Neuronal Subtype-Specific Genes that Control 2005 full cover.pdf · Neuronal Subtype-Specific Genes that Control ... to the analysis of neuronal subtype-specific genes in 1R ... and

Neuron, Vol. 45, 207–221, January 20, 2005, Copyright ©2005 by Elsevier Inc. DOI 10.1016/j.neuron.2004.12.036

Neuronal Subtype-Specific Genes that ControlCorticospinal Motor Neuron Development In Vivo

2002). In contrast, much less is known about the geneticprograms controlling the later specification and differen-tiation of distinct neuronal subtypes (Rallu et al., 2002).

Paola Arlotta,1,3 Bradley J. Molyneaux,1,3

Jinhui Chen,1,4 Jun Inoue,2 Ryo Kominami,2

and Jeffrey D. Macklis1,*Recent success elucidating the genetic determinants1MGH-HMS Center for Nervous System Repairof neuronal subtype specification has been limited toDepartments of Neurosurgery and Neurologydistinct regions of the mammalian CNS, principally spi-Program in Neuroscience andnal cord (Jessell, 2000) and retina (Cepko, 1999).Harvard Stem Cell Institute

In the mammalian neocortex, the identification ofHarvard Medical Schoolgenes that are determinants of neuronal subtypes isMassachusetts General Hospitalcomplicated by greater cellular complexity comparedBoston, Massachusetts 02114to many CNS regions. Here, many different classes of2 Department of Molecular Geneticsprojection neurons are born in a tightly controlled tem-Graduate School of Medical and Dental Sciencesporal order and position themselves in the developingNiigata Universitycortex to produce the six-layered structure seen in theAsahimachi 1-757adult mammal (Bayer and Altman, 1991). Each corticalNiigata 951-8122layer contains one or more distinct subtypes of projec-Japantion neurons that project to different ipsilateral or con-tralateral cortical, subcortical, or subcerebral targets.Although decades of elegant studies into cortical devel-Summaryopment have provided remarkable knowledge about theanatomical and cellular organization of the mammalianWithin the vertebrate nervous system, the presencecortex, the genetic mechanisms that control its complexof many different lineages of neurons and glia compli-neuronal development and diversity are much lesscates the molecular characterization of single neu-known.ronal populations. In order to elucidate molecular

In this report, we focus on the critical population ofmechanisms underlying the specification and devel-corticospinal motor neurons (CSMN), located primarilyopment of corticospinal motor neurons (CSMN), wein cortical layer V, and directly identify genetic determi-purified CSMN at distinct stages of developmentnants of this specific neuronal subtype. CSMN (“upperin vivo and compared their gene expression to twomotor neuron”) degeneration is a key component ofother pure populations of cortical projection neurons:motor neuron degenerative diseases, including amyo-

callosal projection neurons and corticotectal projec-trophic lateral sclerosis (ALS), and CSMN injury contrib-

tion neurons. We found genes that are potentially in- utes critically to the loss of motor function in spinal cordstructive for CSMN development, as well as genes that injury. The anatomical and morphological developmentare excluded from CSMN and are restricted to other of CSMN has been extensively characterized (Jones etpopulations of neurons, even within the same cortical al., 1982; Terashima, 1995), but strategies to repairlayer. Loss-of-function experiments in null mutant CSMN are limited by a lack of understanding of themice for Ctip2 (also known as Bcl11b), one of the newly molecular controls over CSMN development, includingcharacterized genes, demonstrate that it plays a criti- neuron type-specific differentiation, survival, and con-cal role in the development of CSMN axonal projec- nectivity.tions to the spinal cord in vivo, confirming that we A few isolated molecules specifically associated withidentified central genetic determinants of the CSMN CSMN and related cortical neurons have been identified.population. These include Otx1, a transcription factor expressed in

layers V and VI (Frantz et al., 1994; Weimann et al.,Introduction 1999); Er81, a transcription factor of unknown function

expressed by multiple neuronal subtypes in layer V(Hevner et al., 2003); and molecules involved in axonalDuring the development of the central nervous system,pathfinding expressed in several types of neurons, in-neuronal progenitors undergo precise stepwise differen-cluding those with projections along the corticospinaltiation to ultimately produce the complex variety of neu-tract (Coonan et al., 2001; Rolf et al., 2002).ronal subtypes that populate the mature brain. Extensive

Gene expression studies used to detect transcriptswork has progressively unraveled the molecular mecha-that are present in only selected neocortical neurons ornisms controlling processes of early neuronal specifica-that are expressed at low levels are typically compli-tion and has identified transcription factors and fatecated by the cellular heterogeneity of the neocortexdetermination genes that mediate early aspects of neu-(Geschwind, 2000). This is a more general problem ofrogenesis in several regions of the CNS (Bertrand et al.,gene expression studies in the CNS, which has led tothe development of approaches aimed at simplifying the

*Correspondence: [email protected] composition of samples via single-cell transcrip-3 These authors contributed equally to this work.tion analysis (Kamme et al., 2003; Tietjen et al., 2003)4 Current address: Department of Anatomy and Neurobiology, Spinaland to the development of sophisticated statistical anal-Cord and Brain Injury Research Center, University of Kentucky, Lex-

ington, Kentucky 40536. ysis of microarray data derived from more complex tis-

Neuron208

sues (Diaz et al., 2002). Application of these approaches after dissociation and then were FACS purified (Figures1R and 1S). After FACS, P14 CSMN retain elements ofto the analysis of neuronal subtype-specific genes in

neocortex is fundamentally limited by the substantial their original in vivo morphology, including a proximalapical dendrite and occasionally a proximal axon (Fig-lack of antigenic markers with which to discriminate

among different neuronal subtypes. ure 1S�).As comparison neuronal populations, we purified in-We overcame these issues by purifying CSMN and

two closely inter-related neuronal subtypes (callosal terhemispheric callosal neurons, a subset of which sharelamina V location with CSMN, providing insight intoprojection neurons and corticotectal projection neu-

rons) from murine neocortex at four distinct stages of genes that are cell type specific rather than laminarspecific; and corticotectal neurons, which share withdevelopment (E18, P3, P6, and P14), using fluorescence-

activated cell sorting (FACS). These stages span critical CSMN both location in lamina V and overlapping earlydevelopmental extension of subcerebral projections.events of CSMN specification, morphologic maturation,

and connectivity. Using microarrays, we compared the Corticotectal neurons might allow the identification ofgenes unique to CSMN among other highly related layergene expression of purified CSMN and these two other

pure neuronal subtypes. We hypothesized that there are V subcerebral projection neurons. These data demon-strate that CSMN and other subtypes of cortical neuronslikely many genes that are common to all projection

neurons and that there is a smaller number of genes can be purified from the complex and heterogeneousneocortex at distinct and critical developmental stagesthat are more specific to closely related subtypes of

projection neurons (e.g., CSMN, corticotectal neurons, in vivo.or other subcerebral projection neurons, but not callosalneurons). Ultimately, the specification of CSMN most Expression Profiling of Projectionlikely derives from the overlapping combinatorial ex- Neuron Subtypespression of a specific program of genes in CSMN. We Gene expression analysis in brain is generally compli-find genes that are progressively restricted and specific cated by the coexistence of many different cell types,to CSMN, as well as genes excluded from CSMN and resulting in high background noise and the inability torestricted to other neuronal populations, even within the detect small differences in cell type-specific gene ex-same cortical layer. Confirmatory analysis of CSMN- pression. The purification of single populations of corti-specific gene expression for selected candidate genes cal projection neurons allows us to overcome these diffi-of particular mechanistic interest and functional analysis culties and compare the expression profiles of distinctin mutant mice in vivo indicate that these genes are part neuronal populations without confounding contamina-of a combinatorial program of novel genetic determi- tion by other cell types. We used pure populations ofnants of the CSMN population. CSMN from E18, P3, P6, and P14 mice and compared

them by microarray (Affymetrix 430A GeneChips) to twoother neuronal types: callosal neurons and corticotectalResultsneurons. To control rigorously for biological sample vari-ability, at each age we used CSMN, callosal neurons,Purification of Corticospinal Motor Neuronsand corticotectal neurons from two independent sam-To identify genes that control cell type specification andples derived from different litters and independentlydifferentiation of CSMN, we compared gene expressionFACS purified and hybridized (true biological replicates)profiles of CSMN to two other pure populations of corti-(Supplemental Figure S1E [http://www.neuron.org/cgi/cal projection neurons: callosal projection neurons andcontent/full/45/2/207/DC1/]). To assess the consistencycorticotectal projection neurons.of the microarray data, we compared gene expressionCSMN were retrogradely labeled by injecting greenbetween independent homotypic samples (biologicalfluorescent microspheres into their axonal projectionreplicates) (Supplemental Figures S1A and S1B) andfields: the pons at E18 and P3 and the cervical spinalbetween heterotypic samples (different neuronal sub-cord at P6 and P14. Embryonic injections were per-types) (Supplemental Figures S1C and S1D). Prior toformed under ultrasound guidance to accurately controlnormalization, correlation coefficients for biological rep-the injection location (Figures 1A–1C). This strategy spe-licates ranged from 0.94 to 0.99 (Supplemental Figurecifically labels CSMN somas in sensorimotor cortex (Fig-S1E) and hybridization replicates were all �0.99. Theseure 1D) based on their axonal projections, at four agescomparisons indicate that gene expression is very highlyranging from early postmitotic (E18) to more differenti-consistent between biological replicates, whereas manyated (P3–P6) to more mature and synaptically integratedgenes are differentially expressed in heterotypic com-neurons (P14) (Figures 1E–1H). Similar techniques wereparisons, supporting the reliability of the dataset.used to label callosal neurons (Figures 1I–1L) and corti-

cotectal neurons (Figure 1M). Dissociated, labeled CSMNwere purified by FACS to typically �99% purity (Fig- Distinct Classes of Cortical Projection Neurons

Share Clusters of Developmentallyures 1N–1S�).CSMN were collected for RNA isolation immediately Regulated Genes

The development of CSMN is likely controlled by a com-following FACS purification. Even though CSMN are veryfragile neurons, acutely FACS-purified E18, P3, and P6 bination of (1) general molecular pathways common to

all projection neurons and (2) a specific combination ofCSMN are viable and can be cultured in vitro (Ozdinleret al., personal communication), confirming their health genes highly enriched or restricted to CSMN. To obtain a

global view of genes in the first set, we used hierarchicalfollowing FACS. Because neurons at P14 are even morefragile, P14 CSMN were fixed in RNAlater immediately clustering to examine changes in gene expression as

