Top Banner
NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN NETWORKS UNDERLYING EXECUTIVE CONTROL A Dissertation submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Neuroscience By Andrew Lee Breeden, B.S. Washington, DC August 23, 2016
98

NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

Nov 07, 2021

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN NETWORKS UNDERLYING EXECUTIVE

CONTROL

A Dissertation

submitted to the Faculty of the

Graduate School of Arts and Sciences

of Georgetown University

in partial fulfillment of the requirements for the

degree of

Doctor of Philosophy

in Neuroscience

By

Andrew Lee Breeden, B.S.

Washington, DC

August 23, 2016

Page 2: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

ii

Copyright 2016 by Andrew Lee Breeden

All Rights Reserved

Page 3: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

iii

NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN NETWORKS

UNDERLYING EXECUTIVE CONTROL

Andrew Lee Breeden, B.S.

Thesis Advisor: Chandan J. Vaidya, Ph.D.

ABSTRACT

Executive control, the goal directed deployment of cognitive resources, depends critically

on a class of neurotransmitters called catecholamines. One such catecholamine, norepinephrine,

is theorized to facilitate cognition through shaping patterns of synaptic communication within

neural networks. Major questions remain, however, if and how norepinephrine regulates the

network structure of the human brain. Two studies were conducted to investigate this question.

The first study simultaneously measured brain activity with functional magnetic resonance

imaging (fMRI) and pupil diameter (a proxy for norepinephrine signaling). This study found that

fluctuations in pupil diameter were synchronous with spontaneous brain activity in many

canonical brain networks. Furthermore, individuals less prone to distractibility in everyday

behavior demonstrated stronger positive coupling between the cingulo-opercular network and

pupil diameter. These results suggest that networks important to executive control likely do not

act in isolation, but coordinate with the NE system in a behaviorally relevant manner. Study II

investigated how potentiated NE signaling at alpha-2a receptors altered distributed brain

networks when executive control was engaged. A NE alpha-2a receptor agonist, guanfacine, was

administered and brain activity was measured with fMRI during two cognitive states: a

cognitively unconstrained resting state and a working memory task. This study found that NE

signaling at alpha-2a receptors during working memory modulated the strength of every

functional network examined. Machine learning analyses utilizing FC patterns across the brain

Page 4: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

iv

predicted drug vs. placebo state with high accuracy (75%, p < .001) when working memory was

engaged, but with only modest accuracy (60%, p = .13) during the resting state. Furthermore, FC

changes during work`ing memory were not specific to any one network. When SVM analyses

were restricted to each functional network in isolation, each network was sufficient to predict

drug state at levels above chance, but no network achieved accuracy as high as that with whole-

brain FC. Collectively, these FC alterations increased the overall modularity of brain networks

during working memory engagement, indicating greater network segregation. Together, these

studies demonstrate that norepinephrine shapes the spatio-temporal structure of functional

networks in the human brain, and suggest that this may be an important neural mechanism

enabling executive control.

Page 5: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

v

TABLE OF CONTENTS

CHAPTER I: GENERAL INTRODUCTION .............................................................................. 1

Introduction .................................................................................................................. 1

CHAPTER II: COUPLING BETWEEN SPONTANEOUS PUPILLARY FLUCTUATIONS AND BRAIN

ACTIVITY RELATES TO INATTENTIVENESS ......................................................................... 9

Introduction .................................................................................................................. 9

Materials and Methods ............................................................................................... 11

Results ........................................................................................................................ 16

Discussion .................................................................................................................. 21

CHAPTER III: NOREPINEPHRINE ALPHA-2A RECEPTOR ACTIVATION ALTERS DISTRIBUTED

FUNCTIONAL NETWORKS DURING WORKING MEMORY ..................................................... 28

Introduction ................................................................................................................ 28

Results ........................................................................................................................ 32

Discussion .................................................................................................................. 42

Materials and Methods ............................................................................................... 50

CHAPTER IV: GENERAL CONCLUSION ............................................................................. 57

Conclusion ................................................................................................................. 57

APPENDIX ....................................................................................................................... 67

Supplementary Materials from Chapter III ................................................................ 67

REFERENCES ................................................................................................................... 68

Page 6: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

vi

LIST OF FIGURES AND TABLES

FIGURES

Figure 1.1 Resting-state brain activity is associated with fluctuations in

pupil diameter .................................................................................................... 19

Figure 1.2 Pupil-brain coupling correlates with trait-level attention in

sympathetic related brain regions. ..................................................................... 21

Figure 2.1 Guanfacine did not Alter N-back Performance. ....................................... 33

Figure 2.2 Guanfacine Induced Wide-Spread FC Changes. ...................................... 35

Figure 2.3 SVM Analyses Confirm Wide-Spread FC Modulation by Guanfacine ... 37

Figure 2.4 Guanfacine Increased the Modularity of Brain Networks. ....................... 42

Supplementary Figure 2.1 Guanfacine Did Not Change Blood Pressure

or Heart Rate ...................................................................................................... 67

TABLES

Table 1.1 Regions positively and negatively related to pupil diameter. .................... 17

Table 1.2 Participant demographics. .......................................................................... 51

Page 7: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

1

CHAPTER I: GENERAL INTRODUCTION

Introduction

Understanding how distributed collections of neurons interact in a networked fashion is

critical to elucidating the neural basis of behavior. In recent years, this endeavor has been greatly

advanced by the ability to map large-scale networks in vivo in the human brain with a technique

called functional connectivity magnetic resonance imaging (fcMRI). The concept behind fcMRI

is simple—areas whose neural activities are highly correlated over time (termed “functional

connectivity”) are thought to be interacting, either through mono- or poly-synaptic connections.

While conceptually straightforward, this idea is validated by animal models (Schölvinck et al.,

2010; Wang et al., 2012; Shen et al., 2015), and computational models suggesting that neural

synchrony may be an important aspect of the neural code (Fries, 2005). By employing fcMRI in

conjunction with mathematical methods such as independent component analyses (Beckmann et

al., 2005) or graph-theory (Power et al., 2011), the functional anatomy of the brain can be

decomposed into distributed networks (termed “functional networks”) whose components are

highly functionally connected to themselves, and sparsely functionally connected to the rest of

the brain.

Strikingly, functional networks resemble the spatial distribution of regions that often co-

activate during common tasks (Smith et al., 2009; Gordon et al., 2012), and remain highly

functionally connected even when someone is in a task-free “resting” state (Fox et al., 2005; De

Luca et al., 2006). For instance, areas that co-activate during cognitive tasks (i.e. fronto-parietal

or cingulo-operuclar regions) also remain highly functionally connected during rest and thus are

said to form “intrinsic” functional networks termed the fronto-parietal and cingulo-operuclar

Page 8: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

2

networks. These networks may also subserve discrete cognitive functions (Dosenbach et al.,

2008), suggesting that networks uncovered with fcMRI are fundamental, functionally meaningful

sub-units of the human brain.

Historically, uncovering inherent levels of brain organization such as cortical layers and

columns, up through areas and systems (Churchland & Sejnowski, 1988; Felleman & Essen,

1991), permitted greater understanding of brain-behavior relationships. Similarly, the discovery

of large-scale functional networks has enabled a better understanding of the circuits underlying

processes as diverse as stress (Hermans et al., 2011), attention (Castellanos et al., 2008), and

memory (Sami et al., 2014). Network level analyses are particularly advantageous to the study of

complex behaviors, because these behaviors require integration of information that spans

multiple cognitive domains and is distributed across cortical areas (Siegel et al., 2015).

Open questions regarding the neurobiological basis of functional networks

Despite recent breakthroughs, basic questions remain regarding the neurobiological basis

of large-scale functional networks. The reliable spatial distribution of functional networks (Chen

et al., 2008; Van Dijk et al., 2010) suggests that they may be based upon anatomical connections.

This assertion is backed up by both human (Honey et al., 2007) and animal studies (Wang et al.,

2013) demonstrating that resting state connectivity patterns and the distribution of anatomical

connections are broadly consistent (although functional connections can be present in the

absence of direct structural connections (Honey et al., 2007)). While the overall spatial pattern of

functional networks may be stable, however, the strength of functional connections within these

networks is highly dynamic at short time-scales (i.e., tens of seconds) (Chang and Glover, 2010;

Hutchison et al., 2013). This dynamism is too rapid for the spatio-temporal structure of

Page 9: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

3

functional networks to be explained by anatomy alone; it is improbable that new connections can

form in the second to minute timeframe. Even within resting state periods, where participants are

typically asked only to hold still, there are ebbs and flows in connectivity patterns that produce

recurring, quasi-stable brain-state configurations (Allen et al., 2012). Rather than noise, these

functional connectivity dynamics may be a core mechanism that allows the brain to adapt to

changing task demands (Bassett et al., 2011; Braun et al., 2015). When participants are asked to

cycle through various tasks like math, memory, and movie watching, their cognitive state is

predictable with almost perfect accuracy on the basis of brain-wide functional connectivity

patterns (Gonzalez-Castillo et al., 2015). Even more subtle variations in cognitive demands

produce reliable, widespread shifts in functional connectivity (Cole et al., 2012; Braun et al.,

2015). This highlights the distributed nature of cognitive processing. It also raises a critical

question: how do anatomical connections that are largely static at short time scales give rise to

varied and dynamic configurations of functional networks in the service of cognition?

The role of neuromodulation

The motivation of this dissertation capitalizes on the framework provided by animal

models and receptor-level studies designed to answer analogous questions in small circuits.

Current theoretical models propose that a class of neurotransmitters called neuromodulators

quickly “reset” distributed cortical networks (Bouret & Sara, 2005; Bargmann & Marder, 2013).

Rather than directly excite or inhibit their post-synaptic targets, neuromodulators often change

synaptic efficacy, and thus regulate neural interaction patterns (Arnsten et al., 2010). Numerous

findings bolster this concept; including that neuromodulation can functionally “silence” present

anatomical connections, can alter intrinsic excitability, and can change the frequency and phase

Page 10: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

4

relationship within entire functional circuits (Bargmann & Marder, 2013). Collectively, this

suggests that anatomical connections provide a minimal structure, and neuromodulation creates

functional circuits in the service of behavior (Marder, 2012). This dissertation aims to expand

this literature based on small circuits, to also understand how neuromodulation impacts large-

scale functional networks in the human brain.

