

making RCs will be used to assess if there is still loss of non-RC FoxP2 cell identity and increased

generation of Cb-expressing cells, indicating that Notch provides an instructive cue to specify RCs from

other V1 INs. To confirm this result, inducible NICD will be used to determine if activating Notch signaling

in early p1 progenitors prevents formation of RCs. If Notch signaling is acting in p1 progenitors to control

timing of neurogenesis, late-born V1 subtypes and glia should be overproduced (Kessaris et al, 2001).

Alternatively, if Notch is acting to directly specify V1 subtype identity, RCs should still be produced in

normal numbers.

Thus, using pharmacological and genetic manipulations of Notch signaling in ESC-derived V1

INs, we can gain deeper understanding of how V1 IN subtype identity is established during embryonic

development. Notch signaling has been shown to be involved in the transformation of other spinal neuron

cell types, including motor neurons and several classes of ventral spinal INs (Tan et al, 2016; Del Barrio

et al, 2007; Peng et al, 2007; Rocha et al, 2009; Ramos et al, 2010 Marklund et al, 2010). In the case of

MNs, Notch signaling has been shown to regulate the timing of progenitor differentiation for the

production of neurons and glia from the pMN domain, as well as to control spinal neuronal diversity by

directly selecting different subtype identities. Thus, an additional possibility is that Notch acts through both

permissive and instructive mechanisms to specify RC identity, first by controlling early p1 progenitor cell

cycle exit and subsequently by directly repressing RC fate. In future studies, I will also perform similar

manipulations to assess the role and mechanism of Notch signaling on other V1 subtypes. It is possible

that Notch may act differently to specify distinct non-RC subtypes. Such complex regulatory mechanisms

might underlie the remarkable cellular diversity apparent in the V1 spinal domain.

Role of retinoid signaling in Renshaw cell specification

In addition to Notch signaling, the involvement of RA signaling in spinal patterning generally and

MN subtype diversification specifically prompted us to consider if RA might also be involved in V1 IN

181

subtype diversification, especially the generation of V1-derived RCs, which form a unique and

stereotyped recurrent inhibitory circuit with MNs. During spinal neurogenesis, Raldh2 is highly expressed

by paraxial mesoderm and LMCm MNs in brachial levels of the spinal cord compared to thoracic and

lumbar spinal cord (Niederreither et al, 1997; Berggren et al, 1999; Swindell et al, 1999; Sockanathan et

al, 2003). Thus, newly generated V1 INs such as RCs in the brachial spinal cord are likely exposed to

higher levels of RA signaling than those in the thoracic cord. In Chapter 4, I showed that retinoid signaling

indeed has a differential effect on the development of Cb-expressing cells from V1 INs in vitro. However,

whether RA is required for the specific development of RCs in vivo is unclear. Furthermore, whether

differential retinoid signals are provided by the paraxial mesoderm or whether Raldh2-expressing, limb-

innervating MNs provide an essential source of RA for RC specialization has also not been demonstrated.

Preliminary evidence from mouse e12.5 analyses shows that the percentage of Cb- and OC2-

expressing RCs is indeed higher at brachial segments compared to cervical and thoracic segments,

concordant with prior reports (Fig 6.2) (Francius et al, 2013). However, RC generation in the lumbar cord,

which also contains limb-innervating MNs, is more similar to cervical and thoracic segments than brachial.

Whether this is due to a delayed developmental effect since the lumbar cord develops more slowly than

the rest of the spinal cord, or a result of differential regulation of RC generation in the lumbar spinal cord

is not known. Nevertheless, this result aligns with earlier reports showing that Raldh2 expression is high

in paraxial mesoderm at brachial, but not thoracic or lumbar levels, at the onset of MN generation in the

chick spinal cord (Sockanathan et al, 2003). Interestingly, preliminary analysis of p0 spinal cord reveals

that Cb- and MafA-expressing RCs are generated in greater numbers at brachial and lumbar spinal cord

compared to thoracic, but the overall percentage is the same given the concomitant two-fold expansion of

En1-derived V1 INs in these regions. Potentially, early enrichment of RC generation in limb-innervating

areas of the spinal cord may be compensated later by developmentally-regulated mechanisms such as

apoptosis or activity-dependent processes of circuit remodeling (Buss et al, 2006; Lowrie & Lawson,

2000; Mennerick & Zorumski, 2000; Oppenheim, 1991; Shen & Scheiffele, 2010). In ongoing studies, I

am examining RC distribution across the rostrocaudal axis of the developing spinal cord during

intervening stages between e12.5 to p0 to determine if there is dynamic regulation of RC generation

during spinal cord development, and if so, identity potential mechanisms controlling these differences.

182

Nonetheless, the total expansion of V1 INs at limb-innervating spinal segments at both e12.5 and

p0 is interestingly correlated with the expansion of MN numbers in the same regions. Indeed, it was

previously reported that LMCm-derived RA signaling increases MN number in the brachial spinal cord,

raising the possibility that MN-derived RA may act to expand spinal progenitor domains associated with

limb-innervating motor circuits, independent of its role in regulating RC subtype identity (Sockanathan &

Jessell, 1998). To test the possibility that MN-derived RA is involved in V1 IN generation overall and/or

RC subtype specification, I will examine spinal cords of mice lacking all MNs (Olig2Cre/Cre), as well as mice

with specific loss of FoxP1-expressing MNs (Olig2::cre x FoxP1fl/fl) (Novitch et al, 2001; Dessaud et al,

2007; Feng et al 2010; Sürmeli et al, 2011). Importantly, Olig2::cre x FoxP1fl/fl mutant mice specifically

lack Raldh2-expressing, limb-innervating MNs.

Finally, to test the functional role of RA signaling in RC genesis in vivo, I am taking two

approaches. Using chick spinal cord electroporation assay, I will ectopically express genetic constructs

for manipulating RA signaling in developing chick embryo and assess their effects on V1 spinal neuron

specification. To inhibit retinoid receptor signaling, I will express RAR403, a potent dominant negative

form of the RA receptor RARα that effectively disrupts RA signaling pathways in vivo (Damm et al, 1993;

Sockanathan et al, 2003). To activate RA receptor signaling constitutively in the developing spinal cord, I

will express RARα receptor fused to the transcriptional activator VP16 (Blumberg et al, 1997; Lipkin et al,

1996; Sockanathan et al, 2003). The chick electroporation approach allows for examination of the

consequences of disrupting or misexpressing retinoid receptor signaling on V1 IN differentiation,

specifically RC generation, at different levels of the developing spinal cord. In parallel, we are analyzing

mouse spinal cords in which the dominant negative RAR403 receptor is expressed only in En1-derived

cells (En1::cre x R26RAR403). These mouse mutants enable investigations of the effects of disrupting RA

signaling specifically in V1 INs and not in the rest of the spinal cord (Rajaii et al, 2008). One caveat to

this approach is that En1 is expressed relatively late during V1 specification,

Altogether, these studies will substantiate and provide additional insights into our in vitro results

showing that RA signaling is particularly required for the generation of RCs during early embryogenesis.

Future experiments will test the precise mechanism of RA signaling in RC specification, including its

downstream transcriptional targets and specific role at progenitor and postmitotic stages of RC

183

development (Fig 6.1). Identifying the source of RA signaling in the brachial spinal cord will also be crucial

for understanding how spinal IN subtype diversity is established. If V1 IN specification is primarily

influenced by paraxial mesoderm-derived RA, we propose that V1 IN subtype diversity along the

rostrocaudal axis is established similarly to MNs, which are specified via an RA (and FGF)-driven Hox

code of TFs (Philippidou & Dasen, 2013). Alternatively, if V1 subtype differentiation depends on MN-

derived RA signals, MNs may be directly involved in the construction of spinal motor circuits, recruiting

specific IN modulatory inputs for establishment of functionally diverse circuits along the rostrocaudal axis.

Finally, although we showed that sustained RA signaling is required for genesis of Cb-, but not FoxP2-

expressing V1 INs, whether RA signaling is also involved in the formation of other non-RC V1 subtypes

has not yet been determined. Future studies will determine if RA is required more globally in the p1

progenitor domain to properly specify V1 subtype diversity.

Confluence of dorsalizing signals on dI4 interneuron subtype specification

In addition to using ESC-derived INs to study V1 subtype generation, I also presented results

indicating that ES-dI4 INs can be co-opted for studying extrinsic signaling pathways involved in dI4 IN

subtype diversification. In particular, I showed that ActivinA signals, likely emanating from the dorsal

spinal cord, is involved in promoting Ebf1-expressing dI4 IN subtype, likely a subset of late-born dILA

inhibitory INs in the superficial dorsal laminae. While I showed that dorsalizing signals such as BMPs and

Wnts do not significantly affect the overall generation of Ptf1a-derived cells from ESCs, whether ActivinA

and other spatially-restricted signaling pathways integrate downstream of general dP4 progenitor domain

patterning to specify molecularly and functionally distinct dI4 IN subtypes has not been examined (Helms

& Johnson, 2003; Lai et al, 2016). Using ES-dP4 progenitors, we can thus test the confluence of these

factors in the subtype specification of postmitotic dI4 INs, including identification of a specific molecular

signaling program for specific development of GABApre INs.

A related hypothesis is that the dP4 progenitor domain giving rise to distinct dI4 IN subtypes is

partitioned into spatial subdomains along multiple axes. For example, recent studies show that the flexor-

extensor motor connectivity of dorsal Lbx1-derived neurons, which include dI4 INs, is determined by their

mediolateral segregation (Tripodi et al, 2011). The specific cell type identity of these Lbx1-expressing

184

neurons has not yet been determined; nevertheless, these findings raise the possibility that distinct dI4 IN

subtypes may be specified through their specification from and/or migration to spatially confined domains,

which ultimately directs their highly precise connectivity patterns onto distinct spinal neuron cell types,

including GABApre presynaptic connections to Ia afferents in the ventral spinal cord.

In addition to spatial patterning programs, it has long been established that dI4 IN subtypes are

also segregated based on their timing of neurogenesis, with intermediate dI4 INs born early and

superficial dILA INs born later (Glasgow et al, 2005). Indeed, Lbx1-derived neurons projecting to different

flexor and extensor motor pools in the spinal cord are also segregated based on their timing of birth,

suggesting that temporal programs might intersect with spatial patterning to define the subtype identity

and synaptic connectivity of dorsally-derived spinal INs such as dI4 INs. Indeed, I showed here that Notch

signaling effectively abolishes the generation of late-born glia, suggesting that Notch signaling is involved

dP4 progenitor patterning, potentially by controlling timing of dI4 neurogenesis (Kessaris et al, 2001).

Whether we can co-opt Notch signaling to bias the formation of early-born dI4 IN subtypes such as

GABApre INs has not yet been demonstrated. Thus, using ESC-derived dI4 INs, we can test the role of

Notch signaling in intradomain diversification Ptf1a-derived neurons in the spinal cord.

Defining the subtype heterogeneity of ES-V1 and dI4 interneurons

The emergence of specialized neuronal cell types during embryonic development is orchestrated

by signaling factors working together to progressively restrict the fates of progenitor cells. Yet, the

complete repertoire of cell types formed by these intricate developmental processes has not been

extensively catalogued. In future studies, I will use single-cell RNA-seq approaches to systematically

examine V1 and dI4 IN subtype heterogeneity under baseline conditions, as well as following

manipulations of intrinsic genetic and/or extrinsic signaling programs (Poulin et al, 2016; Wichterle et al,

2013). Indeed, single-cell transcriptional profiling at different stages of ESC-to-IN differentiation might

reveal key transcriptional programs underlying cell fate decisions, as well as factors required to generate

specific subtypes. Finally, single-cell data for classifying neuronal subtypes will likely provide novel

insights linking molecular expression profiles of spinal INs to distinct functional properties, including

morphology, electrophysiology and synaptic connectivity.

185

Cell-intrinsic programs directing interneuron subtype-specific synaptogenesis

The considerable number of potential synaptic partners available in the CNS poses a significant

challenge during the establishment of stereotyped neural circuits. In many parts of the CNS, attractive

and repulsive signals, often in the form of cell adhesion or secreted molecules, have been shown to

mediate synaptic specificity between target neurons (Sanes & Yamagata, 2009). What is the identity of

these recognition signals, and what are the regulatory programs controlling cell type-specific expression

of these cues in RC-MN or GABApre-pSN circuits in the spinal cord?

In the spinal cord, there is substantial evidence that transcriptional programs regulate the

organization of MNs and SNs into functional circuits (Dalla Torre di Sanguinetto et al, 2008). For example,

the ETS TFs Er81 and Pea3, which are expressed in group Ia pSNs as well as specific MN pools, are

crucial for the establishment of monosynaptic sensory-motor circuits. Er81-/- mutant mice have reduced

formation of synaptic connections between Ia afferents and MNs due to the failure of pSN afferents to

grow axon collaterals into the ventral spinal cord (Arber et al, 2000). Meanwhile, Pea3-/- mice exhibit

abnormal MN axonal arborization at target muscles, as well as aberrant cell body migration and settling

position in the spinal cord (Livet et al, 2002). Moreover, induction of Pea3 expression in MNs by muscle-

derived signals has also been shown to be essential for MN subtype-specific dendritic arborization and

selective connectivity of Ia SN afferents with discrete MN pools (Vrieseling et al, 2006). As such, TFs

jointly expressed by pre- and postsynaptic partners in the spinal cord coordinate cell type-specific

synaptic connectivity by controlling the growth of Ia afferent axons towards MNs as well as the formation

of appropriate Ia connections with motor pool targets.

Similar transcriptional programs may operate to define RC and GABApre subtype-specific

circuitry. RCs have been shown to express a specialized, temporally-regulated transcriptional program

that is required for the establishment and maintenance of recurrent inhibitory connections (Stam et al,

2012). However, whether any of these factors are directly involved in RC synaptogenesis with MNs

remains to be determined. MafB-/- mutant mice have reduced numbers of RCs, but those RCs remaining

are still able to provide recurrent inhibition to MNs, suggesting that MafB does not control the formation of

functional synaptic connectivity between RCs and MNs (Stam et al, 2012). Transcriptional regulators of

GABApre IN-specific synaptogenesis have also not been identified. Recent studies suggest that the

186

specific formation and maintenance of axo-axonic GABApre synapses onto primary sensory afferent

terminals relies on interactions of complementary immunoglobulin family cell adhesion molecules on

GABApre INs and pSNs, as well as dual retrograde signals from pSN terminals (i.e., brain-derived

neurotrophic factor, BDNF, and glutamate); however, whether TFs such as Ptf1a encode a transcriptional

program determining the expression of synaptic specificity genes in GABApre INs has not been

determined (Betley et al, 2009; Ashrafi et al, 2014; Mende et al, 2016).

Thus, I established ESC-derived V1 and dI4 INs with two main objectives: First, to identify V1 and

dI4 IN intrinsic transcriptional programs mediating subtype-specific functional properties, especially

synaptogenesis; and second, to use V1 and dI4 INs in co-culture with MNs and SNs to screen and

manipulate cell signaling pathways involved in promoting synaptic specificity between discrete subclasses

of INs and their target postsynaptic partners. In the next section, I will describe the progress I have made

towards these goals, as well as outline ongoing and future studies:

Renshaw cell and GABApre subtype-specific transcriptional programs

To identify transcriptional programs mediating RC-MN synaptic specificity, we can take

advantage of our newfound ability to generate large amounts of RCs using DAPT-mediated Notch

inhibition, as described in Chapter 3. Based on RNA-sequencing results, DAPT-treated ES-V1 INs

specifically upregulate genes associated with the RC differentiation program (i.e., Oc1, Oc2, MafA, MafB,

Calb1), while downregulating TFs associated with non-RC V1 IN subtypes in vivo (Bikoff et al, 2016).

Using data from ES-RCs, we can thus identify genes that are highly induced in RCs relative to other V1

INs and which might underlie RC recurrent inhibition with MNs, especially candidate TFs, transmembrane

cell adhesion molecules and secreted signals (Yamagata & Sanes, 2009).

Alternatively, since some non-RC V1 INs also synapse on MNs and receive monosynaptic

cholinergic inputs, although with significantly less efficiency than RCs, V1 IN-MN synaptic connectivity

might not be established through subtype-specific transcriptional programs, but rather through the

transformation of shared transcriptional programs by activity-dependent and/or cell non-autonomous

mechanisms. For example, different V1 IN subtypes might express the same molecular code for

establishing synaptic connectivity with MNs (e.g. En1, as discussed in Chapter 4), but activity-dependent

187

mechanisms may then act to selectively remove or strengthen synaptic connections during development

to shape the final circuit configuration (Shen & Scheiffele, 2010). Conversely, differential interactions with

neighboring cells may provide instructive signals influencing synaptogenesis of distinct V1 subtypes. For

example, synaptogenesis of bipolar cells (BC) onto retinal ganglion cells partially depends on axon-axon

interactions between neighboring BCs, with partial ablation of BC populations resulting in increased

axonal territories and synaptogenesis (Okawa et al, 2014). Similarly, matching between presynaptic V1

subtypes and postsynaptic MNs and/or maintenance of synaptic connections might not be controlled by

cell intrinsic mechanisms, but rather through cell non-autonomous signaling and interactions.

To test these hypotheses, we can use RNA-sequencing data from ES-V1 and dI4 INs to identify

genes selectively enriched in En1-derived V1 INs compared to Ptf1a-derived dI4 INs, which are not

supposed to synapse onto MNs in the spinal cord (Betley et al, 2009). One caveat to this approach is

that, based on RABV tracing, dI4 INs provide many monosynaptic inputs onto MNs. However, whether

synaptic connections between in vitro V1 IN-MNs and dI4 INs-MNs are established through similar

molecular programs is unclear. Furthermore, since dI4 INs receive significantly fewer cholinergic inputs

from MNs compared to V1-derived RCs, this analysis would still be useful for identifying genes controlling

MN recognition of appropriate IN synaptic partners. To validate candidate factors enriched in V1 INs, we

can also perform cross-comparison of RNA-seq data from ESC-derived neurons with microarray

expression profiling data from embryonic and early postnatal spinal V1 and dI4 INs (Betley et al, 2009).

Spinal neuron co-cultures for screening synaptic specificity molecules

In the studies presented here, I showed that we can selectively generate, purify and then co-

culture distinct V1 IN subtypes such as RCs with MNs to assess their synaptic specificity. In addition, I

demonstrated progress in establishing dI4-derived GABApre INs and pSN co-cultures for similar purpose.

Co-culture assays of RCs and GABApre INs with either MNs or SNs provide a potentially powerful tool for

rapidly and efficiently testing the effect of candidate specificity factors identified from analyses of gene

expression profiling data. Additionally, we can use the co-culture assay for high-throughput, unbiased

screening of known synaptic specificity molecules, or for performing screens using RNA interference

(RNAi) or CRISPR-based gene editing methods to knockdown expression of synaptic specificity genes

188

(Paradis et al, 2007; Linhoff et al, 2009; Takahashi et al, 2011; Wang et al, 2015; Sharma et al, 2013).

For example, misexpression of RC specificity factors in SNs might induce RCs to form ectopic synapses

onto SNs, while selective silencing of the same factors in MNs should impair RC synaptic connectivity

with MNs. Moreover, RCs and GABApre INs can be co-cultured together to assess the specific effects of

these molecules on their synaptogenesis with either MNs or SNs. For example, since RCs never form

synapses with SNs in the spinal cord, molecules mediating GABApre synaptic specificity should not

promote synaptogenesis of RCs onto pSN terminals (Alvarez & Fyffe, 2007; Bhumbra et al, 2014).

In addition to selective expression or knockdown of transcriptional programs and/or molecular

signals mediating synaptic specificity between distinct IN cell types and MNs or SNs, we can also use the

co-culture system to test the cellular organization required for proper sensorimotor circuit formation. For

example, I have shown that the subtype composition of ES-V1 and dI4 cultures can be altered by

manipulations of Shh or ActivinA signals, respectively. Additionally, Notch inhibition enriches for Cb-

expressing RCs over other V1 subtypes. Development of protocols for specific V1 or dI4 IN subtypes will

allow assessment of the influence of the cellular milieu, if any, on V1 and dI4 IN synaptic connectivity with

MNs and/or SNs. Furthermore, co-culture of V1 with GABApre INs, which form synapses on different

postsynaptic targets in the same local environment in vivo, together with MNs and/or pSNs, might reveal

the involvement of other neuron cell types in the establishment of V1-MN and dI4-SN circuits.

Finally, ES-MNs generated through normal differentiation protocols (RA and high Shh signaling)

acquire rostrocervical identity (Wichterle et al, 2002; Peljto et al, 2010). Whether more caudalized MN

subtypes, such as limb-innervating LMC MNs, are able to selectively recruit V1 IN inhibitory inputs has

not been definitively shown. As described in Chapter 3, limb-innervating MNs producing RA may

selectively promote the differentiation of RCs compared to other V1 IN subtypes. Furthermore, some

evidence suggests anatomical segregation of recurrent inhibition, which is found in the majority of

proximal limb muscles involved in stereotyped motor outputs, but not in the most distal forelimb or

hindlimb muscles, perianal sphincters, jaw or eye muscles (Hultborn et al, 1979; Hultborn et al, 1988a;

Hultborn et al, 1988b; Moore et al, 2015). Thus, different MN subtypes may exhibit different synaptic

connectivity patterns with V1 IN subtypes. Similarly, pSNs generated at different levels of the spinal cord

189

may evoke different connectivity response from GABApre INs. Using ESC-derived neurons, we can study

the effects of changing postsynaptic neuron identity on V1 or dI4 IN synapse formation.

Ultimately, our ability to establish co-cultures of select spinal neuron cell types will enable many

future investigations to probe the identity and mechanism of synaptic specificity factors involved in the

formation of precise and stereotyped spinal circuits, allowing for better understanding of how neuronal

identity is transformed into specific patterns of synaptic connectivity.

Subtype-specific synaptic connectivity of spinal inhibitory interneurons

In the spinal cord, RCs provide postsynaptic inhibition of MNs and receive monosynaptic inputs

from MN collaterals. Ia inhibitory INs in the spinal cord, which are derived from both V1 and V2b spinal

domains, also provide inhibitory inputs onto MNs, but whether all other V1-derived INs provide selective

synaptic connections with MNs is not known. Using RABV and optogenetics assays, I showed that a

subset of non-Cb-expressing V1 INs also provide monosynaptic inputs onto MNs and depolarize in

response to to MN cholinergic inputs. Some of these V1 INs express the TF FoxP2, but in the lumbar

spinal cord at p0, FoxP2-expressing V1 INs can be further segregated into 19 molecularly distinct

subtypes based on combinatorial expression of other TFs, indicating that FoxP2-expression alone does

not provide enough information about the molecular and functional identity of these V1 subtypes (Bikoff et

al, 2016). Furthermore, our results show that a subset of V1 INs lacking expression of both Cb and FoxP2

were significantly less likely to synapse onto MNs or receive MN cholinergic inputs. The molecular identity

of non-Cb and non-FoxP2-expressing V1 INs in vitro is not known.

Interestingly, Sp8-expressing V1 INs in the spinal cord provide fewer monosynaptic connections

to MNs compared to Cb-expressing V1 INs (J. Bikoff & T. Jessell, unpublished data). However, using

standard ESC-to-V1 IN differentiation conditions, Sp8-expressing INs are a rare subpopulation in vitro

(<5% of all V1 INs). Meanwhile, non-Cb and non-FoxP2-expressing cells account for ~50-55% of V1 INs.

In Chapter 3, I showed that Pou6f2, FoxP2, MafA, and Sp8 non-overlapping subpopulations of V1 INs

comprise ~50% of all En1-derived reporter cells in vitro, indicating that a large subset of ES-V1 INs are

still unaccounted for both molecularly and functionally. Using gene expression profiling of both spinal and

ES-derived V1 INs, we can potentially identify other molecular markers, TFs and otherwise, that are

190

selectively expressed in these cells. For example, although Bikoff et al (2016) used 19 differentially

expressed TFs to define V1 IN subtype diversity in vivo, another 15 TFs were also significantly enriched

in V1 INs compared to dI4 INs and expressed in embryonic and early postnatal ventral spinal cord.

Furthermore, analysis of gene expression of ES-derived V1 INs might yield distinct candidate

factors from the spinal microarray expression data, not only because V1 INs in vitro are not exposed to

identical extrinsic signals as found in the spinal cord, but also because ES-V1 INs likely adopt cervical

and/or brachial spinal cord identity (as revealed from their Hox expression profiles), whereas spinal V1

INs were dissected and profiled from the lumbar cord. Interestingly, recent evidence points to a role for

Hox genes in regulating the formation of limb-specific V1 IN subtypes, suggesting that different

repertoires of V1 IN subtypes might exist at different spinal segments (L. Sweeney & T. Jessell, personal

communication). Indeed, another recent study shows that molecularly distinct subsets of V1 INs are

differentially distributed along the rostrocaudal axis of the developing spinal cord (Francius et al, 2013).

Co-culture assay for testing synaptic connectivity of interneuron subtypes

Once candidate subtype markers have been validated using immunostaining in EBs and spinal

cord tissue, the overarching goal is to determine which V1 IN subtypes are competent to provide

inhibitory inputs to MNs and/or receive cholinergic collaterals from MNs. Our results show that a subset of

V1 INs lacking expression of both Cb and FoxP2 provide relatively fewer monosynaptic inputs onto MNs.

Furthermore, while only RCs receive MN collaterals in the spinal cord, we observed that a small subset of

non-RC V1 INs also depolarize in response to MN photoactivation in vitro.

To assign function to novel V1 IN subtypes, I will perform additional monosynaptic RABV tracing

assays, potentially with the CVS-N2cΔG variant, which exhibits enhanced retrograde synaptic transfer

and neuronal survival compared to the original SADB19ΔG variant (Reardon et al, 2016; Wickersham et

al, 2007a, Wickersham et al, 2007b). While I showed that we can use optogenetic stimulation of MNs

combined with whole-cell patch-clamp recordings of V1 INs to assess their synaptic connectivity

electrophysiologically, in order to perform more high-throughput assessment of V1 subtype-specific

response to MN inputs, I will use calcium imaging methods to detect depolarization-induced calcium

transients in synaptically-connected V1 INs, followed by post hoc immunostaining with subtype markers

191

(Ohki et al, 2005). To carry out these experiments, I generated En1-tdTomato reporter lines carrying the

ultrasensitive genetically encoded calcium indicator GCaMP6 (Fig 6.3) (Chen et al, 2013). Co-culture of

GCaMP6-expressing V1 INs with MNs will enable population-level assessment of V1 IN subtypes

receiving evoked MN inputs, including providing information about their relative position to MNs for

determining spatial organization of subtype-specific connectivity.