Molecular Development of Corticospinal Neurons209

Figure 1. Population-Specific FACS Purification of CSMN, Callosal Neurons, and Corticotectal Neurons during Development In Vivo

(A–C) In utero ultrasound-guided microinjection of fluorescent microspheres into the pons of an E17 mouse embryo showing (A) the initialpositioning of the glass micropipette (arrowheads), (B) injection at the pons/midbrain junction (arrow), and (C) the pons postinjection. (D)Dorsal view of a P14 brain retrogradely labeled from the C5 level of the cervical spinal cord, showing labeling of CSMN in sensorimotor cortex.(E–H) CSMN and (I–L) callosal projection neurons (CPN) labeled with green fluorescent microspheres in E18, P3, P6, and P14 neocortex. (M)Sagittal P14 brain section, showing labeling of CSMN (red; arrowheads) and corticotectal projection neurons (CTPN; green; arrows), in thesame mouse. Labels II/III and V indicate cortical laminae; pia, pial surface; ob, olfactory bulb; cb, cerebellum. (N and Q) Sample FACS plotof the population of CSMN selected; CSMN are selected as (N) a highly fluorescent population (R2; right peak) and (Q) based on size (forwardscatter) and surface characteristics (side scatter). (O and R) Mixed cortical cells before FACS purification; only a very small percentage ofdissociated cells are CSMN (arrows). (P and S) FACS purification of CSMN results in an essentially pure, retrogradely labeled population. (S�)FACS purified P14 CSMN fixed in RNAlater often retain short proximal dendritic and/or axonal processes. Scale bars, (A–C) 500 �m, (E–M)100 �m, (O, P, R, and S) 20 �m, (S�) 10 �m.

cortical projection neurons differentiate and mature opment of distinct classes of cortical projection neuronsis controlled by a large number of genes common to(Supplemental Figure S2A [http://www.neuron.org/cgi/

content/full/45/2/207/DC1/]). As expected, we found all projection neuron subtypes, in combination with asmaller set of subtype-restricted molecules.that the majority of genes are expressed with similar

profiles in different projection neuron subtypes (Supple- This analysis also shows that P14 CSMN are molecu-larly very similar to corticotectal neurons, while both aremental Figures S2B–S2D and B.J.M. et al., unpublished

data). These data support the hypothesis that the devel- distinct from callosal neurons (data not shown). This

Neuron210

likely reflects the fact that both CSMN and corticotectal suo-Takasaki et al., 2000); Pcp4 (Sangameswaran et al.,1989); S100a10 (Saris et al., 1987); Mu-Crystallin (Sego-neurons may have similar requirements for survival andvia et al., 1997); Netrin-G1 (Yin et al., 2002); Cadherinsubcerebral connectivity.13 (Huang et al., 2003); Cadherin 22 (Sugimoto et al.,1996); and one novel EST that we name Csmn1. TheseIdentification of CSMN-Specific Genesare all largely undescribed molecules in cortex that haveTo identify CSMN-specific genes, we assessed the sig-microarray expression profiles strongly indicating sub-nificance of differences in gene expression among neu-type-specific expression in CSMN and/or other subcere-ronal subtypes by pairwise comparisons at each age,bral projection neurons (Figures 3B� and 3D�–3P�).using the SAM method. We selected the 100 most signif-

We find that all of these genes have high levels oficant genes from each pairwise comparison of all threeexpression in layer V of cortex, where they are stronglyneuronal populations performed at each age (total ofexpressed in morphologically identified CSMN (Figures884 unique genes) and further analyzed the trend of3A–3P), confirming and extending the microarray re-expression of each individual gene to define a smallersults. By in situ hybridization, these genes show differentset of molecules of potentially high biological relevance.degrees of restriction to CSMN. Three genes—Diap3,We identified genes specifically expressed in CSMNIgfbp4, and Crim1—demonstrate particularly interestingas well as genes specific to callosal neurons and cortico-and very restricted patterns of expression that distin-tectal neurons that, importantly, serve as negative mo-guish CSMN from other subcerebral projection neurons.lecular markers of CSMN. Some of the most biologicallyDiap3 is expressed only in sensorimotor layer V whereand statistically significant genes are described in Fig-CSMN are located, while it is not expressed in moreure 2. These genes were classified into one of six groupslateral (Figure 3A) or caudal (Figure 3B) areas of layer Vbased on expression profiles suggestive of a specificwhere other subcerebral projection neurons (e.g., corti-role in distinct aspects of CSMN development: (1) genescotectal neurons) are located. Igfbp4 exhibits a similarthat are expressed at higher levels in CSMN at all stagesdegree of restriction to CSMN in sensorimotor layer Vof development (Figure 2A) and might be important for(Figures 3C and 3D), although it is also expressed inthe establishment and maintenance of CSMN identity;other populations in layers II/III and VI. Crim1 is re-(2) genes that are highly expressed in CSMN early instricted to layer V, with high level expression in rostraldevelopment (Figure 2B) and might be important forsensorimotor cortex (Figure 3E). These three genes ap-early CSMN specification; (3) and (4) genes that exhibitpear to be area-specific markers that identify the loca-increasing levels of expression as CSMN develop andtion of CSMN in layer V along mediolateral and rostro-might control intermediate and later aspects of CSMNcaudal axes.differentiation, such as process outgrowth and synapse

A larger group of genes—Ctip2, Encephalopsin, Clim1,formation (Figures 2C and 2D); (5) genes that are ex-Fez, Pcp4, and S100a10—appear to be expressed inpressed at higher levels in CSMN compared to the highlyCSMN and the broader class of closely related subcere-related population of corticotectal neurons (Figure 2E)bral projection neurons in layer V (Figures 3F–3K). Inand are representative of the small class of genes thatcontrast, Mu-Crystallin and Netrin-G1 appear to be ex-

differentiate CSMN from other subcerebral projectionpressed only in some CSMN and subcerebral neurons

neurons of layer V; (6) genes that are negative markers ofof layer V (Figures 3L and 3M) and may delineate distinct

CSMN but that are expressed in callosal or corticotectalfunctional classes. Consistent with the microarray data,

neurons (Figure 2F). Expression trends of these genes other genes—Csmn1, Cadherin 13, and Cadherin 22—are shown in Figure 3 and Supplemental Figure S3 show less restricted patterns of expression but are ex-(http://www.neuron.org/cgi/content/full/45/2/207/ pressed at much higher levels in subcerebral neuronsDC1/). (Figures 3N–3P). Together, these data support the hy-

The CSMN genes identified include transcription acti- pothesis that a small number of CSMN restricted genes,vators and repressors (e.g., Ctip2 [also known as along with a larger group of genes that are also ex-Bcl11b], Bcl6, Sox5); zinc finger domain-containing pro- pressed in other subcerebral neurons, define the molec-teins (e.g., Fez); cell surface proteins and receptors (e.g., ular phenotype of CSMN.Encephalopsin, Itm2a, Daf1); calcium signaling mole- To further confirm the cellular identity and projectioncules (e.g., Pcp4, S100a10); genes involved in neuronal neuron type of labeled cells identified by in situ hybrid-specification (e.g., Crim1), cell adhesion (e.g., Cdh22, ization, we combined in situ hybridization with retro-Cdh13, Cntn6), and axon guidance (e.g., Neto1, Netrin grade labeling in the same tissue. We retrogradely la-G1); as well as genes involved in critical pathways like beled CSMN with DiI, photoconverted the DiI to a visiblethe thyroid hormone and IGF signaling cascades (e.g., cytoplasmic precipitate, and performed nonradioactiveMu-Crystallin, Igfbp4). in situ hybridization. Combining these methods allows

To confirm CSMN-specific gene expression and vali- the colocalization of the DiI photoconverted precipitatedate the microarray data, we performed in situ hybridiza- and the in situ hybridization signal, enabling us to iden-tion or immunocytochemistry for 14 selected genes (Fig- tify CSMN expressing individual genes. We investigatedure 2, bolded) chosen for their particularly interesting four genes from Figure 3 and confirmed that CSMNpatterns of expression and potential function based on express Mu-Crystallin (Supplemental Figures S4A, S4B,protein domains or literature from other systems. We and S4B� [http://www.neuron.org/cgi/content/full/45/2/chose Diap3 (Olson, 2003); Igfbp4 (Stenvers et al., 1994); 207/DC1/]), Fez (Supplemental Figures S4C, S4D, andCrim1 (Kolle et al., 2000); Ctip2 (Avram et al., 2000); S4D�), Encephalopsin (Supplemental Figures S4E, S4F,Encephalopsin (Blackshaw and Snyder, 1999); Clim1 and S4F�), and Crim1 (Supplemental Figures S4G, S4H,

and S4H�).(also known as Ldb2) (Bulchand et al., 2003); Fez (Mat-

Molecular Development of Corticospinal Neurons211

Figure 2. A Subset of CSMN-Specific Genes from Microarray Analysis, Classified Based on Expression Profiles Suggesting Biological Rolesduring CSMN Development

A subset of biologically interesting genes is shown, selected from a larger group of differentially expressed genes. Each group is representedby a prototypical expression profile shown at left. The genes shown in bold are those selected for further analysis in this study. (A) Genesthat are expressed at higher levels in CSMN at all stages of development; (B) genes that are highly expressed in CSMN early in development;(C and D) genes that exhibit increasing levels of expression as CSMN develop; (E) genes that are expressed at higher levels in CSMN comparedto the closely related population of corticotectal neurons; and (F) genes that are expressed at high levels in other populations of corticalprojection neurons, but not in CSMN, thus serving as negative markers for CSMN. Graphic gene expression profiles are shown for all othergenes listed either in Figure 3 (for those listed in bold) or in Supplemental Figure S3.