Neuromodulation and executive control

In addition to regulating neural circuits, neuromodulation is also critical for executive

control, a constellation of processes—including working memory, attentional or task-set

switching, and inhibitory control—that support the goal-driven deployment of cognitive

resources. Understanding the neural basis of executive control is important not only to the basic

science of brain networks, but also because executive deficits are common to many psychiatric

conditions (Raffard & Bayard, 2012; Rosenthal et al., 2013; Shanmugan et al., 2016) and are

highly detrimental to adaptive function (Szatmari et al., 1989; Pugliese et al., 2016). In certain

disorders, such as autism spectrum disorders, even when primary symptoms are ameliorated,

executive deficits still persist and cause marked impairment (Troyb et al., 2014). Despite being a

source of major behavioral impairment, there remains a dearth of effective treatment options for

executive dysfunctions, further highlighting the importance of understanding their brain bases.

As with any cognitive process, executive control is likely subserved by numerous and

distributed neurotransmitters acting in conjunction. In particular, however, the neuromodulatory

influence of one class of neurotransmitters called catecholamines—which include dopamine

(DA), epinephrine (E), and norepinephrine (NE)—are critical to the functioning of the networks

linked to executive control. Indeed, depletion of catecholamines from the dorsolateral prefrontal

Page 11: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

5

cortex (dlPFC) can be as detrimental to its function as lesions to the tissue itself (Brozoski et al.,

1979). While DA’s role in executive control has been more extensively studied, recent work

points to a distinct role of NE (Arnsten & Contant, 1992; Arnsten, 1997, 2011; Sara, 2009; Gamo

et al., 2010; Enge et al., 2012). Varying levels of NE signaling from the brainstem locus

coeruleus (LC), that occur across different behavioral states, can flexibly regulate executive

circuits to bring them online and strengthen their top-down influence for goal-directed behavior,

or to lessen their influence and facilitate a higher fidelity of bottom-up signals that promote task

disengagement (Arnsten, 2009, 2011).

The best studied example of NE’s role in executive control comes from studies

demonstrating NE’s effect on the capacity to temporarily store and manipulate information,

termed working memory (Arnsten, 1997, 2011). Working memory is enabled by local

dorsolateral prefrontal (dlPFC) networks that maintain mental information through recurrent

activation (Goldman-Rakic, 1995; Wang et al., 2013). NE facilitates this process by acting at

alpha-2a receptors to increase the local synaptic efficacy of dlPFC networks, and promotes their

sustained activity during working memory (Arnsten, 2011). This has led to the use of

noradrenergic agonists such as guanfacine for the treatment of disorders with executive deficits,

such as attention deficit hyperactivity disorder.

An important theoretical advancement has been the recognition that NE does not just

facilitate local processing; it is also thought to regulate the overall state of distributed cortical

networks (Bouret & Sara, 2005; Corbetta et al., 2008; Hermans et al., 2011; Eldar et al., 2013;

Schwarz et al., 2015). NE neurons project diffusely throughout most of the cortex (Schwarz et

al., 2015) and are theorized to gate a broader network of connections in the service of cognition

Page 12: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

6

(Arnsten et al., 2010). This concurs with evidence that the NE-dependent executive control

functions, such as working memory, require precise synchronization and processing across

distributed networks (Palva et al., 2010; Salazar et al., 2012). While there is strong theoretical

and experimental evidence to support the fact that NE exerts large-scale influence across cortical

networks, major questions remain regarding how it regulates the network architecture of the

human brain. These questions are a major barrier to understanding the neurobiological origin of

large-scale functional networks and the behaviors they subserve, including executive control.

Dissertation Goals

This dissertation includes two studies investigating how NE modulates brain networks

that are important for executive control. The first study tested for basic relationships between

resting brain activity and the LC-NE system. It also tested if noradrenergic coordination of

activity within functional networks relates to individual differences in inattentiveness, which is

associated with executive dysfunction. While investigating human neurotransmitter function in

vivo often requires more invasive procedures such as positron emission tomography, or

pharmaco-neuroimaging, pupil diameter can non-invasively index LC-NE activity. There is a

close link between non-luminance mediated changes in pupil diameter and the LC-NE system

(Aston-Jones & Cohen, 2005; Murphy et al., 2014), which may be established through common

brainstem efferents that influence both the LC and autonomic pupil control centers (Nieuwenhuis

et al., 2011) or through a direct pathway (Samuels & Szabadi, 2008). This has been specifically

corroborated in the context of resting-state activity, wherein spontaneous fluctuations in pupil

diameter are reliably preceded by LC spiking activity (Joshi et al., 2016). Study I thus utilized

concurrent pupillometry and resting state fMRI, as well as trait-level behavioral questionnaires to

Page 13: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

7

address two important gaps in the literature. Firstly, a limited number of studies (Fan et al., 2012;

Murphy et al., 2014; Yellin et al., 2015) have shown that functional networks may couple with

psychophysiological measures – such as galvanic skin response and pupil diameter – but this

fundamental relationship is not well validated. Study I is the largest, multi-site investigation to

date validating this proposal. Secondly, it is unknown if noradrenergic coordination of functional

network activity relates to executive control as demonstrated in everyday behavior (such as trait-

level inattentiveness). Study I therefore also tested for relationships between pupil-brain coupling

and individual differences in self-reported attentional abilities.

While Study I could make important strides towards delineating the spatial extent and

behavioral relevance of LC-NE related networks, its reliance on pupil diameter precludes causal

inference. Pupil diameter in constant luminance is sensitive to LC-NE signaling, but is likely not

specific to it. Furthermore, Study I cannot disambiguate if pupil-brain correlations resulted from

top-down regulation of the LC-NE system, or this system’s bottom up regulation of brain

activity. Hence, to more causally demonstrate noradrenergic impacts on large-scale brain

networks, Study II was a double-blind pharmaco-fMRI investigation utilizing the NE alpha-2a

receptor agonist guanfacine, a pharmaceutical agent commonly used to treat attentional and

executive dysfunction. Because brain-wide connectivity patterns (Gonzalez-Castillo et al., 2015;

Siegel et al., 2015) and their collective effects on overall network topology (Stevens et al., 2012;

Sadaghiani et al., 2015) are both thought to be important to cognition, we tested for distributed

impact on functional connectivity and network topology. Furthermore, NE neurons’ projection

patterns are highly consistent with a role in regulating overall brain state. A single NE neuron,

originating in the LC, can integrate synaptic inputs from, and project back to, numerous brain

Page 14: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

8

regions (Schwarz et al., 2015). Here we examined NE’s brain-wide impact during the resting

state and a working memory task. We utilized machine learning and graph-theory techniques

which are capable of meaningfully capturing distributed patterns of functional connectivity.

Together, studies I and II provide convincing evidence that NE regulates the spatio-temporal

structure of functional networks in the human brain, and suggest that this may be an important

neural mechanism enabling executive control.

Page 15: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

9

CHAPTER II: COUPLING BETWEEN SPONTANEOUS PUPILLARY FLUCTUATIONS AND BRAIN

ACTIVITY RELATES TO INATTENTIVENESS

Introduction

Dynamic interactions between the central and autonomic nervous systems are

hypothesized to be essential for attentional function. Salient environmental cues evoke a cascade

of autonomic changes, including pupil dilation, increased skin conductance and changes in heart

rate, which are posited to optimize adaptive behavior by facilitating response preparation

(Nieuwenhuis et al., 2011). Pupil diameter, in particular, which is controlled by sympathetic and

parasympathetic inputs (Loewenfeld, 1999), has been used to study diverse attentional processes

(Geva et al., 2013; Unsworth & Robison, 2014). Select regions of the central nervous system

such as the anterior cingulate cortex (ACC) are thought to regulate pupil-linked arousal systems

to facilitate task performance. The ACC is part of a broader network called the cingulo-opercular

network that is a convergent site of attentional and autonomic control (Menon & Uddin, 2010).

This network is thought to modulate autonomic reactivity (Menon & Uddin, 2010) and alertness

(Sadaghiani & D’Esposito, 2015) in response to attentionally demanding stimuli, perhaps by

interfacing with systems indexed by the pupil such as the sympathetic nervous system (Beissner

et al., 2013) and the locus-coruleus norepinephrine (LC-NE) system (Aston-Jones & Cohen,

2005).While previous studies have demonstrated that functional connectivity between cingulo-

opercular regions relates to individual differences in attentional abilities (Vaidya & Gordon,

2013), it remains unclear if these regions’ coordination with autonomic/arousal systems also

relates to individual differences in attentional abilities.

Page 16: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

10

Coupling between central and autonomic systems appears to be maintained in the absence

of a directed task, called the resting state. Both pupil diameter (Murphy et al., 2014; Yellin et al.,

2015) and galvanic skin response (Fan et al., 2012) correlate with resting activity in brain regions

associated with cognitive, attentional, and autonomic functions. Resting-state brain activity is

organized into networks (De Luca et al., 2006) that recapitulate task-evoked neural activity,

possibly reflecting a lifetime of co-activation (Dosenbach et al., 2007; Harmelech et al., 2013).

Furthermore, the connectivity strength of specific networks (e.g., cingulo-opercular, fronto-

parietal) predicts both task performance in the cognitive domain that the networks subserve (e.g.,

executive control) and properties of that domain in everyday life (e.g., trait-level inattentiveness)

(Vaidya & Gordon, 2013). Therefore, if coupling between select central brain regions and lower

autonomic systems is characteristic of adaptive attentional control, its strength may reflect the

history of coordinated usage of central and autonomic systems. Because attentional abilities rely

on adaptive regulation of autonomic activity, we predicted that they should relate to the strength

of resting state coupling between central brain regions regulating attention and lower

autonomic/arousal systems.

We tested this prediction by measuring functional coupling between pupil diameter and

concurrent resting fMRI brain activity, and examining its relationship to behavioral traits

symptomatic of attentional dysfunction. Since the pupil is controlled by both sympathetic and

parasympathetic subdivisions of the autonomic nervous system (which may diverge in their

relationships to both functional networks [Beissner et al., 2013] and to attentional function

[Negrao et al., 2011]), we examined relationships with inattentiveness separately within brain

regions relating to each autonomic subdivision.