Overall, by taking advantage of the monosynaptic RABV tracing and calcium imaging methods

described here, we can more definitively assign function, particularly synaptic connectivity, to molecularly

distinct V1 IN subtypes in vitro to better understand the cellular organization of V1-MN circuits. While I

focused here on V1-MN circuits, similar analyses can be performed to assess the subtype-specific

connectivity of dI4 INs, including determining the identity of the small subset of dI4 INs receiving VAChT+

MN collaterals in vitro (see Chapter 5).

Implications for studying and treating neurological disease

Thus far, I have shown that ESC-derived spinal INs can be used to study basic developmental

processes of neuronal specification, subtype diversification, and synaptic connectivity. However,

understanding the processes of neuronal subtype diversification and synaptic specificity will likely also

yield insights into why certain neuronal cell types are selectively vulnerable in certain neurodevelopmental

and neurodegenerative diseases. Furthermore, stem cell-derived neurons have long been sought for their

potential therapeutic use, especially for cellular replacement studies in which dysfunctional cells are

replaced with functionally equivalent ESC- or iPSC-derived neurons to reverse disease pathology.

Indeed, our ability to efficiently generate highly specific IN cell types from ESCs will not only be

immensely useful for modeling the cell type-specific outcomes of neurological disease, but also for

designing regenerative medicine strategies for replacement of specific cell types.

Although specific loss of V1 or dI4 INs has not been implicated in any known neurological

diseases, increasing attention has turned to the role of synaptic dysfunction within neural circuits as a

precipitating event for a growing number of neurodegenerative diseases (Palop et al, 2006). Indeed,

many neurological diseases have been associated with alterations of synaptic transmission, especially

dysregulation of excitatory and inhibitory balance in neural circuits. For example, in the MN disease spinal

192

muscular atrophy (SMA), reductions in excitatory, but not inhibitory, inputs on SMA MNs results in altered

sensory-motor neurotransmission and increased intrinsic excitability of MNs, eventually leading to the

premature death of vulnerable MNs (Ling et al, 2010; Mentis et al, 2011; Imlach et al 2012). Notably,

increased neuronal excitability has also been observed in other neurodegenerative diseases, including

Huntington’s, Parkinson’s and Alzheimer’s diseases (Zeron et al, 2002; Chan et al, 2007; Palop et al,

2007). Yet, whether neuronal hyperexcitability is directly responsible for degeneration of vulnerable

neuronal populations has not been adequately determined. Alternatively, recent studies suggest that

rather being detrimental, increased neuronal excitability may instead function as a protective mechanism

to balance ongoing degenerative processes (Saxena et al, 2013; Saxena & Caroni, 2011).

In the neurodegenerative disease amyotrophic sclerosis (ALS), patient MNs also exhibit

progressively increased membrane hyperexcitability (Kiernan et al, 2011; Cleveland & Rothstein, 2001;

Wainger et al, 2014). The abnormal firing behavior of MNs has been suggested to result from changes in

their intrinsic membrane properties over disease progression, but more recently it has been posited that

changes in premotor circuits might contribute to MN hyperexcitability (Schutz, 2005; Sunico et al, 2011;

Wootz et al, 2013; Hossaini et al, 2011; Ramírez-Jarquín et al, 2014). Indeed, recent work suggests that

RC recurrent inhibitory circuits are specifically altered in the superoxide dismutase type 1 (SOD1-G93A)

mouse model of ALS (Wootz et al, 2013; Hossaini et al, 2011). In particular, while Cb-expressing RCs

survive to disease end stage, loss of VAChT-immunoreactive MN collateral inputs onto RCs occurs at

early stages of ALS to disconnect the recurrent inhibitory circuit. Thus, early dysfunction of RCs might

make MNs more susceptible to glutamatergic toxicity in ALS, leading to a hyperexcitability state and

eventual MN loss (Wootz et al, 2013). While the role of RCs in MN dysfunction in ALS has not yet been

settled, these results provide support for a closer examination of the role of different IN cell types in the

dysregulation of spinal circuits during neurodegenerative disease processes.

Therefore, we can ESC-derived V1 and dI4 INs to study both the consequences of impaired

excitatory/inhibitory balance, as well as to study the role of these neurons specifically in

neurodegenerative diseases such as ALS. While previous studies have indicated selective vulnerability of

RC-MN circuits in ALS, whether other V1-derived neurons are also affected during disease progression

has not been determined. Furthermore, a subset of dI4 INs also indirectly regulates MN circuits via

193

presynaptic inhibition of Ia sensory afferents – whether dI4 INs are also affected during

neurodegenerative disease processes has not been shown. Using the co-culture system established

through the studies described in this thesis, we can test if loss of RCs or other V1 and dI4-derived

inhibitory cell types is a cause or consequence of MN pathology in ALS. For example, co-culture of ES-

MNs derived from SOD1-93A mutants with increasing numbers of V1 or dI4 INs might help to determine if

increased inhibitory inputs is generally sufficient to rescue excitability phenotypes in MNs to promote

ALS-MN survival. Alternatively, MN abnormalities might only be rescued through co-culture with specific

IN cell types such as RCs, suggesting that distinct MN circuits are differentially affected in ALS and

therefore central to our understanding of disease pathogenesis. Furthermore, we can also use the co-

culture system to test if specific IN subtypes such as RCs are also selectively vulnerable to oxidative and

endoplasmic reticulum (ER) stress, calcium dysregulation, or astrocyte-derived toxic species, as has

been shown for MNs (Nagai et al, 2007; Kaus & Sareen, 2015). If so, this might suggest that, similar to an

increasing number of neurologic and psychiatric diseases, MN diseases such as ALS might henceforth

also be considered an “interneuronopathy,” characterized not only by primary dysfunction of MNs but also

by secondary insults to circuit components (Kato et al, 2005; Southwell et al, 2014). Moreover, the ability

to generate and perform gene expression profiling specific IN cell types such as RCs allows us to identify

the potential molecular pathways involved in their differential susceptibility. Finally, ESC-derived IN co-

cultures with MNs can also provide a useful platform for screening of drug compounds and small

molecules that improve the function of these neurons during neurodegenerative processes.

Beyond disease modeling, direct IN precursor transplantation has been explored as a strategy for

restoring inhibition to neural circuits affected in conditions which exhibit imbalances in neural excitation

and inhibition, including epilepsy, autism, Huntington’s disease, and Parkinson’s disease, among others.

In addition, IN transplantation has also been proposed as a therapeutic agent for functional recovery and

pain alleviation after neurologic injury, such as spinal cord injury, or for neuropathic pain, a sensory

disorder of the spinal cord. In these latter cases, IN transplantation would be used to facilitate brain

plasticity, promoting the reorganization of spinal neural networks to compensate for damaged and/or

dysfunctional neurons and circuits (Fandel et al, 2016; Southwell et al, 2014). Thus, the ability to generate

functional spinal inhibitory INs from pluripotent stem cells may prove beneficial for multiple therapeutic

194

approaches, spanning from disease modeling to regenerative medicine. Additionally, replacement of

specific IN cell types such as RCs that are loss in neurodegenerative processes including ALS would

prove immensely useful for reconstruction of normal spinal circuits.

Conclusion

In my dissertation studies, I developed and optimized differentiation of ESCs to V1 and dI4 INs,

showing that they recapitulate normal spinal development, including acquiring distinct molecular and

functional subtype identities. Remarkably, through manipulation of developmentally relevant signaling

pathways, we showed here that we can steer ESCs towards a highly specific neuronal cell fate. Renshaw

cells have long been of central interest to physiologists studying spinal circuits, while more recently

gaining recognition for their proposed role in MN pathologies such as ALS. Furthermore, I have

established an experimentally accessible in vitro model of sensory-motor circuitry that can be used to

deconstruct the wiring of spinal circuits, as well as to reveal the role of specific classes of inhibitory INs in

normal spinal physiology and in neurological diseases such as ALS. Thus, using ES-V1 and dI4 INs, we

can gain unprecedented insights into molecular mechanisms generating V1 and dI4 subtype diversity and

synaptic specificity, as well as ways to more efficiently differentiate pluripotent stem cells into clinically-

relevant cell types for modeling disease, drug discovery, and cellular replacement therapy.

195

Figure 6.1

Figure 6.1 Model for Notch and retinoid signaling regulation of Renshaw cell specific development V1 interneurons (IN) are born from e9.5 to e12.5 in the embryonic mouse spinal cord in two waves, the first generating early-born subtypes such as Renshaw cells (RC) and the second wave generating non-RC subtypes such as FoxP2-expressing V1 INs, with more FoxP2-expressing cells generated compared to RCs. On the bottom axis is the corresponding days of V1 IN development in vitro, with the blue bar depicting when RA is required for V1 development (dark blue = RA is required for all V1 specification; light blue = RA is required for specific subtype development); and the red bar indicating when Notch signaling is involved in V1 specification. RA/SAG (blue text) is required for the initial specification of the p1 progenitor domain and development of “generic” V1 interneurons. Subsequently, RA, but not Shh (SAG), is required for the specific development of RCs, acting either on late progenitor/early postmitotic V1 interneurons to control regulation of RC specific transcriptional programs. Meanwhile, Notch signaling (red), inhibits neuronal differentiation to maintain p1 progenitors in their proliferative, undifferentiated state. Later, Notch may also be involved in alternative cell fate specification of distinct V1 subtype identities, potentially in postmitotic V1 INs. For example, Notch signaling may directly repress RC identity to promote other V1 subtype fates.

196

Figure 6.2

Figure 6.2 Renshaw cell distribution along the rostrocaudal axis of e12.5 mouse spinal cord (Top panel) Immunostaining for RC markers in e12.5 mouse spinal cord at brachial levels using Calbindin, Onecut2, and En1. Scale bars = 100 µm. (Bottom panel) Quantification of cells expressing one or a combination of those markers at cervical, brachial, thoracic, and lumbar levels of the spinal cord, with left graph being total number of cells expressing those markers and right graph showing numbers normalized to spinal cord area. (n=4, ANOVA, *p<0.05, **p<0.01, ***p<0.001).

197

Figure 6.3

Figure 6.3 GCaMP6-expressing V1 interneurons for recording subtype-specific activity GCaMP6f and s (fast and slow, respectively) variants were cloned into Tol2 expression vector under control of CAGGS promoter and nucleofected into En1-tdTomato ESC lines for generation of stable ESC lines. Here, variable expression of GCaMP6f expression under baseline conditions without depolarization.

198

Experimental procedures

Mouse ESC derivation

Mouse embryonic stem cell lines were derived from En1::cre or Ptf1a::cre mice crossed to Rosa-

LSL-tdTomato (Ai9 or Ai14), Rosa-LSL-eYFP, Thy1-LSL-YFP (line 2 or 15) fluorescent reporter mice

(Kimmel et al, 2000; Kawaguchi et al, 2002; Madisen et al, 2010; Srinivas et al, 2001; Buffelli et al, 2003).

For derivation of stem cell lines, we bred three mating age females and harvested blastocysts at timed

pregnancy age e3.5 according to established protocols (Abbondanzo et al, 1993; Wong et al, 2010).

Genotyping was performed using primers detecting Cre (Chen et al, 2011) and GFP/YFP (A. Joyner Lab,

Memorial Sloan Kettering) or Rosa (Jackson Laboratory).

Generation of transgenic ESC lines

TVA-G-expressing Hb9::GFP motor neurons

To generate the monosynaptic tracing allele, cDNA encoding TVA and Rabies-G was excised

from pBOB-synP-HTB (Addgene plasmid #30195 Miyamichi et al, 2011) and subcloned into a custom

pminiTol2 expression construct (Addgene plasmid #31829, Balciunas et al, 2006) harboring a PGK

promoter (pPL451, Liu et al, 2003), hygromycin resistance cassette (pCEP4, ThermoFisher Scientific)

and BGH poly-A sequence (pcdDNA3.1, Invitrogen) using high-fideity Phusion polymerase (New England

Biolabs) and In-Fusion HD Cloning System (Clontech).

The TVA-G Tol2 transfer vector was nucleofected into Hb9::GFP ESCs with pCMV-Tol2

transposase (Addgene plasmid #31823, Balciunas et al, 2006) using Amaxa Nucleofector Kit for Neural

Stem Cells (Lonza). Following 50-150 ug/mL hygromycin selection (Sigma-Aldrich), 10 clones were

picked, propagated and differentiated using RA/SAG to MNs (see below). A red fluorescent protein

(dsRed) variant of SADB19∆G RABV was added at low titer to Day 6 MN EBs to assess efficiency of viral

transfer (Osakada et al, 2011).

Inducible Ngn2-Isl1/2-Pou4f1 (PIN) sensory neurons

Inducible ESC lines were generated as previously described (Mazzoni et al, 2012). Open reading

frames (ORF) of select genes (Ngn2, Isl1/2, Pou4f1, FoxS1, Runx3) were cloned using Phusion

199

polymerase, and inserted into the Gateway pENTR/D-TOPO vector (ThermoFisher Scientific), linked by

T2A and P2A peptide sequences for co-expression of multiple transgenes. The 5’ primer always

contained the addition of the CACC sequence to ensure directional integration. Subsequently, LR clonase

was used to transfer ORFs from the pENTR vector into a modified p2Lox plasmid in which GFP is

replaced with L1-L2 cassette from pDEST40 Gateway Destination vector. Included in the cassette is a

downstream V5-His sequence in-frame with the L2 recombination site and followed by a stop codon. For

generation of inducible lines with V5-tag, the initial PCR reaction was performed so that the stop codon is

removed from the original ORF.

We utilized an ESC line harboring the reverse-tetracycline controlled transactivator (rtTA) at the

Rosa-26 locus and the Tet-on inducible promoter at the HPRT locus. Downstream of the Tet-on promoter

is a floxed Cre recombinase cassette (Iacovino et al, 2011). ESCs were pre-treated with doxycycline (dox,

1 µg/mL, Sigma-Aldrich) for 24 hours to robustly induce Cre expression, then the final p2Lox plasmid

containing the ORFs was nucleofected into Cre-induced ESCs. Recombination resulted in replacement of

Cre with the desired construct and neomycin resistance gene driven by the PGK promoter. ESCs were

treated with 150 ng/mL G418 (Invivogen) to select for recombined colonies. After 1-2 weeks of selection,

8-10 ESC colonies were picked and propagated. Recombination was confirmed by inducing putative ESC

line with dox (1-5 ug/mL) and immunostaining for V5 and/or construct-specific antibodies (e.g. Isl1/2).

ESC to spinal neuron differentiation and culture

ESC culture

ESC differentiation to spinal neurons was optimized based on previously published protocols

(Wichterle et al, 2002; Wichterle & Peljto, 2008). ESCs were cultured in ES cell media containing ES D-

MEM (EMD Millipore), 15% ES-grade fetal bovine serum (FBS) (ThermoScientific), supplemented with

1% nucleosides (EMD Millipore), 1% non-essential amino acids (EMD Millipore), 0.1 mM ß-

mercaptoethanol (Sigma-Aldrich), 2 mM L-glutamine (Life Technologies), 1% penicillin-streptomycin (pen-

strep, Life Technologies) and 1000 U/mL Leukemia Inhibitory Factor (LIF, EMD Millipore). All ESC

expansions were carried out on T-25/T-75 tissue culture flasks (Nunc) coated with 0.1% gelatin (EMD

Millipore). For inducible transgenic lines (PIN, HoxC8, DnMaml, NICD), tetracycline-free FBS was used

200

(Clontech). For most experiments, ESCs were expanded on a monolayer of irradiated or mitomycin-C

treated primary mouse embryonic fibroblasts (MEFs) (Globalstem, EMD Millipore), grown in media

containing: ES-MEM (EMD Millipore), 10% FBS (ThermoScientific), L-glutamine, and pen-strep. Media

was exchanged every 2-3 days.

Motor neuron differentiation

ESC-to-MN differentiation was performed as described from Wichterle et al (2002). Hb9::GFP

(Wichterle et al, 2002), Hb9::Cd14-IRES-GFP/ChR2-YFP (Bryson et al, 2014) or iHoxC8-V5 (Tan et al,

2016) ESCs were plated at 2.5x105 cells/10 cm adherent tissue culture dishes (Nunc) in Differentiation

Media containing: Advanced D-MEM/F-12 (Life Technologies) and Neurobasal Medium (Life

Technologies) (1:1 ratio), 10% Knockout Serum Replacement (Life Technologies), 0.1 mM ß-

mercaptoethanol, 2 mM L-glutamine, and pen-strep. On Day 2 of differentiation, EBs were collected, spun

down and split 1:4 into 10 cm non-adherent tissue culture dishes (Corning) and supplemented with 1 µM

RA (Sigma-Aldrich) and 500 nM smoothened agonist (SAG, Calbiochem). Media was exchanged on Day

4 of differentiation with RA and SAG replacement. For induction of HoxC8-V5, 1 or 3 µg/mL dox was

added on late Day 3 of differentiation. The endpoint of MN differentiation was Day 6, when MNs were

dissociated and FACS purified for co-culture studies, or combined with V1 INs for RA signaling studies.

V1 interneuron differentiation

ESC-to-V1 IN differentiation was performed largely as described for MN differentiation, with the

following modifications: En1-tdTomato or En1-GFP ESCs were plated at density of 7.0x105 cells/10 cm

Nunc adherent culture dish. On Day 2 of differentiation, EBs were collected, spun down and split 1:6 into

10 cm Corning dishes and supplemented with 1 µM RA and 5 nM SAG. Media was exchanged on Day 4

and 6 of differentiation with RA and SAG replacement, except in RA signaling studies. The endpoint of V1

IN differentiation was generally Day 8, when EBs were collected for immunostaining or dissociated and

sorted using FACS for RNA-seq or co-culture studies.

201

dI4 interneuron differentiation

ESC-to-dI4 IN differentiation was performed as described for V1 IN differentiation, with the

following modifications: Ptf1a-tdTomato or Ptf1a-Thy1YFP ESCs were plated at density of 8.0x105

cells/10 cm on Nunc adherent culture dish. On Day 2 of differentiation, EBs were collected, spun down

and split 1:4 into 10 cm Corning dishes and supplemented with 1 µM RA and 25 ng/mL ActivinA

(Peprotech). Media was exchanged on Day 4 and 6 of differentiation with RA and ActivinA replacement.

The endpoint of dI4 IN differentiation was also generally Day 8, when EBs were collected for

immunostaining or dissociated and sorted using FACS for RNA-seq or co-culture studies.

Induced proprioceptive sensory neurons

PIN (Pou4f1-Ngn2-Isl1/2) ESCs were plated at density of 2.5x105 cells/10 cm on Nunc adherent

culture dish. On Day 2 of differentiation, EBs were collected, spun down and split 1:4 into 10 cm Corning

dishes and supplemented with 1 µM RA. Dox (1 or 3 µg/mL) and NT-3 (10 ng/mL, Peprotech) was added

on Day 2 of differentiation to induce expression of PIN transgenes and promote pSN differentiation,

respectively. Media was exchanged on Days 4 and 6 with dox/NT-3 replacement. The endpoint of pSN

differentiation was Day 6, when EBs were collected for IHC or dissociated for long-term culture.

DAPT treatment

To inhibit Notch signaling, DAPT was added on early Day 4 for p1 and Day 5 for dP4 progenitors

at concentration 5 µg/mL in DMSO (Selleck Chemicals).

Flow cytometry of reporter cells

To quantify differentiation efficiency, EBs from V1 and dI4 INs were dissociated with 0.05%

trypsin (Life Technologies), manually triturated in Dissociation Buffer: 1X PBS without Ca2+/Mg2+ (Fisher

Scientific), 3% FBS, 2.5% 1M glucose (Sigma-Aldrich), 0.5% 1M MgCl2 (Sigma-Aldrich), 2 mM L-

glutamine, and pen-strep, supplemented with DNase I (1 µg/mL, Sigma-Aldrich). Following filtering

through 40 µM cell strainer, En1-tdTomato/GFP or Ptf1a-tdTomato cells were sorted from the cell

suspension with BD Biosciences FACSCalibur cell analyzer using CellQuest Pro Software.

202

Retinoic acid experiments

Early Day 5 V1 IN EBs and Day 6 Hb9::GFP or iHoxC8-V5 EBs were spun down, washed 3x in

1X PBS and mixed in 1:1 ratio in 10 cm non-adherent dishes in Differentiation Media supplemented with

Gdnf and dox as needed. 1 µM RA and/or 5 nM SAG was replaced as needed. EBs were collected on

Day 8 and processed for immunostaining. For dissociation experiments, EBs were trypsinized briefly in

0.05% Trypsin, manually triturated to single cells, followed by spinning down of V1 INs and MNs together

to aggregate the cells. For vitamin A-depletion studies, EB co-cultures were grown in media containing:

Advanced D-MEM/F-12 and Neurobasal (1:1 ratio), 2% B-27 Supplement without vitamin A (Life

Technologies), 1% N-2 Supplement (Life Technologies), 0.1 mM ß-mercaptoethanol, and pen-strep,

supplemented with Gdnf and dox. For media only replacement studies, MN EBs were spun down and

conditioned media was collected, filtered through 20 µM syringe filter (Fisher Scientific) and mixed with

fresh Differentiation media in 1:1 ratio.

Embedding of embryoid bodies for immunohistochemistry

EBs were generally collected on Day 8 of V1 and dI4 IN differentiation and fixed in 4%

paraformaldehyde (PFA, ThermoScientific) for 15 minutes at 4ºC, followed by 10 min 1X PBS wash steps

(3x). EBs were incubated in 30% sucrose (in 1X PBS with 0.05% NaN3, Sigma-Aldrich) at 4ºC until they

equilibrated. EBs were then deposited into square molds (Polysciences), filled with Optimal Cutting

Temperature (OCT, Tissue-Tek) embedding media, frozen in dry ice, and stored at -80ºC until

cryosectioning. EBs were cryosectioned at 14 µM and collected on SuperFrost Plus glass slides

(ThermoScientific) for immunostaining.

BrdU labeling

To perform BrdU birthdating of V1 INs, 1 µM BrdU (Life Technologies) was added on each day of

differentiation from Days 3-7 for 10 hrs, then washed off to prevent toxicity. To detect BrdU-labeled cells,

EBs were stained as described above, then post-fixed for 10 minutes in 4% PFA at room temperature

(RT), followed by 1X PBS wash steps. Subsequently, 1N hydrochloric acid (HCl) (Sigma-Aldrich) was

applied onto slides for 20-30 min at 37ºC, followed by wash steps in 1X PBS. Rat anti-BrdU antibody was

applied at 1:2000 (Accurate Chemical).

203

Chick spinal cord transplants

Chick transplantations were performed as described in Wichterle et al, 2002 and Peljto et al,

2011. Fertilized eggs were ordered from Charles River and incubated at 38.5ºC in 55-70% humidified

chamber. Early EBs from V1 (Day 5) and dI4 (Day 6) IN differentiations were transplanted into

Hamburger-Hamilton (HH) Stage 16 (embryonic day 3, E3) chick spinal cord, which was suction-lesioned

to accommodate transplanted tissue. Half-to-one EB equivalent was transplanted into a segment

spanning 2-5 somites at varying levels of the spinal cord. Embryos were harvested 4 days later (HH

Stage 30) and examined for successful graft placement under fluorescent microscope.

Dissection of chick and mouse embryonic spinal cords

Embryonic mouse and chick spinal cords were dissected into cervical, brachial, thoracic and

lumbar segments, fixed with 4% PFA for 1-1.5 hr, cryoprotected using 30% sucrose, and embedded into

OCT-filled plastic molds before freezing with dry ice. Tissue blocks were stored at -80ºC until

cryosectioning, which was performed at 20 µM thickness and collected on slides for antibody staining.

Immunohistochemistry/immunocytochemistry

Tissues were first blocked with 10% donkey serum (DS) (EMD Millipore), 0.2% Triton X-100

(Sigma-Aldrich) in 1X PBS, and 0.05% NaN3 for 20-30 min at RT, followed by primary antibody staining

overnight or 4 hrs at RT. The same solution was used to dilute primary and secondary antibodies: 2% DS,

0.2% Triton X-100 in 1X PBS and 0.05% NaN3. After washing in 0.2% Triton X-100, secondary antibody

(1:800, Cy3/Cy5/Alexa488/Alexa405 dyes, Jackson Immunoresearch Laboratories) was applied for 2 hrs

at RT. After washes, coverslips were mounted with Fluoromount-G (ThermoScientific) and imaged.

204

Primary Antibodies

Antibody Species Source

Antibody Species Source Ascl1 Mouse BD Pharmingen

Lhx3 Rabbit T. Jessell

BrdU Rat Accurate Chemical

Lmx1b Guinea Pig T. Jessell

Calbindin Rabbit Swant

MafA Rabbit Novus Biologicals

Calbindin Mouse Swant

MafA Rat Bikoff et al, 2016

ChAT Goat EMD Millipore

MafB Rabbit Bethyl Laboratories

Chx10 Rabbit T. Jessell

NeuN Mouse Millipore

Ebf1 Goat R&D Systems

NeuN Rabbit EMD Millipore

En1 Mouse DSHB

NF-H Mouse T. Jessell

En1 Guinea Pig T. Jessell

Nkx6.2 Guinea Pig Vallstedt et al, 2001

Evx1/2 Mouse T. Jessell

Nr5a2 Rat Bikoff et al, 2016

FoxD3 Rabbit Tompers et al, 2005

Nr5a2 Goat Santa Cruz

FoxP1 Rabbit T. Jessell

Olig3 Mouse R&D Systems

FoxP1 Guinea Pig T. Jessell

Onecut2 Sheep R&D Systems

FoxP2 Rabbit Abcam

Parvalbumin Chick T. Jessell

FoxP2 Goat Santa Cruz

Parvalbumin Rabbit Swant

Gad6 Mouse DSHB

Pax2 Rabbit Zymed/Life Tech

Gad65 Rabbit Betley et al, 2009

Pax6 Mouse T. Jessell

Gad65 Rabbit Chemicon Int'l

Pax7 Rabbit T. Jessell

Gad67 Mouse Chemicon Int'l

Pou4f1 Mouse Santa Cruz

GFAP Rabbit Dako

Pou6f2 Rat Bikoff et al, 2016

GFP Rabbit Invitrogen

Ptf1a Rabbit C. Wright

GFP Sheep Serotec

Rabies-N Mouse EMD Millipore

GlyT2 Guinea Pig Chemicon Int'l

Raldh2 Rabbit Sockanathan & Jessell, 1998

Hb9 Mouse T. Jessell

RFP Rabbit Chemicon Int'l

Hb9 Guinea Pig T. Jessell

Runx3 Rabbit T. Jessell

HoxA2 Rabbit Geisen et al, 2008?