Molecular Development of Corticospinal Neurons213

These data provide strong evidence that we identified To test the hypothesis that CTIP2 controls some as-pects of CSMN and subcerebral projection neuron de-novel or previously uncharacterized genes that are spe-

cific to CSMN. Of interest, while all are expressed by velopment, we first defined its cell type-specific patternof expression in cortex and confirmed that it is ex-CSMN, each has a different pattern of expression, and

it is likely their combinatorial interaction that defines pressed at high levels in CSMN and corticotectal neu-rons, but not callosal neurons. In addition to the predom-CSMN. Together with negative markers, they allow pro-

gressive definition of the molecular phenotype of inant population of subcerebral projection neurons inlayer V that express CTIP2 at very high levels, we alsoCSMN in vivo.observed a much lower level of CTIP2 expression incorticothalamic neurons and GABAergic neurons (B.J.M.Lmo4 Is Not Expressed in CSMN and Is Restrictedet al., unpublished data).to Callosal Neurons in Layer V

We found that CTIP2 is expressed at high levels inAs cell fate specification and maturation of CSMN willlayer V of cortex in a pattern that extends across thelikely depend on both positive and negative molecularentire rostrocaudal (Figure 4A) and mediolateral (Figuredeterminants, we further characterized one additional5D) aspects of cortex. This is consistent with our mi-gene, Lmo4 (Supplemental Figure S5A [http://www.croarray data, showing high levels of expression in bothneuron.org/cgi/content/full/45/2/207/DC1/]), a LIMCSMN and corticotectal neurons. Since high level CTIP2domain-containing protein known to be expressed inexpression extends beyond the boundaries of motorlayers II/III and V (Bulchand et al., 2003), although itscortex but is restricted to layer V, we hypothesized thatcell type-specific expression within these layers was notCTIP2 is expressed at high levels in all subcerebral pro-previously known. Lmo4 and other genes expressed injection neurons, but not in cortico-cortical projectioncallosal or corticotectal neurons but not in CSMN (Figureneurons (e.g., callosal neurons) or in other locally inte-2F) can serve as negative markers of CSMN.grated neurons of layer V.Microarray analysis indicated that Lmo4 is specifically

To test this hypothesis, we performed a series of ex-expressed in callosal neurons but not in CSMN (Figureperiments in which we selectively retrogradely labeled2F). We confirmed and extended these results and foundindividual populations of projection neurons. First, weby immunocytochemistry that LMO4 is expressed in allinjected FluoroGold (FG) into the pyramidal tract at thecallosal neurons but not in CSMN (Supplemental Figurespons-midbrain junction to label subcerebral projectionS5A–S5L [http://www.neuron.org/cgi/content/full/45/2/neurons (Figure 4B). We found that all subcerebral pro-207/DC1/]). These data demonstrate that LMO4 is a neu-jection neurons located in layer V expressed CTIP2 atronal subtype-restricted gene and, most importantly tohigh levels (Figures 4C and 4H–4K). Next, we labeledthe present study, is a negative marker of CSMN thatCSMN specifically via FG injections into the cervicaldistinguishes CSMN from other neuronal types withinspinal cord and confirmed that all CSMN express CTIP2layer V.at high levels (Figures 4D–4G). Conversely, FG labelingof callosal neurons via injection in contralateral cortex

CTIP2 Is Expressed in CSMN but Notrevealed that CTIP2 was not expressed by the relatively

in Callosal Neurons in Layer Vsmall number of callosal neurons in layer V, where CTIP2

To begin to understand the functional roles of selectedexpression is highest (Figures 4L–4O), nor by callosal

CSMN-specific molecules, we characterized more pre-neurons in layer II/III or layer VI (B.J.M. et al., unpublished

cisely the CSMN expression of Ctip2, a gene of yetdata). Together, these data demonstrate that CTIP2 is

unknown function in the brain that shows a very higha neuronal subtype-specific, not simply a layer-specific,

level of expression in layer V in both CSMN and cortico-marker of CSMN and other evolutionarily related popula-

tectal neurons (Figures 3F, 3F�, and 4A). Very recenttions of neurons with subcerebral projections.

studies have shown that COUP-TF1 interacting protein2 (CTIP2) has critical roles in the immune system, con-trolling T cell subtype specification and survival in the CTIP2 Is Expressed in Developing Cortical Plate

and in CSMN in Layer Vdeveloping thymus (Wakabayashi et al., 2003). Thesedata suggested to us that it might have similar, yet undis- To better understand the functional role of CTIP2 in

CSMN development, we investigated its temporalcovered, roles in specification, maintenance, and/orconnectivity of distinct neuronal populations in the ner- course of expression through embryonic and postnatal

cortical development. At E12, when early cortical pro-vous system, specifically of CSMN and other subcere-bral projection neurons. genitors are dividing, CTIP2 is expressed in only a small

Figure 3. Genes Identified from the Microarray Analysis Are Expressed in CSMN

(A–P) In situ hybridization in coronal (A, C, and E–P) or sagittal (B and D) sections of cortex, showing specific expression of all 14 genesselected in the morphologically distinct population of CSMN (insets, enlarged from boxed areas; small arrows) in layer V. Red arrows indicatethe limit of gene expression in the mediolateral (A, C, and E) and rostrocaudal (B and D) axes. Black arrows in (B) and (D) indicate sensorimotorcortex, where Diap3 and Igfbp4 are expressed; arrowheads indicate visual cortex where Diap3 and Igbp4 expression was not detected. Agesare P0 (Pcp4), P3 (CTIP2, Cadherin 13, S100a10), P6 (Crim1, Clim1), P14 (Diap3, Igfbp4, Fez, Encephalopsin, Mu-Crystallin, Netrin-G1, Csmn1,Cadherin 22). (B�, and D�–P�) Temporal profiles of gene expression from microarray analysis of each selected gene in CSMN (blue) and callosalneurons (red). Bars indicate standard errors of the mean. Expression in corticotectal neurons (CTPN) closely resembles that in CSMN (datanot shown), with the exception of a restricted set of genes that discriminate between these closely related projection neuron populations(e.g., Diap3, Igfbp4, and Crim1). Scale bars, (A–P) 100 �m, ([A–P], inset) 20 �m.

Neuron214

Figure 4. CTIP2 Is Expressed in CSMN and Subcerebral Projection Neurons of Layer V but Not in Callosal Neurons

(A–C) Sagittal mouse brain section at P6, showing (A) labeling of large projection neurons in layer V with anti-CTIP2 antibody (arrows) and (B)FG labeling of subcerebral projection neurons in layer V. (C) Merge of (A) and (B), showing CTIP2 expression in subcerebral projection neurons.(D) Coronal section of cortex at P6, showing high levels of CTIP2 expression in layer V (red) and FG staining of CSMN in the same layer(green). (E) High-magnification FG labeling of CSMN and (F) CTIP2 expression in the boxed area in (D). (G) Merged image of (E) and (F),showing CTIP2 expression in all CSMN. (H) Coronal section of cortex at P6, showing high levels of CTIP2 expression in layer V (red) and FG-labeled subcerebral projection neurons in layer V (green). (I) High-magnification image of FG labeling of subcerebral projection neurons and(J) CTIP2 expression in the boxed area in (H). (K) Merged image of (I) and (J), showing CTIP2 expression in essentially all subcerebral projectionneurons. (L) Coronal section of cortex at P6, showing high levels of CTIP2 expression in layer V (red) and FG-labeled callosal neurons (green).(M) High-magnification FG labeling of callosal neurons and (N) CTIP2 expression in the boxed area in (L). (O) Merged image of (M) and (N),showing exclusion of CTIP2 from callosal neurons. Scale bars, (A–C) 100 �m, (D, H, and L) 50 �m, (E–G, I–K, and M–O) 10 �m.

cluster of cells in ventrolateral cortex (Figure 5A). In and other subcerebral projection neurons once theyreach the cortical plate (Figures 5B and 5C). These datacontrast, no cells expressing CTIP2 are visible in either

the ventricular or subventricular zones, where neural suggested that CTIP2 might control final CSMN posi-tioning in layer V, or, alternatively, CSMN postmitoticprecursors are located that give rise to cortical projec-

tion neurons. This suggests that CTIP2 is not involved in differentiation, including process outgrowth and path-finding and/or survival. The second alternative appearsearly specification of cortical precursors. At E14, during

peak production of CSMN, and at E16, when the majority more likely because (1) CTIP2 is not expressed in otherneurons that also take position in developing layer Vof CSMN have reached the cortical plate, CTIP2 is highly

expressed by cells in the cortical plate, but not by cells (i.e., callosal neurons); rather, CTIP2 exhibits restrictedexpression to CSMN and other related neurons within the ventricular or subventricular zones, suggesting

that CTIP2 begins to be expressed in postmitotic CSMN similar long-distance subcerebral connections (Figure

Molecular Development of Corticospinal Neurons215

pathfinding, resulting in failure of CSMN to connect tothe spinal cord. In addition, in Ctip2�/� mice, reducedCTIP2 expression results in abnormal developmentalpruning of corticospinal axons.

To investigate whether the neocortex of Ctip2�/� miceis abnormal, we first compared the cortical architectureof P0 Ctip2�/� mice to wild-type littermate controls. Welabeled cortical layers using markers of distinct corticallaminae: LMO4 (layers II/III and V); ER81 (layer V); andTBR1 (layers I and VI and subplate). At the level of cellularresolution of these markers, the cortex of Ctip2�/� miceappears normal, suggesting that lack of Ctip2 does notresult in widespread death of neuronal populations incortex, nor in neuronal lamination defects (data notshown). While more subtle cytoarchitectural abnormali-ties might be present and/or cortical neuronal cell deathmight occur at later developmental stages, as sug-gested by our observation of widespread neuronal deathin several other CNS areas where CTIP2 is expressed(B.J.M. et al., unpublished data), the early neonatal le-thality of Ctip2�/� mice makes it currently difficult to fullytest these possibilities.