Page 17: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

11

Materials and Methods

Participants

Fifty-one healthy volunteers were drawn from two sites, Georgetown University (N =

23) and University of Pittsburgh (N = 28). Sixteen participants were excluded for quality control

issues described later, leaving a final sample of 35 (10 males, 25 females; Age Mean ± SD =

29.2 ± 10.3). Participants were screened by self-report for the use of psychotropic medication,

MRI contra-indications, and psychiatric or neurological disorders. The Georgetown sample (N =

20) completed the Adult ADHD Self-Report Scale v1.0 (ASRS) (Kessler et al., 2005) to provide

a trait-level measure of attention (Inattention Scale mean = 12.3 ± 3.3; range = 8-21;

Hyperactive/Impulsive Scale mean = 11.2 ± 4.9; range = 0-20); these data could not be collected

on the Pittsburgh sample. The protocols and consent procedures were approved by the

Georgetown University Institutional Review Board and the University of Pittsburgh Institutional

Review Board; data were aggregated at the Georgetown site with no transfer of identifying

information. All study procedures conformed to World Medical Association Declaration of

Helsinki.

Data Acquisition

We combined data from the Pittsburgh and Georgetown sites to create a pupil-linked map

delineating regions related to pupil diameter. All participants included in analyses of trait-level

attention were scanned at Georgetown University with identical scanning parameters to ensure

that these analyses were not biased by scanner type or scanning parameters. Participants were

scanned while viewing a central fixation cue (black cross at Georgetown, red dot at Pittsburgh)

on a gray background (E-Prime, Psychology Software Tools) while maintaining constant

Page 18: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

12

luminance to eliminate pupillary light reflexes. Participants were told to relax, keep their eyes

open, and fixate on the central cue.

The scans acquired at Georgetown University included 288 functional images acquired

on a Siemens Trio 3T scanner (Erlangen, Germany) using a gradient echo pulse sequence with

the parameters: 47 slices with 3.2 mm thickness, TR = 2500 ms, TE = 30 ms, 90° flip angle. The

first two images were discarded to allow for signal stabilization. Further, a high-resolution T1-

weighted structural scan (magnetization prepared rapid gradient echo) was acquired with the

parameters: 176 sagittal slices with 1.0 mm thickness, TR/TE = 1900/2.52 ms, TI = 900 ms, 9°

flip angle. Pupil diameter was recorded continuously at 60 Hz with MR compatible goggles

equipped with an integrated infrared camera over the right eye (Mag Design and Engineering)

and ViewPoint EyeTracker® software (Arrington Research, Inc.).

The scans acquired at the University of Pittsburgh included 280 functional images

acquired on either a Siemens Trio 3T scanner (Erlangen, Germany), or a Siemens Allegra 3T

scanner (Erlangen, Germany). The scanning parameters were: 29 slices with 3.2 mm thickness,

TR = 1500 ms, TE = 25 ms, 73° flip angle. The first two images were discarded to allow for

signal stabilization. A high-resolution T1-weighted structural scan was acquired with the

parameters: either 224 sagittal slices with 1.0 mm thickness, TR/TE = 1630/2.48 ms, TI = 800

ms, 8° flip angle; or 175 axial slices with 1.0 mm thickness, TR/TE = 2100/3.31ms, TI = 1050

ms, 8° flip angle. Pupil diameter was recorded continuously at 60 Hz with a wall-mounted

infrared camera recording from the left eye via a hot-mirror on the coil (Applied Science

Laboratories (ASL) 5000 Eyetracker with long range optics, collected via ASL EyeTracker®

software).

Page 19: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

13

Data Analysis

Pupil Preprocessing

Pupil data were preprocessed with custom Matlab software (Siegle, 2000-2015) as

described in (Siegle et al., 2003, 2008). Similar to Siegle et al. (2003), the proportion of each

participant’s maximal dilation (95th percentile) was measured, rather than absolute pupil

diameter, because the distance between participants’ eyes and the infrared camera / hot-mirror

varied slightly. First, eye blinks and other artifacts were identified as pupillary changes too large

to represent actual dilation or contraction (criteria described in Siegle et al, 2008), and replaced

by linear interpolation. Data were then smoothed by averaging each time point with the

preceding and following time points. Regressors for subsequent analyses relating pupil and brain

data were created by convolving the artifact-corrected pupillary time courses with a canonical

hemodynamic response function (HRF) and downsampling to one value per fMRI volume. fMRI

volumes acquired during periods of greater than 20% pupil artifact were identified for later

removal. Thirteen participants with less than 100 usable fMRI volumes due to pupil artifact were

eliminated from further analysis. Within the final sample the mean usable fMRI volumes was

183.1 ± 55.8.

fMRI preprocessing

Functional images were preprocessed using SPM 8 (Wellcome Department of Cognitive

Neurology, London, UK). Images were realigned, slice time corrected, coregistered to T1

anatomical scans, normalized using parameters calculated during segmentation of T1 scans, and

smoothed with a Gaussian kernel with full-width at half-maximum of 8 mm. Any participant

showing high mean frame-wise displacement (FD) (> 0.5) in the included fMRI volumes was

Page 20: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

14

excluded from the final sample. Three participants violated this criterion and were removed from

further analysis. The average mean FD in the final sample (N = 35) was 0.16 ± 0.1 mm. Average

FD did not relate to the amount of useable pupil data (r = 0.02 p = .89) or standard deviation of

pupil time courses (r = -.01, p = .95).

Creation of pupil-linked map

To identify a pupil-linked map, two general linear models (GLMs) were performed for

each participant in SPM12 – an initial GLM to remove artifacts due to motion, physiological

noise, and estimate serial auto-correlations; and, using the residuals from the initial GLM, a

subsequent GLM in order to identify voxels significantly co-varying with pupil diameter. The

initial GLM contained nuisance regressors including the mean ventricle and white-matter signals

extracted from subject-specific masks from T1 segmentation, and the six realignment parameters

generated during image realignment. Low-frequency drifts were removed using a high-pass filter

with a 128 s cutoff. Serial auto-correlation was estimated using an autoregressive AR(1) model.

Grand mean scaling was applied with global normalization to remove nonspecific noise (Van

Dijk et al., 2010). To ensure appropriate filtering and estimation of serial auto-correlations, all

fMRI images were included in the initial GLM. The subsequent GLM tested the effects of pupil-

time course against the fMRI volumes retained after removing those occurring during periods of

excessive pupil artifact (described above). We refer to the beta value created at each voxel from

this GLM as pupil-brain coupling. In order to define a group-level pupil-linked map, each

participant’s individual pupil-linked map was included in a second-level one sample t-test, with

site as a covariate of no interest.

Page 21: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

15

Correlation with individual differences in attention

Across the 20 participants from the Georgetown site, we assessed the correlation between

pupil-brain coupling strength and inattention and hyperactivity/impulsivity ASRS scores (higher

scores indicate worse function), separately within brain regions relating to sympathetic and

parasympathetic subdivisions, using anatomical templates provided by (Beissner et al., 2013).

These templates were derived by ALE meta-analysis of previous studies using sympathetic (e.g.,

skin conductance) and parasympathetic (e.g., high frequency heart-rate variability) measures in

conjunction with neuroimaging. We created masks that included the overlap between our group

level pupil-linked map, and template sympathetic or parasympathetic related brain regions. We

then averaged the pupil-brain coupling separately within each mask and tested for correlation

with ASRS scores. We then assessed whether p-values survived Bonferroni correction for the

number of templates and ASRS sub-scales tested. To ensure these results were not driven by the

amount of useable pupil data or head motion, the analyses were also repeated after entering the

amount of usable data and average FD as covariates of no interest. We also tested if number of

usable fMRI volumes or head motion independently related to pupil-brain coupling. Finally, we

assessed potential confounds posed by individual differences in mental state during fMRI

scanning, by using metrics derived from the pupillary signal. Given evidence that variability in

spontaneous pupil fluctuations may relate to mind wandering (Grandchamp et al., 2014) we

tested if the standard deviation of pupil time-courses related to pupil brain coupling or inattention

scores. Additionally, given evidence that the spectral power of pupil fluctuations in specific

frequencies relates to alertness (Wilhelm et al., 1998), we tested if the power of pupil time

courses related to pupil brain coupling or inattention scores. We used a modified version of the

Page 22: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

16

pupillary unrest index (PUI; Lüdtke et al., 1998), a validated measure of alertness. Ludke et al

(1998) have shown similarities of the standard PUI to an index derived using power in the 0-

0.6Hz range via a Fourier Transform. To establish a dynamic analog of this measure, we applied

a continuous wavelet transform (Morelet) to the 60Hz data, smoothed with a 3.86Hz filter

(equivalent of a 10 point moving average applied twice to yield a center-weighted moving

average), and summed the power in the 0-0.6 Hz range at each sample. All TRs with less than

20% pupil artifact were then averaged to create an overall measure of alertness. In other

comparably large samples from our lab, the mean value of this index across an entire resting

state study has correlations of r>0.75 with the standard PUI index.

Results

Group pupil-linked map

The second-level analysis revealed a positive relationship between resting-state brain

activity and pupil diameter in bilateral frontal, parietal, and temporal lobes, cerebellum, and

thalamus (FDR corrected p < .05) (Figure 1.1, Table 1.1). Frontal lobe regions included regions

included dorsolateral (BA 9), ventrolateral including anterior insula and inferior frontal gyrus,

and mid cingulate extending into anterior cingulate (BA 24, 31, 32) cortex. Parietal lobe regions

included inferior parietal lobe (BA 40), and bilateral medial regions including posterior cingulate

(BA 30) extending into precuneus (BA 31). Temporal lobe regions included middle and inferior

temporal regions (BA 37), and hippocampus. A negative correlation was observed in primary

visual areas (BA 17), extending laterally along the ventral pathway into extrastriate areas (BA

18, 19), fusiform gyrus (BA 37, 19), and middle temporal gyrus (BA 21, 22, 37). The negative

Page 23: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

17

correlation was also observed in sensorimotor (BA 3, 4), parietal (BA 7) and cerebellar regions

(Figure 1.1, Table 1.1).

Table 1.1. Regions positively and negatively related to pupil diameter. Brodmann areas

(BAs) are labeled to the nearest gray matter, when available.