Runx3 Guinea Pig T. Jessell

HoxC6 Mouse Liu et al, 2001

Sp8 Rat Bikoff et al, 2016

HoxC6 Guinea Pig Liu et al, 2001

Sp8 Goat Santa Cruz

HoxC8 Mouse Liu et al, 2001

Synapsin Rabbit Millipore

HoxC8 Guinea Pig Liu et al, 2001

TFAP2b Mouse Abnova

Isl1 Goat Neuromics

TrkA Rabbit T. Jessell, S. Morton

Isl1/2 Guinea Pig T. Jessell

TrkC Rabbit L.F. Reichardt

Isl1/2 Mouse T. Jessell

V5 Mouse Life Technologies

Lbx1 Guinea Pig Müller et al, 2002

VAChT Goat Millipore

Lhx1/5 Mouse DSHB

VGAT Mouse Synaptic Systems

Lhx2/9 Rabbit T. Jessell

Vglut2 Guinea Pig Chemicon Int'l

205

Imaging

All EB and mouse/chick spinal cord images were acquired with 20X objective using confocal laser

scanning microscope (LSM Zeiss Meta 510 or 780). For dissociated neuron cultures, images were

acquired with Zeiss AxioObserver with Coolsnap HQ2 camera (Photometrics) or LSM Zeiss Meta 510 or

780 with 10, 20, and 40X objectives. Images were analyzed off-line using Image-J.

RNA purification for qRT-PCR and RNA-sequencing

For qRT-PCR, EBs were collected on Day 6 for MNs (DnMaml and NICD) and Day 8 for V1 or dI4

INs and spun down, followed by RNA extraction. For RNA-seq, EBs were dissociated for FACS, followed

by RNA extraction. TRIzol or TRIzol LS Reagent (Life Technologies) was used for RNA extraction,

followed by flash freezing in liquid nitrogen and storage at -80°C until further processing. RNA as

separated from DNA and protein fractions using chloroform (Sigma-Aldrich) with Phase Lock Gel columns

(5 Prime). RNA supernatant was further purified using RNeasy (Qiagen) spin column mini kits. For qRT-

PCR, RNA concentration and quality was determined by Nanodrop spectrophotometer and stored at -

80°C. cDNA was prepared using Superscript III Reverse Transcriptase (Life Technologies). cDNA was

then mixed with 0.5 µM primers and 2X SYBR Green qRT-PCR mix (Stratagene) with ROX reference and

analyzed using MX300P qPCR system (Strategene). For RNA-sequencing, RNA concentration and

quality was screened using Bioanalyzer (Agilent) and then submitted to JP Sulzberger Genome Center

(Columbia University Medical Center) for Illumina TruSeq RNA prep with poly-A pull-down to enrich

mRNA. Libraries were then sequenced using Illumina HiSeq2000 (30 million, single-end 100 bp reads).

RNA-seq data analysis

RTA (Illumina) was used for base calling and bcl2fastq (version 1.8.4) for converting BCL to fastq

format, coupled with adaptor trimming. Reads were then mapped to reference genome (UCSC/mm9)

using Tophat (version 2.1.0) with 4 mismatches and 10 maximum multiple hits. The relative abundance

(aka expression level) of genes and splice isoforms was estimated using Cufflinks (version 2.0.2) with

default settings. Differentially expressed genes under various conditions using DEseq2. It is an R

package based on a negative binomial distribution that models the number reads from RNA-seq

206

experiments and test for differential expression. Differentially expressed genes were identified by >2

log2fold change and p adjusted value <0.01. Heatmaps and scatterplots were generated with R packages.

FACS and culture of ESC-derived neurons

Day 8 V1 and dI4 IN and Day 6 MN EBs were dissociated using 0.05% trypsin, manually

triturated in Disssociation Buffer, filtered through 40 µM cell strainer, and sorted using FACS (BD

FACSAria Cell Sorter) to collect highly-expressing tdTomato or GFP cells. Sorted cells were plated on 15

mm coverslips (Fisher Scientific) in 4-well dishes (Nunc). Coverslips were first sterilized by incubation for

2 min in plasma cleaner (Harrick Plasma), coated overnight in 0.0001% poly-L-ornithine (Sigma-Aldrich)

diluted in sterile water at 37ºC and then 3 hrs at 37ºC with mouse laminin (5 ng/mL, Fisher Scientific) and

fibronectin (10 ng/mL, Sigma-Aldrich). For synapses assays, a monolayer of primary cortical astrocytes

was plated on top of laminin only-coated coverslips. Dissociated neurons were plated in Neuronal Media

containing: Neurobasal media, 2% B-27 Supplement, 2% FBS, 0.1 mM ß-mercaptoethanol, L-glutamine,

and pen-strep, supplemented with 10 ng/mL glial-derived neurotrophic factor (Gdnf, R&D Systems) and 1

µM 5-fluoro-2’-deoxyuridine (5-FDU, Sigma-Aldrich) to inhibit non-neuronal and dividing cells, with half or

full media exchange every 2 days. One to two weeks post-culture, cells were fixed in 4% PFA for 10 min

at RT and processed for ICC.

Primary cortical astrocyte culture

Astrocytes were prepared from p0-5 wildtype C57BL/6 mice as described in Albuquerque et al,

2009. Following expansion of astrocyte cultures, ~20-25,000 cells were seeded on laminin-coated glass

coverslips in Astrocyte Media containing Advanced D-MEM/F-12, 10% FBS, L-glutamine, and pen-strep.

Astrocytes were expanded to confluence before neuronal cultures were added.

Motor neuron and interneuron co-culture

Day 6 Hb9::GFP or Hb9::CD14-IRES-GFP EBs were dissociated using trypsin and FACS

(Hb9::GFP) or MACS (Hb9::Cd14-IRES-GFP) (Miltenyi-Biotec) purified. MACS purification was performed

as described in Bryson et al, 2014, using mouse IgG anti-human CD14 (clone UCHM-1, AbD Serotec)

207

primary antibody (5 µg/mL) and 1:10 goat anti-mouse IgG microbeads (Miltenyi-Biotec). For VAChT-

immunoreactivity and optogenetics/electrophysiology studies, V1 INs were differentiated with DAPT

added on Day 4 and replaced on Day 6 to enrich for RC generation. For all other studies V1 and dI4 INs

were differentiated using standard protocols, collected on Day 8 for dissociation and FACS purification

and plated with MNs in varying densities and ratios. For 4-well co-cultures, 50-100,000 sorted cells were

plated on astrocyte monolayer, with survival of ~1/10 cells after 1-2 weeks culture. For electrophysiology

recordings, 1:3 ratio MN: V1 INs was used for most experiments. Changing the densities and ratios of

cells within a certain range (1:1 up to 1:20 ratio and 25,000 to 150,000 cells) produced small, but not

statistically significant changes.

Monosynaptic rabies virus tracing in vitro

Day 6 Hb9::GFP MN EBs carrying TVA-G transgene for RABV initial infection and Day 8 En1-

tdTomato or Ptf1a-tdTomato EBs were FACS purified and cultured for one week on astrocyte monolayer,

with half-media exchange every other day. A small volume (1 µL) of low titer SADB19∆G-GFP RABV

(Wickersham et al, 2007a; Wickersham et al, 2007b) was added directly to media, then removed after 2

days. After RABV treatment, cells were cultured for an additional 4 or 7 days, then fixed in 4% PFA and

prepared for ICC. To assess efficiency of RABV transfer, anti-Rabies N nucleoprotein (1:50, clone C18-

62-143-2, EMD Millipore) was used in conjunction with other antibody stainings.

V1 interneuron electrophysiology

Glass coverslips containing En1-tdTomato and Hb9::GFP or Hb9::CD14-IRES-GFP/ChR2-YFP

MNs on astrocyte monolayer were transferred to a customized recording chamber under an SP5 Leica

confocal microscope equipped with 4 single laser lines (405, 488, 543, 650nm). The recording chamber

was filled with HEPES buffer consisting of: 145mM NaCl, 3mM KCl, 1mM MgCl2, 1.5mM CaCl2, 10mM

glucose, and 10mM HEPES adjusted to pH 7.4 with NaOH. Electrodes were pulled with a P1000 puller

(Sutter) to resistance of 10-15MΩ. Recording electrodes were filled with intracellular solution containing:

10mM NaCl, 130mM K-Gluconate, 10mM HEPES, 1mM EGTA, 1mM MgCl2, 0.1mM CaCl2, 1mM

208

Na2ATP, and fluorescent dye (Cascade Blue, Neurobiotin), pH adjusted to 7.2-7.3 with KOH (the final

osmolarity of the intracellular solution was 295-299 mOsm).

V1 interneurons were visually targeted by their endogenous tdTomato fluorescence. Whole-cell

patch-clamp recordings were performed in current-clamp setting to characterize the intrinsic membrane

properties of En1-tdTomato FP cells, as well as their firing patterns in response to injected increments of

current steps, using standard patch-clamp protocols. Recordings were accepted for analysis if they had a

resting membrane potential of -35 mV or lower and overshooting action potentials. The passive

membrane properties of V1 INs were assessed by injection of negative and positive steps of current (100-

300 ms duration) at -60 mV holding membrane potential. The input resistance was calculated from the

slope of the linear current/voltage relationship. Cascade Blue/Neurobiotin-filled recorded cells were

subsequently fixed in 4% PFA for 10 min for post hoc ICC to detect for V1 IN subtype-specific markers.

Optogenetic stimulation

Hb9::CD14-IRES-GFP/ChR2-YFP MNs were photostimulated with 470nm light pulses (25 ms

duration) from a LED source (CoolLED) while whole-cell patch-clamp recordings of En1-tdTomato FP

cells were performed to assess for depolarization in response to MN synaptic inputs. In some

experiments, the cholinergic receptor antagonists atropine (5µM) and mecamylamine (50µM) were

applied to the bath solution. We calculated the latency of V1 IN response relative to the onset of the MN

action potential. To confirm that the response was monosynaptic, we subjected V1 INs to multiple trials at

different stimulation frequencies (0.1, 1, and 10 Hz) to determine the jitter, or variability, of V1 IN

response onset, calculated as the coefficient of variation (Shneider et al, 2009). Cascade

Blue/Neurobiotin-filled En1-tdTomato FP cells were fixed with 4% PFA and prepared for post hoc ICC.

Quantifications/Statistical Analysis

V1 and dI4 IN molecular marker analysis

For quantification of markers expressed in En1-FP or Ptf1a-FP EBs, confocal images of EBs

were taken at 20X objective and processed in Image-J for counting using the Cell Counter plugin for

manual counting. The number of cells that co-expressed the marker and reporter was counted and

209

divided by the total number of reporter cells in the EB. At least five EBs were counted for each

experiment, with at least 3 independent experiments for calculation of mean ± standard error of the mean

(SEM). Statistics were performed using two-tailed, unpaired Student’s t-test or one-way ANOVA. Relevant

p-values: *p<0.05, **p<0.01, ***p<0.001

Transplants

To quantify migration of transplanted En1-GFP and Ptf1a-tdTomato FP cells, spinal cord images

were taken at 20X objective on confocal microscope and processed in Image-J. At least four transplants

were used for either V1 or dI4 IN quantifications. Each image was aligned dorsoventrally and then divided

into 6 equal bins and the number of reporter cells in each bin was counted to plot the fraction of total

reporter cells in each of the bins. To quantify subtype migration of Cb or FoxP2-expressing V1 INs, at

least four transplants stained for both Cb and FoxP2 were examined. Multiple spinal cord images were

overlayed and then an average spinal cord area was calculated. Subsequently, positional coordinates of

Cb and FoxP2-expressing cells were determined using Image-J and normalized to the average spinal

cord area, with the most dorsal/lateral positions represented as 1 and the most ventral/medial as 0. A

scatterplot of subtype distribution was generated in Matlab, while quantification of subtype position was

performed by generating 20 different mediolateral or dorsoventral bins.

Dissociated cells

Cultured neurons on coverslips were immunostained and imaged using Zeiss AxioObserver

inverted microscope with 20X or 40X objective. Typically, the entire glass coverslip was counted for total

number of reporter cells and cells co-expressing protein of interest. At least 2 coverslips from each

experiment, with at least 3 independent experiments for calculation of mean ± standard error of the mean

(SEM). Statistics were performed using two-tailed, unpaired Student’s t-test or ANOVA. Relevant p-

values: *p<0.05, **p<0.01, ***p<0.001.

210

References Abbondanzo SJ, Gadi I, Stewart CL. “Derivation of embryonic stem cell lines.” Methods Enzymol. 1993;225:803-23. Agalliu D, Takada S, Agalliu I, McMahon AP, Jessell TM. “Motor neurons with axial muscle projections specified by Wnt4/5 signaling.” Neuron. 2009 Mar 12;61(5):708-20. Aguila JC, Hedlund E, Sanchez-Pernaute R. “Cellular programming and reprogramming: sculpting cell fate for the production of dopamine neurons for cell therapy.” Stem Cells Int. 2012:412040. Alaynick WA, Jessell TM, Pfaff SL. “SnapShot: spinal cord development.” Cell. 2011 Jul 8;146(1):178-178.e1. Albuquerque C, Joseph DJ, Choudhury P, MacDermott AB. “Dissection, plating and maintenance of cortical astrocyte cultures.” Cold Spring Harb Protoc. 2009 Aug;2009(8):pdb.prot5273. Alder J, Cho NK, Hatten ME. “Embryonic precursor cells from the rhombic lip are specified to a cerebellar granule neuron identity.” Neuron. 1996 Sep;17(3):389-99. Allodi I, Hedlund E. “Directed midbrain and spinal cord neurogenesis from pluripotent stem cells to model development and disease in a dish.” Front Neurosci. 2014 May 20;8:109. Alvarez FJ, Dewey DE, Harrington DA, Fyffe RE. “Cell-type specific organization of glycine receptor clusters in the mammalian spinal cord.” J Comp Neurol. 1997 Mar 3;379(1):150-70. Alvarez FJ, Dewey DE, McMillin P, Fyffe RE. “Distribution of cholinergic contacts on Renshaw cells in the rat spinal cord: a light microscopic study.” J Physiol. 1999 Mar 15;515 ( Pt 3):787-97. Alvarez FJ, Jonas PC, Sapir T, Hartley R, Berrocal MC, Geiman EJ, Todd AJ, Goulding M. “Postnatal phenotype and localization of spinal cord V1 derived interneurons.” J Comp Neurol. 2005 Dec 12;493(2):177-92. Alvarez FJ, Fyffe RE. “The continuing case for the Renshaw cell.” J Physiol. 2007 Oct 1;584(Pt 1):31-45.

Alvarez FJ, Benito-Gonzalez A, Siembab VC. “Principles of interneuron development learned from Renshaw cells and the motoneuron recurrent inhibitory circuit.” Ann N Y Acad Sci. 2013 Mar;1279:22-31.

Andersson E, Tryggvason U, Deng Q, Friling S, Alekseenko Z, Robert B, Perlmann T, Ericson J. “Identification of intrinsic determinants of midbrain dopamine neurons.” Cell. 2006 Jan 27;124(2):393-405.

Arber S, Ladle DR, Lin JH, Frank E, Jessell TM. “ETS gene Er81 controls the formation of functional connections between group Ia sensory afferents and motor neurons.” Cell. 2000 May 26;101(5):485-98.

Arber S. “Motor circuits in action: specification, connectivity, and function.” Neuron. 2012 Jun 21;74(6):975-89. Arenas E, Denham M, Villaescusa JC. “How to make a midbrain dopaminergic neuron.” Development. 2015 Jun 1;142(11):1918-36.

Artavanis-Tsakonas S, Rand MD, Lake RJ. “Notch signaling: cell fate control and signal integration in development.” Science. 1999 Apr 30;284(5415):770-6.

Aruga J, Inoue T, Hoshino J, Mikoshiba K. “Zic2 controls cerebellar development in cooperation with Zic1.” J Neurosci. 2002 Jan 1;22(1):218-25.

211

Aruga J, Tohmonda T, Homma S, Mikoshiba K. “Zic1 promotes the expansion of dorsal neural progenitors in spinal cord by inhibiting neuronal differentiation.” Dev Biol. 2002 Apr 15;244(2):329-41.

Ashrafi S, Betley JN, Comer JD, Brenner-Morton S, Bar V, Shimoda Y, Watanabe K, Peles E, Jessell TM, Kaltschmidt JA. “Neuronal Ig/Caspr recognition promotes the formation of axoaxonic synapses in mouse spinal cord.” Neuron. 2014 Jan 8;81(1):120-9.

Au E, Ahmed T, Karayannis T, Biswas S, Gan L, Fishell G. “A modular gain-of-function approach to generate cortical interneuron subtypes from ES cells.” Neuron. 2013 Dec 4;80(5):1145-58.

Austin CP, Feldman DE, Ida JA Jr, Cepko CL. “Vertebrate retinal ganglion cells are selected from competent progenitors by the action of Notch.” Development. 1995 Nov;121(11):3637-50.

Bai CB, Stephen D, Joyner AL. “All mouse ventral spinal cord patterning by hedgehog is Gli dependent and involves an activator function of Gli3.” Dev Cell. 2004 Jan;6(1):103-15.

Balciunas D, Wangensteen KJ, Wilber A, Bell J, Geurts A, Sivasubbu S, Wang X, Hackett PB, Largaespada DA, McIvor RS, Ekker SC. “Harnessing a high cargo-capacity transposon for genetic applications in vertebrates.” PLoS Genet. 2006 Nov 10;2(11):e169.

Barber RP, Phelps PE, Houser CR, Crawford GD, Salvaterra PM, Vaughn JE. “The morphology and distribution of neurons containing choline acetyltransferase in the adult rat spinal cord: an immunocytochemical study.” J Comp Neurol. 1984 Nov 1;229(3):329-46. Barth KA, Kishimoto Y, Rohr KB, Seydler C, Schulte-Merker S, Wilson SW. “Bmp activity establishes a gradient of positional information throughout the entire neural plate.” Development. 1999 Nov;126(22):4977-87.

Bassett EA, Korol A, Deschamps PA, Buettner R, Wallace VA, Williams T, West-Mays JA. “Overlapping expression patterns and redundant roles for AP-2 transcription factors in the developing mammalian retina.” Dev Dyn. 2012 Apr;241(4):814-29. Batista MF, Lewis KE. “Pax2/8 act redundantly to specify glycinergic and GABAergic fates of multiple spinal interneurons.” Dev Biol. 2008 Nov 1;323(1):88-97.

Baumgardt M, Miguel-Aliaga I, Karlsson D, Ekman H, Thor S. “Specification of neuronal identities by feedforward combinatorial coding.” PLoS Biol. 2007 Feb;5(2):e37. Baumgardt M, Karlsson D, Terriente J, Díaz-Benjumea FJ, Thor S. “Neuronal subtype specification within a lineage by opposing temporal feed-forward loops.” Cell. 2009 Nov 25;139(5):969-82. Benito-Gonzalez A, Alvarez FJ. “Renshaw cells and Ia inhibitory interneurons are generated at different times from p1 progenitors and differentiate shortly after exiting the cell cycle.” J Neurosci. 2012 Jan 25;32(4):1156-70.

Beres TM, Masui T, Swift GH, Shi L, Henke RM, MacDonald RJ. “PTF1 is an organ-specific and Notch-independent basic helix-loop-helix complex containing the mammalian Suppressor of Hairless (RBP-J) or its paralogue, RBP-L.” Mol Cell Biol. 2006 Jan;26(1):117-30.

Berggren K, McCaffery P, Dräger U, Forehand CJ. “Differential distribution of retinoic acid synthesis in the chicken embryo as determined by immunolocalization of the retinoic acid synthetic enzyme, RALDH-2.” Dev Biol. 1999 Jun 15;210(2):288-304. Betley JN, Wright CV, Kawaguchi Y, Erdélyi F, Szabó G, Jessell TM, Kaltschmidt JA. “Stringent specificity in the construction of a GABAergic presynaptic inhibitory circuit.” Cell. 2009 Oct 2;139(1):161-74.

212

Bhumbra GS, Bannatyne BA, Watanabe M, Todd AJ, Maxwell DJ, Beato M. “The recurrent case for the Renshaw cell.” J Neurosci. 2014 Sep 17;34(38):12919-32.

Bikoff JB, Gabitto MI, Rivard AF, Drobac E, Machado TA, Miri A, Brenner-Morton S, Famojure E, Diaz C, Alvarez FJ, Mentis GZ, Jessell TM. “Spinal inhibitory interneuron diversity delineates variant motor microcircuits.” Cell. 2016 Mar 24;165(1):207-19.

Björnfors ER, El Manira A. “Functional diversity of excitatory commissural interneurons in adult zebrafish.” Elife. 2016 Aug 25;5. pii: e18579. Blanchard JW, Eade KT, Szűcs A, Lo Sardo V, Tsunemoto RK, Williams D, Sanna PP, Baldwin KK. “Selective conversion of fibroblasts into peripheral sensory neurons.” Nat Neurosci. 2015 Jan;18(1):25-35. Blatow M, Caputi A, Burnashev N, Monyer H, Rozov A. “Ca2+ buffer saturation underlies paired pulse facilitation in calbindin-D28k-containing terminals.” Neuron. 2003 Apr 10;38(1):79-88.

Blumberg B, Bolado J Jr., Morena TA, Kintner C, Evans RM, Papalopulu N. “An essential role for retinoid signaling in anteroposterior neural patterning.” Development. 1997 Jan;124(2):373-9.

Bonanomi D, Pfaff SL. “Motor axon pathfinding.” Cold Spring Harb Perspect Biol. 2010 Mar;2(3):a001735.

Bonner J, Gribble SL, Veien ES, Nikolaus OB, Weidinger G, Dorsky RI. “Proliferation and patterning are mediated independently in the dorsal spinal cord downstream of canonical Wnt signaling.” Dev Biol. 2008 Jan 1;313(1):398-407.

Borromeo MD, Meredith DM, Castro DS, Chang JC, Tung KC, Guillemot F, Johnson JE. “A transcription factor network specifying inhibitory versus excitatory neurons in the dorsal spinal cord.” Development. 2014 Jul;141(14):2803-12. Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K. “Millisecond-timescale, genetically targeted optical control of neural activity.” Nat Neurosci. 2005 Sep;8(9):1263-8. Brink EE, Suzuki I. “Recurrent inhibitory connexions among neck motoneurones in the cat.” J Physiol. 1987 Feb;383:301-26. Briscoe J, Sussel L, Serup P, Hartigan-O'Connor D, Jessell TM, Rubenstein JL, Ericson J. “Homeobox gene Nkx2.2 and specification of neuronal identity by graded Sonic hedgehog signaling.” Nature. 1999 Apr 15;398(6728):622-7.

Briscoe J, Pierani A, Jessell TM, Ericson J. “A homeodomain protein code specifies progenitor cell identity and neuronal fate in the ventral neural tube.” Cell. 2000 May 12;101(4):435-45. Briscoe J, Ercison J. “Specification of neuronal fates in the ventral neural tube.” Curr Opin Neurobiol. 2001 Feb;11(1):43-9. Briscoe J, Chen Y, Jessell TM, Struhl G. “A hedgehog-insensitive form of patched provides evidence for direct long-range morphogen activity of sonic hedgehog in the neural tube.” Mol Cell. 2001 Jun;7(6):1279-91. Britanova O, de Juan Romero C, Cheung A, Kwan KY, Schwark M, Gyorgy A, Vogel T, Akopov S, Mitkovski M, Agoston D, Sestan N, Molnár Z, Tarabykin V. “Satb2 is a postmitotic determinant for upper-layer neuron specification in the neocortex.” Neuron. 2008 Feb 7;57(3):378-92 Britz O, Zhang J, Grossmann KS, Dyck J, Kim JC, Dymecki S, Gosgnach S, Goulding M. “A genetically defined asymmetry underlies the inhibitory control of flexor-extensor locomotor movements.” Elife. 2015 Oct 14;4.

213

Broadus J, Doe CQ. “Extrinsic cues, intrinsic cues and microfilaments regulate asymmetric protein localization in Drosophila neuroblasts.” Curr Biol. 1997 Nov 1;7(11):827-35. Bröhl D, Strehle M, Wende H, Hori K, Bormuth I, Nave KA, Müller T, Birchmeier C. “A transcriptional network coordinately determines transmitter and peptidergic fate in the dorsal spinal cord.” Dev Biol. 2008 Oct 15;322(2):381-93. Brown CR, Butts JC, McCreedy DA, Sakiyama-Elbert SE. “Generation of v2a interneurons from mouse embryonic stem cells.” Stem Cells Dev. 2014 Aug 1;23(15):1765-76. Brownstone RM, Bui TV. “Spinal interneurons providing input to the final common path during locomotion.” Prog Brain Res. 2010;187:81-95. Bryson JB, Machado CB, Crossley M, Stevenson D, Bros-Facer V, Burrone J, Greensmith L, Lieberam I. “Optical control of muscle function by transplantation of stem cell-derived motor neurons in mice.” Science. 2014 Apr 4;344(6179):94-7. Buffelli M, Burgess RW, Feng G, Lobe CG, Lichtman JW, Sanes JR. “Genetic evidence that relative synaptic efficacy biases the outcome of synaptic competition.” Nature. 2003 Jul 24;424(6947):430-4. Bui TV, Cushing S, Dewey D, Fyffe RE, Rose PK. “Comparison of the morphological and electrotonic properties of Renshaw cells, Ia inhibitory interneurons, and motoneurons in the cat.” J Neurophysiol. 2003 Nov;90(5):2900-18. Burke RE, Fedina L, Lundberg A. “Spatial synaptic distribution of recurrent and group Ia inhibitory systems in cat spinal motoneurones.” J Physiol. 1971 Apr;214(2):305-26. Burrill JD, Moran L, Goulding MD, Saueressig H. “PAX2 is expressed in multiple spinal cord interneurons, including a population of EN1+ interneurons that require PAX6 for their development.” Development. 1997 Nov;124(22):4493-503.

Bushong EA, Martone ME, Ellisman MH. “Maturation of astrocyte morphology and the establishment of astrocyte domains during postnatal hippocampal development.” Int J Dev Neurosci. 2004 Apr;22(2):73-86.