Very interestingly, however, despite the fact thatCtip2�/� mice die at P0, well before CSMN axons connectto final targets in the spinal cord, Ctip2�/� mice displaystriking abnormalities of axonal fiber tracts exiting neo-cortex and forming the internal capsule. Specifically,Ctip2�/� mice have substantial disorganization of thecortical axon fascicles that normally perforate the stria-tum to form the internal capsule (Figure 6). The effectof CTIP2 in such axonal fasciculation and extensiondefects appears specific to only distinct types of subcer-ebral and/or subcortical axons, since other fiber tracts(e.g., corpus callosum) appear normal (Figures 6A andFigure 5. CTIP2 Is Expressed in the Developing Cortical Plate and6D). This is consistent with our findings that callosalin Neocortical Layer Vneurons do not normally express CTIP2 (Figure 4).(A) At E12, no expression of CTIP2 is detected in the preplate (PP);

expression is limited to far lateral developing cortex. (B) At E14, To more closely examine axonal projections of CSMNCTIP2 is expressed at high levels (arrows) in the developing cortical in Ctip2�/� mice, we first compared the corticospinalplate (CP) and developing striatum (asterisk), but not in the ventricu- fiber tracts of Ctip2�/� mice to those of wild-type lit-lar zone or overlying subventricular zone (dashed line near ventricle, termates at P0, using immunocytochemistry for L1, aLV). (C) At E16, CTIP2 is expressed in the early developing neurons

member of the CAM family of cell adhesion moleculesof deep cortical layers (arrows) and in the striatum (asterisks). (D)known to be expressed by CSMN projections and se-Expression is maintained at high levels in layer V of cortex andlected fiber tracts in the CNS (Fujimori et al., 2000). Westriatum at P3. (E) Sagittal section at P6, showing high-level expres-

sion of CTIP2 in layer V of neocortex along the rostral to caudal found that P0 Ctip2�/� mice lacked the typical fascicu-axis and in the striatum (asterisk), hippocampus (hp), and olfactory lated bundles of subcerebral projection fibers that nor-bulb (ob). Scale bars, (A–E) 100 �m. Dotted lines indicate pial surface mally form the internal capsule (n � 5). This abnormality(Pia), corpus callosum (cc), and ventricular margin (LV). is quite distinct along the entire rostrocaudal axis of the

internal capsule (Figures 6B, 6C, 6E–6H, 6J, and 6K).Interestingly, some highly disorganized and nonfascicu-

4); (2) high levels of CTIP2 expression are observed in lated Ctip2�/� axonal projections deviated dramaticallypostmitotic immature neurons that have just started to from their normal path (Figure 6K), coursing obliquelyextend an axon (E14–E18); and (3) mice with a targeted and transverse to other axons of the internal capsule.deletion of COUP-TF1 (a major interacting protein of To illuminate the fine axonal architecture of abnormalCTIP2) display axonal pathfinding defects (Zhou et al., nonfasciculated internal capsule fibers in Ctip2�/� mice,1999). we performed anterograde DiI tracing of these fiber

tracts by placing DiI crystals in cortex of E18 Ctip2�/�

Ctip2�/� Mice Fail to Form Cortical Connections mice and wild-type controls. Confocal analysis of DiI-to the Spinal Cord labeled fibers in the internal capsule further confirmedTo test these alternative hypotheses and to define the this axon growth and fasciculation defect by showingfunction of CTIP2 in vivo, we investigated homozygous that Ctip2�/� axons were present as individual, nonfas-Ctip2�/� and heterozygous Ctip2�/� mice (Wakabayashi ciculated, and disorganized fibers (Figures 6I and 6L).et al., 2003). Ctip2�/� mice are born alive but die soon Some disorganized axons possess what appear to beafter birth (P0). We found that, in Ctip2�/� mice, CSMN abnormal, bulbous varicosities and dysmorphic growth

cones (Figure 6L) suggestive of those first described byaxons exhibit defects in fasciculation, outgrowth, and

Neuron216

Figure 6. Ctip2�/� Mice Display Defects inSubcerebral Axon Extension and Fascicula-tion in the Internal Capsule

(A) Wild-type brain section at P0, stained withcresyl violet, showing the typical axonal fasci-cles of the internal capsule (arrows) and cor-pus callosum. (D) Matched section from aCtip2�/� null mutant brain, demonstrating thestriking absence of these internal capsulefascicles (arrows), while the corpus callosumappears normal. L1-expressing axons in theinternal capsule of P0 wild-type mice ([B andC]; arrows) are highly fasciculated and tightlybundled compared to internal capsule axonsof Ctip2�/� mice ([E and F]; arrows), whichshow distinct lack of fasciculation and strik-ing disorganization. This abnormality is evi-dent through the entire rostrocaudal extentof the internal capsule, shown here at bothrostral (B and E) and caudal (C and F) loca-tions and in sagittal sections (G and J); DAPInuclear staining (blue). (H and K) High-magni-fication images from the boxed areas in (G)and (J), respectively, reveal the fine details ofthe nonfasciculated Ctip2 null mutant axons([K]; arrows) compared to large fascicles inwild-type controls ([H]; arrows); arrowheadsin (K) indicate highly disorganized axonal pro-jections deviating from their normal path inthe Ctip2�/� null mutant mice. Anterograde DiItracing of axons through the internal capsuleof E18 wild-type (I) and matched Ctip2�/� mice(L). Many of the disorganized axons in theCtip2�/� mice possess bulbous varicositiessuggestive of dysmorphic growth cones(arrows in [L]). Scale bars, (B, C, E, and F) 100�m; (H and K) 50 �m; (I and L) 10 �m. ctx,cortex; cc, corpus callosum; ic, internal cap-sule; str, striatum; hp, hippocampus.

Ramon y Cajal (Ramon y Cajal, 1928) and more recently Given the extremely high levels of CTIP2 expressionin subcerebral projection neurons, we examined hetero-highlighted and investigated by Silver and colleagues

(Silver, 2004). zygous Ctip2�/� mice to determine whether there is agene dosage effect on the observed abnormalities.Given these striking abnormalities, we further exam-

ined the outgrowth of subcerebral axonal projections These experiments allowed us to investigate the role ofCTIP2 into adulthood, much later than the P0 age atand the formation of the corticospinal tract in detail.

We performed in vivo anterograde DiI tracing at P0 by which Ctip2�/� mice die. Interestingly, we found subtledefects in fasciculation in the internal capsule in Ctip2�/�injecting DiI into developing sensorimotor cortex of

Ctip2�/� mice and matched wild-type littermate con- mice (data not shown), indicating a gene dosage effect.To investigate the ability of Ctip2�/� CSMN to properlytrols. Close examination of axons along the length of

the developing corticospinal tract revealed that, while establish and maintain projections to the spinal cord,we injected FG into the cervical spinal cord of 3- andapproximately normal numbers of axons extend as far as

the hypothalamus, they are disorganized, not normally 10-week-old Ctip2�/� mice and quantified labeled CSMNin the entire cortex. During normal development, subcer-fasciculated, and located dorsal to their normal position.

Outgrowing axons also exhibit striking deviations from ebral neurons in layer V of lateral sensory cortex initiallyextend an axon to the spinal cord, but only a smalltheir normal path and extend toward ectopic targets

(Figures 7A–7G). Only a small number of axons were percentage of these neurons maintain corticospinal pro-jections into adulthood (Polleux et al., 2001). Quite re-observed caudal to the hypothalamus, and these were

frequently extending in the wrong direction (Figures 7C markably, we find that, at 3 and 10 weeks of age, a largenumber of neurons in lateral sensory cortex of Ctip2�/�and 7G). Most notably, no CSMN axons extended past

the pons in Ctip2�/� mice (n � 7) (Figure 7H), while CSMN mice aberrantly maintain ectopic projections to the spi-nal cord (at 3 weeks: wt 272 � 39, n � 5; Ctip2�/� 623 �axons in all wild-type and heterozygous littermates ana-

lyzed (n � 20) extended normally through the medulla 23, n � 4; p � 0.0002; at 10 weeks: wt 333 � 27, n �4; Ctip2�/� 1088 � 403, n � 5; p � 0.14) (Figure 8). Intoward the pyramidal decussation and in some cases

had already entered the spinal cord by P0 (Figure 7D). contrast, the number of neurons with spinal projectionsin sensorimotor cortex is the same in wild-type andTaken together, these data demonstrate that Ctip2 is

critical and necessary for CSMN to extend projections Ctip2�/� mice (at 3 weeks: wt 4767 � 507, n � 5; Ctip2�/�

4004 � 223, n � 4; p � 0.25; at 10 weeks: wt 3977 �to the spinal cord.

Molecular Development of Corticospinal Neurons217

Figure 7. CSMN in Ctip2�/� Mice DisplayPathfinding Defects and Fail to Extend to theSpinal Cord

(A and E) Schematic representations of sagit-tal views of the brain and proximal spinal cordin wild-type and Ctip2�/� mice, respectively,showing the location of CSMN somas in thecortex (red triangles) and their axonal projec-tions toward the spinal cord (red lines). (B–Dand F–H) Photomicrographs of boxed areasin (A) and (E), respectively. (B and F) Axonalprojections by subcerebral projection neu-rons showing that (B) P0 wild-type axons areorganized in typical axon fascicles (arrows),but (F) matched P0 Ctip2�/� null mutant axonsare very disorganized, nonfasciculated (arrow),and display axonal projections that deviatefrom the normal pathway and extend to ec-topic targets (arrowhead). (C and G) The sameaxonal fibers as (B) and (F), at a more caudallocation. (C) Wild-type axons are highly orga-nized in tight bundles of fibers progressingunidirectionally toward the pons (arrow),while (G) Ctip2�/� axons are strikingly reduced

in numbers with many individual fibers extending to ectopic sites (arrowheads). (D and H) Photomicrographic montages demonstrating (D)that P0 wild-type axons are abundant through the pons (arrows) and have already reached the pyramidal decussation entering the spinalcord (arrowhead). (H) A much smaller number of axons in Ctip2�/� mice enters the pons (arrows) and no axons extend into the medulla orreach the pyramidal decussation. Scale bars, 100 �m.

216, n � 4; Ctip2�/� 3911 � 454, n � 5; p � 0.91). axonal projections, taking advantage of an intrinsic ana-tomical property (i.e., distant axonal fields) shared byThese data indicate that ctip2 plays an important role

in directing the developmental pruning and refinement many other classes of projection neurons. These ap-proaches could thus be used to purify other neuronalof projections to the spinal cord.