MNI

Coordinates

Region Name

Brodmann

Area

Extent

(voxels) t-value x y z

Regions Positively Related to Pupil Diameter

L Middle Cingulate 31 23678 10.6572 0 -42 40

L Middle Cingulate 31 4.4068 -16 -30 42

L Middle Cingulate 8.8812 -8 -8 6

R Middle Cingulate 24 4.9289 2 10 42

R Thalamus 9.7115 16 -12 12

R Putamen 7.6425 14 4 -10

L Anterior Cingulate 32 6.4617 2 28 32

R Middle Frontal 24 6.3036 24 6 40

R Middle Frontal 8 3.6307 24 46 36

R Precuneus 29 5.7705 20 -46 12

R Precuneus 31 3.9389 10 -62 30

R Precuneus 7 2.6139 6 -68 50

L Insula 5.5935 -28 12 -4

L Hippocampus 5.2046 -28 -22 -8

R Hippocampus 3.0758 34 -14 -10

R Inferior Parietal 40 4.7056 46 -52 38

L Inferior Frontal 4.5144 -30 2 32

L Middle Frontal 8 546 5.697 -30 34 36

Cerebellum (vermis) 297 4.5515 -2 -68 -12

L Cerebellum 4.2041 -16 -56 -20

Middle Frontal 393 4.0147 30 52 -12

R Inferior Temporal 37 179 3.9373 56 -48 -12

L Middle Frontal 23 3.6213 -28 42 -4

L Inferior Parietal 40 150 3.4478 -50 -54 48

R Anterior Cingulate 24 180 3.3912 12 40 -4

L Anterior Cingulate 24 23 3.3428 -10 34 -4

Page 24: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

18

Table 1.1 (cont.)

MNI

Coordinates

Region Name

Brodmann

Area

Extent

(voxels) t-value x y z

L Middle Frontal 9 45 3.2638 -28 52 30

L Cerebellum 30 3.1061 -24 -38 -22

L Precuneus 19 28 2.9632 -42 -80 34

L Inferior Frontal 46 12 2.8149 -44 30 10

R Middle Frontal 46 11 2.7537 46 36 18

L Insula 13 8 2.7059 -34 24 10

Regions Negatively Related to Pupil Diameter

L Fusiform Gyrus 19 11866 -11.0163 -32 -84 -10

L Fusiform Gyrus 37 -4.6885 -38 -54 -14

R Fusiform Gyrus 37 -6.1609 40 -52 -14

L Middle Occipital 17 -9.2198 -20 -86 18

L Middle Occipital 19 -6.8927 -48 -78 8

R Middle Occipital 18 -7.4165 26 -86 20

R Inferior Occipital 18 -8.6605 40 -84 2

R Linual Gyrus 18 -7.0746 8 -70 2

R Cerebellum -5.2967 16 -48 -4

R Superior Occipital 19 -4.9172 26 -78 46

R Postcentral Gyrus 40 289 -5.579 50 -24 52

R Inferior Parietal 7 127 -5.3195 30 -54 56

L Superior Parietal 7 151 -5.0968 -28 -54 56

L Postcentral Gyrus 3 643 -4.7076 -46 -16 44

L Middle Temporal 22 569 -4.6467 -62 -34 4

L Middle Temporal 22 -4.0755 -60 -10 -4

R Superior Temporal 22 220 -4.1114 54 8 -14

R Superior Temporal 22 -3.6781 66 -14 0

R ParaHippocampal 22 -3.8877 28 -16 -22

R Middle Temporal 21 11 -3.294 62 -6 -16

Page 25: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

19

Figure 1.1. Resting-state brain activity is associated with fluctuations in pupil diameter

(FDR corrected p < .05). The red color indicates voxels positively related to pupil diameter,

the blue color indicates voxels negatively related to pupil diameter.

Correlation with Attention Traits

Applying anatomical templates for sympathetic and parasympathetic related brain regions

from Beissner et al. (2014) to our positive group pupil-linked map, the sympathetic division

overlapped with right supramarginal gyrus, middle and anterior cingulate, thalamus and anterior

insula/inferior frontal gyrus (total k = 1,706 voxels), whereas the parasympathetic division

overlapped with posterior cingulate, precuneus, and inferior parietal lobe (total k = 800 voxels)

(Figure 1.2A). Applying templates to our negative group linked-pupil map revealed minimal

overlap with the sympathetic division (in right postcentral gyrus, k = 97 voxels) or the

parasympathetic division (in left fusiform gyrus, k = 75 voxels). Therefore, in order to reduce the

Page 26: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

20

number of comparisons, we only tested for correlation with attentional traits within regions

positively related to pupil diameter.

Participants with overall stronger pupil-brain coupling in positive sympathetic related-

regions reported lower inattention on the ASRS (r = -0.72, p = 0.0003) (Figure 1.2B), but not

significantly different hyperactivity/impulsivity (r = -0.30, p = 0.20). A partial correlation,

controlling for each participants’ amount of useable data and mean frame-wise displacement in

head motion, showed a similar relationship with inattention (r = -0.67, p = 0.002). By contrast,

pupil-brain coupling within positive parasympathetic regions did not significantly correlate with

inattention (r = -0.15, p = 0.52), or hyperactivity/impulsivity (r = -0.38, p = 0.10). A test of

dependent correlations (Steiger, 1980) revealed that coupling strengths in sympathetic and

parasympathetic regions significantly differed in their relationship to inattention t(17) = -3.10,

(p=.003). Neither mean frame-wise displacement in head motion, nor the amount of usable data

significantly related to pupil-brain coupling in sympathetic regions or parasympathetic regions

(all ps > 0.1). Furthermore, the standard deviation of pupil time-courses did not relate to pupil-

brain coupling in sympathetic related regions (r = -0.24, p = 0.31) or inattention (r = 0.17, p =

0.48). Finally, individual differences in alertness, as measured by the pupillary unrest index, did

not relate to pupil-brain coupling in sympathetic related regions (r = 0.12, p = 0.61) or inattention

(r = -0.09, p = 0.71). All statistically significant correlations survived Bonferroni correction for

multiple comparisons.

Page 27: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

21

Figure 1.2. Pupil-brain coupling correlates with trait-level attention in sympathetic related

brain regions. A) The group level map of regions positively associated with pupil diameter was

subdivided into regions that overlapped with sympathetic (red) and parasympathetic (blue) brain

areas, as defined by a recent meta-analysis (Beissner et al., 2013). For each participant, the mean

strength of relationship between brain activity and pupil diameter was computed separately

within these sub-divisions. B) Coupling between the pupil and sympathetic regions related to

ASRS inattention scores.

Discussion

We found a positive association between spontaneous fluctuations in pupil diameter and

the activity of widespread brain regions, and the strength of this positive coupling in regions

associated with the sympathetic system predicted inattentiveness. Positive pupil-linked coupling

spanned regions involved in both externally oriented processing (ACC, dlPFC, anterior insula,

supramarginal gyrus) (Fox et al., 2005) and regions of the default mode network (PCC, IPL)

often linked to internally oriented processing (Buckner et al., 2008). All these regions have been

associated with autonomic indices during the performance of tasks in cognitive or emotional

domains (Beissner et al., 2013). Parsing our pupil-linked map by Beissner et al.’s (2013)

sympathetic/parasympathetic anatomical subdivision, we observed that stronger positive

coupling in regions associated with the sympathetic system, including ACC, anterior insula, and

Page 28: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

22

supramarginal gyrus, was associated with lower inattentiveness. Together, these results indicate

that central brain regions known to interface with autonomic systems during task performance

(Beissner et al., 2013) maintain their coupling to pupil diameter during the resting-state, and this

coupling within specific regions is behaviorally relevant. We also found a negative association

between spontaneous pupil diameter and resting-state activity in primary visual and extrastriate

cortices, and primary sensorimotor areas.

While pupil diameter has long been known to spontaneously fluctuate at low frequencies

(Stark et al., 1958), its association with spontaneous brain activity has been noted only recently.

One study found positive pupil-linked coupling in regions similar to ours (Murphy et al., 2014)

and another in slightly different regions, including IPL/precuneus, but not the ACC, insula, or

thalamus (Yellin et al., 2015). Direct neural recordings from the ACC, however, also find

spontaneous coupling with pupil diameter (Joshi et al., 2016). Furthermore, markedly similar

regions also couple with spontaneous fluctuations in skin conductance (Fan et al., 2012),

indicating that resting-state central-autonomic coupling is not limited to pupil diameter. Our data

extend these findings to demonstrate that central-autonomic coupling within specific regions is

behaviorally relevant. Correlation with inattentiveness was observed in only regions associated

with the sympathetic system, which largely comprise the cingulo-opercular resting-state network

(Dosenbach et al., 2007; Seeley et al., 2007). The cingulo-opercular network is associated with

the maintenance of tonic alertness (Sadaghiani & D’Esposito, 2015), monitoring the external

environment for salient cues and, in response, initiating global network control signals in the

service of attention (Sridharan et al., 2008; Menon & Uddin, 2010). Cingulo-opercular regions

are also critical for the monitoring and control of autonomic activity (Critchley et al., 2011).

Page 29: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

23

Thus, stronger intrinsic coordination between autonomic systems and the cingulo-opercular

network may enable adaptive autonomic shifts in response to salient environmental cues, and the

maintenance of alertness during focused attention. Such attentional functions have been

associated with task-evoked pupillary responsivity (Kristjansson et al., 2009; Geva et al., 2013).

Indeed, whether individuals with stronger resting-state pupil-brain coordination also demonstrate

more adaptive task-evoked pupillary responses would be of great interest to examine in future

work.

Resting-state brain-autonomic coupling can be viewed at three related but different

levels. First, at a phenomenological level, visceral autonomic information is thought to be

constantly relayed to regions of the pupil-linked map such as ACC and insula, as a necessary

component of any subjective experience (Park & Tallon-Baudry, 2014). By this view, the

coupling in these regions may be intrinsic to subjective experience, regardless of the resting/task

state of the individual. Second, the coupling may reflect the nature of mental activity. Mind

wandering, which is commonly reported in the resting state (Diaz et al., 2013), is associated with

pupil diameter (Grandchamp et al., 2014). Since spontaneous mind wandering is associated with

ADHD symptoms, even in healthy participants (Seli et al., 2015), our observed association with

inattentiveness may reflect individual differences in mind wandering. We did not measure mind

wandering, but pupil variability, which is associated with mind wandering (Grandchamp et al.,

2014), did not correlate with inattentiveness or pupil-brain coupling in the present sample.