Buss, RR, Sun, W, Oppenheim, RW. “Adaptive roles of programmed cell death during nervous system development.” Annu. Rev. Neurosci. 2006;29:1-35.

Butt SJ, Fuccillo M, Nery S, Noctor S, Kriegstein A, Corbin JG, Fishell G. “The temporal and spatial origins of cortical interneurons predict their physiological subtype.” Neuron. 2005 Nov 23;48(4):591-604. Butts T, Green MJ, Wingate RJ. “Development of the cerebellum: simple steps to make a 'little brain'.” Development. 2014 Nov;141(21):4031-41. Cajal, RyS. “Histologie du systéme nerveux de l' homme et des vertébrés.” Paris: Maloine; c1911. Callaway EM. “Transneuronal circuit tracing with neurotropic viruses.” Curr Opin Neurobiol. 2008 Dec;18(6):617-23. Callaway EM, Luo L. “Monosynaptic Circuit Tracing with Glycoprotein-Deleted Rabies Viruses.” J Neurosci. 2015 Jun 17;35(24):8979-85. Camardo J, Proshansky E, Schacher S. “Identified Aplysia neurons form specific chemical synapses in culture.” J Neurosci. 1983 Dec;3(12):2614-20. Carletti B, Grimaldi P, Magrassi L, Rossi F. “Specification of cerebellar progenitors after heterotopic-

214

heterochronic transplantation to the embryonic CNS in vivo and in vitro.” J Neurosci. 2002 Aug 15;22(16):7132-46.

Carr PA, Alvarez FJ, Leman EA, Fyffe RE. “Calbindin D28k expression in immunohistochemically identified Renshaw cells.” Neuroreport. 1998 Aug 3;9(11):2657-61.

Casas C, Herrando-Grabulosa M, Manzano R, Mancuso R, Osta R, Navarro X. “Early presymptomatic cholinergic dysfunction in a murine model of amyotrophic lateral sclerosis.” Brain Behav. 2013 Mar;3(2):145-58. Cash S, Chiba A, Keshishian H. “Alternate neuromuscular target selection following the loss of single muscle fibers in Drosophila.” J Neurosci. 1992 Jun;12(6):2051-64. Caspary T, Anderson KV. “Patterning cell types in the dorsal spinal cord: what the mouse mutants say.” Nat Rev Neurosci. 2003 Apr;4(4):289-97. Catela C, Shin MM, Dasen JS. “Assembly and function of spinal circuits for motor control.” Annu Rev Cell Dev Biol. 2015;31:669-98. Cepko C. “Intrinsically different retinal progenitor cells produce specific types of progeny.” Nat Rev Neurosci. 2014 Sep;15(9):615-27. Chan CS, Guzman JN, Ilijic E, Mercer JN, Rick C, Tkatch T, Meredith GE, Surmeier DJ. “'Rejuvenation' protects neurons in mouse models of Parkinson's disease.” Nature. 2007 Jun 28;447(7148):1081-6. Chang JC, Meredith DM, Mayer PR, Borromeo MD, Lai HC, Ou YH, Johnson JE. “Prdm13 mediates the balance of inhibitory and excitatory neurons in somatosensory circuits.” Dev Cell. 2013 Apr 29;25(2):182-95 Chen JA, Huang YP, Mazzoni EO, Tan GC, Zavadil J, Wichterle H. “Mir-17-3p controls spinal neural progenitor patterning by regulating Olig2/Irx3 cross-repressive loop.” Neuron 2011 Feb 24;69(4):721-35. Chen TW, Wardill TJ, Sun Y, Pulver SR, Renninger SL, Baohan A, Schreiter ER, Kerr RA, Orger MB, Jayaraman V, Looger LL, Svoboda K, Kim DS. “Ultrasensitive fluorescent proteins for imaging neuronal activity.” Nature. 2013 Jul 18;499(7458):295-300. Cheng L, Arata A, Mizuguchi R, Qian Y, Karunaratne A, Gray PA, Arata S, Shirasawa S, Bouchard M, Luo P, Chen CL, Busslinger M, Goulding M, Onimaru H, Ma Q. “Tlx3 and Tlx1 are post-mitotic selector genes determining glutamatergic over GABAergic cell fates.” Nat Neurosci. 2004 May;7(5):510-7. Cheng L, Samad OA, Xu Y, Mizuguchi R, Luo P, Shirasawa S, Goulding M, Ma Q. “Lbx1 and Tlx3 are opposing switches in determining GABAergic versus glutamatergic transmitter phenotypes.” Nat Neurosci. 2005 Nov;8(11):1510-5. Chesnutt C, Burrus LW, Brown AM, Niswander L. “Coordinate regulation of neural tube patterning and proliferation by TGFbeta and WNT activity.” Dev Biol. 2004 Oct 15;274(2):334-47. Chiang C, Litingtung Y, Lee E, Young KE, Corden JL, Westphal H, Beachy PA. “Cyclopia and defective axial patterning in mice lacking Sonic hedgehog gene function.” Nature. 1996 Oct 3;383(6599):407-13. Chih B, Engelman H, Scheiffele P. “Control of excitatory and inhibitory synapse formation by neuroligins.” Science. 2005 Feb 25;307(5713):1324-8. Chung WS, Allen NJ, Eroglu C. “Astrocytes Control Synapse Formation, Function, and Elimination.” Cold Spring Harb Perspect Biol. 2015 Feb 6;7(9):a020370.

215

Clarke LE, Barres BA. “Emerging roles of astrocytes in neural circuit development.” Nat Rev Neurosci. 2013 May;14(5):311-21. Cleveland DW, Rothstein JD. “From Charcot to Lou Gehrig: deciphering selective motor neuron death in ALS.” Nat Rev Neurosci. 2001 Nov;2(11):806-19. Conradi S. “Ultrastructure and distribution of neuronal and glial elements on the motoneuron surface in the lumbosacral spinal cord of the adult cat.” Acta Physiol Scand Suppl. 1969;332:5-48. Conradi S, Skoglund S. “Observations on the ultrastruture and distribution of neuronal and glial elements on the motoneuron surface in the lumbosacral spinal cord of the cat during postnatal development.” Acta Physiol Scand Suppl. 1969;333:5-52. Corbin JG, Butt SJ. “Developmental mechanisms for the generation of telencephalic interneurons.” Dev Neurobiol. 2011 Aug;71(8):710-32. Corbin JG, Rutlin M, Gaiano N, Fishell G. “Combinatorial function of the homeodomain proteins Nkx2.1 and Gsh2 in ventral telencephalic patterning.” Development. 2003 Oct;130(20):4895-906. Coulon P, Bras H, Vinay L. “Characterization of last-order premotor interneurons by transneuronal tracing with rabies virus in the neonatal mouse spinal cord.” J Comp Neurol. 2011 Dec 1;519(17):3470-87. Cox WG, Hemmati-Brivanlou A. “Caudalization of neural fate by tissue recombination and bFGF” Development. 1995 Dec;121(12):4349-58. Craig AM, Blackstone CD, Huganir RL, Banker G. “Selective clustering of glutamate and gamma-aminobutyric acid receptors opposite terminals releasing the corresponding neurotransmitters.” Proc Natl Acad Sci U S A. 1994 Dec 20;91(26):12373-7. Crawford TQ, Roelink H. “The notch response inhibitor DAPT enhances neuronal differentiation in embryonic stem cell-derived embryoid bodies independently of sonic hedgehog signaling.” Dev Dyn. 2007 Mar;236(3):886-92. Cullheim S, Kellerth JO. “Two kinds of recurrent inhibition of cat spinal alpha-motoneurones as differentiated pharmacologically.J Physiol. 1981 Mar;312:209-24. Curtis DR, Eccles JC. “The time courses of excitatory and inhibitory synaptic actions.” J Physiol. 1959 Mar 12;145(3):529-46. Curtis DR, Phillis JW, Watkins JC. “The chemical excitation of spinal neurones by certain acidic amino acids.” J Physiol. 1960 Mar;150:656-82. Dalla Torre di Sanguinetto SA, Dasen JS, Arber S. “Transcriptional mechanisms controlling motor neuron diversity and connectivity.” Curr Opin Neurobiol. 2008 Feb;18(1):36-43. Danjo T, Eiraku M, Muguruma K, Watanabe K, Kawada M, Yanagawa Y, Rubenstein JL, Sasai Y. “Subregional specification of embryonic stem cell-derived ventral telencephalic tissues by timed and combinatory treatment with extrinsic signals.” J Neurosci. 2011 Feb 2;31(5):1919-33. Dasen JS, Liu JP, Jessell TM. “Motor neuron columnar fate imposed by sequential phases of Hox-c activity.” Nature. 2003 Oct 30;425(6961):926-33. Dasen JS, Tice BC, Brenner-Morton S, Jessell TM. “A Hox regulatory network establishes motor neuron pool identity and target-muscle connectivity.” Cell. 2005 Nov 4;123(3):477-91. Dasen JS, De Camilli A, Wang B, Tucker PW, Jessell TM. “Hox repertoires for motor neuron diversity and

216

connectivity gated by a single accessory factor, FoxP1.” Cell. 2008 Jul 25;134(2):304-16. Dasen JS, Jessell TM. “Hox networks and the origins of motor neuron diversity.” Curr Top Dev Biol. 2009;88:169-200.

De Marco Garcia NV, Jessell TM. “Early motor neuron pool identity and muscle nerve trajectory defined by postmitotic restrictions in Nkx6.1 activity.” Neuron. 2008 Jan 24;57(2):217-31.

Dean C, Scholl FG, Choih J, DeMaria S, Berger J, Isacoff E, Scheiffele P. “Neurexin mediates the assembly of presynaptic terminals.” Nat Neurosci. 2003 Jul;6(7):708-16.

Deguchi Y, Donato F, Galimberti I, Cabuy E, Caroni P. “Temporally matched subpopulations of selectively interconnected principal neurons in the hippocampus.” Nat Neurosci. 2011 Apr;14(4):495-504.

Del Barrio MG, Taveira-Marques R, Muroyama Y, Yuk DI, Li S, Wines-Samuelson M, Shen J, Smith HK, Xiang M, Rowitch D, Richardson WD. “A regulatory network involving Foxn4, Mash1 and delta-like 4/Notch1 generates V2a and V2b spinal interneurons from a common progenitor pool.” Development. 2007 Oct;134(19):3427-36.

Del Barrio MG, Bourane S, Grossmann K, Schüle R, Britsch S, O'Leary DD, Goulding M. “A transcription factor code defines nine sensory interneuron subtypes in the mechanosensory area of the spinal cord.” PLoS One. 2013 Nov 4;8(11):e77928.

Delaspre F, Massumi M, Salido M, Soria B, Ravassard P, Savatier P, Skoudy A. “Directed pancreatic acinar differentiation of mouse embryonic stem cells via embryonic signalling molecules and exocrine transcription factors.” PLoS One. 2013;8(1):e54243.

Dessaud E, Yang LL, Hill K, Cox B, Ulloa F, Ribeiro A, Mynett A, Novitch BG, Briscoe J. “Interpretation of the sonic hedgehog morphogen gradient by a temporal adaptation mechanism.” Nature. 2007 Nov 29;450(7170):717-20. Dessaud E, McMahon AP, Briscoe J. “Pattern formation in the vertebrate neural tube: a sonic hedgehog morphogen-regulated transcriptional network.” Development. 2008 Aug;135(15):2489-503.

Dickinson ME, Krumlauf R, McMahon AP. “Evidence for a mitogenic effect of Wnt-1 in the developing mammalian central nervous system.” Development. 1994 Jun;120(6):1453-71.

Diez del Corral R, Olivera-Martinez I, Goriely A, Gale E, Maden M, Storey K. “Opposing FGF and retinoid pathways control ventral neural pattern, neuronal differentiation, and segmentation during body axis extension.” Neuron. 2003 Sep 25;40(1):65-79 Diez del Corral R, Storey KG. “Opposing FGF and retinoid pathways: a signaling switch that controls differentiation and patterning onset in the extending vertebrate body axis.” Bioessays. 2004 Aug;26(8):857-69. Ding YQ, Yin J, Kania A, Zhao ZQ, Johnson RL, Chen ZF. “Lmx1b controls the differentiation and migration of the superficial dorsal horn neurons of the spinal cord.” Development. 2004 Aug;131(15):3693-703. Durston AJ, Jansen HJ, In der Rieden P, Hooiveld MH. “Hox collinearity - a new perspective.” Int J Dev Biol. 2011;55(10-12):899-908. Dykes IM, Tempest L, Lee SI, Turner EE. “Brn3a and Islet1 act epistatically to regulate the gene expression program of sensory differentiation.” J Neurosci. 2011 Jul 6;31(27):9789-99.

217

Eccles, JC, Fatt, P, Koketsu, K. “Cholinergic and inhibitory synapses in a pathway from motor-axon collaterals to motoneurones.” J Physiol. 1954 Dec 10;126(3):524-62. Eccles, JC, Fatt P, Landgren S. “The inhibitory pathway to motoneurones.” Prog Neurobiol. 1956;(2):72-82. Eccles JC, Eccles RM, Lundberg A. “The convergence of monosynaptic excitatory afferents on to many different species of alpha motoneurones.” J Physiol. 1957 Jun 18;137(1):22-50. Eccles JC, Eccles RM, Iggo A, Lundberg A. “Electrophysiological investigations on Renshaw cells.” J Physiol. 1961 Dec;159:461-78. Eckert D, Buhl S, Weber S, Jäger R, Schorle H. “The AP-2 family of transcription factors.” Genome Biol. 2005;6(13):246. Ernfors P, Lee KF, Kucera J, Jaenisch R. “Lack of neurotrophin-3 leads to deficiencies in the peripheral nervous system and loss of limb proprioceptive afferents.” Cell. 1994 May 20;77(4):503-12 Ericson J, Morton S, Kawakami A, Roelink H, Jessell TM. “Two critical periods of Sonic Hedgehog signaling required for the specification of motor neuron identity.” Cell. 1996 Nov 15;87(4):661-73. Ericson J, Rashbass P, Schedl A, Brenner-Morton S, Kawakami A, van Heyningen V, Jessell TM, Briscoe J. “Pax6 controls progenitor cell identity and neuronal fate in response to graded Shh signaling.” Cell. 1997 Jul 11;90(1):169-80. Espinosa JS, Luo, L. “Timing Neurogenesis and Differentiation: Insights from Quantitative Clonal Analyses of Cerebellar Granule Cells.” J Neurosci. 2008 Mar 5; 28(10): 2301–2312. Fandel TM, Trivedi A, Nicholas CR, Zhang H, Chen J, Martinez AF, Noble-Haeusslein LJ, Kriegstein AR. “Transplanted Human Stem Cell-Derived Interneuron Precursors Mitigate Mouse Bladder Dysfunction and Central Neuropathic Pain after Spinal Cord Injury.” Cell Stem Cell. 2016 Oct 6;19(4):544-557. Feldman AG, Orlovsky GN. “Activity of interneurons mediating reciprocal 1a inhibition during locomotion.” Brain Res. 1975 Feb 7;84(2):181-94. Feng X, Ippolito GC, Tian L, Wiehagen K, Oh S, Sambandam A, Willen J, Bunte RM, Maika SD, Harriss JV, Caton AJ, Bhandoola A, Tucker PW, Hu H. “Foxp1 is an essential transcriptional regulator for the generation of quiescent naive T cells during thymocyte development.” Blood. 2010 Jan 21;115(3):510-8. Fishell G. “Perspectives on the developmental origins of cortical interneuron diversity.” Novartis Found Symp. 2007;288:21-35; discussion 35-44, 96-8. Fink AJ, Croce KR, Huang ZJ, Abbott LF, Jessell TM, Azim E. “Presynaptic inhibition of spinal sensory feedback ensures smooth movement.” Nature. 2014 May 1;509(7498):43-8.

Floyd TL, Ladle DR. “Characterization of Calbindin Positive Interneurons within the Ventral Horn of the Mouse Spinal Cord.” 2015: http://corescholar.libraries.wright.edu/cgi/viewcontent.cgi?article=1010&context=urop_celebration.

Fornaro M, Lee JM, Raimondo S, Nicolino S, Geuna S, Giacobini-Robecchi M. “Neuronal intermediate filament expression in rat dorsal root ganglia sensory neurons: an in vivo and in vitro study.” Neuroscience. 2008 Jun 2;153(4):1153-63. Foucher I, Mione M, Simeone A, Acampora D, Bally-Cuif L, Houart C. “Differentiation of cerebellar cell identities in absence of Fgf signalling in zebrafish Otx morphants.” Development. 2006

218

May;133(10):1891-900. Francius C, Harris A, Rucchin V, Hendricks TJ, Stam FJ, Barber M, Kurek D, Grosveld FG, Pierani A, Goulding M, Clotman F. “Identification of multiple subsets of ventral interneurons and differential distribution along the rostrocaudal axis of the developing spinal cord.” PLoS One. 2013 Aug 15;8(8):e70325. Frank F, Fuortes M. “Presynaptic and postsynaptic inhibition of monosynaptic reflexes.” Fed Proc 1957:19, 39-40. Frank E. “The formation of specific synaptic connections between muscle sensory and motor neurons in the absence of coordinated patterns of muscle activity.” J Neurosci. 1990 Jul;10(7):2250-60. Freeman MR. “Specification and morphogenesis of astrocytes.” Science. 2010 Nov 5;330(6005):774-8. Fritschy JM, Panzanelli P, Kralic JE, Vogt KE, Sassoè-Pognetto M. “Differential dependence of axo-dendritic and axo-somatic GABAergic synapses on GABAA receptors containing the alpha1 subunit in Purkinje cells.” J Neurosci. 2006 Mar 22;26(12):3245-55. Fujitani Y, Fujitani S, Luo H, Qiu F, Burlison J, Long Q, Kawaguchi Y, Edlund H, MacDonald RJ, Furukawa T, Fujikado T, Magnuson MA, Xiang M, Wright CV. “Ptf1a determines horizontal and amacrine cell fates during mouse retinal development.” Development. 2006 Nov;133(22):4439-50. Furukawa T, Mukherjee S, Bao ZZ, Morrow EM, Cepko CL. “rax, Hes1, and notch1 promote the formation of Müller glia by postnatal retinal progenitor cells.” Neuron. 2000 May;26(2):383-94. Fujiyama T, Yamada M, Terao M, Terashima T, Hioki H, Inoue YU, Inoue T, Masuyama N, Obata K, Yanagawa Y, Kawaguchi Y, Nabeshima Y, Hoshino M. “Inhibitory and excitatory subtypes of cochlear nucleus neurons are defined by distinct bHLH transcription factors, Ptf1a and Atoh1.” Development. 2009 Jun;136(12):2049-58. Fyffe RE. “Evidence for separate morphological classes of Renshaw cells in the cat's spinal cord.” Brain Res. 1990 Dec 17;536(1-2):301-4. Fyffe RE. “Spatial distribution of recurrent inhibitory synapses on spinal motoneurons in the cat.” J Neurophysiol. 1991 May;65(5):1134-49. Gabitto MI, Pakman A, Bikoff JB, Abbott LF, Jessell TM, Paninski L. “Bayesian sparse regression analysis documents the diversity of spinal inhibitory interneurons.” Cell. 2016 Mar 24;165(1):220-33. Gao WQ, Hatten ME. “Immortalizing oncogenes subvert the establishment of granule cell identity in developing cerebellum.” Development. 1994 May;120(5):1059-70. Gaiano N, Nye JS, Fishell G. “Radial glial identity is promoted by Notch1 signaling in the murine forebrain.” Neuron. 2000 May;26(2):395-404. Garcia-Dominguez M, Poquet C, Garel S, Charnay P. “Ebf gene function is required for coupling neuronal differentiation and cell cycle exit.” Development. 2003 Dec;130(24):6013-25. Garel S, Marín F, Grosschedl R, Charnay P. “Ebf1 controls early cell differentiation in the embryonic striatum.” Development. 1999 Dec;126(23):5285-94. Garel S, Garcia-Dominguez M, Charnay P. “Control of the migratory pathway of facial branchiomotor neurones.” Development. 2000 Dec;127(24):5297-307.

219

Garel S, Yun K, Grosschedl R, Rubenstein JL. “The early topography of thalamocortical projections is shifted in Ebf1 and Dlx1/2 mutant mice.” Development. 2002 Dec;129(24):5621-34. Gaspard N, Bouschet T, Hourez R, Dimidschstein J, Naeije G, van den Ameele J, Espuny-Camacho I, Herpoel A, Passante L, Schiffmann SN, Gaillard A, Vanderhaeghen P. “An intrinsic mechanism of corticogenesis from embryonic stem cells.” Nature. 2008 Sep 18;455(7211):351-7. Gaspard N, Bouschet T, Herpoel A, Naeije G, van den Ameele J, Vanderhaeghen P. “Generation of cortical neurons from mouse embryonic stem cells.” Nat Protoc. 2009;4(10):1454-63. Gaspard N, Vanderhaeghen P. “From stem cells to neural networks: recent advances and perspectives for neurodevelopmental disorders.” Dev Med Child Neurol. 2011 Jan;53(1):13-7. Gazave E, Lapébie P, Richards GS, Brunet F, Ereskovsky AV, Degnan BM, Borchiellini C, Vervoort M, Renard E. “Origin and evolution of the Notch signalling pathway: an overview from eukaryotic genomes.” BMC Evol Biol. 2009 Oct 13;9:249 Geiman EJ, Knox MC, Alvarez FJ. “Postnatal maturation of gephyrin/glycine receptor clusters on developing Renshaw cells.” J Comp Neurol. 2000 Oct 9;426(1):130-42. Geiman EJ, Zheng W, Fritschy JM, Alvarez FJ. “Glycine and GABA(A) receptor subunits on Renshaw cells: relationship with presynaptic neurotransmitters and postsynaptic gephyrin clusters.” J Comp Neurol. 2002 Mar 12;444(3):275-89. Geling A, Steiner H, Willem M, Bally-Cuif, Haass C. “A gamma-secretase inhibitor blocks Notch signaling in vivo and causes a severe neurogenic phenotype in zebrafish.” EMBO Rep. 2002 Jul;3(7):688-94.

Glasgow SM, Henke RM, Macdonald RJ, Wright CV, Johnson JE. “Ptf1a determines GABAergic over glutamatergic neuronal cell fate in the spinal cord dorsal horn.” Development. 2005 Dec;132(24):5461-9.

Gómez-Skarmeta JL, Campuzano S, Modolell J. “Half a century of neural prepatterning: the story of a few bristles and many genes.” Nat Rev Neurosci. 2003 Jul;4(7):587-98.

Gonzalez-Forero D, Pastor AM, Geiman EJ, Benítez-Temiño B, Alvarez FJ. “Regulation of gephyrin cluster size and inhibitory synaptic currents on Renshaw cells by motor axon excitatory inputs.” J Neurosci. 2005 Jan 12;25(2):417-29. Gosgnach S, Lanuza GM, Butt SJ, Saueressig H, Zhang Y, Velasquez T, Riethmacher D, Callaway EM, Kiehn O, Goulding M. “V1 spinal neurons regulate the speed of vertebrate locomotor output.” Nature. 2006 Mar 9;440(7081):215-9.

Gottlieb DI, Huettner JE. “An in vitro pathway from embryonic stem cells to neurons and glia.” Cells Tissues Organs. 1999;165(3-4):165-72.

Goulburn AL, Stanley EG, Elefanty AG, Anderson SA. “Generating GABAergic cerebral cortical interneurons from mouse and human embryonic stem cells.” Stem Cell Res. 2012 May;8(3):416-26.

Goulding, M. “Circuits controlling vertebrate locomotion: moving in a new direction.” Nat Rev Neurosci. 2009 Jul;10(7):507-18. Gouti M, Tsakiridis A, Wymeersch FJ, Huang Y, Kleinjung J, Wilson V, Briscoe J. “In vitro generation of neuromesodermal progenitors reveals distinct roles for wnt signalling in the specification of spinal cord and paraxial mesoderm identity.” PLoS Biol. 2014 Aug 26;12(8):e1001937. Gouti M, Metzis V, Briscoe J. “The route to spinal cord cell types: a tale of signals and switches.” Trends Genet. 2015 Jun;31(6):282-9.

220

Gowan K, Helms AW, Hunsaker TL, Collisson T, Ebert PJ, Odom R, Johnson JE. “Crossinhibitory activities of Ngn1 and Math1 allow specification of distinct dorsal interneurons.” Neuron. 2001 Aug 2;31(2):219-32.

Gray EG. “A morphological basis for pre-synaptic inhibition?” Nature. 1962 Jan 6;193:82-3.

Green YS, Vetter ML. “EBF factors drive expression of multiple classes of target genes governing neuronal development.” Neural Dev. 2011 Apr 30;6:19.

Gregori N, Pröschel C, Noble M, Mayer-Pröschel M. “The tripotential glial-restricted precursor (GRP) cell and glial development in the spinal cord: generation of bipotential oligodendrocyte-type-2 astrocyte progenitor cells and dorsal-ventral differences in GRP cell function.” J Neurosci. 2002 Jan 1;22(1):248-56. Greig LC, Woodworth MB, Galazo MJ, Padmanabhan H, Macklis JD. “Molecular logic of neocortical projection neuron specification, development and diversity.” Nat Rev Neurosci. 2013 Nov;14(11):755-69. Griener A, Zhang W, Kao H, Wagner C, Gosgnach S. “Probing diversity within subpopulations of locomotor-related V0 interneurons.” Dev Neurobiol. 2015 Nov;75(11):1189-203. Grillner S, Jessell TM. “Measured motion: searching for simplicity in spinal locomotor networks.” Curr Opin Neurobiol. 2009 Dec;19(6):572-86 Gross MK, Dottori M, Goulding M. “Lbx1 specifies somatosensory association interneurons in the dorsal spinal cord.” Neuron. 2002 May 16;34(4):535-49. Grosskortenhaus R, Pearson BJ, Marusich A, Doe CQ. “Regulation of temporal identity transitions in Drosophila neuroblasts.” Dev Cell. 2005 Feb;8(2):193-202. Greenwald I, Rubin GM. “Making a difference: the role of cell-cell interactions in establishing separate identities for equivalent cells.” Cell. 1992 Jan 24;68(2):271-81. Greenwald I. “LIN-12/Notch signaling: lessons from worms and flies.” Genes Dev. 1998 Jun 15;12(12):1751-62. Hahne M, Illert M, Wietelmann D. “Recurrent inhibition in the cat distal forelimb.” Brain Res. 1988 Jul 19;456(1):188-92. Hamm TM, Sasaki S, Stuart DG, Windhorst U, Yuan CS. “Distribution of single-axon recurrent inhibitory post-synaptic potentials in a single spinal motor nucleus in the cat.” J Physiol. 1987 Jul;388:653-64. Hanley O, Zewdu R, Cohen LJ, Jung H, Lacombe J, Philippidou P, Lee DH, Selleri L, Dasen JS. “Parallel Pbx-Dependent Pathways Govern the Coalescence and Fate of Motor Columns.” Neuron. 2016 Sep 7;91(5):1005-20. Hatten ME, Alder J, Zimmerman K, Heintz N. “Genes involved in cerebellar cell specification and differentiation.” Curr Opin Neurobiol. 1997 Feb;7(1):40-7. Haase G, Dessaud E, Garcès A, de Bovis B, Birling M, Filippi P, Schmalbruch H, Arber S, deLapeyrière O. “GDNF acts through PEA3 to regulate cell body positioning and muscle innervation of specific motor neuron pools.” Neuron. 2002 Aug 29;35(5):893-905. Heitzler P, Simpson P.” Altered epidermal growth factor-like sequences provide evidence for a role of Notch as a receptor in cell fate decisions.” Development. 1993 Mar;117(3):1113-23.