Together, these results with both Ctip2�/� and Ctip2�/� subclasses in a systematic fashion. Additionally, at eachdevelopmental stage studied, we were able to purifymice support the hypothesis that CTIP2 is centrally in-

volved in orchestrating the complex extension, fascicu- relatively homogeneous neuronal populations in fairlylarge numbers. This reduces artifacts of RNA amplifica-lation, and refinement of subcerebral axonal projections

and particularly the ability of CSMN to extend projec- tion and enhances the probability of identifying genesthat are true genetic determinants of the neuronal popu-tions to the spinal cord during the formation of the corti-

cospinal tract. lation sampled rather than differentially expressedgenes in only some of the neurons within a heteroge-neous population. The depth and robustness of dataDiscussionobtained using the neuronal populations that we purifiedin this manner is demonstrated by the fact that all differ-Attempts to study developmental controls over neuronal

subtypes have been hampered by the inability to distin- entially expressed genes that we further investigatedwere confirmed by in situ hybridization or immunocyto-guish different types of projection neurons with distinct

and specific molecular markers. A detailed molecular chemistry (Figure 3).anatomy of neuronal subtypes would facilitate identifi-cation of molecular programs specifying different pro-

Identification of CSMN-Specific Genesjection neuron subtypes.We hypothesized that, during CSMN development, thereHere, we report a molecular characterization of theexist both (1) genes that are used by all cortical projec-clinically important population of CSMN at differenttion neurons to control general aspects of early projec-stages during their development in vivo. We identify ation neuron specification and later morphologic differen-distinct set of largely uncharacterized genes that exhibittiation and (2) genes that are neuronal subtype restricteda range of progressive restriction to CSMN among otherand contribute to define the specific population ofrelated subcerebral projection neurons. These genesCSMN. Combinatorial interactions of both of theseencode critically important molecules for proper CSMNclasses of genes, in the correct temporal order, is likelydifferentiation, as demonstrated by our experiments re-necessary to instruct neural precursors toward avealing that lack of CTIP2 results in striking abnormali-CSMN-specific fate.ties of axonal outgrowth and pathfinding by CSMN and

Our investigation of genes in the second class, genessubcerebral projection neurons, leading to the failure ofspecific to CSMN, identified many genes that were notCSMN to connect to the spinal cord.previously known to be expressed in CSMN (or in otherspecific classes of cortical neurons) and, thus, are novelPurification and Genetic Analysis of Distinctgenetic determinants of this neuronal subtype. TheseNeuronal Subpopulationsmolecules are of particular interest, as they includeDistinct classes of cortical projection neurons were puri-

fied at different stages of development based on their genes that might be involved in different aspects of

Neuron218

subcortically, although they are located within the samecortical layer. Others of the genes reported here havea gradient of expression in layer V, suggesting that theyare expressed by many but not all subcerebral projec-tion neurons (e.g., Crystallin-Mu and Netrin G1).

Together, our data support a model in which CSMNand other subcerebral projection neurons share com-mon genetic programs that determine their specific lami-nar position and initial outgrowth to subcerebral targets.This model is consistent with elegant studies by O’Learyand colleagues showing that subcerebral projectionneurons extend axons in a stereotypic pattern towardthe spinal cord with collateral branching from the pri-mary axons and only later refine their differentiation bypruning collateral projections and only maintaining pro-jections to the appropriate subcortical targets (O’Learyand Koester, 1993). We now provide direct molecularevidence that supports and extends this hypothesis.

Importantly, in further support of a model of combina-torial gene expression to delineate CSMN, we find thatLMO4 is expressed by callosal neurons but excludedfrom CSMN, thus representing a critical negative markerof CSMN and other subcerebral projection neurons. To-gether, these genes provide the foundation for definingthe molecular anatomy of CSMN.

CSMN-Specific Genes Are Biologically ImportantConfirming that the set of genes differentially expressedby CSMN have important functions during CSMN devel-

Figure 8. Heterozygous Ctip2�/� Mice Fail to Correctly Prune Sub- opment, we find that Ctip2�/� mice exhibit striking abnor-cerebral Projections malities in their axonal projections to subcerebral tar-(A and D) FG-labeled layer V CSMN in sensorimotor cortex (asterisks) gets and that Ctip2�/� CSMN fail to extend projectionsand lateral sensory cortex (orange boxes) in (A) wild-type and (D)

to the spinal cord. Ctip2 encodes a zinc finger DNACtip2�/� heterozygous mice. (B) Higher-magnification image of thebinding protein that acts as a transcriptional repressorarea boxed in (A), showing the typical small number of residual(Avram et al., 2000; Senawong et al., 2003). While CTIP2CSMN in lateral sensory cortex of 3-week-old wild-type mice. (E)

Higher-magnification image of the area boxed in (D), showing the was initially discovered as an interacting partner ofmarked increase in the number of residual CSMN in littermate COUP-TF orphan nuclear receptors, it is unclear whether3-week-old Ctip2�/� heterozygous mice, suggesting that reduced CTIP2 interaction with COUP-TFs in vivo is required forlevels of CTIP2 limit the ability of subcerebral projection neurons to

CTIP2-mediated gene expression. No role for this geneproperly prune ectopic connections to the spinal cord. (C and F)in the nervous system was previously known. However,Camera lucida drawings of (B) and (E), respectively. (G) At 3 weeksprior loss-of-function experiments in vivo highlight anof age, Ctip2�/� mice (blue) retain more than double the number of

CSMN in lateral sensory cortex compared to controls (red); at 3 important role for CTIP2 in cell type specification in theweeks, p � 0.0002; at 10 weeks, p � 0.14. Neuron counts are shown immune system (Wakabayashi et al., 2003).as the mean � SEM of the number of CSMN in every sixth section We find that the absence of CTIP2 in vivo results inof lateral sensory cortex of both hemispheres.

defects in the organization and fasciculation of subcere-bral fiber tracts, including CSMN axonal projections.This phenotype is dramatic in the internal capsule, theCSMN development, from fate specification and migra-

tion to process outgrowth and axon guidance to cell path that CSMN axons follow during their initial out-growth toward distal targets in the spinal cord. Mostadhesion and survival. These are the critical develop-

mental events that should be controlled in a subtype- importantly, we found that the absence of CTIP2 resultsin the inability of corticospinal neurons to extend projec-specific fashion.

Of interest, the CSMN- and subcerebral-specific tions to the spinal cord, with striking pathfinding errorsalong the corticospinal pathway. In addition, our obser-genes that we identify have various levels of restricted

expression. A small number appear to be restricted to vation that decreased levels of CTIP2 in heterozygousmice results in abnormal pruning of axon collaterals tosensorimotor cortex (e.g., Diap3, Igfbp4, and Crim1),

distinguishing CSMN from other subcerebral projection the spinal cord further highlights the importance of Ctip2during the development of CSMN connectivity. Our dataneurons. Many genes (e.g., Ctip2, Enchephalopsin, Fez,

Clim1, Pcp4, and S100a10) exhibit broader layer V ex- are supported by the observation that Coup-Tf1 nullmutant mice also have defects in axon guidance (Qiupression, strongly suggesting restriction to subcerebral

projection neurons. Indeed, we found that CTIP2 is ex- et al., 1997; Zhou et al., 1999). Interestingly, COUP-TF1exhibits a gradient of expression across cortical areaspressed in CSMN and in closely related subtypes of

subcerebral projection neurons but is excluded from (Liu et al., 2000). We speculate that, if COUP-TF1 inter-acts with CTIP2 in vivo, as has been shown in vitrocallosal neurons, which do not send axonal projections

Molecular Development of Corticospinal Neurons219

(Avram et al., 2000), the two molecules could contribute CSMN population via specific combinatorial expression.The data presented here support the idea that a preciseimportantly in defining the areal identity of subcere-

bral neurons. molecular classification of distinct classes of projectionneurons is possible and provide a foundation for increas-ingly sophisticated analysis of stage-specific genes con-CSMN-Specific Genes Potentially Control Distincttrolling corticospinal motor neuron development.Aspects of CSMN Development

At least two of the CSMN-specific genes, Fez (Matsuo-Takasaki et al., 2000; Hirata et al., 2004) and Clim1 (Bach Experimental Procedures

et al., 1997), may be novel early instructive signals ofNeuronal Subtype Labeling, Dissociation, and PurificationCSMN fate specification, as suggested by their specificAll neuronal subtypes were purified from C57BL/6 mice (Charlesexpression in CSMN and by recent reports suggestingRiver Laboratories, MA). CSMN were retrogradely labeled with green

roles in fate specification in other organisms (Becker et fluorescent microspheres (Lumafluor Corp., FL) injected into theal., 2002; Levkowitz et al., 2003). In agreement with the pons-midbrain junction (E18), pons (P3), or cervical spinal cord at

the C2-3 or C5 level for P6 and P14, respectively. CPN were labeledCSMN-specific data that we report here, both genesat E18, P3, P6, and P14 by injection into contralateral cortex (E17,were recently shown to be restricted to layer V, furtherP1, P4, P12), as previously described (Catapano et al., 2001). CTPNsupporting our in situ and microarray data (Bulchand etwere labeled by injection into the superior colliculus of P11 pups. FGal., 2003; Inoue et al., 2004). Fez and Clim1, togetherwas injected into cervical spinal cord or contralateral sensorimotor

with other differentially expressed molecules identified cortex to label CSMN or callosal neurons, respectively, as previouslyhere (Figure 2 and Supplemental Figure S2 [http://www. described (Fricker-Gates et al., 2002). All embryonic and neonatal

pontine injections were performed using a Vevo 660 ultrasoundneuron.org/cgi/content/full/45/2/207/DC1/]), are likelysystem (VisualSonics, Toronto). All animal studies were approvedto be critical for specifying CSMN fate. We hypothesizeby the Massachusetts General Hospital Institutional Animal Carethat at least some of these molecules might help identifyand Use Committee and performed in accordance with institutionalsubtype-specific progenitors (if they exist) or cells com-and federal guidelines.

mitted to CSMN fate soon after mitosis. This could con- Sensorimotor cortex (CSMN and CPN) or visual cortex (CTPN)nect the pathways we present here with the extensive were dissociated essentially as described (Catapano et al., 2001),

and microsphere-labeled CSMN, CPN, or CTPN were purified byliterature on initial neuronal specification (Rallu et al.,FACS directly into RNAlater.2002).