Additionally the pupillary unrest index, a validated measure of alertness, did not relate to pupil

brain coupling or inattention. Thus, greater or lesser mind wandering/alertness across subjects

during the resting state are likely not the source of the observed correlation of pupil-brain

Page 30: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

24

coupling and inattention. Third, the strength of coupling may reflect the accumulated history of

brain-autonomic interactions through a lifetime of task-evoked activity. Prior task-evoked

activity shapes subsequent resting-state network coupling at short (Stevens et al., 2010; Gordon

et al., 2014) and long (Ma et al., 2011) time-scales. As central-autonomic coupling is inherent to

the attentional processes one engages in throughout life experience, its strength may reflect the

history of coordinated usage of central and autonomic systems.

There are multiple potentially concurrent pathways that may underlie pupil-brain

coupling. Insular, prefrontal, and cingulate areas are thought to contribute to the generation of

autonomic activity patterns to meet behavioral demands (Cechetto & Saper, 1990; Critchley et

al., 2011). These regions communicate – possibly directly (Ward & Reed, 1946; Lowenstein,

1955), or through the thalamus, hypothalamus, and brainstem (Loewenfeld, 1999; Andreassi,

2000) – with the sympathetic with parasympathetic nuclei mediating pupil diameter.

Additionally, visual areas are speculated to send direct excitatory input to the Edinger-Westphal

nucleus, the parasympathetic nucleus whose activity constricts the pupil (Loewenfeld, 1999).

This may be the reason we observed negative pupil-brain coupling in visual areas, which was

also reported by (Yellin et al., 2015). Direct neural recordings in mice help contextualize our

negative coupling; in the absence of visual stimulation V1 cells decrease their firing rate during

pupil dilation (Vinck et al., 2015). Luminance was held constant in the present study, thereby

minimizing visual stimulation, and V1 activity may have been suppressed when pupils dilated,

yielding the observed negative coupling.

In addition to direct anatomical pathways, locus-coerulues norepinephrine (LC-NE)

system may also contribute to pupil-brain coupling. A close link between non-luminance

Page 31: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

25

mediated changes in pupil diameter and the LC-NE system has been posited (Aston-Jones &

Cohen, 2005; Murphy et al., 2014), which may be established through common brainstem

efferents that influence both the LC and autonomic pupil control centers (Nieuwenhuis et al.,

2011) or through a direct pathway (Samuels & Szabadi, 2008). In any case, the LC acts in close

conjunction with the autonomic nervous system (Sara & Bouret, 2012), has a pronounced

neuromodulatory influence on the cortex (Bouret & Sara, 2005; Castro-Alamancos & Gulati,

2014), and may optimize neural gain to meet attentional demands (Aston-Jones & Cohen, 2005).

Given that spontaneous fluctuations in pupil diameter are reliably preceded by LC spiking

activity (Joshi et al., 2016; Varazzani et al., 2015) and co-vary with LC activity measured with

resting-state fMRI (Murphy et al., 2014), the strength of pupil-brain coupling may reflect the

efficacy of noradrenergic modulation of brain activity. Pupil-brain coupling may have predicted

trait-level inattention in part due to individual differences in the LC-NE system. This is

consistent with the suggestion that pupil-brain coupling within regions such as the ACC reflects

LC-mediated coordination of brain activity (Joshi et al., 2016), and with theories that autonomic

disruptions in disorders like ADHD relate to noradrenergic dysregulation (Beauchaine, 2001).

The fact that our attentional correlation was specific to sympathetic-related brain areas also

supports this notion, because the sympathetic nervous system and the LC-NE system are thought

to often act in an integrated manner (Nieuwenhuis et al., 2011).

We attempted to account for multiple potential confounds that may influence pupil-brain

coupling. Specifically, fMRI activity is affected by physiological processes like respiration and

heartbeat, and pupil diameter is exogenously influenced by luminance and eye fixation. To

control for physiological processes, we included mean white matter and CSF signal regressors

Page 32: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

26

(Dagli et al., 1999), a common procedure in resting-state studies, and our pupil-linked map was

similar to regions co-varying with pupil diameter, after measuring and controlling for those

functions (Murphy et al., 2014). To control for exogenous influences on pupil diameter, we kept

screen luminance constant and asked participants to maintain gaze on a central fixation. In

previous work, pupil-brain coupling was still observed when no fixation stimuli was present, and

manipulating luminance did not induce fMRI activations in pupil related regions (Yellin et al.,

2015), indicating that slow fluctuations in screen luminance and eye fixation are unlikely to drive

pupil-brain coupling. Finally, our observed correlation with inattentiveness is also unlikely to be

an artifact of head motion or useable pupil data because our findings persisted after controlling

for these factors.

Overall, our results suggest that resting-state brain-autonomic coupling may prove a

useful index in understanding the broader neural networks underlying cognition. A growing body

of resting state studies has shaped current understanding of coordinated central brain circuits, and

how these circuits subserve behavior. Psychophysiology research, however, has long indicated

that autonomic changes are also important components of cognition. Our results emphasize the

contribution of brain-autonomic system coordination to individual differences in cognitive

processes subserving behavior such as attention. Additionally, our results indicate that studying

resting-state brain-autonomic coupling may also help elucidate the neural basis of autonomic

abnormalities, which are common to many psychiatric and neurological disorders. Individuals

with autism spectrum disorders exhibit decreased resting-state coupling between skin

conductance and cingulo-opercular areas (Eilam-Stock et al., 2014), regions that our results

indicate have behaviorally meaningful links to pupil diameter. Hence, functional disconnection

Page 33: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

27

between higher brain areas and autonomic systems may be a contributing factor to autonomic

abnormalities (Williams et al., 2004, 2007). Resting-state coupling may be a viable way to

measure this functional disconnection and track outcomes of behavioral or pharmacological

interventions targeting autonomic dysfunction, or attention more broadly.

Page 34: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

28

CHAPTER III: NOREPINEPHRINE ALPHA-2A RECEPTOR ACTIVATION ALTERS DISTRIBUTED

FUNCTIONAL NETWORKS DURING WORKING MEMORY

Introduction

Emerging work suggests that cognitive processes—from basic sensory awareness

(Godwin et al., 2015) to higher level attention (Rosenberg et al., 2016)—rely on highly diffuse

neural interactions. This non-localized account of brain function is supported by studies showing

that cognitive state can be reliably predicted using patterns of neural synchrony (termed

functional connectivity, FC) across the whole brain (Shirer et al., 2012; Cole et al., 2013;

Gonzalez-Castillo et al., 2015; Milazzo et al., 2016). When participants are asked to cycle

through various tasks like math, memory, and movie watching, their cognitive state is predictable

with near perfect accuracy on the basis of brain-wide FC patterns (Gonzalez-Castillo et al.,

2015). Spatially restricting these analyses—even by excluding the least predictive brain

regions—decreases their accuracy. This raises the possibility that brain-wide FC patterns may

play a role in cognitive processing.

Working memory, the active maintenance of mental information, may be especially

reliant on distributed FC patterns. Lesion (Butters & Pandya, 1969), single-unit recording

(Funahashi et al., 1989), and functional Magnetic Resonance Imaging (fMRI) (Owen, 1997)

studies have traditionally linked working memory to the dorsolateral prefrontal cortex (dlPFC).

By modeling the whole brain as a network, however, recent FC fMRI and MEG studies reveal

that engaging working memory adjusts the overall network topology of the brain (Kitzbichler et

al., 2011; Vatansever et al., 2015; Liang et al., 2016). When working memory tasks increase in

difficulty, FC changes occur across multiple canonical networks such as default-mode,

Page 35: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

29

executive, and salience networks (Liang et al., 2016). Large-scale FC patterns not only

dynamically change during working memory, they also reliably distinguish individuals who

perform better on working memory tasks (Stevens et al., 2012; Stanley et al., 2015; Vatansever

et al., 2015) and improve their performance after training (Yamashita et al., 2015). Whole-brain

network interactions are thought to enable working memory by facilitating the integration of

disparate information (Kitzbichler et al., 2011; Vatansever et al., 2015; Liang et al., 2016). This

aligns with theoretical accounts of a “global mental workspace” underlying working memory

(Baars, 2002). Together, these studies raise the possibility that working memory may rely on

task-specific and widespread patterns of FC. The neurochemical mechanisms that adaptively

regulate large-scale FC patterns during working memory, however, remain unclear. This poses a

barrier not only to the understanding of working memory, but also to the utilization of FC

network measures to inform clinical decision making when working memory is impaired.

Norepinephrine (NE) signaling is one potential mechanism that regulates both distributed

neural interactions and working memory. Theoretical models and the anatomy of NE neurons

suggest that NE modulates overall state of cortical networks in the service of arousal, attention,

and mood (Sara & Bouret, 2012). A single NE neuron, originating in the locus-coeruleus (LC),

can integrate synaptic inputs from, and project back to, diverse brain regions (Schwarz et al.,

2015). Models such as the “network reset” theory of NE posit that the signaling of these LC

projections can interrupt the activity of neural networks and reorganize them into configurations

that determine behavioral output (Bouret & Sara, 2005). In the context of working memory,

animal models show NE acts at diffuse (Scheinin et al., 1994; Tavares et al., 1996) post-synaptic

alpha-2a receptors to quickly alter their synaptic weights and thereby regulate the strength of

Page 36: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

30

neural interaction patterns (Wang et al., 2007; Arnsten et al., 2010). Blocking these receptors

leads to profound impairments in working memory (Li & Mei, 1994), which suggests they play a

critical role in working memory. These findings have led to the use of NE alpha-2a agonists such

as guanfacine for the treatment of disorders with attentional and cognitive control deficits, such

as attention deficit hyperactivity disorder (ADHD) (Hnatko, 2002). NE’s role in working

memory, taken together with network models of NE function, lead us to hypothesize that NE

signaling at alpha-2a receptors plays an important role in regulating large-scale patterns of FC

during working memory.

We tested this hypothesis in a double-blind pharmaco-fMRI study utilizing the NE alpha-

2a receptor agonist guanfacine. On two separate visits, a group of 20 healthy adult volunteers

were scanned with fMRI during a resting state (to determine NE alpha-2a receptors’ impact on

baseline network architecture) and a verbal N-back task (to determine NE alpha-2a receptors’

impact when working memory is required). Prior to scanning, participants received 1.5 mg

guanfacine on one of the visits, and placebo on the other visit, in counterbalanced order. To test

for changes in FC we used univariate approaches, and two techniques that are well-suited for

examining whole-brain patterns of FC – multivariate machine learning and graph theory.