221

Heitzler P, Bourouis M, Ruel L, Carteret C, Simpson P. “Genes of the Enhancer of split and achaete-scute complexes are required for a regulatory loop between Notch and Delta during lateral signalling in Drosophila.” Development. 1996 Jan;122(1):161-71.

Helms AW, Johnson JE. “Specification of dorsal spinal cord interneurons.” Curr Opin Neurobiol. 2003 Feb;13(1):42-9.

Helms AW, Battiste J, Henke RM, Nakada Y, Simplicio N, Guillemot F, Johnson JE. “Sequential roles for Mash1 and Ngn2 in the generation of dorsal spinal cord interneurons.” Development. 2005 Jun;132(12):2709-19. Hemmati-Brivanlou A, Melton DA. “A truncated activin receptor inhibits mesoderm induction and formation of axial structures in Xenopus embryos.” Nature. 1992 Oct 15;359(6396):609-14. Hemmati-Brivanlou A, Melton DA. “Inhibition of activin receptor signaling promotes neuralization in Xenopus.” Cell. 1994 Apr 22;77(2):273-81. Hemmati-Brivanlou A, Kelly OG, Melton DA. “Follistatin, an antagonist of activin, is expressed in the Spemann organizer and displays direct neuralizing activity.” Cell. 1994 Apr 22;77(2):283-95. Henke RM, Savage TK, Meredith DM, Glasgow SM, Hori K, Dumas J, MacDonald RJ, Johnson JE. Neurog2 is a direct downstream target of the Ptf1a-Rbpj transcription complex in dorsal spinal cord.” Development. 2009 Sep;136(17):2945-54. Hidalgo-Sánchez M, Simeone A, Alvarado-Mallart RM. “Fgf8 and Gbx2 induction concomitant with Otx2 repression is correlated with midbrain-hindbrain fate of caudal prosencephalon.” Development. 1999 Jun;126(14):3191-203. Higashijima S, Masino MA, Mandel G, Fetcho JR. “Engrailed-1 expression marks a primitive class of inhibitory spinal interneuron.” J Neurosci. 2004 Jun 23;24(25):5827-39. Hilde KL, Levine AJ, Hinckley CA, Hayashi M, Montgomery JM, Gullo M, Driscoll SP, Grosschedl R, Kohwi Y, Kohwi-Shigematsu T, Pfaff SL. “Satb2 Is Required for the Development of a Spinal Exteroceptive Microcircuit that Modulates Limb Position.” Neuron. 2016 Aug 17;91(4):763-76. Hilger-Eversheim K, Moser M, Schorle H, Buettner R. “Regulatory roles of AP-2 transcription factors in vertebrate development, apoptosis and cell-cycle control.” Gene. 2000 Dec 30;260(1-2):1-12. Hippenmeyer S, Vrieseling E, Sigrist M, Portmann T, Laengle C, Ladle DR, Arber S. “A developmental switch in the response of DRG neurons to ETS transcription factor signaling.” PLoS Biol. 2005 May;3(5):e159. Hirabayashi Y, Itoh Y, Tabata H, Nakajima K, Akiyama T, Masuyama N, Gotoh Y. “The Wnt/beta-catenin pathway directs neuronal differentiation of cortical neural precursor cells.” Development. 2004 Jun;131(12):2791-801. Ho R, Sances S, Gowing G, Amoroso MW, O'Rourke JG, Sahabian A, Wichterle H, Baloh RH, Sareen D, Svendsen CN. “ALS disrupts spinal motor neuron maturation and aging pathways within gene co-expression networks.” Nat Neurosci. 2016 Sep;19(9):1256-67. Hochgreb T, Linhares VL, Menezes DC, Sampaio AC, Yan CYI, Cardoso WV, Rosenthal N, Xavier-Neto J. “A caudorostral wave of RALDH2 conveys anteroposterior information to the cardiac field.” Development 2003 130: 5363-5374. Hochstim C, Deneen B, Lukaszewicz A, Zhou Q, Anderson DJ. “Identification of positionally distinct

222

astrocyte subtypes whose identities are specified by a homeodomain code.” Cell. 2008 May 2;133(3):510-22. Hogan BL. “Bone morphogenetic proteins in development.” Curr Opin Genet Dev. 1996 Aug;6(4):432-8. Hol EM, Pekny M. “Glial fibrillary acidic protein (GFAP) and the astrocyte intermediate filament system in diseases of the central nervous system.” Curr Opin Cell Biol. 2015 Feb;32:121-30. Hollyday M, Hamburger V. “An autoradiographic study of the formation of the lateral motor column in the chick embryo.” Brain Res. 1977 Aug 26;132(2):197-208. Hollyday M, Jacobson RD. “Location of motor pools innervating chick wing.” J Comp Neurol. 1990 Dec 15;302(3):575-88. Hori K, Cholewa-Waclaw J, Nakada Y, Glasgow SM, Masui T, Henke RM, Wildner H, Martarelli B, Beres TM, Epstein JA, Magnuson MA, Macdonald RJ, Birchmeier C, Johnson JE. “A nonclassical bHLH Rbpj transcription factor complex is required for specification of GABAergic neurons independent of Notch signaling.” Genes Dev. 2008 Jan 15;22(2):166-78. Hori K, Hoshino M. “GABAergic neuron specification in the spinal cord, the cerebellum, and the cochlear nucleus.” Neural Plast. 2012;2012:921732. Hori K, Sen A, Artavanis-Tsakonas S. “Notch signaling at a glance.” J Cell Sci. 2013 May 15;126(Pt 10):2135-40.

Hossaini M, Cardona Cano S, van Dis V, Haasdijk ED, Hoogenraad CC, Holstege JC, Jaarsma D. “Spinal inhibitory interneuron pathology follows motor neuron degeneration independent of glial mutant superoxide dismutase 1 expression in SOD1-ALS mice.” J Neuropathol Exp Neurol. 2011 Aug;70(8):662-77. Hoshino M, Nakamura S, Mori K, Kawauchi T, Terao M, Nishimura YV, Fukuda A, Fuse T, Matsuo N, Sone M, Watanabe M, Bito H, Terashima T, Wright CV, Kawaguchi Y, Nakao K, Nabeshima Y. “Ptf1a, a bHLH transcriptional gene, defines GABAergic neuronal fates in cerebellum.” Neuron. 2005 Jul 21;47(2):201-13. Hu J, Huang T, Li T, Guo Z, Cheng L. “c-Maf is required for the development of dorsal horn laminae III/IV neurons and mechanoreceptive DRG axon projections.” J Neurosci. 2012 Apr 18;32(16):5362-73. Huang A, Noga BR, Carr PA, Fedirchuk B, Jordan LM. “Spinal cholinergic neurons activated during locomotion: localization and electrophysiological characterization.” J Neurophysiol. 2000 Jun;83(6):3537-47. Huang C, Chan JA, Schuurmans C. “Proneural bHLH genes in development and disease.” Curr Top Dev Biol. 2014;110:75-127. Huber AB, Kania A, Tran TS, Gu C, De Marco Garcia N, Lieberam I, Johnson D, Jessell TM, Ginty DD, Kolodkin AL. “Distinct roles for secreted semaphorin signaling in spinal motor axon guidance.” Neuron. 2005 Dec 22;48(6):949-64. Hughes DI, Mackie M, Nagy GG, Riddell JS, Maxwell DJ, Szabó G, Erdélyi F, Veress G, Szucs P, Antal M, Todd AJ. “P boutons in lamina IX of the rodent spinal cord express high levels of glutamic acid decarboxylase-65 and originate from cells in deep medial dorsal horn.” Proc Natl Acad Sci U S A. 2005 Jun 21;102(25):9038-43. Hultborn H, Lindstrom S, Wigstrom H. “On the function of recurrent inhibition in the spinal cord.” Exp Brain Res. 1979: 37:399-403.

223

Hultborn H, Pierrot-Deseilligny E. “Input-output relations in the pathway of recurrent inhibition to motoneurones in the cat.” J Physiol. 1979 Dec;297(0):267-87.

Hultborn H, Katz R, Mackel R. “Distribution of recurrent inhibition within a motor nucleus. II. Amount of recurrent inhibition in motoneurones to fast and slow units.” Acta Physiol Scand. 1988a:134:363-374.

Hultborn H, Lipski J, Mackel R, Wigstrom H. “Distribution of recurrent inhibition within a motor nucleus. I. Contribution from slow and fast motor units to the excitation of Renshaw cells.” Acta Physiol Scand. 1988b:134:347-361.

Hultborn, H. “Spinal reflexes, mechanisms and concepts: from Eccles to Lundberg and beyond.” Prog Neurobiol. 2006 Feb-Apr;78(3-5):215-32.

Iacovino M, Bosnakovski D, Fey H, Rux D, Bajwa G, Mahen E, Mitanoska A, Xu Z, Kyba M. “Inducible cassette exchange: a rapid and efficient system enabling conditional gene expression in embryonic stem and primary cells.” Stem Cells. 2011 Oct;29(10):1580-8. Ideguchi M, Palmer TD, Recht LD, Weimann JM. “Murine embryonic stem cell-derived pyramidal neurons integrate into the cerebral cortex and appropriately project axons to subcortical targets.” J Neurosci. 2010 Jan 20;30(3):894-904. Illert M, Kümmel H. “Reflex pathways from large muscle spindle afferents and recurrent axon collaterals to motoneurones of wrist and digit muscles: a comparison in cats, monkeys and humans.” Exp Brain Res. 1999 Sep;128(1-2):13-9. Imamura F, Ayoub AE, Rakic P, Greer CA. “Timing of neurogenesis is a determinant of olfactory circuitry.” Nat Neurosci. 2011 Mar;14(3):331-7. Imayoshi I, Sakamoto M, Yamaguchi M, Mori K, Kageyama R. “Essential roles of Notch signaling in maintenance of neural stem cells in developing and adult brains.” J Neurosci. 2010 Mar 3;30(9):3489-98. Imlach WL, Beck ES, Choi BJ, Lotti F, Pellizzoni L, McCabe BD. “SMN is required for sensory-motor circuit function in Drosophila.” Cell. 2012 Oct 12;151(2):427-39. Inaki M, Yoshikawa S, Thomas JB, Aburatani H, Nose A. “Wnt4 is a local repulsive cue that determines synaptic target specificity.” Curr Biol. 2007 Sep 18;17(18):1574-9.

Inoue K, Ozaki S, Shiga T, Ito K, Masuda T, Okado N, Iseda T, Kawaguchi S, Ogawa M, Bae SC, Yamashita N, Itohara S, Kudo N, Ito Y. “Runx3 controls the axonal projection of proprioceptive dorsal root ganglion neurons.” Nat Neurosci. 2002 Oct;5(10):946-54

Inoue K, Ito K, Osato M, Lee B, Bae SC, Ito Y. “The transcription factor Runx3 represses the neurotrophin receptor TrkB during lineage commitment of dorsal root ganglion neurons.” J Biol Chem. 2007 Aug 17;282(33):24175-84.

Ishii K, Wong JK, Sumikawa K. “Comparison of alpha2 nicotinic acetylcholine receptor subunit mRNA expression in the central nervous system of rats and mice.” J Comp Neurol. 2005 Dec 12;493(2):241-60. Isshiki T, Pearson B, Holbrook S, Doe CQ. “Drosophila neuroblasts sequentially express transcription factors which specify the temporal identity of their neuronal progeny.” Cell. 2001 Aug 24;106(4):511-21. Ito S, Takeichi M. “Dendrites of cerebellar granule cells correctly recognize their target axons for synaptogenesis in vitro.” Proc Natl Acad Sci U S A. 2009 Aug 4;106(31):12782-7.

224

Iyer NR, Huettner JE, Butts JC, Brown CR, Sakiyama-Elbert SE. “Generation of highly enriched V2a interneurons from mouse embryonic stem cells.” Exp Neurol. 2016 Mar;277:305-16. Jacob J, Briscoe J. “Gli proteins and the control of spinal-cord patterning.” EMBO Rep. 2003 Aug;4(8):761-5. Jacobson M. “Development of specific neuronal connections.” Science. 1969 Feb 7;163(3867):543-7. Jadhav AP, Mason HA, Cepko CL. “Notch 1 inhibits photoreceptor production in the developing mammalian retina.” Development. 2006 Mar;133(5):913-23. Jadhav AP, Cho SH, Cepko CL. “Notch activity permits retinal cells to progress through multiple progenitor states and acquire a stem cell property.” Proc Natl Acad Sci U S A. 2006 Dec 12;103(50):18998-9003. Jankovski A, Rossi F, Sotelo C. “Neuronal precursors in the postnatal mouse cerebellum are fully committed cells: evidence from heterochronic transplantations.” Eur J Neurosci. 1996 Nov;8(11):2308-19. Jankowska E, Smith DO. “Antidromic activation of Renshaw cells and their axonal projections.” Acta Physiol Scand. 1973 Jun;88(2):198-214. Jankowska E. “Interneuronal relay in spinal pathways from proprioceptors.” Prog Neurobiol. 1992;38(4):335-78. Jankowska E, Puczynska A. “Interneuronal activity in reflex pathways from group II muscle afferents is monitored by dorsal spinocerebellar tract neurons in the cat.”J Neurosci. 2008 Apr 2;28(14):3615-22. Jean D, Bernier G, Gruss P. “Six6 (Optx2) is a novel murine Six3-related homeobox gene that demarcates the presumptive pituitary/hypothalamic axis and the ventral optic stalk.” Mech Dev. 1999 Jun;84(1-2):31-40. Jessell TM. “Neuronal specification in the spinal cord: inductive signals and transcriptional codes.” Nat Rev Genet. 2000 Oct;1(1):20-9.

Jin K, Jiang H, Xiao D, Zou M, Zhu J, Xiang M. “Tfap2a and 2b act downstream of Ptf1a to promote amacrine cell differentiation during retinogenesis.” Mol Brain. 2015 May 13;8:28.

John A, Brylka H, Wiegreffe C, Simon R, Liu P, Jüttner R, Crenshaw EB 3rd, Luyten FP, Jenkins NA, Copeland NG, Birchmeier C, Britsch S. “Bcl11a is required for neuronal morphogenesis and sensory circuit formation in dorsal spinal cord development.” Development. 2012 May;139(10):1831-41. Johnson MA, Weick JP, Pearce RA, Zhang SC. “Functional neural development from human embryonic stem cells: accelerated synaptic activity via astrocyte coculture.” J Neurosci. 2007 Mar 21;27(12):3069-77. Jonas P, Bischofberger J, Sandkühler J. “Corelease of two fast neurotransmitters at a central synapse.” Science. 1998 Jul 17;281(5375):419-24. Jovanovic K, Pastor AM, O'Donovan MJ. “The use of PRV-Bartha to define premotor inputs to lumbar motoneurons in the neonatal spinal cord of the mouse.” PLoS One. 2010 Jul 23;5(7):e11743. Jung H, Lacombe J, Mazzoni EO, Liem KF Jr, Grinstein J, Mahony S, Mukhopadhyay D, Gifford DK, Young RA, Anderson KV, Wichterle H, Dasen JS. “Global control of motor neuron topography mediated by the repressive actions of a single hox gene.” Neuron. 2010 Sep 9;67(5):781-96.

225

Kania, A., Johnson, R.L. and Jessell, T.M. (2000). Coordinate roles for LIM homeobox genes in directing the dorsoventral trajectory of motor axons in the vertebrate limb. Cell. 102, 161-173. Kania A, Jessell TM. “Topographic motor projections in the limb imposed by LIM homeodomain protein regulation of ephrin-A:EphA interactions.” Neuron. 2003 May 22;38(4):581-96. Kanning KC, Kaplan A, Henderson CE. “Motor neuron diversity in development and disease.” Annu Rev Neurosci. 2010;33:409-40. Kardon AP, Polgár E, Hachisuka J, Snyder LM, Cameron D, Savage S, Cai X, Karnup S, Fan CR, Hemenway GM, Bernard CS, Schwartz ES, Nagase H, Schwarzer C, Watanabe M, Furuta T, Kaneko T, Koerber HR, Todd AJ, Ross SE. “Dynorphin acts as a neuromodulator to inhibit itch in the dorsal horn of the spinal cord.” Neuron. 2014 May 7;82(3):573-86. Kato M, Dobyns WB. “X-linked lissencephaly with abnormal genitalia as a tangential migration disorder causing intractable epilepsy: Proposal for a new term, “interneuronopathy” J Child Neurol. 2005;20:392–397.” Katz LC, Shatz CJ. “Synaptic activity and the construction of cortical circuits.” Science. 1996 Nov 15;274(5290):1133-8. Kaus A, Sareen D. “ALS Patient Stem Cells for Unveiling Disease Signatures of Motoneuron Susceptibility: Perspectives on the Deadly Mitochondria, ER Stress and Calcium Triad.” Front Cell Neurosci. 2015 Nov 19;9:448. Kawaguchi Y, Cooper B, Gannon M, Ray M, MacDonald RJ, Wright CV. “The role of the transcriptional regulator Ptf1a in converting intestinal to pancreatic progenitors.” Nat Genet. 2002 Sep;32(1):128-34. Kessaris N, Pringle N, Richardson WD. “Ventral neurogenesis and the neuron-glial switch. Neuron. 2001 Sep 13;31(5):677-80.

Kiecker C, Niehrs C.“A morphogen gradient of Wnt/beta-catenin signalling regulates anteroposterior neural patterning in Xenopus.” Development. 2001 Nov;128(21):4189-201. Kiehn O. “Locomotor circuits in the mammalian spinal cord.” Annu Rev Neurosci. 2006;29:279-306. Kiehn O. “Decoding the organization of spinal circuits that control locomotion.” Nat Rev Neurosci. 2016 Apr;17(4):224-38. Kiernan MC, Vucic S, Cheah BC, Turner MR, Eisen A, Hardiman O, Burrell JR, Zoing MC. “Amyotrophic lateral sclerosis.” Lancet. 2011 Mar 12;377(9769):942-55. Kim M, Habiba A, Doherty JM, Mills JC, Mercer RW, Huettner JE. “Regulation of mouse embryonic stem cell neural differentiation by retinoic acid.” Dev Biol. 2009 Apr 15;328(2):456-71. Kimura Y, Satou C, Higashijima S. “V2a and V2b neurons are generated by the final divisions of pair-producing progenitors in the zebrafish spinal cord.” Development. 2008 Sep;135(18):3001-5. Kimble J, Simpson P. “The LIN-12/Notch signaling pathway and its regulation.” Annu Rev Cell Dev Biol. 1997;13:333-61. Kimmel RA, Turnbull DH, Blanquet V, Wurst W, Loomis CA, Joyner AL. “Two lineage boundaries coordinate vertebrate apical ectodermal ridge formation.” Genes Dev. 2000;14, 1377-1389.

226

Kinameri E, Inoue T, Aruga J, Imayoshi I, Kageyama R, Shimogori T, Moore AW. “Prdm proto-oncogene transcription factor family expression and interaction with the Notch-Hes pathway in mouse neurogenesis.” PLoS One. 2008;3(12):e3859. Kingsley DM. “The TGF-beta superfamily: new members, new receptors, and new genetic tests of function in different organisms.” Genes Dev. 1994 Jan;8(2):133-46. Kmita M, Duboule D. “Organizing axes in time and space; 25 years of colinear tinkering.” Science. 2003 Jul 18;301(5631):331-3. Kohl A, Marquardt T, Sela-Donenfeld D. “Control of axon guidance and neurotransmitter phenotype of dB1 hindbrain interneurons by Lim-HD code.” J Neurosci. 2015 Feb 11;35(6):2596-611. Kramer ER, Knott L, Su F, Dessaud E, Krull CE, Helmbacher F, Klein R. “Cooperation between GDNF/Ret and ephrinA/EphA4 signals for motor-axon pathway selection in the limb.” Neuron. 2006 Apr 6;50(1):35-47. Kriks S, Lanuza GM, Mizuguchi R, Nakafuku M, Goulding M. “Gsh2 is required for the repression of Ngn1 and specification of dorsal interneuron fate in the spinal cord.” Development. 2005 Jul;132(13):2991-3002.

Kubo A, Shinozaki K, Shannon JM, Kouskoff V, Kennedy M, Woo S, Fehling HJ, Keller G. “Development of definitive endoderm from embryonic stem cells in culture.” Development. 2004 Apr;131(7):1651-62.

Kurusu M, Cording A, Taniguchi M, Menon K, Suzuki E, Zinn K. “A screen of cell-surface molecules identifies leucine-rich repeat proteins as key mediators of synaptic target selection.” Neuron. 2008 Sep 25;59(6):972-85.

Kyba M, Perlingeiro RC, Daley GQ. “HoxB4 confers definitive lymphoid-myeloid engraftment potential on embryonic stem cell and yolk sac hematopoietic progenitors.” Cell. 2002 Apr 5;109(1):29-37.

Lagerbäck PA, Ronnevi LO, Cullheim S, Kellerth JO. “An ultrastructural study of the synaptic contacts of alpha-motoneurone axon collaterals. I. Contacts in lamina IX and with identified alpha-motoneurone dendrites in lamina VII.” Brain Res. 1981 Mar 2;207(2):247-66. Lagerbäck PA, Kellerth JO. “Light microscopic observations on cat Renshaw cells after intracellular staining with horseradish peroxidase. I. The axonal systems.” J Comp Neurol. 1985 Oct 22;240(4):359-67. Lai HC, Seal RP, Johnson JE. “Making sense out of spinal cord somatosensory development.” Development. 2016 Oct 1;143(19):3434-3448. Lamar E, Deblandre G, Wettstein D, Gawantka V, Pollet N, Niehrs C, Kintner C. “Nrarp is a novel intracellular component of the Notch signaling pathway.” Genes Dev. 2001 Aug 1;15(15):1885-99. Lamotte d'Incamps B, Ascher P. “Four excitatory postsynaptic ionotropic receptors coactivated at the motoneuron-Renshaw cell synapse.” J Neurosci. 2008 Dec 24;28(52):14121-31.

Lanier J, Dykes IM, Nissen S, Eng SR, Turner EE. “Brn3a regulates the transition from neurogenesis to terminal differentiation and represses non-neural gene expression in the trigeminal ganglion.” Dev Dyn. 2009 Dec;238(12):3065-79.

Lanuza GM, Gosgnach S, Pierani A, Jessell TM, Goulding M. “Genetic identification of spinal interneurons that coordinate left-right locomotor activity necessary for walking movements.” Neuron. 2004 May 13;42(3):375-86.

227

Lallemend F, Ernfors P. “Molecular interactions underlying the specification of sensory neurons.” Trends Neurosci. 2012 Jun;35(6):373-81

Lamb TM, Harland RM. “Fibroblast growth factor is a direct neural inducer, which combined with noggin generates anterior-posterior neural pattern.” Development. 1995 Nov;121(11):3627-36.

Landmesser L. “The distribution of motoneurones supplying chick hind limb muscles.” J Physiol. 1978 Nov;284:371-89.

Lanford PJ, Lan Y, Jiang R, Lindsell C, Weinmaster G, Gridley T, Kelley MW. “Notch signalling pathway mediates hair cell development in mammalian cochlea.” Nat Genet. 1999 Mar;21(3):289-92.

Lee KJ, Mendelsohn M, Jessell TM. “Neuronal patterning by BMPs: a requirement for GDF7 in the generation of a discrete class of commissural interneurons in the mouse spinal cord.” Genes Dev. 1998 Nov 1;12(21):3394-407.

Lee KJ, Jessell TM. “The specification of dorsal cell fates in the vertebrate central nervous system.” Annu Rev Neurosci. 1999;22:261-94.

Lee KJ, Dietrich P, Jessell TM. “Genetic ablation reveals that the roof plate is essential for dorsal interneuron specification.” Nature. 2000 Feb 17;403(6771):734-40.

Lee SK, Pfaff SL. “Transcriptional networks regulating neuronal identity in the developing spinal cord.” Nat Neurosci. 2001 Nov;4 Suppl:1183-91.

Lei Q, Zelman AK, Kuang E, Li S, Matise MP. “Transduction of graded Hedgehog signaling by a combination of Gli2 and Gli3 activator functions in the developing spinal cord.” Development. 2004 Aug;131(15):3593-604.

Leto K, Carletti B, Williams IM, Magrassi L, Rossi F. “Different types of cerebellar GABAergic interneurons originate from a common pool of multipotent progenitor cells.” J Neurosci. 2006 Nov 8;26(45):11682-94. Leto K, Bartolini A, Yanagawa Y, Obata K, Magrassi L, Schilling K, Rossi F. “Laminar fate and phenotype specification of cerebellar GABAergic interneurons.” J Neurosci. 2009 May 27;29(21):7079-91. Leto K, Rossi F. “Specification and differentiation of cerebellar GABAergic neurons.” Cerebellum. 2012 Jun;11(2):434-5. Levanon D, Bettoun D, Harris-Cerruti C, Woolf E, Negreanu V, Eilam R, Bernstein Y, Goldenberg D, Xiao C, Fliegauf M, Kremer E, Otto F, Brenner O, Lev-Tov A, Groner Y. “The Runx3 transcription factor regulates development and survival of TrkC dorsal root ganglia neurons.” EMBO J. 2002 Jul 1;21(13):3454-63. Levine AJ, Hinckley CA, Hilde KL, Driscoll SP, Poon TH, Montgomery JM, Pfaff SL. “Identification of a cellular node for motor control pathways.” Nat Neurosci. 2014 Apr;17(4):586-93. Li WC, Higashijima S, Parry DM, Roberts A, Soffe SR. “Primitive roles for inhibitory interneurons in developing frog spinal cord.” J Neurosci. 2004 Jun 23;24(25):5840-8. Li XJ, Du ZW, Zarnowska ED, Pankratz M, Hansen LO, Pearce RA, Zhang SC. “Specification of motoneurons from human embryonic stem cells.” Nat Biotechnol. 2005 Feb;23(2):215-21. Li XJ, Zhang X, Johnson MA, Wang ZB, Lavaute T, Zhang SC. “Coordination of sonic hedgehog and Wnt signaling determines ventral and dorsal telencephalic neuron types from human embryonic stem cells.” Development. 2009 Dec;136(23):4055-63.