A critical aspect of CSMN development is the abilityAffymetrix Microarraysto extend an extremely long axon to precise locationsRNA was extracted using the StrataPrep Total RNA Micro Kit (Stra-within the spinal cord. Here, we identify several mole-tagene), and RNA quality was assessed using a bioanalyzer (Agilentcules specifically expressed in CSMN that might playTechnologies). RNA was amplified per Affymetrix small sample pro-

important roles in CSMN axonal growth and guidance. tocol, using two consecutive rounds of linear in vitro transcriptionIn addition to Ctip2, these include Netrin-G1, Neto1, to obtain 15–20 �g of amplified and labeled cRNA for each hybridiza-Contactin 6, Cadherin 13, and Cadherin 22. tion (Eberwine et al., 1992). To ensure reproducibility and biological

significance, RNA samples were collected from two independentOur genetic analysis also identified molecules whoseFACS purifications at each age (biological replicates). Data fromexpression profile and identity suggests they could bemicroarrays were normalized using two independent methods withinvolved in later CSMN maturation and survival. IGFBP4Rosetta Resolver software. Statistical significance of gene expres-

binds insulin-like growth factors and directly modulates sion differences between neuronal subtypes was determined byIGF stability and action (Zhou et al., 2003). The estab- pairwise comparisons at each age using Significance Analysis oflished role of IGF on cell survival (Stewart and Rotwein, Microarrays (SAM) (Tusher et al., 2001). Microarray data from the

two biological replicates were combined in Rosetta Resolver for1996), combined with the fact that we found that IGFBP4trend plots. Additional details on microarray methods and analysishas a defined area-specific pattern of expression in cor-are in the Suppplemental Data (http://www.neuron.org/cgi/content/tex, suggests that IGFBP4 may be a mediator of thefull/45/2/207/DC1/). All microarray data have been deposited in the

effects of IGF in sensorimotor cortex. Mu-Crystallin has Gene Expression Omnibus database at NCBI (Accession GSE2039).a direct role in controlling T3 mediated gene transactiva-tion (Mori et al., 2002). Because thyroid hormone con-

Immunocytochemistry, In Situ Hybridization,trols important aspects of neuronal differentiation and and DiI Photoconversionsurvival in the CNS (Oppenheimer and Schwartz, 1997), Brains were fixed and stained using standard methods (Fricker-

Gates et al., 2002). Primary antibodies and dilutions are detailed inMu-Crystallin could play a central role in CSMN survival.the Supplemental Data (http://www.neuron.org/cgi/content/full/45/Of additional interest, human Mu-Crystallin maps to2/207/DC1/). Nonradioactive in situ hybridization was performedchromosome 16 at a location near a newly identifiedusing reported methods (Berger and Hediger, 2001). Sense probeslocus for hereditary ALS (Sapp et al., 2003). Thewere used as negative controls in all experiments. cDNA clones are

CSMN-specific expression of this gene, together with listed in Supplemental Table S1. DiI photoconversion combined withthe central involvement of CSMN in ALS, suggests Mu- in situ hybridization was performed as previously described (Sandell

and Masland, 1988; Fujimori et al., 2000).Crystallin as an interesting candidate gene for subtypesof hereditary ALS.

DiI Tracing and FG Labeling in Ctip2�/� and Ctip2�/� MiceAnterograde DiI tracing in vivo and in fixed tissue was performedConclusionas previously described (Godement et al., 1987; O’Leary and Tera-In this report, we identify neuronal subtype specificshima, 1988). CSMN in sensorimotor and lateral sensory cortex weregenes that control the development of corticospinal mo-retrogradely labeled via FG injections in the cervical spinal cord.

tor neurons in vivo. We further study 15 particularly inter- Mice were injected at P14–P15 and sacrificed at P21 or were injectedesting CSMN genes at the cellular, anatomic, and func- and sacrificed as 10-week-old adults. Brains were sectioned coro-

nally at 40 �m, and all CSMN (in sensorimotor and in lateral sensorytional levels. We propose that these genes delineate the

Neuron220

cortex) were counted in both hemispheres on every sixth section, Eberwine, J., Yeh, H., Miyashiro, K., Cao, Y., Nair, S., Finnell, R.,Zettel, M., and Coleman, P. (1992). Analysis of gene expression inacross the entire rostrocaudal extent of the cortex.

Additional methods are given in the Supplemental Experimental single live neurons. Proc. Natl. Acad. Sci. USA 89, 3010–3014.Procedures (http://www.neuron.org/cgi/content/full/45/2/207/DC1/). Frantz, G.D., Bohner, A.P., Akers, R.M., and McConnell, S.K. (1994).

Regulation of the POU domain gene SCIP during cerebral corticalAcknowledgments development. J. Neurosci. 14, 472–485.

Fricker-Gates, R.A., Shin, J.J., Tai, C.C., Catapano, L.A., andWe thank J. Menezes for insightful comments and advice on experi-Macklis, J.D. (2002). Late-stage immature neocortical neurons re-mental design; I. Kohane, M. Ramoni, and L. Page for advice onconstruct interhemispheric connections and form synaptic contactsmicroarray analysis; L. Catapano, D. Dombkowski, and H. Ozdinlerwith increased efficiency in adult mouse cortex undergoing targetedfor advice on FACS methods; S. Gustincich for advice on microar-neurodegeneration. J. Neurosci. 22, 4045–4056.rays; K. MacQuarrie, A. Eswar, and D. Herman for technical assis-Fujimori, K.E., Takeuchi, K., Yazaki, T., Uyemura, K., Nojyo, Y., andtance; U. Berger for assistance with in situ hybridization; J. CougetTamamki, N. (2000). Expression of L1 and TAG-1 in the corticospinal,for assistance with microarrays; M. Leid, T. Jessell, R. Hevner, F.callosal, and hippocampal commissural neurons in the developingRathjen, I. Bach, and T. Sargent for generous sharing of antibodiesrat telencephalon as revealed by retrograde and in situ hybridizationand cDNA clones; and R. Masland, Z.Y. Chen, and other membersdouble labeling. J. Comp. Neurol. 417, 275–288.of the Macklis lab for critical reading of the manuscript. This work

was partially supported by grants from the NIH, Christopher Reeve Geschwind, D.H. (2000). Mice, microarrays, and the genetic diversityParalysis Foundation, and ALS Association (to J.D.M.). P.A. was of the brain. Proc. Natl. Acad. Sci. USA 97, 10676–10678.supported by a Wills Foundation Postdoctoral Fellowship. B.J.M. Godement, P., Vanselow, J., Thanos, S., and Bonhoeffer, F. (1987).was supported by the Harvard M.S.T.P. A study in developing visual systems with a new method of staining

neurons and their processes in fixed tissue. Development 101,Received: March 4, 2004 697–713.Revised: September 8, 2004

Hevner, R.F., Daza, R.A., Rubenstein, J.L., Stunnenberg, H., Olavar-Accepted: December 15, 2004ria, J.F., and Englund, C. (2003). Beyond laminar fate: toward aPublished: January 19, 2005molecular classification of cortical projection/pyramidal neurons.Dev. Neurosci. 25, 139–151.ReferencesHirata, T., Suda, Y., Nakao, K., Narimatsu, M., Hirano, T., and Hibi,M. (2004). Zinc finger gene fez-like functions in the formation ofAvram, D., Fields, A., Pretty On Top, K., Nevrivy, D.J., Ishmael, J.E.,subplate neurons and thalamocortical axons. Dev. Dyn. 230,and Leid, M. (2000). Isolation of a novel family of C(2)H(2) zinc finger546–556.proteins implicated in transcriptional repression mediated by

chicken ovalbumin upstream promoter transcription factor (COUP- Huang, Z.Y., Wu, Y., Hedrick, N., and Gutmann, D.H. (2003). T-cad-TF) orphan nuclear receptors. J. Biol. Chem. 275, 10315–10322. herin-mediated cell growth regulation involves G2 phase arrest andBach, I., Carriere, C., Ostendorff, H.P., Andersen, B., and Rosenfeld, requires p21(CIP1/WAF1) expression. Mol. Cell. Biol. 23, 566–578.M.G. (1997). A family of LIM domain-associated cofactors confer Inoue, K., Terashima, T., Nishikawa, T., and Takumi, T. (2004). Fez1transcriptional synergism between LIM and Otx homeodomain pro- is layer-specifically expressed in the adult mouse neocortex. Eur.teins. Genes Dev. 11, 1370–1380. J. Neurosci. 20, 1–8.Bayer, S.A., and Altman, J. (1991). Neocortical Development (New Jessell, T.M. (2000). Neuronal specification in the spinal cord: induc-York: Raven Press). tive signals and transcriptional codes. Nat. Rev. Genet. 1, 20–29.Becker, T., Ostendorff, H.P., Bossenz, M., Schluter, A., Becker, C.G.,

Jones, E.G., Schreyer, D.J., and Wise, S.P. (1982). Growth and matu-Peirano, R.I., and Bach, I. (2002). Multiple functions of LIM domain-

ration of the rat corticospinal tract. Prog. Brain Res. 57, 361–379.binding CLIM/NLI/Ldb cofactors during zebrafish development.

Kamme, F., Salunga, R., Yu, J., Tran, D.T., Zhu, J., Luo, L., Bittner,Mech. Dev. 117, 75–85.A., Guo, H.Q., Miller, N., Wan, J., and Erlander, M. (2003). Single-

Berger, U.V., and Hediger, M.A. (2001). Differential distribution ofcell microarray analysis in hippocampus CA1: demonstration and

the glutamate transporters GLT-1 and GLAST in tanycytes of thevalidation of cellular heterogeneity. J. Neurosci. 23, 3607–3615.

third ventricle. J. Comp. Neurol. 433, 101–114.Kolle, G., Georgas, K., Holmes, G.P., Little, M.H., and Yamada, T.Bertrand, N., Castro, D.S., and Guillemot, F. (2002). Proneural genes(2000). CRIM1, a novel gene encoding a cysteine-rich repeat protein,and the specification of neural cell types. Nat. Rev. Neurosci. 3,is developmentally regulated and implicated in vertebrate CNS de-517–530.velopment and organogenesis. Mech. Dev. 90, 181–193.

Blackshaw, S., and Snyder, S.H. (1999). Encephalopsin: a novelLevkowitz, G., Zeller, J., Sirotkin, H.I., French, D., Schilbach, S.,mammalian extraretinal opsin discretely localized in the brain. J.Hashimoto, H., Hibi, M., Talbot, W.S., and Rosenthal, A. (2003). ZincNeurosci. 19, 3681–3690.finger protein too few controls the development of monoaminergic

Bulchand, S., Subramanian, L., and Tole, S. (2003). Dynamic spatio- neurons. Nat. Neurosci. 6, 28–33.temporal expression of LIM genes and cofactors in the embryonic

Liu, Q., Dwyer, N.D., and O’Leary, D.D. (2000). Differential expressionand postnatal cerebral cortex. Dev. Dyn. 226, 460–469.of COUP-TFI, CHL1, and two novel genes in developing neocortex

Catapano, L.A., Arnold, M.W., Perez, F.A., and Macklis, J.D. (2001).identified by differential display PCR. J. Neurosci. 20, 7682–7690.