Applying univariate approaches is important as they are commonly applied and therefore allow

comparison with past findings. It is conceivable, however, that NE may have subtle whole-brain

impact that require higher sensitivity to detect. Previous work has demonstrated that meaningful

overall patters of FC are readily detectable with high sensitivity using multivariate (Cole et al.,

2013; Gonzalez-Castillo et al., 2015; Milazzo et al., 2016), and graph-theory methods (Giessing

& Thiel, 2012). We therefore used machine learning (support vector machine; SVM) to predict

Page 37: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

31

drug vs. placebo state based on the FC between a set of previously defined ROIs that covered the

entire cortical surface (Gordon et al., 2016). To test for the specificity of these findings we also

derived data-driven functional networks with community detection (Blondel et al., 2008) and

conducted two types of follow up SVM analyses. Firstly we conducted one SVM per network

that included only that network’s connections (i.e. the connections of the default mode network

only), and secondly we conducted one SVM per network with that network removed from

consideration (i.e. the connections of all other networks except the default mode network). This

allowed us to test if guanfacine’s network effects were restricted to specific networks. Based on

theories that working memory is subserved by multiple functional networks, as well as the

diffusivity of the noradrenergic system, we hypothesized that drug state would be best detected

on the basis of multivariate patterns of FC across all networks.

We next tested for changes in overall topology of FC patterns caused by guanfacine using

modularity, a metric from graph theory, which quantifies the overall balance of integration and

segregation between distributed brain networks. While there are numerous graph theoretic

measures that capture neurobiologically meaningful whole-brain network parameters, we utilized

modularity because it has been associated with cognitive functions, particularly working

memory. Intra-individual changes in brain-wide network modularity are associated with

variations in WM performance between days (Stevens et al., 2012), differences in working

memory load (Vatansever et al., 2015), and trial-to-trial variations in stimulus detection

(Sadaghiani et al., 2015). We therefore further hypothesized that alpha-2a receptor related

alterations in FC ought to collectively increase brain network modularity.

Page 38: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

32

Results

Guanfacine’s effect on behavior and physiology

Working memory performance was measured with accuracy and reaction time on a

verbal N-back task with four loads (1-back through 4-back) performed during fMRI scanning.

Drug (placebo, guanfacine) x Load (1-, 2-, 3-, 4-back) repeated-measures ANOVAs revealed no

main effects of guanfacine on accuracy, F(1,19) = 1.1, p = .31, or reaction time, F (1,19) = 1.2, p

= .28, and no significant Drug x Load interactions on accuracy, F(3,57) = 1.8, p = .20, or reaction

time, F(3,57) = 1.5,p = .22. As expected, there were significant main effects of N-back load on

accuracy, F(3,57) = 19.1, p = .00, and reaction time, F(3,57) = 11.0, p = .00 (Figure 2.1A&B);

participants showed less accurate and slower performance at higher loads. These results are in

line with previous work demonstrating a lack of working memory differences on guanfacine in

healthy adults (Müller et al., 2005 although see Jäkälä et al., 1999). Guanfacine has more

pronounced impact on WM, however, in those with lower working memory ability, such as

individuals with traumatic brain injury (McAllister et al., 2011). In our sample, task accuracy

was high in all loads, but it is feasible that those with lower accuracy on placebo may have

accrued benefits. Consistent with this possibility, there was a significant negative relationship

between placebo accuracy and drug-related accuracy changes (guanfacine - placebo accuracy

across all loads), B=-.54, p=.01, indicating that individuals with lower placebo accuracy showed

a larger magnitude of performance improvement on guanfacine (Figure 2.1C). To ensure this

was not driven by order effects, order of visits (i.e. placebo or drug first) was included as a

covariate of no interest. As evident in Figure 2.1A, drug-related performance change was small

in magnitude and differed among the sample with only a subset showing improvement. Across

Page 39: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

33

the full sample, however, the lack of significant drug-related change in average performance

allows for interpretation of our brain network findings without confounds of performance

differences.

Figure 2.1. Guanfacine did not Alter N-back Performance. (A-B) Mean reaction time and

accuracy for all N-back loads. Error bars represent standard error of the mean. (C) The

relationship between placebo accuracy and the change in accuracy on guanfacine (guanfacine –

placebo).

We also tested for basic physiological effects of guanfacine. Systolic and diastolic blood

pressure, and pulse rate were not significantly different from placebo before or after dosing (all

ps >.25; Supplementary Figure 2.1). The lack of physiological side effects suggests that our brain

network findings were not due to hemodynamic effects driven by blood pressure or pulse rate.

Page 40: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

34

Guanfacine’s effects on FC detected with univariate methods

We next tested if guanfacine induced brain-wide changes in functional connectivity using

univariate methods. Functional connectivity was defined as the fisher-transformed temporal

correlation between a set of 333 previously defined ROIs that encompassed the entire cortical

surface (Gordon et al., 2016). Guanfacine’s effects were assessed using paired-t tests (drug vs.

placebo) conducted on all functional connections, false discovery rate (FDR) corrected for

multiple comparisons. No connections during the resting state or N-back task survived correction

for multiple comparisons.

Uncorrected results, however, suggested that, during both the resting state and N-back

task, guanfacine altered distributed functional connectivity patterns in a manner that was not

spatially restricted to any specific region. Results at the p < .001 uncorrected level are shown in

Figure 2.2, and the t-scores of all comparisons are shown in figure 4E&F.

To determine if guanfacine’s effects during working memory varied by load, for each

connection we also conducted a repeated measures Drug X Load ANOVA. To increase power,

we averaged across 1- and 2-back loads to create a low load condition, and 3- and 4-back loads

to create a high load condition. Significant Drug x Load interactions (Figure 2.2) revealed that

the effect of guanfacine on many connections was load-dependent. Again, no connections

survived FDR correction for multiple comparisons. Overall, univariate analyses revealed

widespread effects of guanfacine during both the resting state and WM. No connections survived

statistical correction for multiple comparisons.

Page 41: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

35

Figure 2.2. Guanfacine Induced Wide-Spread FC Changes. The top panels show connections

significantly altered by guanfacine, detected with paired t-tests (p<.001, uncorrected). The

bottom panel shows connections whose drug effects depended on N-back load, detected with

repeated measures ANOVAs (p<.001, uncorrected). The color of connections indicates t or F

values. The color of nodes indicates their network affiliation, as determined with community

detection.

Guanfacine’s effects on FC detected with multivariate methods

Our univariate analyses suggested that NE alpha-2a receptor activation induced

distributed and complex changes in FC, consistent with our hypothesis and the known

distribution of NE alpha-2a receptors across the entire brain (Scheinin et al., 1994; Tavares et al.,

1996). These analyses were difficult to definitively interpret, however, because no changes

survived correction for multiple comparisons.

Page 42: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

36

Previous work has demonstrated that meaningful overall patterns of FC are readily

detectable with high sensitivity using multivariate methods (Cole et al., 2013; Gonzalez-Castillo

et al., 2015; Milazzo et al., 2016). We utilized linear support vector machine (SVM) to predict

drug vs. placebo state using all ROI-to-ROI FC strengths as features. Although machine learning

analyses are sometimes restricted to connections expected to best differentiate the condition of

interest, we included FC strengths from across the brain to address our hypothesis regarding

noradrenergic influence on wide-spread networks.

We trained and validated SVM models using leave one out cross validation (LOOCV);

for each person, a model was trained excluding their drug and placebo data, and the trained

model was then used to predict their drug vs. placebo state from connectivity strengths. SVM

analyses detected drug vs. placebo during the resting state with 60% accuracy (55% sensitivity,

65% specificity; Figure 2.3A). Although this result was above chance accuracy (50%), testing for

statistical significance against a null distribution ( LOOCV accuracy computed 1,000 times with

drug and placebo labels randomly switched), revealed that SVM accuracy during rest did not

reach statistical significance (p=.13). SVM analyses during the N-back task, however, detected

drug state with 75% accuracy (75% sensitivity, 75% specificity; Figure 2.3A) and with high

statistical significance (p < .001). A direct comparison of rest vs. N-back SVM accuracies

through permutation testing revealed a marginally significant accuracy difference between the

conditions (p = .055). Although our univariate analyses above suggested that drug effects during

the N-back may have been load dependent, LOOCV accuracy was low when restricting SVM

Page 43: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

37

analyses to either the low load blocks (50%), or high load blocks (55%), perhaps due to lower

power.

Figure 2.3. SVM Analyses Confirm Wide-Spread FC Modulation by Guanfacine. (A) The

prediction accuracy (drug vs. placebo) of SVM analyses utilizing all functional connections. (B)

Specificity of N-back SVM results was assessed by restricting analyses to, and excluding,

connections detected with univariate analyses at p thresholds of .001-.05. (C) Specificity was

also assessed by systematically restricting analyses to, and excluding, each network. (D) The

connections most significantly driving overall SVM analyses were detected through permutation

testing on SVM feature weights. The color of connections indicates their SVM feature weight.

The color of nodes indicates their network affiliation. (E) The distribution of feature weights

amongst networks.

Follow up analyses assessing the specificity of FC changes during working memory

We tested for the specificity of our working memory findings in several ways. Firstly, we

tested if guanfacine’s effects were limited only to the regions detected with traditional univariate

Page 44: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

38

approaches. Restricting the SVM to only include significant connections from univariate

analyses decreased LOOCV accuracy, and this was true across many different p-value thresholds

(Figure 2.3B). Furthermore, excluding significant connections from the SVM entirely did not

result in marked decreases in LOOCV accuracy (Figure 2.3B). Together, these results suggest

that NE alpa-2a activation altered diffuse patterns of functional connectivity even beyond

connections detected with traditional univariate methods.

We next tested if select networks or connections drove our N-back SVM classifier

results. We derived data-driven functional networks that were common across participants and

visits during the N-back task. We utilized the Louvain community detection algorithm (Blondel

et al., 2008) on weighted functional connectivity matrices (Rubinov & Sporns, 2011), which sub-

divides brain areas into partitions that are highly connected within themselves, and sparsely

connected to other partitions. We used this method because it has been previously used to define

networks both during rest (Rubinov & Sporns, 2011) and working memory (Vatansever et al.,

2015; Liang et al., 2016), and it also quantifies how modular these networks are – the graph-

theory metric we utilized in later analyses. Defining a common set of networks used in both

follow-on SVM and later graph theory analyses facilitated clearer interpretability of both

analyses. It allowed us to first to assess the specificity of network changes, and then quantify the

collective effects of guanfacine on these same networks.