228

Lichtman JW, Colman H. “Synapse elimination and indelible memory.” Neuron. 2000 Feb;25(2):269-78.

Liem KF Jr, Tremml G, Roelink H, Jessell TM. “Dorsal differentiation of neural plate cells induced by BMP-mediated signals from epidermal ectoderm.” Cell. 1995 Sep 22;82(6):969-79.

Liem KF Jr, Tremml G, Jessell TM. “A role for the roof plate and its resident TGFbeta-related proteins in neuronal patterning in the dorsal spinal cord.” Cell. 1997 Oct 3;91(1):127-38.

Lin JH, Saito T, Anderson DJ, Lance-Jones C, Jessell TM, Arber S. “Functionally related motor neuron pool and muscle sensory afferent subtypes defined by coordinate ETS gene expression.” Cell. 1998 Oct 30;95(3):393-407.

Ling KK, Lin MY, Zingg B, Feng Z, Ko CP. “Synaptic defects in the spinal and neuromuscular circuitry in a mouse model of spinal muscular atrophy.” PLoS One. 2010 Nov 11;5(11):e15457. Linhoff MW, Lauren J, Cassidy RM, Dobie FA, Takahashi H, Nygaard HB, Airaksinen MS, Strittmatter SM, Craig AM. “An unbiased expression screen for synaptogenic proteins identifies the LRRTM protein family as synaptic organizers.” Neuron. 2009;61:734–749. Lipkin SM, Grider TL, Heyman RA, Glass CK, Gage FH. “Constitutive retinoid receptors expressed from adenovirus vectors that specifically activate chromosomal target genes required for differentiation of promyelocytic leukemia and teratocarcinoma cells.” J Virol. 1996 Oct;70(10):7182-9. Lipski J, Fyffe RE, Jodkowski J. “Recurrent inhibition of cat phrenic motoneurons.” J Neurosci. 1985 Jun;5(6):1545-55. Liu P, Wakamiya M, Shea MJ, Albrecht U, Behringer RR, Bradley A. “Requirement for Wnt3 in vertebrate axis formation. Nat Genet. 1999 Aug;22(4):361-5. Liu JP, Laufer E, Jessell TM. “Assigning the positional identity of spinal motor neurons: rostrocaudal patterning of Hox-c expression by FGFs, Gdf11, and retinoids.” Neuron. 2001 Dec 20;32(6):997-1012. Liu P, Jenkins NA, Copeland NG. “A highly efficient recombineering-based method for generating conditional knockout mutations.” Genome Res. 2003 Mar;13(3):476-84. Liu A, Niswander LA. “Bone morphogenetic protein signalling and vertebrate nervous system development.” Nat Rev Neurosci. 2005 Dec;6(12):945-54. Liu JP. “The function of growth/differentiation factor 11 (Gdf11) in rostrocaudal patterning of the developing spinal cord.” Development. 2006 Aug;133(15):2865-74. Liu Y, Ma Q. “Generation of somatic sensory neuron diversity and implications on sensory coding. Curr Opin Neurobiol.” 2011 Feb;21(1):52-60. Liu Z, Hu X, Huang C, Zheng K, Takebayashi H, Cao C, Qiu M. “Olig3 is not involved in the ventral patterning of spinal cord.” PLoS One. 2014 Oct 28;9(10):e111076. Livesey FJ, Cepko CL. “Vertebrate neural cell-fate determination: lessons from the retina.” Nat Rev Neurosci. 2001 Feb; 2(2);109-18.

Livet J, Sigrist M, Stroebel S, De Paola V, Price SR, Henderson CE, Jessell TM, Arber S. “ETS gene Pea3 controls the central position and terminal arborization of specific motor neuron pools.” Neuron. 2002 Aug 29;35(5):877-92. Lodato S, Rouaux C, Quast KB, Jantrachotechatchawan C, Studer M, Hensch TK, Arlotta P. “Excitatory

229

projection neuron subtypes control the distribution of local inhibitory interneurons in the cerebral cortex.” Neuron. 2011 Feb 24;69(4):763-79. Louvi A, Artavanis-Tsakonas. “Notch signaling in vertebrate neural development.” Nat Rev Neurosci. 2006 Feb;7:93-102.

Lowrie, MB, Lawson, SJ. “Cell death of spinal interneurons.” Prog Neurobiol. 2000 Aug;61(6):543-55.

Lu QR, Sun T, Zhu Z, Ma N, Garcia M, Stiles CD, Rowitch DH. “Common developmental requirement for Olig function indicates a motor neuron/oligodendrocyte connection.” Cell. 2002 Apr 5;109(1):75-86. Lumsden A, Krumlauf R. “Patterning the vertebrate neuraxis.” Science. 1996 Nov 15;274(5290):1109-15. Lu DC, Niu T, Alaynick WA. “Molecular and cellular development of spinal cord locomotor circuitry.” Front Mol Neurosci. 2015 Jun 16;8:25. Lumsden A, Krumlauf R. “Patterning the vertebrate neuraxis.” Science. 1996 Nov 15;274(5290):1109-15. Lundell TG, Zhou Q, Doughty ML. “Neurogenin1 expression in cell lineages of the cerebellar cortex in embryonic and postnatal mice.” Dev Dyn. 2009 Dec;238(12):3310-25. Luo L, Callaway EM, Svoboda K. “Genetic Dissection of Neural Circuits.” Neuron. 2008 Mar 13; 57(5): 634–660. Ma Q, Sommer L, Cserjesi P, Anderson DJ. “Mash1 and neurogenin1 expression patterns define complementary domains of neuroepithelium in the developing CNS and are correlated with regions expressing notch ligands.” J Neurosci. 1997 May 15;17(10):3644-52. Ma Q, Fode C, Guillemot F, Anderson DJ. “Neurogenin1 and neurogenin2 control two distinct waves of neurogenesis in developing dorsal root ganglia.” Genes Dev. 1999 Jul 1;13(13):1717-28. Machado CB, Kanning KC, Kreis P, Stevenson D, Crossley M, Nowak M, Iacovino M, Kyba M, Chambers D, Blanc E, Lieberam I. “Reconstruction of phrenic neuron identity in embryonic stem cell-derived motor neurons.” Development. 2014 Feb;141(4):784-94. Machold R, Fishell G. “Math1 is expressed in temporally discrete pools of cerebellar rhombic-lip neural progenitors.” Neuron. 2005 Oct 6;48(1):17-24. Maden M, Gale E, Kostetskii I, Zile M. “Vitamin A-deficient quail embryos have half a hindbrain and other neural defects.” Curr Biol. 1996 Apr 1;6(4):417-26. Maden M. “Retinoid signalling in the development of the central nervous system.” Nat Rev Neurosci. 2002 Nov;3(11):843-53. Maden M. “Retinoic acid in the development, regeneration and maintenance of the nervous system.” Nat Rev Neurosci. 2007 Oct;8(10):755-65. Madisen L, Zwingman TA, Sunkin SM, Oh SW, Zariwala HA, Gu H, Ng LL, Palmiter RD, Hawrylycz MJ, Jones AR, Lein ES, Zeng H. “A robust and high-throughput Cre reporting and characterization system for the whole mouse brain.” Nat Neurosci. 2010 Jan;13(1):133-40. Maltenfort MG, McCurdy ML, Phillips CA, Turkin VV, Hamm TM. “Location and magnitude of conductance changes produced by Renshaw recurrent inhibition in spinal motoneurons.”J Neurophysiol. 2004 Sep;92(3):1417-32.

230

Mariani J, Crepel F, Mikoshiba K, Changeux JP, Sotelo C. “Anatomical, physiological and biochemical studies of the cerebellum from Reeler mutant mouse.” Philos Trans R Soc Lond B Biol Sci. 1977 Nov 2;281(978):1-28. Maricich SM, Herrup K. “Pax-2 expression defines a subset of GABAergic interneurons and their precursors in the developing murine cerebellum.” J Neurobiol. 1999 Nov 5;41(2):281-94. Marklund U, Hansson EM, Sundstrom E, de Angelis MH, Przemeck GK, Lendahl U, Muhr J, Ericson J. “Domain-specific control of neurogenesis achieved through patterned regulation of Notch ligand expression.” Development. 2010 Feb;137(3):437-45.

Marmigère F, Ernfors P. “Specification and connectivity of neuronal subtypes in the sensory lineage.” Nat Rev Neurosci. 2007 Feb;8(2):114-27.

Maroof AM, Brown K, Shi SH, Studer L, Anderson SA. “Prospective isolation of cortical interneuron precursors from mouse embryonic stem cells.” J Neurosci. 2010 Mar 31;30(13):4667-75.

Maroof AM, Keros S, Tyson JA, Ying SW, Ganat YM, Merkle FT, Liu B, Goulburn A, Stanley EG, Elefanty AG, Widmer HR, Eggan K, Goldstein PA, Anderson SA, Studer L. “Directed differentiation and functional maturation of cortical interneurons from human embryonic stem cells.” Cell Stem Cell. 2013 May 2;12(5):559-72.

Martí E, Bumcrot DA, Takada R, McMahon AP. “Requirement of 19K form of Sonic hedgehog for induction of distinct ventral cell types in CNS explants.” Nature. 1995 May 25;375(6529):322-5.

Martinat C, Bacci JJ, Leete T, Kim J, Vanti WB, Newman AH, Cha JH, Gether U, Wang H, Abeliovich A. “Cooperative transcription activation by Nurr1 and Pitx3 induces embryonic stem cell maturation to the midbrain dopamine neuron phenotype.” Proc Natl Acad Sci U S A. 2006 Feb 21;103(8):2874-9

Martinez S, Andreu A, Mecklenburg N, Echevarria D. “Cellular and molecular basis of cerebellar development.” Front Neuroanat. 2013 Jun 26;7:18. Masahira N, Takebayashi H, Ono K, Watanabe K, Ding L, Furusho M, Ogawa Y, Nabeshima Y, Alvarez-Buylla A, Shimizu K, Ikenaka K. “Olig2-positive progenitors in the embryonic spinal cord give rise not only to motoneurons and oligodendrocytes, but also to a subset of astrocytes and ependymal cells.” Dev Biol. 2006 May 15;293(2):358-69. Matise MP, Joyner AL. “Expression patterns of developmental control genes in normal and Engrailed-1 mutant mouse spinal cord reveal early diversity in developing interneurons.” J Neurosci. 1997 Oct 15;17(20):7805-16. Matise MP, Lustig M, Sakurai T, Grumet M, Joyner AL. “Ventral midline cells are required for the local control of commissural axon guidance in the mouse spinal cord.” Development. 1999 Aug;126(16):3649-59. Matsumoto K, Ieda T, Saito N, Ono T, Shimada K. “Role of retinoic acid in regulation of mRNA expression of CaBP-D28k in the cerebellum of the chicken.” Comp Biochem Physiol A Mol Integr Physiol. 1998 Jun;120(2):237-42. Maury Y, Côme J, Piskorowski RA, Salah-Mohellibi N, Chevaleyre V, Peschanski M, Martinat C, Nedelec S. “Combinatorial analysis of developmental cues efficiently converts human pluripotent stem cells into multiple neuronal subtypes.” Nat Biotechnol. 2015 Jan;33(1):89-96. Maxwell DJ, Christie WM, Short AD, Brown AG. “Direct observations of synapses between GABA-immunoreactive boutons and muscle afferent terminals in lamina VI of the cat's spinal cord.” Brain Res. 1990 Oct 22;530(2):215-22.

231

Mazurier N, Parain K, Parlier D, Pretto S, Hamdache J, Vernier P, Locker M, Bellefroid E, Perron M. “Ascl1 as a novel player in the Ptf1a transcriptional network for GABAergic cell specification in the retina.” PLoS One. 2014 Mar 18;9(3):e92113. Mazzocchio R, Rossi A. “Role of Renshaw cells in amyotrophic lateral sclerosis.” Muscle Nerve. 2010 Apr;41(4):441-3. Mazzoni EO, Mahony S, Iacovino M, Morrison CA, Mountoufaris G, Closser M, Whyte WA, Young RA, Kyba M, Gifford DK, Wichterle H. “Embryonic stem cell-based system for mapping developmental transcriptional programs.” Nat Methods. 2011 Dec; 8(12): 1056–1058. Mazzoni EO, Mahony S, Closser M, Morrison CA, Nedelec S, Williams DJ, An D, Gifford DK, Wichterle H. “Synergistic binding of transcription factors to cell-specific enhancers programs motor neuron identity.” Nat Neurosci. 2013 Sep; 16(9): 1219-27. Mazzoni EO, Mahony S, Peljto M, Patel T, Thornton SR, McCuine S, Reeder C, Boyer LA, Young RA, Gifford DK, Wichterle H. “Saltatory remodeling of Hox chromatin in response to rostrocaudal patterning signals.” Nat Neurosci. 2013 Sep;16(9):1191-8.

McCaffery P, Dräger UC. “Hot spots of retinoic acid synthesis in the developing spinal cord.” Proc Natl Acad Sci U S A. 1994 Jul 19;91(15):7194-7. McConnell SK, Kaznowski CE. “Cell cycle dependence of laminar determination in developing neocortex.” Science. 1991 Oct 11;254(5029):282-5. McLean DL, Fetcho JR. “Spinal interneurons differentiate sequentially from those driving the fastest swimming movements in larval zebrafish to those driving the slowest ones.” J Neurosci. 2009 Oct 28;29(43):13566-77. Mears SC, Frank E. “Formation of specific monosynaptic connections between muscle spindle afferents and motoneurons in the mouse.” J Neurosci. 1997 May 1;17(9):3128-35. Meinhardt A, Eberle D, Tazaki A, Ranga A, Niesche M, Wilsch-Bräuninger M, Stec A, Schackert G, Lutolf M, Tanaka EM. “3D reconstitution of the patterned neural tube from embryonic stem cells. Stem Cell Reports.” 2014 Dec 9;3(6):987-99. Megason SG, McMahon AP. “A mitogen gradient of dorsal midline Wnts organizes growth in the CNS.” Development. 2002 May;129(9):2087-98. Mende M, Fletcher EV, Belluardo JL, Pierce JP, Bommareddy PK, Weinrich JA, Kabir ZD, Schierberl KC, Pagiazitis JG, Mendelsohn AI, Francesconi A, Edwards RH, Milner TA, Rajadhyaksha AM, van Roessel PJ, Mentis GZ, Kaltschmidt JA. “Sensory-derived glutamate regulates presynaptic inhibitory terminals in the mouse spinal cord.” Neuron. 2016 Jun 15;90(6):1189-202.

Mendelson B, Frank E. “Specific monosynaptic sensory-motor connections form in the absence of patterned neural activity and motoneuronal cell death.” J Neurosci. 1991 May;11(5):1390-403.

Mennerick, S, Zorumski, CF. “Neural activity and survival in the developing nervous system.” Mol Neurobiol. 2000 Aug-Dec;22(1-3):41-54.

Mentis GZ, Alvarez FJ, Bonnot A, Richards DS, Gonzalez-Forero D, Zerda R, O'Donovan MJ. “Noncholinergic excitatory actions of motoneurons in the neonatal mammalian spinal cord.” Proc Natl Acad Sci U S A. 2005 May 17;102(20):7344-9.

Mentis GZ, Siembab VC, Zerda R, O'Donovan MJ, Alvarez FJ. “Primary afferent synapses on developing and adult Renshaw cells.” J Neurosci. 2006 Dec 20;26(51):13297-310.

232

Mentis GZ, Blivis D, Liu W, Drobac E, Crowder ME, Kong L, Alvarez FJ, Sumner CJ, O’Donovan MJ. “Early functional impairment of sensory-motor connectivity in a mouse model of spinal muscular atrophy.” Neuron. 2011 Feb 10;69(3):453-67. Meredith DM, Masui T, Swift GH, MacDonald RJ, Johnson JE. “Multiple transcriptional mechanisms control Ptf1a levels during neural development including autoregulation by the PTF1-J complex.” J Neurosci. 2009 Sep 9;29(36):11139-48. Meredith DM, Borromeo MD, Deering TG, Casey BH, Savage TK, Mayer PR, Hoang C, Tung KC, Kumar M, Shen C, Swift GH, Macdonald RJ, Johnson JE. “Program specificity for Ptf1a in pancreas versus neural tube development correlates with distinct collaborating cofactors and chromatin accessibility.” Mol Cell Biol. 2013 Aug;33(16):3166-79. Meshul CK, Seil FJ, Herndon RM. “Astrocytes play a role in regulation of synaptic density.” Brain Res. 1987 Jan 27;402(1):139-45. Miles GB, Yohn DC, Wichterle H, Jessell TM, Rafuse VF, Brownstone RM.J “Functional properties of motoneurons derived from mouse embryonic stem cells.” Neurosci. 2004 Sep 8;24(36):7848-58. Millonig JH, Millen KJ, Hatten MJ. “The mouse Dreher gene Lmx1a controls formation of the roof plate in the vertebrate CNS. Nature. 2000 Feb 17;403(6771):764-9. Minaki Y, Nakatani T, Mizuhara E, Inoue T, Ono Y. “Identification of a novel transcriptional corepressor, Corl2, as a cerebellar Purkinje cell-selective marker.” Gene Expr Patterns. 2008 Jul;8(6):418-23. Miyamichi K, Amat F, Moussavi F, Wang C, Wickersham I, Wall NR, Taniguchi H, Tasic B, Huang ZJ, He Z, Callaway EM, Horowitz MA, Luo L. “Cortical representations of olfactory input by trans-synaptic tracing.” Nature. 2011 Apr 14;472(7342):191-196. Miyoshi G, Butt SJ, Takebayashi H, Fishell G. “Physiologically distinct temporal cohorts of cortical interneurons arise from telencephalic Olig2-expressing precursors.” J Neurosci. 2007 Jul 18;27(29):7786-98. Misra K, Luo H, Li S, Matise M, Xiang M. “Asymmetric activation of Dll4-Notch signaling by Foxn4 and proneural factors activates BMP/TGFβ signaling to specify V2b interneurons in the spinal cord.” Development. 2014 Jan;141(1):187-98. Miyoshi G, Butt SJ, Takebayashi H, Fishell G. “Physiologically distinct temporal cohorts of cortical interneurons arise from telencephalic Olig2-expressing precursors.” J Neurosci. 2007 Jul 18;27(29):7786-98. Mizeracka K, DeMaso CR, Cepko CL. “Notch1 is required in newly postmitotic cells to inhibit the rod photoreceptor fate.” Development. 2013 Aug;140(15):3188-97.

Mizuguchi R, Kriks S, Cordes R, Gossler A, Ma Q, Goulding M. “Ascl1 and Gsh1/2 control inhibitory and excitatory cell fate in spinal sensory interneurons.” Nat Neurosci. 2006 Jun;9(6):770-8. Epub 2006 May 21.

Molotkova N, Molotkov A, Sirbu IO, Duester G. “Requirement of mesodermal retinoic acid generated by Raldh2 for posterior neural transformation.” Mech Dev. 2005 Feb;122(2):145-55. Molyneaux BJ, Arlotta P, Menezes JRL, Macklis JD. “Neuronal subtype specification in the cerebral cortex.” Nat Rev Neurosci. June 2007; 8:427-37. Montelius A, Marmigère F, Baudet C, Aquino JB, Enerbäck S, Ernfors P. “Emergence of the sensory

233

nervous system as defined by Foxs1 expression.” Differentiation. 2007 Jun;75(5):404-17. Monyer H, Markram H. “Interneuron Diversity series: Molecular and genetic tools to study GABAergic interneuron diversity and function.” Trends Neurosci. 2004 Feb;27(2):90-7. Moore NJ, Bhumbra GS, Foster JD, Beato M. “Synaptic Connectivity between Renshaw Cells and Motoneurons in the Recurrent Inhibitory Circuit of the Spinal Cord.” J Neurosci. 2015 Oct 7;35(40):13673-86.

Moran-Rivard L, Kagawa T, Saueressig H, Gross MK, Burrill J, Goulding M. “Evx1 is a postmitotic determinant of v0 interneuron identity in the spinal cord.” Neuron. 2001 Feb;29(2):385-99. Morikawa Y, Hisaoka T, Senba E. “Characterization of Foxp2-expressing cells in the developing spinal cord.” Neuroscience. 2009 Sep 15;162(4):1150-62. Muhr J, Graziano E, Wilson S, Jessell TM, Edlund T. “Convergent inductive signals specify midbrain, hindbrain, and spinal cord identity in gastrula stage chick embryos.” Neuron. 1999 Aug;23(4):689-702.

Müller T, Brohmann H, Pierani A, Heppenstall PA, Lewin GR, Jessell TM, Birchmeier C. “The homeodomain factor lbx1 distinguishes two major programs of neuronal differentiation in the dorsal spinal cord.” Neuron. 2002 May 16;34(4):551-62.

Müller T, Anlag K, Wildner H, Britsch S, Treier M, Birchmeier C. “The bHLH factor Olig3 coordinates the specification of dorsal neurons in the spinal cord.” Genes Dev. 2005 Mar 15;19(6):733-43.

Muñoz-Sanjuán I, Brivanlou AH. “Neural induction, the default model and embryonic stem cells.” Nat Rev Neurosci. 2002 Apr;3(4):271-80 Murashov AK, Pak ES, Hendricks WA, Owensby JP, Sierpinski PL, Tatko LM, Fletcher PL. “Directed differentiation of embryonic stem cells into dorsal interneurons.” FASEB J. 2005 Feb;19(2):252-4. Muroyama Y, Fujihara M, Ikeya M, Kondoh H, Takada S. “Wnt signaling plays an essential role in neuronal specification of the dorsal spinal cord.” Genes Dev. 2002 Mar 1;16(5):548-53. Myers CP, Lewcock JW, Hanson MG, Gosgnach S, Aimone JB, Gage FH, Lee KF, Landmesser LT, Pfaff SL. “Cholinergic input is required during embryonic development to mediate proper assembly of spinal locomotor circuits.” Neuron. 2005 Apr 7;46(1):37-49. Nagai M, Re DB, Nagata T, Chalazonitis A, Jessell TM, Wichterle H, Przedborski S. “Astrocytes expressing ALS-linked mutated SOD1 release factors selectively toxic to motor neurons.” Nat Neurosci. 2007 May;10(5):615-22.

Nagel G, Szellas T, Huhn W, Kateriya S, Adeishvili N, Berthold P, Ollig D, Hegemann P, Bamberg E. “Channelrhodopsin-2, a directly light-gated cation-selective membrane channel.” Proc Natl Acad Sci U S A. 2003 Nov 25;100(24):13940-5. Nakamura S, Senzaki K, Yoshikawa M, Nishimura M, Inoue K, Ito Y, Ozaki S, Shiga T. “Dynamic regulation of the expression of neurotrophin receptors by Runx3. Development. 2008 May;135(9):1703-11. Nakhai H, Sel S, Favor J, Mendoza-Torres L, Paulsen F, Duncker GI, Schmid RM. “Ptf1a is essential for the differentiation of GABAergic and glycinergic amacrine cells and horizontal cells in the mouse retina.” Development. 2007 Mar;134(6):1151-60. Nedelec S, Peljto M, Shi P, Amoroso MW, Kam LC, Wichterle H. “Concentration-dependent requirements

234

for local protein synthesis in motor neuron subtype-specific response to axon guidance cues.” J Neurosci. 2012 Jan 25:32(4):1496-506. Nelson BR, Hartman BH, Georgi SA, Lan MS, Reh TA. “Transient inactivation of Notch signaling synchronizes differentiation of neural progenitor cells.” Dev Biol. 2007 Apr 15;304(2):479-98. Nelson BR, Hartman BH, Georgi SA, Lan MS, Reh TA. “Transient inactivation of Notch signaling synchronizes differentiation of neural progenitor cells.” Dev Biol. 2007 Apr 15;304(2):479-98. Nery S, Fishell G, Corbin JG. “The caudal ganglionic eminence is a source of distinct cortical and subcortical cell populations.” Nat Neurosci. 2002 Dec;5(12):1279-87.

Nguyen VH, Trout J, Connors SA, Andermann P, Weinberg E, Mullins MC. “Dorsal and intermediate neuronal cell types of the spinal cord are established by a BMP signaling pathway.” Development. 2000 Mar;127(6):1209-20.

Ni Y, Nawabi H, Liu X, Yang L, Miyamichi K, Tedeschi A, Xu Bengang X, Wall NR, Callaway EM, He Z. “Characterization of long descending premotor propriospinal neurons in the spinal cord.” J Neurosci. 2014 July 9: 34(28):9404-17. Nicholas CR, Chen J, Tang Y, Southwell DG, Chalmers N, Vogt D, Arnold CM, Chen YJ, Stanley EG, Elefanty AG, Sasai Y, Alvarez-Buylla A, Rubenstein JL, Kriegstein AR. “Functional maturation of hPSC-derived forebrain interneurons requires an extended timeline and mimics human neural development.” Cell Stem Cell. 2013 May 2;12(5):573-86. Niederreither K, McCaffery P, Drager UC, Chambon P, Dolle P. “Restricted expression and retinoic acid-induced downregulation of the retinaldehyde dehydrogenase type 2 (RALDH-2) gene during mouse development.” Mech Dev. 1997 Feb;62(1):67-78. Niederreither K, Vermot J, Schuhbaur B, Chambon P, Dollé P. “Retinoic acid synthesis and hindbrain patterning in the mouse embryo.” Development. 2000 Jan;127(1):75-85. Nishimaru H, Restrepo CE, Ryge J, Yanagawa Y, Kiehn O. “Mammalian motor neurons corelease glutamate and acetylcholine at central synapses.” Proc Natl Acad Sci U S A. 2005 Apr 5;102(14):5245-9.