Specific neurotrophic factors support the survival of cortical projec-Matsuo-Takasaki, M., Lim, J.H., Beanan, M.J., Sato, S.M., andtion neurons at distinct stages of development. J. Neurosci. 21,Sargent, T.D. (2000). Cloning and expression of a novel zinc finger8863–8872.gene, Fez, transcribed in the forebrain of Xenopus and mouse em-Cepko, C.L. (1999). The roles of intrinsic and extrinsic cues andbryos. Mech. Dev. 93, 201–204.bHLH genes in the determination of retinal cell fates. Curr. Opin.Mori, J., Suzuki, S., Kobayashi, M., Inagaki, T., Komatsu, A., Takeda,Neurobiol. 9, 37–46.T., Miyamoto, T., Ichikawa, K., and Hashizume, K. (2002). Nicotin-Coonan, J.R., Greferath, U., Messenger, J., Hartley, L., Murphy, M.,amide adenine dinucleotide phosphate-dependent cytosolic T(3)Boyd, A.W., Dottori, M., Galea, M.P., and Bartlett, P.F. (2001). Devel-binding protein as a regulator for T(3)-mediated transactivation. En-opment and reorganization of corticospinal projections in EphA4docrinology 143, 1538–1544.deficient mice. J. Comp. Neurol. 436, 248–262.O’Leary, D.D., and Koester, S.E. (1993). Development of projectionDiaz, E., Ge, Y., Yang, Y.H., Loh, K.C., Serafini, T.A., Okazaki, Y.,neuron types, axon pathways, and patterned connections of theHayashizaki, Y., Speed, T.P., Ngai, J., and Scheiffele, P. (2002).mammalian cortex. Neuron 10, 991–1006.Molecular analysis of gene expression in the developing pontocere-

bellar projection system. Neuron 36, 417–434. O’Leary, D.D., and Terashima, T. (1988). Cortical axons branch to

Molecular Development of Corticospinal Neurons221

multiple subcortical targets by interstitial axon budding: implications et al. (2003). Bcl11b is required for differentiation and survival ofalphabeta T lymphocytes. Nat. Immunol. 4, 533–539.for target recognition and “waiting periods”. Neuron 1, 901–910.

Olson, M.F. (2003). Dispatch. GTPase signalling: new functions for Weimann, J.M., Zhang, Y.A., Levin, M.E., Devine, W.P., Brulet, P.,and McConnell, S.K. (1999). Cortical neurons require Otx1 for theDiaphanous-related formins. Curr. Biol. 13, R360–R362.refinement of exuberant axonal projections to subcortical targets.Oppenheimer, J.H., and Schwartz, H.L. (1997). Molecular basis ofNeuron 24, 819–831.thyroid hormone-dependent brain development. Endocr. Rev. 18,

462–475. Yin, Y., Miner, J.H., and Sanes, J.R. (2002). Laminets: laminin- andnetrin-related genes expressed in distinct neuronal subsets. Mol.Polleux, F., Dehay, C., Goffinet, A., and Kennedy, H. (2001). Pre-Cell. Neurosci. 19, 344–358.and post-mitotic events contribute to the progressive acquisition

of area-specific connectional fate in the neocortex. Cereb. Cortex Zhou, C., Qiu, Y., Pereira, F.A., Crair, M.C., Tsai, S.Y., and Tsai,M.J. (1999). The nuclear orphan receptor COUP-TFI is required for11, 1027–1039.differentiation of subplate neurons and guidance of thalamocorticalQiu, Y., Pereira, F.A., DeMayo, F.J., Lydon, J.P., Tsai, S.Y., andaxons. Neuron 24, 847–859.Tsai, M.J. (1997). Null mutation of mCOUP-TFI results in defects in

morphogenesis of the glossopharyngeal ganglion, axonal projec- Zhou, R., Diehl, D., Hoeflich, A., Lahm, H., and Wolf, E. (2003). IGF-binding protein-4: biochemical characteristics and functional conse-tion, and arborization. Genes Dev. 11, 1925–1937.quences. J. Endocrinol. 178, 177–193.Rallu, M., Corbin, J.G., and Fishell, G. (2002). Parsing the prosen-

cephalon. Nat. Rev. Neurosci. 3, 943–951.

Ramon y Cajal, S. (1928). In Recuerdos de mi vida (New York:Hafner).

Rolf, B., Bastmeyer, M., Schachner, M., and Bartsch, U. (2002).Pathfinding errors of corticospinal axons in neural cell adhesionmolecule-deficient mice. J. Neurosci. 22, 8357–8362.

Sandell, J.H., and Masland, R.H. (1988). Photoconversion of somefluorescent markers to a diaminobenzidine product. J. Histochem.Cytochem. 36, 555–559.

Sangameswaran, L., Hempstead, J., and Morgan, J.I. (1989). Molec-ular cloning of a neuron-specific transcript and its regulation duringnormal and aberrant cerebellar development. Proc. Natl. Acad. Sci.USA 86, 5651–5655.

Sapp, P.C., Hosler, B.A., McKenna-Yasek, D., Chin, W., Gann, A.,Genise, H., Gorenstein, J., Huang, M., Sailer, W., Scheffler, M., etal. (2003). Identification of two novel loci for dominantly inheritedfamilial amyotrophic lateral sclerosis. Am. J. Hum. Genet. 73,397–403.

Saris, C.J., Kristensen, T., D’Eustachio, P., Hicks, L.J., Noonan, D.J.,Hunter, T., and Tack, B.F. (1987). cDNA sequence and tissue distri-bution of the mRNA for bovine and murine p11, the S100-relatedlight chain of the protein-tyrosine kinase substrate p36 (calpactinI). J. Biol. Chem. 262, 10663–10671.

Segovia, L., Horwitz, J., Gasser, R., and Wistow, G. (1997). Tworoles for mu-crystallin: a lens structural protein in diurnal marsupialsand a possible enzyme in mammalian retinas. Mol. Vis. 3, 9.

Senawong, T., Peterson, V.J., Avram, D., Shepherd, D.M., Frye, R.A.,Minucci, S., and Leid, M. (2003). Involvement of the histone deacety-lase SIRT1 in chicken ovalbumin upstream promoter transcriptionfactor (COUP-TF)-interacting protein 2-mediated transcriptional re-pression. J. Biol. Chem. 278, 43041–43050.

Silver, J. and Miller, J.H. (2004). Regeneration beyond the glial scar.Nat. Rev. Neurosci. 5, 146–156.

Stenvers, K.L., Zimmermann, E.M., Gallagher, M., and Lund, P.K.(1994). Expression of insulin-like growth factor binding protein-4and -5 mRNAs in adult rat forebrain. J. Comp. Neurol. 339, 91–105.

Stewart, C.E., and Rotwein, P. (1996). Growth, differentiation, andsurvival: multiple physiological functions for insulin-like growth fac-tors. Physiol. Rev. 76, 1005–1026.

Sugimoto, K., Honda, S., Yamamoto, T., Ueki, T., Monden, M., Kaji,A., Matsumoto, K., and Nakamura, T. (1996). Molecular cloning andcharacterization of a newly identified member of the cadherin family,PB-cadherin. J. Biol. Chem. 271, 11548–11556.

Terashima, T. (1995). Anatomy, development and lesion-inducedplasticity of rodent corticospinal tract. Neurosci. Res. 22, 139–161.

Tietjen, I., Rihel, J.M., Cao, Y., Koentges, G., Zakhary, L., and Dulac,C. (2003). Single-cell transcriptional analysis of neuronal progeni-tors. Neuron 38, 161–175.

Tusher, V.G., Tibshirani, R., and Chu, G. (2001). Significance analysisof microarrays applied to the ionizing radiation response. Proc. Natl.Acad. Sci. USA 98, 5116–5121.

Wakabayashi, Y., Watanabe, H., Inoue, J., Takeda, N., Sakata, J.,Mishima, Y., Hitomi, J., Yamamoto, T., Utsuyama, M., Niwa, O.,

Supplemental Data

Neuronal Subtype-Specific Genes that Control

Corticospinal Motor Neuron Development In Vivo Paola Arlotta, Bradley J. Molyneaux, Jinhui Chen, Jun Inoue, Ryo Kominami, and Jeffrey D. Macklis Supplemental Experimental Procedures Neuronal Subtype Labeling, Dissociation, and Purification For embryonic injections, the day of vaginal plug was designated E0. E17 pregnant mice were deeply anesthetized with Avertin, and each embryo was injected through the uterine wall into the pons, using a VisualSonics Vevo 660 ultrasound-guided microinjection system (VisualSonics, Toronto, Canada) to precisely control the position of the injection site (Figures 1A–1C). Two injections per embryo were performed, with a total of 60–80 nl of green fluorescent microspheres per injection site. The pregnant dam was deeply anesthetized 1 day after surgery, embryos were removed, and sensorimotor cortex was microdissected from both cerebral hemispheres using a fluorescence dissecting microscope to precisely visualize the labeled region (Figure 1D). Similarly, for postnatal injections, P1, P4, and P11 mice were anesthetized by hypothermia (P1, P4) or Avertin (P11) and injected in the pons (P1) or the cervical spinal cord (P4 and P11). Four injections per pup were performed, with a total of 60–80 nl of green fluorescent microspheres per injection site. Pups were returned to the care of their mother and deeply anesthetized at P3, P6, and P14, respectively. Sensorimotor cortex was microdissected as described above. Retrogradely labeled cortex was enzymatically digested in dissociation medium (glucose [20 mM], Kynurenic acid [0.8 mM], APV [0.05 mM], penicillin-streptomycin [50 u/ml and 0.05 mg/ml, respectively], Na2SO4 [0.09 M], K2SO4 [0.03 M], and MgCl2 [0.014 M]) containing L-cysteine HCl (0.016 µg/µL) and papain (10 u/ml; Worthington Biochemical Corp., NJ) at 37°C for 30 min. Papain digestion was stopped with ovomucoid (10 mg/ml) and BSA (10 mg/ml) in dissociation medium at room temperature. Neurons were mechanically dissociated to obtain a single-cell suspension by gentle trituration in iced OptiMem containing glucose (20 mM), kynurenic acid (0.4 mM), and APV (0.025 mM). All chemicals were purchased from Sigma, and all media were purchased from GIBCO-BRL, unless stated otherwise. Viability, assessed with trypan blue staining, was greater than 95% for E18, P3, and P6 and greater than 85% for P14 neurons. P14 CSMN, callosal neurons, and corticotectal neurons were purified using a modified protocol that we developed for use with the more fragile late stage neurons. P14 tissue was enzymatically digested in papain (20 U/ml) for 45 min with constant stirring. DNase was added for the last 5 min of enzymatic digestion and to the media during trituration at 10 U/ml. Projection neurons were dissociated to a single cell suspension, as described above, and immediately preserved in