The results of the Louvain algorithm can vary from run to run, so we also used a

consensus clustering approach (Lancichinetti & Fortunato, 2012). For each participant, we

generated 1,000 network partitions, and used the consensus approach to group nodes together

that were commonly assigned to the same network across the 1,000 runs. We then used the

Page 45: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

39

consensus approach again to find networks present across participants and visits (Figure 2.4A).

These networks corresponded well to known functional networks: including a network

encompassing the posterior cingulate, inferior parietal lobe, and medial frontal cortex (default

mode network, red in Figure 2.4A); a network encompassing sensory and motor regions (sensori-

motor network, gray in Figure 2.4A); a network encompassing lateral frontal, parietal, and

cingulo-insular regions (cognitive control network, yellow in Figure 2.4A); and a network

encompassing primary and secondary visual areas (visual network, blue in Figure 2.4A).

After deriving these networks, we removed all of the ROIs of each network, one network

at a time, and conducted a SVM on the remaining connections. The removal of each network

resulted in a decrease in LOOCV accuracy (Figure 2.3C), indicating that each network

contributed to overall classification accuracy. Conversely, conducting SVM analyses on the

connections of each network in isolation revealed that each network alone was sufficient to

predict drug state with above chance accuracy (Figure 2.3C), but no single network achieved

accuracy as high as when all networks were included. Together, these results suggest that the

effects of NE alpa-2a activation were not restricted to any large-scale network.

In linear SVM analyses, feature weights are sometimes used to provide a more detailed

interpretation of SVM models (although see Haufe et al., 2014). Therefore, lastly, to confirm if

functional connectivity differences were confined to restricted networks, we computed an SVM

model from all participants’ data, and summed feature weights by network (Figure 2.3E). While

the summed weights indicated which networks were more important in an absolute sense, larger

networks may have contributed more because they contained more features. We therefore also

averaged feature weights by network (Figure 2.3E). Both the summed and averaged feature

Page 46: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

40

weights were distributed amongst networks, again indicating that guanfacine induced widespread

connectivity changes that were not specific to any network.

For a more fine-grained analysis of which connections were most strongly weighted in

the SVM model, we created a null feature weight distribution for each connection. We randomly

permuted drug and placebo labels 15,000 times, and obtained SVM feature weights for each

iteration. The true feature weight of each connection was then compared against its own null

distribution to compute a p-value. The most strongly predictive connections (p < .001) are shown

in Figure 2.3D. Again, the connections most strongly altered by guanfacine were distributed

across the brain.

NE alpha-2a receptor activation’s effects functional network topology

SVM analyses revealed that NE alpha-2a receptor activation induced distributed changes

in functional connectivity that were more readily detectable when cognitive control was required.

We next tested if, cumulatively, these changes altered overall brain network topology. Because

network modularity is a topological feature thought to be important for cognitive control, we

tested if guanfacine altered network modularity during the resting state and N-back task. The

Louvain method above that was used to generate functional network partitions (Figure 2.4A) was

also used to quantify how modular these network partitions were. We averaged modularity

values across each participant’s 1,000 iterations of the Louvain algorithm for drug and placebo

during each state (rest and N-back) and conducted paired t-tests to examine the effect of

guanfacine. Guanfacine did not alter modularity during rest (guanfacine: Q = .32; placebo: Q =

.32), t(19) = .07, p = 0.943, but significantly increased modularity during working memory

(guanfacine: Q = .32; placebo: Q = .34), t(19) = 2.42, p = 0.037 (Figure 2.4B&C). A repeated

Page 47: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

41

measures ANOVA did not reveal a significant drug x state interaction, F(1,19) = 1.3, p = .25.

Furthermore, changes in modularity did not correlate with changes in N-back accuracy (r = -.20,

p = .40).

We reasoned that two different scenarios were capable of producing this pattern of

results; modularity during the N-back could have differed amongst fundamentally re-organized

networks, or the same networks could have become more modular. To adjudicate between these

possibilities, we tested if the networks during the N-back re-organized between drug and placebo

visits. We derived networks separately for each visit using the consensus clustering approach,

and quantified how different these network structures were using the normalized variation of

information (VIn). This metric quantifies how much information is lost and gained when

comparing two network partitions (Meilă, 2007). Next, to determine if network structure was

significantly altered by guanfacine, we constructed a null distribution by randomly permuting

drug and placebo labels 1,000 times, creating consensus network partitions, and comparing these

partitions with VIn (Dwyer et al., 2014). This analysis revealed that network structure on

guanfacine and placebo visits was not significantly different (p = .38) (Figure 2.4D).

Collectively, these findings indicate that the same core functional networks became more

modular on guanfacine, and this effect was specific to when cognitive control was engaged.

Page 48: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

42

Figure 2.4. Guanfacine Increased the Modularity of Brain Networks. (A) Consensus

clustering applied to the results of Louvain community detection revealed four networks that

were present across participants and drug and placebo visits. (B) Spring-graph visualizations of

N-back network structure in a representative participant. Colors represent network affiliation of

each node. (C) Comparison of network modularity values across all drug and task states. (D) A

box plot showing VIn values. Network structure between drug and placebo visits was similar for

all participants. (E) N-back functional connectivity matrices on placebo and guanfacine. (F) A

matrix of t-values for paired t-tests of guanfacine vs. placebo.

Discussion

The present findings demonstrate that NE alpha-2a receptor activation induces wide-

spread FC changes that are more readily detectable during working memory. We found that drug

state (guanfacine vs. placebo) was predictable at above chance levels during both the resting

state (60%) and working memory (75%) on the basis of whole-brain FC. Only predictions during

working memory, however, reached statistical significance upon permutation testing.

Page 49: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

43

Furthermore, we found that FC changes were not restricted to any one network during working

memory. The FC of each functional network in isolation was sufficient to predict drug state at

levels above chance, but no network achieved accuracy as high as that with whole-brain FC. We

also found that FC changes extended beyond regions detected with traditional univariate

analysis; drug state could be predicted even when exclusively utilizing connections not showing

significant univariate drug vs. placebo differences. Finally, we found that the FC changes

induced by NE alpha-2a activation collectively altered the overall topology of functional brain

networks. Guanfacine increased whole-brain network modularity during working memory, a

metric which indexes the overall balance of network integration and segregation, and has been

linked to better working memory abilities (Stevens et al., 2012). Although guanfacine did not

affect average working memory performance, its effects varied across individuals such that those

with lower performance on placebo showed more drug-related improvement. Together, these

findings indicate that NE alpha-2a receptor activation during working memory engagement has

brain-wide effects that drive networks into a more modular configuration, a state thought to

facilitate executive control.

NE neuromodulation shapes FC across the brain

FC fMRI studies have increasingly linked a variety of cognitive processes to whole brain

network structure. When cognitive functions— such as working memory—are engaged, there are

subtle but distributed changes in network FC (Cole et al., 2014). It is important to consider what

neurochemical mechanisms may drive these changes. Animal models suggest that the anatomical

connectome provides a stable structure whose activity is constantly modified by

neuromodulation (Marder, 2012). NE is one such neuromodulator that is thought to interrupt the

Page 50: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

44

activity of neural networks and reorganize them into configurations that determine behavioral

output (Bouret & Sara, 2005). Our data is consistent with this notion, demonstrating that, when

working memory is engaged, NE alpha-2a receptors modify FC across the cortex. Other

neuromodulators such as acetylcholine (Giessing et al., 2013) and dopamine (Carbonell et al.,

2014) also produce whole-brain network changes. Our data, in conjunction with these studies,

suggests that neuromodulation may allow for flexible and adaptive modification of human large-

scale brain network structure.

We utilized multivariate methods to study noradrenergic modification of network

structure in the present investigation for several reasons. We utilized modularity because this

metric may reveal whole-brain communication patterns, and therefore capture cognitive

processes that rely on signaling across broad extents of sensory and association cortex (Godwin

et al., 2015), like working memory. It has also been argued that whole brain metrics like

modularity are well suited to studying cholinergic and noradrenergic systems, which affect broad

cognitive processes and whose neural impact are subtle and highly non-focal (Giessing & Thiel,

2012). Hence, our finding that guanfacine increased modularity during working memory is an

important first step to confirming our hypothesis that large-scale network communication is

altered by noradrenergic signaling. However, a disadvantage of this metric is that it cannot reveal

which specific patterns of FC were modulated by the drug, as distinct patterns could produce

similar modularity values. Therefore, future studies are needed to characterize the precise nature

of the observed noradrenergic effects. Hence, while our graph theory analyses are well targeted

to capture distributed noradrenergic effects, follow-up studies are needed to probe granular

noradrenergic changes within specific networks or regions. Our SVM analyses were capable of

Page 51: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

45

producing network specific effects, however, these also indicated that guanfacine induced wide-

spread FC changes in all networks. Furthermore, SVM analyses can detect complex patterns, but

their output makes it difficult to parse the specific nature of these patterns. One potential method

to gain insight into the patterns underlying SVM accuracy is to examine feature weights, as we

have done here. It is not possible to unequivocally assign neurobiological relevance to SVM

feature weights, however, because non-zero feature weights can be observed for regions or

connections that are statistically independent of the examined brain process (Haufe et al., 2014).

Therefore, while our SVM analyses confirmed the widespread nature of noradrengeric effects,

the difficulty of interpreting the pattern underlying their accuracy reinforces the importance of

future studies aimed at probing granular noradrenergic changes within specific networks or

regions.

It should also be noted that, while our results indicate that NE alpha-2a receptor

activation induces large-scale network changes, we cannot definitively ascribe behavioral

relevance to all of these changes. There are several possible reasons why we did not observe a

significant working memory performance difference on guanfacine. Firstly, repeated or higher

dosing may be needed to produce behavioral effects. Daily guanfacine treatment is sometimes

required before behavioral effects manifest in ADHD populations (Scahill et al., 2001).

Secondly, heterogeneity in behavioral drug response may necessitate larger sample sizes to

detect effects, whereas functional brain measures may be more sensitive. Our results indicate

some heterogeneity may result from baseline performance differences, because placebo accuracy

was predictive of drug effects on working memory performance. Hence, those with lower

baseline working memory may accrue more cognitive benefits from guanfacine. Thirdly, our N-

Page 52: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

46

back task may be have been insensitive to working memory differences because of ceiling

effects, or insufficient difficulty of our cognitive loads.