Nordström U, Jessell TM, Edlund T. “Progressive induction of caudal neural character by graded Wnt signaling.” Nat Neurosci. 2002 Jun;5(6):525-32.

Novitch BG, Chen AI, Jessell TM. “Coordinate regulation of motor neuron subtype identity and pan-neuronal properties by the bHLH repressor Olig2.” Neuron. 2001 Sep 13;31(5):773-89.

Novitch BG, Wichterle H, Jessell TM, Sockanathan S. “A requirement for retinoic acid-mediated transcriptional activation in ventral neural patterning and motor neuron specification.” Neuron. 2003 Sep 25;40(1):81-95.

Novotny T, Eiselt R, Urban J. “Hunchback is required for the specification of the early sublineage of neuroblast 7-3 in the Drosophila central nervous system.” Development. 2002 Feb;129(4):1027-36. Nye JS, Kopan R, Axel R. “An activated Notch suppresses neurogenesis and myogenesis but not gliogenesis in mammalian cells.” Development. 1994 Sep;120(9):2421-30. Oh EC, Khan N, Novelli E, Khanna H, Strettoi E, Swaroop A. “Transformation of cone precursors to functional rod photoreceptors by bZIP transcription factor NRL.” Proc Natl Acad Sci U S A. 2007 Jan 30;104(5):1679-84.

235

Ohki K, Chung S, Ch'ng YH, Kara P, Reid RC. “Functional imaging with cellular resolution reveals precise micro-architecture in visual cortex.” Nature. 2005 Feb 10; 433(7026):597-603. Okawa H, Della Santina L, Schwartz Gw, Rieke F, Wong RO. “Interplay of cell-autonomous and nonautonomous mechanisms tailors synaptic connectivity of converging axons in vivo.” Neuron. 2014 Apr 2;82(1):125-37. Okigawa S, Mizoguchi T, Okano M, Tanaka H, Isoda M, Jiang YJ, Suster M, Higashijima S, Kawakami K, Itoh M. “Different combinations of Notch ligands and receptors regulate V2 interneuron progenitor proliferation and V2a/V2b cell fate determination.” Dev Biol. 2014 Jul 15;391(2):196-206. Oppenheim, RW. “Cell death during development of the nervous system.” Annu. Rev. Neurosci. 1991: 14,453 -501. Osakada F, Mori T, Cetin AH, Marshel JH, Virgen B, Callaway EM. “New rabies virus variants for monitoring and manipulating activity and gene expression in defined neural circuits.” Neuron. 2011 Aug 25; 71(4): 617–631. Palay SL, Chan-Palay V. “A guide to the synaptic analysis of the neuropil.” Cold Spring Harb Symp Quant Biol. 1976;40:1-16. Palop JJ, Chin J, Mucke L. “A network dysfunction perspective on neurodegenerative diseases.” Nature. 2006 Oct 19;443(7113):768-73. Palop JJ, Chin J, Roberson ED, Wang J, Thwin MT, Bien-Ly N, Yoo J, Ho KO, Yu GQ, Kreitzer A, Finkbeiner S, Noebels JL, Mucke L. “Aberrant excitatory neuronal activity and compensatory remodeling of inhibitory hippocampal circuits in mouse models of Alzheimer’s disease.” Neuron. 2007 Sep 6;55(5):697-711.

Panchision DM1, Pickel JM, Studer L, Lee SH, Turner PA, Hazel TG, McKay RD. “Sequential actions of BMP receptors control neural precursor cell production and fate.” Genes Dev. 2001 Aug 15;15(16):2094-110.

Panman L, Andersson E, Alekseenko Z, Hedlund E, Kee N, Mong J, Uhde CW, Deng Q, Sandberg R, Stanton LW, Ericson J, Perlmann T. “Transcription factor-induced lineage selection of stem-cell-derived neural progenitor cells.” Cell Stem Cell. 2011 Jun 3;8(6):663-75.

Papalopulu N, Kintner C. “A posteriorising factor, retinoic acid, reveals that anteroposterior patterning controls the timing of neuronal differentiation in Xenopus neuroectoderm.” Development. 1996 Nov;122(11):3409-18.

Papaneri AB, Wirblich C, Cann JA, Cooper K, Jahrling PB, Schnell MJ, Blaney JE. “A replication-deficient rabies virus vaccine expressing Ebola virus glycoprotein is highly attenuated for neurovirulence.” Virology. 2012 Dec 5;434(1):18-26. Paradis S, Harrar DB, Lin Y, Koon AC, Hauser JL, Griffith EC, Zhu L, Brass LF, Chen C, Greenberg ME. “An RNAi-based approach identifies molecules required for glutamatergic and GABAergic synapse development.” Neuron. 2007;53:217–232. Paschaki M, Lin SC, Wong RL, Finnell RH, Dollé P, Niederreither K. “Retinoic acid-dependent signaling pathways and lineage events in the developing mouse spinal cord.” PLoS One. 2012;7(3):e32447. Pascual M, Abasolo I, Mingorance-Le Meur A, Martínez A, Del Rio JA, Wright CV, Real FX, Soriano E. “Cerebellar GABAergic progenitors adopt an external granule cell-like phenotype in the absence of Ptf1a transcription factor expression.” Proc Natl Acad Sci U S A. 2007 Mar 20;104(12):5193-8.

236

Pasinelli P, Brown RH. “Molecular biology of amyotrophic lateral sclerosis: insights from genetics.” Nat Rev Neurosci. 2006 Sep;7(9):710-23.

Patel TD, Kramer I, Kucera J, Niederkofler V, Jessell TM, Arber S, Snider WD. “Peripheral NT3 signaling is required for ETS protein expression and central patterning of proprioceptive sensory afferents.” Neuron. 2003 May 8;38(3):403-16.

Patrizi A, Scelfo B, Viltono L, Briatore F, Fukaya M, Watanabe M, Strata P, Varoqueaux F, Brose N, Fritschy JM, Sassoè-Pognetto M. “Synapse formation and clustering of neuroligin-2 in the absence of GABAA receptors.” Proc Natl Acad Sci U S A. 2008 Sep 2;105(35):13151-6. Pearson BJ, Doe CQ. “Regulation of neuroblast competence in Drosophila.” Nature. 2003 Oct 9;425(6958):624-8 Pearson BJ, Doe CQ. “Specification of temporal identity in the developing nervous system.” Annu Rev Cell Dev Biol. 2004;20:619-47. Pecho-Vrieseling E, Sigrist M, Yoshida Y, Jessell TM, Arber S. “Specificity of sensory-motor connections encoded by Sema3e-Plxnd1 recognition.” Nature. 2009 Jun 11;459(7248):842-6. Peljto M, Dasen JS, Mazzoni EO, Jessell TM, Wichterle H. “Functional diversity of ESC-derived motor neuron subtypes revealed through intraspinal transplantation.” Cell Stem Cell. 2010 Sep 3;7(3):355-66. Peljto M, Wichterle H. “Programming embryonic stem cells to neuronal subtypes.” Curr Opin Neurobiol. 2011 Feb;21(1):43-51. Peng CY, Yajima H, Burns CE, Zon LI, Sisodia SS, Pfaff SL, Sharma K. “Notch and MAML signaling drives Scl-dependent interneuron diversity in the spinal cord.” Neuron. 2007 Mar 15;53(6):813-27. Perry S, Gezelius H, Larhammar M, Hilscher MM, Lamotte d'Incamps B, Leao KE, Kullander K. “Firing properties of Renshaw cells defined by Chrna2 are modulated by hyperpolarizing and small conductance ion currents Ih and ISK.” Eur J Neurosci. 2015 Apr;41(7):889-900. Peters A, Feldman ML. “The projection of the lateral geniculate nucleus to area 17 of the rat cerebral cortex. I. General description.” J Neurocytol. 1976 Feb;5(1):63-84. Petros TJ, Tyson JA, Anderson SA. “Pluripotent stem cells for the study of CNS development.” Front Mol Neurosci. 2011 Oct 12;4:30. Phelps PE, Barber RP, Houser CR, Crawford GD, Salvaterra PM, Vaughn JE. “Postnatal development of neurons containing choline acetyltransferase in rat spinal cord: an immunocytochemical study.” J Comp Neurol. 1984 Nov 1;229(3):347-61. Philippidou P, Dasen JS. “Hox Genes: Choreographers in Neural Development, Architects of Circuit Organization.” Neuron. 2013 Oct 2;80(1):12-34. Pierani A, Brenner-Morton S, Chiang C, Jessell TM. “A sonic hedgehog-independent, retinoid-activated pathway of neurogenesis in the ventral spinal cord.” Cell. 1999 Jun 25;97(7):903-15. Pierani A, Moran-Rivard L, Sunshine MJ, Littman DR, Goulding M, Jessell TM. “Control of interneuron fate in the developing spinal cord by the progenitor homeodomain protein Dbx1.” Neuron. 2001 Feb;29(2):367-84. Pillai A, Mansouri A, Behringer R, Westphal H, Goulding M. “Lhx1 and Lhx5 maintain the inhibitory-

237

neurotransmitter status of interneurons in the dorsal spinal cord.” Development. 2007 Jan;134(2):357-66. Epub 2006 Dec 13. Pituello F, Yamada G, Gruss P. “Activin A inhibits Pax-6 expression and perturbs cell differentiation in the developing spinal cord in vitro.” Proc Natl Acad Sci U S A. 1995 Jul 18;92(15):6952-6. Polgár E, Sardella TC, Tiong SY, Locke S, Watanabe M, Todd AJ. “Functional differences between neurochemically defined populations of inhibitory interneurons in the rat spinal dorsal horn.” Pain. 2013 Dec;154(12):2606-15. Poulin J-F, Tasic B, Hjerling-Leffler J, Trimarchi JM, Awatramani R. “Disentangling neural cell diversity using single-cell transcriptomics.” Nat Neurosci. 2016 Aug 26;19(9):1131-41. Pringle NP, Guthrie S, Lumsden A, Richardson WD. “Dorsal spinal cord neuroepithelium generates astrocytes but not oligodendrocytes.” Neuron. 1998 May;20(5):883-93. Qian Y, Shirasawa S, Chen CL, Cheng L, Ma Q. “Proper development of relay somatic sensory neurons and D2/D4 interneurons requires homeobox genes Rnx/Tlx-3 and Tlx-1.” Genes Dev. 2002 May 15;16(10):1220-33. Rajaii F, Bitzer ZT, Xu Q, Sockanathan S. “Expression of the dominant negative retinoid receptor, RAR403, alters telencephalic progenitor proliferation, survival, and cell fate specification.” Dev Biol. 2008 Apr 15;316(2):371-82. Ramírez-Jarquín UN, Lazo-Gómez R, Tovar-Y-Romo LB, Tapia R. “Spinal inhibitory circuits and their role in motor neuron degeneration.” Neuropharmacology. 2014 Jul;82:101-7. Ramos C, Rocha S, Gaspar C, Henrique D. “Two notch ligands, dll1 and jag1, are differently restricted in their range of action to control neurogenesis in the mammalian spinal cord”. PLoS One. 2010 Nov 24;5(11):e15515. Rao MS, Noble M, Mayer-Pröschel M. “A tripotential glial precursor cell is present in the developing spinal cord.” Proc Natl Acad Sci U S A. 1998 Mar 31;95(7):3996-4001. Rao A, Cha EM, Craig AM. “Mismatched appositions of presynaptic and postsynaptic components in isolated hippocampal neurons.” J Neurosci. 2000 Nov 15;20(22):8344-53. Raynor EM, Shefner JM. “Recurrent inhibition is decreased in patients with amyotrophic lateral sclerosis.” Neurology. 1994 Nov;44(11):2148-53. Reardon TR, Murray AJ, Turi GF, Wirblich C, Croce KR, Schnell MJ, Jessell TM, Losonczy A. “Rabies Virus CVS-N2c(ΔG) Strain Enhances Retrograde Synaptic Transfer and Neuronal Viability.” Neuron. 2016 Feb 17;89(4):711-24. Renier N, Schonewille M, Giraudet F, Badura A, Tessier-Lavigne M, Avan P, De Zeeuw CI, Chédotal A. “Genetic dissection of the function of hindbrain axonal commissures.” PLoS Biol. 2010 Mar 9;8(3):e1000325. Renshaw, B. “Influence of discharge of motoneurons upon excitation of neighboring motoneurons.” J Neurophysiol. 1941 4, 167-183. Renshaw, B. “Central effects of centripetal impulses in axons of spinal ventral roots. J Neurophysiol.” 1946 May;9:191-204. Rhee HS, Closser M, Guo Y, Bashkirova EV, Tan GC, Gifford DK, Wichterle H. “Expression of terminal effector genes in maturing neurons is maintained by a dynamic relay of transient enhancers.” Neuron.

238

2016 in press. Ribes V, Briscoe J. “Establishing and interpreting graded Sonic Hedgehog signaling during vertebrate neural tube patterning: the role of negative feedback.” Cold Spring Harb Perspect Biol. 2009 Aug;1(2):a002014. Richards DS, Griffith RW, Romer SH, Alvarez FJ. “Motor axon synapses on renshaw cells contain higher levels of aspartate than glutamate.” PLoS One. 2014 May 9;9(5):e97240. Rocha SF, Lopes SS, Gossler A, Henrique D. “Dll1 and Dll4 function sequentially in the retina and pV2 domain of the spinal cord to regulate neurogenesis and create cell diversity.” Dev Biol. 2009 Apr 1;328(1):54-65 Rockhill RL, Daly FJ, MacNeil MA, Brown SP, Masland RH. “The diversity of ganglion cells in a mammalian retina.” J Neurosci. 2002 May 1;22(9):3831-43. Roelink H, Porter JA, Chiang C, Tanabe Y, Chang DT, Beachy PA, Jessell TM. “Floor plate and motor neuron induction by different concentrations of the amino-terminal cleavage product of sonic hedgehog autoproteolysis.” Cell. 1995 May 5;81(3):445-55. Rose D, Chiba A. “Synaptic target recognition at Drosophila neuromuscular junctions.” Microsc Res Tech. 2000 Apr 1;49(1):3-13. Rossant J, Zirngibl R, Cado D, Shago M, Giguère V. “Expression of a retinoic acid response element-hsplacZ transgene defines specific domains of transcriptional activity during mouse embryogenesis.” Genes Dev. 1991 Aug;5(8):1333-44. Rousso DL, Gaber ZB, Wellik D, Morrisey EE, Novitch BG. “Coordinated actions of the forkhead protein Foxp1 and Hox proteins in the columnar organization of spinal motor neurons.” Neuron. 2008 Jul 31;59(2):226-40. Rowitch DH, Lu QR, Kessaris N, Richardson WD. “An 'oligarchy' rules neural development.” Trends Neurosci. 2002 Aug;25(8):417-22. Rudomin P, Schmidt RF. “Presynaptic inhibition in the vertebrate spinal cord revisited.” Exp Brain Res. 1999 Nov;129(1):1-37. Ryall RW. “Renshaw cell mediated inhibition of Renshaw cells: patterns of excitation and inhibition from impulses in motor axon collaterals.” J Neurophysiol. 1970 Mar;33(2):257-70. Ryu JH, Kong HJ, Park JY, Lim KE, An CM, Lee J, Yeo SY. “Generation of late-born neurons in the ventral spinal cord requires the coordination of retinoic acid and Notch signaling.” Neurosci Lett. 2015 Aug 18;602:95-8. Sabharwal P, Lee C, Park S, Rao M, Sockanathan S. “GDE2 regulates subtype-specific motor neuron generation through inhibition of Notch signaling.” Neuron. 2011 Sep 22;71(6):1058-70. Salero E, Hatten ME. “Differentiation of ES cells into cerebellar neurons.” Proc Natl Acad Sci U S A. 2007 Feb 20;104(8):2997-3002. Sances S, Bruijn LI, Chandran S, Eggan K, Ho R, Klim JR, Livesey MR, Lowry E, Macklis JD, Rushton D, Sadegh C, Sareen D, Wichterle H, Zhang SC, Svendsen CN. “Modeling ALS with motor neurons derived from human induced pluripotent stem cells.” Nat Neurosci. 2016 Apr;19(4):542-53. Sander M, Paydar S, Ericson J, Briscoe J, Berber E, German M, Jessell TM, Rubenstein JL. “Ventral neural patterning by Nkx homeobox genes: Nkx6.1 controls somatic motor neuron and ventral interneuron

239

fates.” Genes Dev. 2000 Sep 1;14(17):2134-9. Sandoe J, Eggan K. “Opportunities and challenges of pluripotent stem cell neurodegenerative disease models.” Nat Neurosci. 2013 Jul;16(7):780-9. Sanes JR, Yamagata M. “Many paths to synaptic specificity.” Annu Rev Cell Dev Biol. 2009;25:161-95 Sapir T, Geiman EJ, Wang Z, Velasquez T, Mitsui S, Yoshihara Y, Frank E, Alvarez FJ, Goulding M. “Pax6 and engrailed 1 regulate two distinct aspects of Renshaw cell development.” J Neurosci. 2004 Feb 4;24(5):1255-64. Sardella TC, Polgár E, Garzillo F, Furuta T, Kaneko T, Watanabe M, Todd AJ. “Dynorphin is expressed primarily by GABAergic neurons that contain galanin in the rat dorsal horn.” Mol Pain. 2011 Sep 29;7:76. Sato T, Joyner AL, Nakamura H. “How does Fgf signaling from the isthmic organizer induce midbrain and cerebellum development?” Dev Growth Differ. 2004 Dec;46(6):487-94. Sato T, Joyner AL. “The duration of Fgf8 isthmic organizer expression is key to patterning different tectal-isthmo-cerebellum structures.” Development. 2009 Nov;136(21):3617-26. Satou C, Kimura Y, Higashijima S. “Generation of multiple classes of V0 neurons in zebrafish spinal cord: progenitor heterogeneity and temporal control of neuronal diversity.” J Neurosci. 2012 Feb 1;32(5):1771-83.

Saueressig H, Burrill J, Goulding M. “Engrailed-1 and netrin-1 regulate axon pathfinding by association interneurons that project to motor neurons.” Development. 1999 Oct;126(19):4201-12.

Saxena S, Caroni P. “Selective neuronal vulnerability in neurodegenerative diseases: from stressor thresholds to degeneration.” Neuron. 2011 Jul 14;71(1):35-48. Saxena S, Roselli F, Singh K, Leptien K, Julien JP, Gros-Louis F, Caroni P. “Neuroprotection through excitability and mTOR required in ALS motoneurons to delay disease and extend survival.” Neuron. 2013 Oct 2;80(1):80-96. Saywell SA, Ford TW, Kirkwood PA. “Axonal projections of Renshaw cells in the thoracic spinal cord.” Physiol Rep. 2013 Nov;1(6):e00161. Scheiffele P, Fan J, Choih J, Fetter R, Serafini T. “Neuroligin expressed in nonneuronal cells triggers presynaptic development in contacting axons.” Cell. 2000 Jun 9;101(6):657-69. Schilling K. “Lineage, development and morphogenesis of cerebellar interneurons.” Prog Brain Res. 2000;124:51-68. Schmid A, Chiba A, Doe CQ. “Clonal analysis of Drosophila embryonic neuroblasts: neural cell types, axon projections and muscle targets.” Development. 1999 Nov;126(21):4653-89. Schneider SP, Fyffe RE. “Involvement of GABA and glycine in recurrent inhibition of spinal motoneurons.” J Neurophysiol. 1992 Aug;68(2):397-406. Schubert FR, Dietrich S, Mootoosamy RC, Chapman SC, Lumsden A. “Lbx1 marks a subset of interneurons in chick hindbrain and spinal cord.” Mech Dev. 2001 Mar;101(1-2):181-5. Schnell MJ, McGettigan JP, Wirblich C, Papaneri A. “The cell biology of rabies virus: using stealth to reach the brain.” Nat Rev Microbiol. 2010 Jan;8(1):51-61. Schütz B. “Imbalanced excitatory to inhibitory synaptic input precedes motor neuron degeneration in an

240

animal model of amyotrophic lateral sclerosis.” Neurobiol Dis. 2005 Oct;20(1):131-40. Sekerková G, Ilijic E, Mugnaini E. “Bromodeoxyuridine administered during neurogenesis of the projection neurons causes cerebellar defects in rat.” J Comp Neurol. 2004 Mar 8;470(3):221-39. Sekerková G, Ilijic E, Mugnaini E. “Time of origin of unipolar brush cells in the rat cerebellum as observed by prenatal bromodeoxyuridine labeling.” Neuroscience. 2004;127(4):845-58. Seto Y, Nakatani T, Masuyama N, Taya S, Kumai M, Minaki Y, Hamaguchi A, Inoue YU, Inoue T, Miyashita S, Fujiyama T, Yamada M, Chapman H, Campbell K, Magnuson MA, Wright CV, Kawaguchi Y, Ikenaka K, Takebayashi H, Ishiwata S, Ono Y, Hoshino M. “Temporal identity transition from Purkinje cell progenitors to GABAergic interneuron progenitors in the cerebellum.” Nat Commun. 2014;5:3337. Sharma K, Choi SY, Zhang Y, Nieland TJ, Long S, Li M, Huganir RL. “High-throughput genetic screen for synaptogenic factors: identification of LRP6 as critical for excitatory synapse development.” Cell Rep. 2013 Dec 12;5(5):1330-41. Shen K, Bargmann CI. “The immunoglobulin superfamily protein SYG-1 determines the location of specific synapses in C. elegans.” Cell. 2003 Mar 7;112(5):619-30. Shen K, Fetter RD, Bargmann CI. “Synaptic specificity is generated by the synaptic guidepost protein SYG-2 and its receptor, SYG-1.” Cell. 2004 Mar 19;116(6):869-81. Shen K, Scheiffele P. “Genetics and cell biology of building specific synaptic connectivity.” Annu Rev Neurosci. 2010;33:473-507. Sherrington, CS. The integrative action of the nervous system. New Haven: Yale University Press; c1906.

Shigenaga Y, Doe K, Suemune S, Mitsuhiro Y, Tsuru K, Otani K, Shirana Y, Hosoi M, Yoshida A, Kagawa K. “Physiological and morphological characteristics of periodontal mesencephalic trigeminal neurons in the cat--intra-axonal staining with HRP.” Brain Res. 1989 Dec 25;505(1):91-110. Shimojo H, Ohtsuka T, Kageyama R. “Oscillations in notch signaling regulate maintenance of neural progenitors.” Neuron. 2008 Apr 10;58(1):52-64. Shirasaki R, Pfaff SL. “Transcriptional codes and the control of neuronal identity.” Annu Rev Neurosci. 2002;25:251-81.

Shishido E, Takeichi M, Nose A. “Drosophila synapse formation: regulation by transmembrane protein with Leu-rich repeats, CAPRICIOUS.” Science. 1998 Jun 26;280(5372):2118-21. Shneider NA, Mentis GZ, Schustak J, O’Donovan MJ. “Functionally Reduced Sensorimotor Connections Form with Normal Specificity Despite Abnormal Muscle Spindle Development: The Role of Spindle-Derived Neurotrophin 3.” J Neurosci. 2009 April 15:29(15):4719-35.

Siembab VC, Smith CA, Zagoraiou L, Berrocal MC, Mentis GZ, Alvarez FJ. “Target selection of proprioceptive and motor axon synapses on neonatal V1-derived Ia inhibitory interneurons and Renshaw cells.” J Comp Neurol. 2010 Dec 1;518(23):4675-701.

Siembab VC, Gomez-Perez L, Rotterman TM, Shneider NA, Alvarez FJ. “Role of primary afferents in the developmental regulation of motor axon synapse numbers on Renshaw cells.” J Comp Neurol. 2016 Jun 15;524(9):1892-919. Simeone A. “Positioning the isthmic organizer where Otx2 and Gbx2 meet.” Trends Genet. 2000 Jun;16(6):237-40.

241

Simpson P. “Notch and the choice of cell fate in Drosophila neuroepithelium.” Trends Genet. 1990 Nov;6(11):343-5. Skaggs K, Martin DM, Novitch BG. “Regulation of spinal interneuron development by the Olig-related protein Bhlhb5 and Notch signaling.” Development. 2011 Aug;138(15):3199-211.

Skeath JB. “At the nexus between pattern formation and cell-type specification: the generation of individual neuroblast fates in the Drosophila embryonic central nervous system.” Bioessays. 1999 Nov;21(11):922-31. Slezak M, Pfrieger FW. “New roles for astrocytes: regulation of CNS synaptogenesis.” Trends Neurosci. 2003 Oct;26(10):531-5. Smith KM, Boyle KA, Madden JF, Dickinson SA, Jobling P, Callister RJ, Hughes DI, Graham BA. “Functional heterogeneity of calretinin-expressing neurons in the mouse superficial dorsal horn: implications for spinal pain processing.” J Physiol. 2015 Oct 1;593(19):4319-39. Sockanathan S, Jessell TM. “Motor neuron-derived retinoid signaling specifies the subtype identity of spinal motor neurons.” Cell. 1998 Aug 21;94(4):503-14. Sockanathan S, Perlmann T, Jessell TM. “Retinoid receptor signaling in postmitotic motor neurons regulates rostrocaudal positional identity and axonal projection pattern.” Neuron. 2003 Sep 25;40(1):97-111. Soghomonian JJ, Martin DL. “Two isoforms of glutamate decarboxylase: why?” Trends Pharmacol Sci. 1998 Dec;19(12):500-5. Solomin L, Johansson CB, Zetterström RH, Bissonnette RP, Heyman RA, Olson L, Lendahl U, Frisén J, Perlmann T. “Retinoid-X receptor signalling in the developing spinal cord.” Nature. 1998 Sep 24;395(6700):398-402. Sotelo C. “Dendritic abnormalities of Purkinje cells in the cerebellum of neurologic mutant mice (weaver and staggerer).” Adv Neurol. 1975;12:335-51

Soundararajan P, Miles GB, Rubin LL, Brownstone RM, Rafuse VF. “Motoneurons derived from embryonic stem cells express transcription factors and develop phenotypes characteristic of medial motor column neurons.” J Neurosci. 2006 Mar 22;26(12):3256-68.