RNAlater (Ambion). Cells were separated from debris by centrifuging at 5000 × g for 45 min, and the pellet was resuspended in RNAlater for FACS purification. Microsphere-labeled neurons were purified using a BD FACS Vantage SE DiVa cell sorter. Cells were gated based on fluorescence, and forward and side scatter gates were set to select the population of large projection neurons shown in Figure 1Q. FACS-sorted neurons from all ages and preparations were collected and stored in RNAlater before RNA extraction (Barrett et al., 2002). Affymetrix Microarrays We used approximately 10,000 to 30,000 FACS-sorted neurons for each biological replicate, with the exception of P14 corticotectal neuron and E18 CSMN samples. In these two cases, a minimum of 1000 neurons were used, because fewer neurons are labeled in these samples, leading to a smaller population of FACS-sorted neurons. Because P14 neurons were fixed prior to FACS purification, while neurons at other ages were fixed after purification, we collected an additional biological replicate for P3 CSMN using the P14 method of fixation prior to sorting. The additional P3 replicates correlated highly with the other P3 samples (pairwise correlation coefficients of 0.95 and 0.96), indicating that fixation of neurons prior to sorting did not affect microarray results. Micrroarray Analysis Homotypic comparisons (biological replicates) and heterotypic comparisons (different neuronal subtypes/ages) were performed using Rosetta Resolver software (Rosetta Inpharmatics) for all differentially expressed genes with >2-fold change and p value <0.01 (Figure S1). Pearson’s correlation coefficients of biological replicates were calculated using SAS software (SAS Institute). Data from all microarrays were normalized using two independent methods with Rosetta Resolver software by (1) global scaling of each array intensity to the mean intensity value of all arrays; (2) the error modeling method that more accurately normalizes transcripts expressed at low levels. Statistical significance of gene expression differences between neuronal subtypes was determined by pairwise comparisons at each age using SAM. Only genes scored as “present” in at least one microarray with a p value <0.01 were included for analysis. Fold change cut off was set at 3 for all comparisons, with the exception of the P14 CSMN versus P14 CTPN comparison, for which it was set at 2.0, because differences between these samples are smaller in magnitude. We selected the 100 most statistically significant genes for each subtype at each age. We further analyzed these genes for trends with likely biological relevance, 36 of which are described in Figures 2, 3, and S3. Cluster Analysis All cluster analysis was performed in Rosetta Resolver using an agglomerative hierarchical clustering method (Eisen et al., 1998). Unique genes scored as “present” in at least one experiment (p value <10−5; total of 5904 genes) were included (Figure S2). Immunocytochemistry Brains were fixed by transcardial perfusion with PBS-heparin (10 U/ml) followed by 4% paraformaldehyde and were postfixed overnight at 4°C in 4% paraformaldehyde. 40 µm coronal or 60 µm sagittal floating sections were blocked in 0.3% BSA (Sigma), 8% goat or donkey

serum, and 0.3% Triton X-100 (Sigma) for 1 hr at room temperature, before incubation in primary antibody. Primary antibodies and dilutions used were rat anti-CTIP2, 1:500, a gift of M. Leid (Senawong et al., 2003); goat anti-LMO4 antibody, 1:200 (Santa Cruz Biotechnology); rabbit anti-TBR1, 1:500, a gift of R. Hevner; rabbit anti-ER81, 1:1000, a gift of T. Jessell (Arber et al., 2000); rabbit anti-L1, 1:500, a gift of F. Rathjen (Brummendorf et al., 1998). Laser confocal analysis was performed using a Biorad Radiance 2100 confocal microscope.

Figure S1. Microarray Data Are Consistent and Reproducible (A-D) Intensity plots of all genes in pair-wise comparisons at P3; unchanged genes are in blue, and differentially expressed genes (p-value < 0.01) are in red (upregulated) and green (downregulated). Red lines mark the 2-fold change interval. (A,B) Comparison of biological replicate samples reveals only (A) 80 of >22,000 genes without concordance (p-value < .01; fold change > 2) between CSMN replicates, and (B) 43 of > 22,000 genes without concordance between callosal replicates. (C,D) In contrast, heterotypic comparisons between CSMN and callosal neuron samples revealed over 650 differentially expressed genes. (E) Summary of microarray experiments and correlation coefficients.

Figure S2. Distinct Subtypes of Projection Neurons Share Common Genetic Determinants at Different Stages of Development (A) Global cluster of all genes present, normalized across all the samples. At each stage distinct sub-clusters of genes are expressed in each of the three neuronal subtypes analyzed (CSMN, callosal, corticotectal). Columns represent samples and rows represent genes. (B-D) Zoomed images of expression data from the selected areas in (A), showing a selection of highly correlated and biologically relevant genes that are differentially expressed at E18 (B), P3-P6 (C), and P14 (D). These represent examples of a set of common genes that are expressed in a distinct stage-specific manner and which may control early, intermediate, or late aspects of cortical projection neuron development within each of the projection neuron subtypes analyzed. Z-score scale indicates the number of standard deviations from the mean of each data set.

Figure S3. Microarray Gene Expression Profiles for All Other Genes Described in Figure 2, but Not Included in Figures 2 or 3 CPN, callosal projection neurons; CTPN, corticotectal projection neurons.

Figure S4. Genes Identified from the Microarray Analysis Are Expressed in DiI Retrogradely Labeled CSMN in Layer V (A,C,E,G) Low-magnification photomicrographs of in situ hybridization analysis of 4 selected genes in coronal sections of P14 cortex. (B,D,F,H and insets) Higher magnification images of the boxed areas in A,C,E,G, showing specific co-localization of each transcript (purple in situ signal) with DiI labeled CSMN (brown photoconversion precipitate). Scale bars: (A,C,E,G) 100 µm, (B,D,F,H) 10 µm, (B’,D’,F’,H’) 5 µm.

Figure S5. LMO4 Is Expressed in Callosal Neurons, but Not CSMN (A-D) Fluorescence photomicrographs showing exclusion of LMO4 from CSMN in layer V of cortex. (A) Low magnification image of a coronal section of cortex at P6, showing no co-localization of LMO4 (red) with FluoroGold labeled CSMN (green). (B) High magnification image of FluoroGold labeling of CSMN in layer V, and (C) LMO4 expression, in the boxed area in A. (D) Merged image of B and C, showing exclusion of LMO4 from CSMN. (E-L) Fluorescence photomicrographs showing expression of LMO4 in callosal neurons (CPN) of layer II/III (E-H) and layer V (I-L). (E) Low magnification image of a coronal section of cortex at P6, showing broad co-localization of LMO4 (red) with FluoroGold labeled callosal neurons (CPN; green). (F) High magnification image of FluoroGold labeling of callosal neurons in layer II/III of cortex, and (G) LMO4 expression, in the boxed area in E. (H) Merged image of F and G, showing LMO4 expression in essentially all callosal neurons. (I-L) Similarly, LMO4 (K) is expressed in essentially all FluoroGold labeled CPN (J) in layer V of cortex, as shown at low magnification in I, and in the merged image in L. Scale bars: (A, E, I) 50 µm, (B-D), (F-H), (J-L) 10 µm.

Table S1. Detailed Information About the Clones Used for In Situ Hybridization Clone Name Source PCR primers Amplicon Size Plasmid Vector

Igfbp4 The I.M.A.G.E. Consortium; clone # 5123738 n/a n/a pCMV-SPORT 6

Crim1 RT-PCR TCTCAAGACTGTTGGTTGCTG TGAACAACCAATGATAGCACAG 442 bp

pCR II TOPO

Fez gift of Dr. T. D. Sargent, NIH n/a n/a pT7T3D-Pac

Encephalopsin RT-PCR CTCACTGTGCTGGCCTATGA GGGTTGTACACAGTGCTCGAT 523 bp pCR II TOPO

Clim1 gift of Dr. I. Bach, Hamburg University, Germany n/a n/a pBluescript KS+

Mu-Crystallin RT-PCR ACTGGCGAGAACTGGATGAC GCCATCACCCCTTAACAGAA 436 bp pCR II TOPO

Netrin G1 RT-PCR TGGCCGTTTCTTAAGACGAA CTCAATTTCCCTGCTGCTGT 410 bp pCR II TOPO

Csmn1 RT-PCR CCCTAGTTGGGACTGTGACC AGCTCTGTTCACCCCACAAG 367 bp pCR II TOPO

Pcp4 RT-PCR GGAGTCAGGCCAACATGAGT TGCAGAGGAATTGTAATGGAGA 432 bp pCR II TOPO

Cadherin 13 RT-PCR CTTCTAGTCGGGCAAGATGC GGGTCTGTTGTCGTTCTGGT 745 bp pCR II TOPO

S100a10 RT-PCR GCCCAGGTTTCGACAGACT CCACTAGTGATAGAAAGCTCTGGA 255 bp pCR II TOPO

Cadherin 22 RT-PCR CAAGCCCTATTCAGGAGCAG CTTGGGGTCCACTGTGAAGT 872 bp pCR II TOPO

Diap3 RT-PCR AGCAGTTCGCTGTTGTGATG GCACGTTTCTCCTTTTCTGC

733 bp pCR II TOPO

Supplemental References

Arber, S., Ladle, D.R., Lin, J.H., Frank, E., and Jessell, T.M. (2000). ETS gene Er81 controls the formation of functional connections between group Ia sensory afferents and motor neurons. Cell 101, 485–498.

Barrett, M.T., Glogovac, J., Prevo, L.J., Reid, B.J., Porter, P., and Rabinovitch, P.S. (2002). High-quality RNA and DNA from flow cytometrically sorted human epithelial cells and tissues. Biotechniques 32, 888–890, 892, 894, 896.

Brummendorf, T., Kenwrick, S., and Rathjen, F.G. (1998). Neural cell recognition molecule L1: from cell biology to human hereditary brain malformations. Curr. Opin. Neurobiol. 8, 87–97.

Eisen, M.B., Spellman, P.T., Brown, P.O., and Botstein, D. (1998). Cluster analysis and display of genome-wide expression patterns. Proc. Natl. Acad. Sci. USA 95, 14863–14868.

Senawong, T., Peterson, V.J., Avram, D., Shepherd, D.M., Frye, R.A., Minucci, S., and Leid, M. (2003). Involvement of the histone deacetylase SIRT1 in chicken ovalbumin upstream promoter transcription factor (COUP-TF)-interacting protein 2-mediated transcriptional repression. J. Biol. Chem. 278, 43041–43050.