Southwell DG, Nicholas CR, Basbaum AI, Stryker MP, Kriegstein AR, Rubenstein JL, Alvarez-Buylla A. “Interneurons from embryonic development to cell-based therapy.” Science. 2014 Apr 11;344(6180):1240622. Spemann H, Mangold H. “Induction of embryonic primordia by implantation of organizers from a different species.” Roux's Arch. Entw. Mech. 1924;100:599–638. Sperry RW. “Chemoaffinity in the orderly growth of nerve fiber patterns and connections.” Proc Natl Acad Sci U S A. 1963 Oct;50:703-10. Srinivas S, Watanabe T, Lin CS, William CM, Tanabe Y, Jessell TM, Costantini F. “Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus.” BMC Dev Biol. 2001;1:4. Stadler MB, Murr R, Burger L, Ivanek R, Lienert F, Schöler A, van Nimwegen E, Wirbelauer C, Oakeley EJ, Gaidatzis D, Tiwari VK, Schübeler D. “DNA-binding factors shape the mouse methylome at distal regulatory regions.” Nature. 2011 Dec 14;480(7378):490-5.

242

Stam FJ, Hendricks TJ, Zhang J, Geiman EJ, Francius C, Labosky PA, Clotman F, Goulding M. “Renshaw cell interneuron specialization is controlled by a temporally restricted transcription factor program.” Development. 2012 Jan;139(1):179-90.

Stepanyants A, Hof PR, Chklovskii DB. “Geometry and structural plasticity of synaptic connectivity.” Neuron. 2002 Apr 11;34(2):275-88. Storm R, Cholewa-Waclaw J, Reuter K, Bröhl D, Sieber M, Treier M, Müller T. “The bHLH transcription factor Olig3 marks the dorsal neuroepithelium of the hindbrain and is essential for the development of brainstem nuclei.” Development. 2009 Jan;136(2):295-305. Sudarov A, Turnbull RK, Kim EJ, Lebel-Potter M, Guillemot F, Joyner AL. “Ascl1 genetics reveals insights into cerebellum local circuit assembly.” J Neurosci. 2011 Jul 27;31(30):11055-69. Sultan F. “Analysis of mammalian brain architecture.” Nature. 2002 Jan 10;415(6868):133-4. Sun Y, Dykes IM, Liang X, Eng SR, Evans SM, Turner EE. “A central role for Islet1 in sensory neuron development linking sensory and spinal gene regulatory programs.” Nat Neurosci. 2008 Nov;11(11):1283-93. Sunico CR, Domínguez G, García-Verdugo JM, Osta R, Montero F, Moreno-López B. “Reduction in the motoneuron inhibitory/excitatory synaptic ratio in an early-symptomatic mouse model of amyotrophic lateral sclerosis.” Brain Pathol. 2011 Jan;21(1):1-15. Sunkin SM, Ng L, Lau C, Dolbeare T, Gilbert TL, Thompson CL, Hawrylycz M, Dang C. “Allen Brain Atlas: an integrated spatio-temporal portal for exploring the central nervous system.” Nucleic Acids Res. 2013 Jan;41(Database issue):D996-D1008. Sürmeli G, Akay T, Ippolito GC, Tucker PW, Jessell TM. “Patterns of spinal sensory-motor connectivity prescribed by a dorsoventral positional template.” Cell. 2011 Oct 28;147(3):653-65 Swaroop A, Kim D, Forrest D. “Transcriptional regulation of photoreceptor development and homeostasis in the mammalian retina.” Nat Rev Neurosci. 2010 Aug;11(8):563-76. Swindell EC, Thaller C, Sockanathan S, Petkovich M, Jessell TM, Eichele G. “Complementary domains of retinoic acid production and degradation in the early chick embryo.” Dev Biol. 1999 Dec 1;216(1):282-96. Takahashi K, Yamanaka S. “Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors.” Cell. 2006 Aug 25;126(4):663-76. Takahashi H, Arstikaitis P, Prasad T, Bartlett TE, Wang YT, Murphy TH, Craig AM. “Postsynaptic TrkC and presynaptic PTPsigma function as a bidirectional excitatory synaptic organizing complex.” Neuron. 2011;69:287–303. Takazawa T, Croft GF, Amoroso MW, Studer L, Wichterle H, Macdermott AB. “Maturation of spinal motor neurons derived from human embryonic stem cells.” PLoS One. 2012;7(7):e40154. Tan GC, Mazzoni EO, Wichterle H. “Iterative role of Notch signaling in spinal motor neuron diversification.” Cell Rep. 2016 Jul 26;16(4):907-16.

Tanabe Y, William C, Jessell TM. “Specification of motor neuron identity by the MNR2 homeodomain protein.” Cell. 1998 Oct 2;95(1):67-80.

Taniguchi H, Lu J, Huang ZJ. “The spatial and temporal origin of chandelier cells in mouse neocortex.” Science. 2013 Jan 4;339(6115):70-4.

243

Tétreault N, Champagne MP, Bernier G. “The LIM homeobox transcription factor Lhx2 is required to specify the retina field and synergistically cooperates with Pax6 for Six6 trans-activation.” Dev Biol. 2009 Mar 15;327(2):541-50.

Thélie A, Desiderio S, Hanotel J, Quigley I, Van Driessche B, Rodari A, Borromeo MD, Kricha S, Lahaye F, Croce J, Cerda-Moya G, Ordoño Fernandez J, Bolle B, Lewis KE, Sander M, Pierani A, Schubert M, Johnson JE, Kintner CR, Pieler T, Van Lint C, Henningfeld KA, Bellefroid EJ, Van Campenhout C. “Prdm12 specifies V1 interneurons through cross-repressive interactions with Dbx1 and Nkx6 genes in Xenopus.” Development. 2015 Oct 1;142(19):3416-28. Timmer JR, Wang C, Niswander L. “BMP signaling patterns the dorsal and intermediate neural tube via regulation of homeobox and helix-loop-helix transcription factors.” Development. 2002 May;129(10):2459-72. Timmer J, Chesnutt C, Niswander L. “The activin signaling pathway promotes differentiation of dI3 interneurons in the spinal neural tube.” Dev Biol. 2005 Sep 1;285(1):1-10. Tozer S, Le Dréau G, Marti E, Briscoe J. “Temporal control of BMP signalling determines neuronal subtype identity in the dorsal neural tube.” Development. 2013 Apr;140(7):1467-74. Tran TS, Rubio ME, Clem RL, Johnson D, Case L, Tessier-Lavigne M, Huganir RL, Ginty DD, Kolodkin AL. “Secreted semaphorins control spine distribution and morphogenesis in the postnatal CNS.” Nature. 2009 Dec 24;462(7276):1065-9.

Tripodi M, Stepien AE, Arber S. “Motor antagonism exposed by spatial segregation and timing of neurogenesis.” Nature. 2011 Oct 19;479(7371):61-6.

Tripodi M, Arber S. “Regulation of motor circuit assembly by spatial and temporal mechanisms.” Curr Opin Neurobiol. 2012 Aug;22(4):615-23.

Tsai HH, Li H, Fuentealba LC, Molofsky AV, Taveira-Marques R, Zhuang H, Tenney A, Murnen AT, Fancy SP, Merkle F, Kessaris N, Alvarez-Buylla A, Richardson WD, Rowitch DH. “Regional astrocyte allocation regulates CNS synaptogenesis and repair.” Science. 2012 Jul 20;337(6092):358-62.

Tsuchida T, Ensini M, Morton SB, Baldassare M, Edlund T, Jessell TM, Pfaff SL. “Topographic organization of embryonic motor neurons defined by expression of LIM homeobox genes.” Cell. 1994 Dec 16;79(6):957-70.

Tsunemoto RK, Eade KT, Blanchard JW, Baldwin KK. “Forward engineering neuronal diversity using direct reprogramming.” EMBO J. 2015 Jun 3;34(11):1445-55 Tyson JA, Goldberg EM, Maroof AM, Xu Q, Petros TJ, Anderson SA. “Duration of culture and sonic hedgehog signaling differentially specify PV versus SST cortical interneuron fates from embryonic stem cells.” Development. 2015 Apr 1;142(7):1267-78. Vallstedt A, Muhr J, Pattyn A, Pierani A, Mendelsohn M, Sander M, Jessell TM, Ericson J. “Different levels of repressor activity assign redundant and specific roles to Nkx6 genes in motor neuron and interneuron specification.” Neuron. 2001 Sep 13;31(5):743-55. Van Keulen LC. “Axon trajectories of Renshaw cells in the lumbar spinal cord of the cat, as reconstructed after intracellular staining with horseradish peroxidase.” Brain Res. 1979 May 5;167(1):157-62. Van Keulen L. “Autogenetic recurrent inhibition of individual spinal motoneurones of the cat.” Neurosci Lett. 1981 Feb 6;21(3):297-300.

244

Vierbuchen T, Ostermeier A, Pang ZP, Kokubu Y, Südhof TC, Wernig M. “Direct conversion of fibroblasts to functional neurons by defined factors.” Nature. 2010 Feb 25;463(7284):1035-41. Villapol S, Wang Y, Adams M, Symes AJ. “Smad3 deficiency increases cortical and hippocampal neuronal loss following traumatic brain injury.” Exp Neurol. 2013 Dec;250:353-65. Vokes SA, Ji H, McCuine S, Tenzen T, Giles S, Zhong S, Longabaugh WJ, Davidson EH, Wong WH, McMahon AP. “Genomic characterization of Gli-activator targets in sonic hedgehog-mediated neural patterning.” Development. 2007 May;134(10):1977-89. Vokes SA, Ji H, Wong WH, McMahon AP. “A genome-scale analysis of the cis-regulatory circuitry underlying sonic hedgehog-mediated patterning of the mammalian limb.” Genes Dev. 2008 Oct 1;22(19):2651-63. Vrieseling E, Arber S. “Target-induced transcriptional control of dendritic patterning and connectivity in motor neurons by the ETS gene Pea3.” Cell. 2006 Dec 29;127(7):1439-52. Vue TY, Kim EJ, Parras CM, Guillemot F, Johnson JE. “Ascl1 controls the number and distribution of astrocytes and oligodendrocytes in the gray matter and white matter of the spinal cord.” Development. 2014 Oct;141(19):3721-31. Wainger BJ, Kiskinis E, Mellin C, Wiskow O, Han SS, Sandoe J, Perez NP, Williams LA, Lee S, Boulting G, Berry JD, Brown RH Jr, Cudkowicz ME, Bean BP, Eggan K, Woolf CJ. “Intrinsic membrane hyperexcitability of amyotrophic lateral sclerosis patient-derived motor neurons.” Cell Rep. 2014 Apr 10;7(1):1-11. Wainger BJ, Buttermore ED, Oliveira JT, Mellin C, Lee S, Saber WA, Wang AJ, Ichida JK, Chiu IM, Barrett L, Huebner EA, Bilgin C, Tsujimoto N, Brenneis C, Kapur K, Rubin LL, Eggan K, Woolf CJ. “Modeling pain in vitro using nociceptor neurons reprogrammed from fibroblasts.” Nat Neurosci. 2015 Jan;18(1):17-24. Walmsley B, Tracey DJ. “An intracellular study of Renshaw cells.” Brain Res. 1981 Oct 26;223(1):170-5. Walmsley B, Wieniawa-Narkiewicz E, Nicol MJ. “Ultrastructural evidence related to presynaptic inhibition of primary muscle afferents in Clarke's column of the cat.” J Neurosci. 1987 Jan;7(1):236-43 Walmsley B, Graham B, Nicol MJ. “Serial E-M and simulation study of presynaptic inhibition along a group Ia collateral in the spinal cord.” J Neurophysiol. 1995 Aug;74(2):616-23. Wang YZ, Christakos S. “Retinoic acid regulates the expression of the calcium binding protein, calbindin-D28K.” Mol Endocrinol. 1995 Nov;9(11):1510-21. Wang X, Penzes P, Napoli JL. “Cloning of a cDNA encoding an aldehyde dehydrogenase and its expression in Escherichia coli.” Recognition of retinal as substrate.” J Biol Chem. 1996 Jul 5;271(27):16288-93. Wang S, Sdrulla AD, diSibio G, Bush G, Nofziger D, Hicks C, Weinmaster G, Barres BA. “Notch receptor activation inhibits oligodendrocyte differentiation.” Neuron. 1998 Jul;21(1):63-75. Wang S, Barres BA. “Up a notch: instructing gliogenesis.” Neuron. 2000 Aug;27(2):197-200. Wang VY, Rose MF, Zoghbi HY. “Math1 expression redefines the rhombic lip derivatives and reveals novel lineages within the brainstem and cerebellum.” Neuron. 2005 Oct 6;48(1):31-43. Wang Z, Gerstein M, Snyder M. “RNA-Seq: a revolutionary tool for transcriptomics.” Nat Rev Genet. 2009 Jan;10(1):57-63.

245

Wang RN, Green J, Wang Z, Deng Y, Qiao M, Peabody M, Zhang Q, Ye J, Yan Z, Denduluri S, Idowu O, Li M, Shen C, Hu A, Haydon RC, Kang R, Mok J, Lee MJ, Luu HL, Shi LL. “Bone Morphogenetic Protein (BMP) signaling in development and human diseases.” Genes Dis. 2014 Sep;1(1):87-105. Wang T, Birsoy K, Hughes NW, Krupczak KM, Post Y, Wei JJ, Lander ES, Sabatini DM. ““Identification and characterization of essential genes in the human genome.” Science. 2015 Nov 27;350(6264):1096-101. Watanabe K, Kamiya D, Nishiyama A, Katayama T, Nozaki S, Kawasaki H, Watanabe Y, Mizuseki K, Sasai Y. “Directed differentiation of telencephalic precursors from embryonic stem cells.” Nat Neurosci. 2005 Mar;8(3):288-96. Wenner P, O’Donovan MJ. “Identification of an interneuronal population that mediates recurrent inhibition of motoneurons in the developing chick spinal cord.” J Neurosci. 1999 Sep 1;19(17):7557-67. Wichterle H, Turnbull DH, Nery S, Fishell G, Alvarez-Buylla A. “In utero fate mapping reveals distinct migratory pathways and fates of neurons born in the mammalian basal forebrain.” Development. 2001 Oct;128(19):3759-71. Wichterle H, Lieberam I, Porter JA, Jessell TM. “Directed differentiation of embryonic stem cells into motor neurons.” Cell. 2002 Aug 9;110(3):385-97.

Wichterle H, Peljto M. “Differentiation of mouse embryonic stem cells to spinal motor neurons.” Curr Protoc Stem Cell Biol. 2008 May;Chapter 1:Unit 1H.1.1-1H.1.9.

Wichterle H, Peljto M, Nedelec S. “Xenotransplantation of embryonic stem cell-derived motor neurons into the developing chick spinal cord.” Methods Mol Biol. 2009;482:171-83.

Wickersham IR, Finke S, Conzelmann KK, Callaway EM. “Retrograde neuronal tracing with a deletion-mutant rabies virus.” Nat Methods. 2007 Jan;4(1):47-9. Wickersham IR, Lyon DC, Barnard RJ, Mori T, Finke S, Conzelmann KK, Young JA, Callaway EM. “Monosynaptic restriction of transsynaptic tracing from single, genetically targeted neurons.” Neuron. 2007 Mar 1;53(5):639-47. Wijgerde M, McMahon JA, Rule M, McMahon AP. “A direct requirement for Hedgehog signaling for normal specification of all ventral progenitor domains in the presumptive mammalian spinal cord.” Genes Dev. 2002 Nov 15;16(22):2849-64. Wildner H, Müller T, Cho SH, Bröhl D, Cepko CL, Guillemot F, Birchmeier C. “dILA neurons in the dorsal spinal cord are the product of terminal and non-terminal asymmetric progenitor cell divisions, and require Mash1 for their development.” Development. 2006 Jun;133(11):2105-13. Wildner H, Das Gupta R, Bröhl D, Heppenstall PA, Zeilhofer HU, Birchmeier C. “Genome-wide expression analysis of Ptf1a- and Ascl1-deficient mice reveals new markers for distinct dorsal horn interneuron populations contributing to nociceptive reflex plasticity.” J Neurosci. 2013 Apr 24;33(17):7299-307. Williams ME, Wilke SA, Daggett A, Davis E, Otto S, Ravi D, Ripley B, Bushong EA, Ellisman MH, Klein G, Ghosh A. “Cadherin-9 regulates synapse-specific differentiation in the developing hippocampus.” Neuron. 2011 Aug 25;71(4):640-55. Wilson L, Gale E, Chambers D, Maden M. “Retinoic acid and the control of dorsoventral patterning in the avian spinal cord.” Dev Biol. 2004 May 15;269(2):433-46. Wilson JJ, Kovall RA. “Crystal structure of the CSL-Notch-Mastermind ternary complex bound to DNA.” Cell. 2006 Mar 10;124(5):985-96.

246

Wilson V, Olivera-Martinez I, Storey KG. “Stem cells, signals and vertebrate body axis extension.” Development. 2009 May;136(10):1591-604. Wilson JM, Blagovechtchenski E, Brownstone RM. “Genetically defined inhibitory neurons in the mouse spinal cord dorsal horn: a possible source of rhythmic inhibition of motoneurons during fictive locomotion.” J Neurosci. 2010 Jan 20;30(3):1137-48. Windhorst U, Adam D, Inbar GF. “The effects of recurrent inhibitory feedback in shaping discharge patterns of motoneurones excited by phasic muscle stretches.” Biol Cybern. 1978 Jun 21;29(4):221-7. Windhorst U. “Activation of Renshaw cells.” Prog Neurobiol. 1990;35(2):135-79. Windhorst U. “On the role of recurrent inhibitory feedback in motor control.” Prog Neurobiol. 1996 Aug;49(6):517-87. Wine-Lee L, Ahn KJ, Richardson RD, Mishina Y, Lyons KM, Crenshaw EB 3rd. “Signaling through BMP type 1 receptors is required for development of interneuron cell types in the dorsal spinal cord.” Development. 2004 Nov;131(21):5393-403. Wingate RJ, Hatten ME. “The role of the rhombic lip in avian cerebellum development.” Development. 1999 Oct;126(20):4395-404. Wong ES, Ban KH, Mutalif R, Jenkins NA, Copeland NG, Stewart CL. “A simple procedure for the efficient derivation of mouse ES cells.” Methods Enzymol. 2010;476:265-83. Wootz H, Fitzsimons-Kantamneni E, Larhammar M, Rotterman TM, Enjin A, Patra K, André E, Van Zundert B, Kullander K, Alvarez FJ. “Alterations in the motor neuron-renshaw cell circuit in the Sod1(G93A) mouse model.” J Comp Neurol. 2013 May 1;521(7):1449-69. Wu S, Wu Y, Capecchi MR. “Motoneurons and oligodendrocytes are sequentially generated from neural stem cells but do not appear to share common lineage-restricted progenitors in vivo.” Development. 2006 Feb;133(4):581-90. Wu Y, Wang G, Scott SA, Capecchi MR. “Hoxc10 and Hoxd10 regulate mouse columnar, divisional and motor pool identity of lumbar motoneurons.” Development. 2008 Jan;135(1):171-82. Xie Z, Chen Y, Li Z, Bai G, Zhu Y, Yan R, Tan F, Chen YG, Guillemot F, Li L, Jing N. “Smad6 promotes neuronal differentiation in the intermediate zone of the dorsal neural tube by inhibition of the Wnt/beta-catenin pathway.” Proc Natl Acad Sci U S A. 2011 Jul 19;108(29):12119-24. Xu H, Whelan PJ, Wenner P. “Development of an inhibitory interneuronal circuit in the embryonic spinal cord.” J Neurophysiol. 2005 May;93(5):2922-33. Xu Q, Guo L, Moore H, Waclaw RR, Campbell K, Anderson SA. “Sonic hedgehog signaling confers ventral telencephalic progenitors with distinct cortical interneuron fates.” Neuron. 2010 Feb 11;65(3):328-40. Xu H, Iyer N, Huettner JE, Sakiyama-Elbert SE. “A puromycin selectable cell line for the enrichment of mouse embryonic stem cell-derived V3 interneurons.” Stem Cell Res Ther. 2015 Nov 10;6:220. Yamada M, Terao M, Terashima T, Fujiyama T, Kawaguchi Y, Nabeshima Y, Hoshino M. “Origin of climbing fiber neurons and their developmental dependence on Ptf1a.” J Neurosci. 2007 Oct 10;27(41):10924-34. Yamada M, Seto Y, Taya S, Owa T, Inoue YU, Inoue T, Kawaguchi Y, Nabeshima Y, Hoshino M. “Specification of spatial identities of cerebellar neuron progenitors by ptf1a and atoh1 for proper

247

production of GABAergic and glutamatergic neurons.” J Neurosci. 2014 Apr 2;34(14):4786-800. Yamagata M, Sanes JR. “Dscam and Sidekick proteins direct lamina-specific synaptic connections in vertebrate retina.” Nature. 2008 Jan 24;451(7177):465-9. Yang X, Tomita T, Wines-Samuelson M, Beglopoulos V, Tansey MG, Kopan R, Shen J. “Notch1 signaling influences v2 interneuron and motor neuron development in the spinal cord.” Dev Neurosci. 2006;28(1-2):102-17. Yang N, Ng YH, Pang ZP, Südhof TC, Wernig M. “Induced neuronal cells: how to make and define a neuron.” Cell Stem Cell. 2011 Dec 2;9(6):517-25. Yaron O, Farhy C, Marquardt T, Applebury M, Ashery-Padan R. “Notch1 functions to suppress cone-photoreceptor fate specification in the developing mouse retina.” Development. 2006 Apr;133(7):1367-78. Ye Z, Mostajo-Radji MA, Brown JR, Rouaux C, Tomassy GS, Hensch TK, Arlotta P. “Instructing Perisomatic Inhibition by Direct Lineage Reprogramming of Neocortical Projection Neurons.” Neuron. 2015 Nov 4;88(3):475-83. Yeo SY, Chitnis AB. “Jagged-mediated Notch signaling maintains proliferating neural progenitors and regulates cell diversity in the ventral spinal cord.” Proc Natl Acad Sci U S A. 2007 Apr 3;104(14):5913-8. Yoon K, Nery S, Rutlin ML, Radtke F, Fishell G, Gaiano N. “Fibroblast growth factor receptor signaling promotes radial glial identity and interacts with Notch1 signaling in telencephalic progenitors.” J Neurosci. 2004 Oct 27;24(43):9497-506. Yoon K, Gaiano N. “Notch signaling in the mammalian central nervous system: insights from mouse mutants.” Nat Neurosci. 2005 Jun;8(6):709-15. Zagoraiou L, Akay T, Martin JF, Brownstone RM, Jessell TM, Miles GB. “A cluster of cholinergic premotor interneurons modulates mouse locomotor activity.” Neuron. 2009 Dec 10;64(5):645-62. Zannino DA, Downes GB, Sagerström CG. “prdm12b specifies the p1 progenitor domain and reveals a role for V1 interneurons in swim movements.” Dev Biol. 2014 Jun 15;390(2):247-60. Zechner D, Müller T, Wende H, Walther I, Taketo MM, Crenshaw EB 3rd, Treier M, Birchmeier W, Birchmeier C. “Bmp and Wnt/beta-catenin signals control expression of the transcription factor Olig3 and the specification of spinal cord neurons.” Dev Biol. 2007 Mar 1;303(1):181-90. Zeilhofer HU, Wildner H, Yévenes GE. “Fast synaptic inhibition in spinal sensory processing and pain control.” Physiol Rev. 2012 Jan;92(1):193-235. Zeron MM, Hansson O, Chen N, Wellington CL, Leavitt BR, Brundin P, Hayden MR, Raymond LA. “Increased sensitivity to N-methyl-D-aspartate receptor-mediated excitotoxicity in a mouse model of Huntington's disease.” Neuron. 2002 Mar 14;33(6):849-60. Zhang JH, Morita Y, Hironaka T, Emson PC, Tohyama M. “Ontological study of calbindin-D28k-like and parvalbumin-like immunoreactivities in rat spinal cord and dorsal root ganglia.” J Comp Neurol. 1990 Dec 22;302(4):715-28. Zhang XM, Ramalho-Santos M, McMahon AP. “Smoothened mutants reveal redundant roles for Shh and Ihh signaling including regulation of L/R symmetry by the mouse node.” Cell. 2001;106:781–792. Zhang Y, Narayan S, Geiman E, Lanuza GM, Velasquez T, Shanks B, Akay T, Dyck J, Pearson K, Gosgnach S, Fan CM, Goulding M. “V3 spinal neurons establish a robust and balanced locomotor rhythm during walking.” Neuron. 2008 Oct 9;60(1):84-96.

248

Zhang J, Lanuza GM, Britz O, Wang Z, Siembab VC, Zhang Y, Velasquez T, Alvarez FJ, Frank E, Goulding M. “V1 and v2b interneurons secure the alternating flexor-extensor motor activity mice require for limbed locomotion.” Neuron. 2014 Apr 2;82(1):138-50. Zhao D, McCaffery P, Ivins KJ, Neve RL, Hogan P, Chin WW, Dräger UC. “Molecular identification of a major retinoic-acid-synthesizing enzyme, a retinaldehyde-specific dehydrogenase.” Eur J Biochem. 1996 Aug 15;240(1):15-22. Zhao Y, Kwan KM, Mailloux CM, Lee WK, Grinberg A, Wurst W, Behringer RR, Westphal H. “LIM-homeodomain proteins Lhx1 and Lhx5, and their cofactor Ldb1, control Purkinje cell differentiation in the developing cerebellum.” Proc Natl Acad Sci U S A. 2007 Aug 7;104(32):13182-6. Zhou Q, Choi G, Anderson DJ. “The bHLH transcription factor Olig2 promotes oligodendrocyte differentiation in collaboration with Nkx2.2.” Neuron. 2001 Sep 13;31(5):791-807. Zhou Q, Anderson DJ. “The bHLH transcription factors OLIG2 and OLIG1 couple neuronal and glial subtype specification.” Cell. 2002 Apr 5;109(1):61-73. Zimmerman LB, De Jesús-Escobar JM, Harland RM. “The Spemann organizer signal noggin binds and inactivates bone morphogenetic protein 4.” Cell. 1996 Aug 23;86(4):599-606. Zou M, Li S, Klein WH, Xiang M. “Brn3a/Pou4f1 regulates dorsal root ganglion sensory neuron specification and axonal projection into the spinal cord.” Dev Biol. 2012 Apr 15;364(2):114-27. Zou M, Luo H, Xiang M. “Selective neuronal lineages derived from Dll4-expressing progenitors/precursors in the retina and spinal cord.” Dev Dyn. 2015 Jan;244(1):86-97.

Subtype diversification and synaptic specificity of stem cell ...

Documents