The absence of a behavioral finding makes it difficult to determine potential cognitive

changes related to the observed functional connectivity modulations. It is possible that our

functional connectivity effects were specific to working memory processes, and with a more

difficult task or higher sample size, behavioral differences would have emerged. It is also

possible that guanfacine altered more general cognitive processes, such as alertness or attention,

during the working memory task. Therefore, it is important for future studies to determine if

guanfacine induces large-scale network changes non-specifically during cognition, or its effects

are specific to tasks requiring working memory.

NE’s network effects depend on behavior state

In the present study, we examined whole brain networks during two behavioral states in

the same participants. Permutation testing revealed that drug state was significantly better

predicted by FC during working memory compared to the resting-state, and our graph theory

findings were also specific to working memory, although our design was limited in detecting a

significant drug x state interaction. It is possible that a larger sample size would be needed to

find graph theory effects that were specific to cognitive state The fact that our SVM accuracy

was significantly higher during working memory, however, suggests that NE’s network effects

can differ by cognitive state. There are likely several reasons for this. Firstly, different NE

receptors are activated in different behavioral states, and these receptors can have very different

cognitive and neural effects (Arnsten, 2009, 2011). Secondly, neurmodoluatory influence on

neural networks depend on numerous exogenous and endogenous factors (Giessing & Thiel,

Page 53: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

47

2012), including the prior history of the network and the other neuromodulators present in the

network (Marder, 2012). Collectively, this indicates that studies seeking to study the large-scale

network effects to neuromodulators, or predict neuromodulatory drug response from network

measures, should consider cognitive state.

The dependence of network effects on cognitive state also indicates that a more complete

model of NE action requires consideration of other behavioral domains. Previous human

neuroimaging work has shown that NE also adjusts functional networks during stress (Hermans

et al., 2011), attention (Coull et al., 1999), emotional control (Schulz et al., 2014), and motor

control (Wang et al., 2011). It remains unclear, however, how NE’s impact on networks differs

as a function of behavioral state. One reason for this uncertainty is that most previous studies

examined isolated networks or behavioral states. Future studies examining different behavioral

domains should test if NE alters global network parameters in a fashion similar to our working

memory findings, or has more restricted effects.

Implications for future pharmaco-neuroimaging studies

In addition to furthering understanding of NE’s network effects, guanfacine’s global

network impacts also have important implications for future pharmacological neuroimaging

studies. Studies combining neuroimaging with acute or long-term drug dosing have lent

important insights into mechanisms of psychiatric drug action and helped predict heterogeneous

clinical responses (Schaefer et al., 2014; Sarpal et al., 2016; Whitfield-Gabrieli et al., 2016). Like

the LC-NE system, however, many neuromodulatory systems targeted by common

pharmaceutical agents – such as the serotonin system (Frazer & Hensler, 1999) – are also widely

dispersed. This suggests that other psychiatric medications may also have global network effects.

Page 54: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

48

Furthermore, the fact that guanfacine’s network effects extended beyond connections detectable

with univariate methods, suggests that univariate methods may be insufficient to capture whole-

brain network changes. For this reason, machine learning and graph theory techniques, which are

suitable for detecting distributed and multivariate patterns of change, may prove useful in future

pharmacological neuroimaging studies.

Mechanisms of guanfacine induced network changes

Although it can be difficult to assign mechanisms of action to neuropsychiatric drugs,

several mechanistic considerations provide important context in interpreting our results. Firstly,

while guanfacine has high binding affinity for NE alpha-2a receptors (Uhlén & Wikberg, 1991),

these receptors can be located both pre-synaptically and post-synaptically (Aoki et al., 1998).

This is important to consider because the neural and behavioral effects of these receptors differ

based on their synaptic location. Whereas guanfacine’s cognitive effects are attributed to post-

synaptic alpha-2a receptors, pre-synaptic receptors can down-regulate the LC (Arnsten, 2011). In

comparison with other alpha-2a receptor agonists like clonidine, however, guanfacine has 10

times weaker effects on pre-synaptic LC inhibition (Engberg & Eriksson, 1991), and 10-100

times stronger effects on post-synaptic actions (Arnsten et al., 1988). Hence, there is evidence

that guanfacine’s network effects may be ascribed predominantly to the cognitively relevant

post-synaptic alpha-2a receptors, but we cannot definitively test this possibility with the present

data.

The second important mechanistic consideration is the possibility that not all FC changes

were caused by direct drug actions within the altered networks. Localized neural activity changes

can propagate widely throughout many networks (Gratton et al., 2013). Guanfacine may have

Page 55: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

49

caused distributed network changes in part through local impact on key network nodes. Animal

(Arnsten et al., 2010) and fMRI studies (Clerkin et al., 2009; Schulz et al., 2013) indicate that

alpha-2a activation may have a particularly strong impact in the dlPFC, where it closes

Hyperpolarization-activated Cyclic Nucleotide-gated (HCN) channels, strengthening the

functional connectivity of local PFC networks (Wang et al., 2007). Furthermore, local processing

in PFC networks is thought to influence behavior through top-down modulation of other brain

regions (Munakata et al., 2011). It is thus possible that guanfacine altered distributed networks

directly, or indirectly through its actions in the PFC and other key network nodes.

The final important mechanistic consideration is that guanfacine may also have other

effects that indirectly affect fMRI network measures. Given that guanfacine is often prescribed to

treat attention deficits and hyperactivity (Hnatko, 2002), it is conceivable that reduced head

motion during on-drug scanning may relate to the observed drug-related FC changes. Even small

head-motion differences can bias estimates of FC (Power et al., 2012). In the present sample,

however, head motion did not differ by drug state and therefore unlikely to have contributed to

FC differences. Additionally, guanfacine can also alter performance on cognitive tasks (Jäkälä et

al., 1999). If task performance was changed by guanfacine, the drug’s impact on the brain could

be attributable to either direct action, or to indirect effects due to performance differences. N-

back performance, however, was not significantly altered on drug, which enables us to interpret

drug effects on the brain, without the confounds of performance differences (Casey, 2002;

Clerkin et al., 2009). Finally, guanfacine can have basic physiological effects, which may

influence the fMRI signal. None of the physiological parameters we measured, including heart

rate and blood pressure, however, were significantly altered by the drug.

Page 56: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

50

Conclusion

In this report we found that NE alpha-2a receptor activation during working memory

affects the overall network structure of the human brain. This lends support to models such as the

network reset theory of NE action, which posits that the LC-NE system can reset brain networks

and reorganize them in a manner that supports behavioral output. More broadly, these results

contribute to a growing body of work demonstrating that neuromodulation may allow for flexible

and adaptive modification of human large-scale brain network structure in the service of

cognition.

Materials and Methods

Participant Screening and Demographics

20 (7 male, 13 female) participants, aged 21-37 (mean 27 years) were recruited from a

Washington, DC area community sample. All participants provided informed consent. The

research protocol and consent procedures were approved by the Georgetown University

Institutional Review Board. Prior to fMRI scanning visits, participants were screened for

exclusionary criteria; including heart, lung, kidney, neurological or psychological disorders;

currently smoking cigarettes; blood pressure or resting heart rate values outside of the 5th - 95th

percentile range (by gender and age); or abnormal EKG results. Screening data were collected by

the study research team and trained nursing staff, and reviewed by our study physician.

Participant characteristics are shown in Table 2.1.

Page 57: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

51

Table 2.1. Participant demographics.

Mean (S.D.)

Height (cm) 169.55 (10.34)

Weight (kg) 73.02 (15.36)

Age (years) 27.05 (4.72)

Male:female ratio 7:13

FMRI scanning protocol and medication dosing

Participants were scanned with fMRI on two separate visits (average of 12 days apart),

and were asked to abstain from caffeine for at least 2 hours prior to each visit. 90 minutes prior

to fMRI scanning, participants once received placebo and once 1.5 mg guanfacine. Order of

dosing was randomized, counter-balanced and double blind. Physiological monitoring was

conducted by trained nursing staff 30 minutes prior to dosing, 45 minutes post-dosing, and after

fMRI scanning (mean time post-dosing = 150 minutes) (Supplementary Figure 2.1).

FMRI scanning was conducted on a 3T Siemens Trio (Erlangen, Germany). 288

functional images were collected during the resting state scan, and 209 functional images were

collected during the N-back task. Both functional scans were collected with gradient echo pulse

sequences with the parameters: 47 slices with 3.2 mm thickness, TR = 2500 ms, TE = 30 ms, 90°

flip angle. The first two images from each functional run were discarded to allow for signal

stabilization. Further, a high-resolution T1-weighted structural scan (magnetization prepared

rapid gradient echo) was acquired with the parameters: 176 sagittal slices with 1.0 mm thickness,

TR/TE = 1900/2.52 ms, TI = 900 ms, 9° flip angle.

Task parameters

During the resting state scan, participants were asked to stay awake and keep their eyes

focused on a central fixation cross. The N-back task consisted of twelve 30 second N-back

Page 58: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

52

blocks (3 blocks each of 1-,2-,3-, and 4-back loads, in pseudorandomized order). Blocks began

with 3000 ms of instructions indicating the N-back condition, and all blocks but the last were

followed by a 14 second fixation. During each block, a series of 9 black consonants were

displayed for 500 ms in the center of a gray background, with an inter-stimulus-interval of 2500

ms. Participants were told to press a button in their right hand if the letter currently displayed

matched n letters ago, and a button in their left hand if it did not match. 18% of all trials were

targets, with 1-2 targets per block. Stimuli were presented with E-prime (Psychology Tools Inc.,

Pittsburgh, PA). Participants practiced the N-back task outside of the scanner on both visits. Both

accuracy and reaction time measures were created by first computing separate means within

target and non-target trials, and then averaging these.

MRI data preprocessing and functional connectivity calculation

Using SPM12 (Wellcome Department of Cognitive Neurology, London, UK)

implemented in MATLAB (Mathworks, Inc., Sherborn, MA), fMRI images were realigned,

slice-time corrected, normalized using parameters calculated through segmentation of EPI

images, and smoothed with a 8 mm FWHM Gaussian kernel. Anatomical volumes were

segmented into grey matter, white matter, and cerebro-spinal fluid, and the resulting masks were

eroded by one voxel to minimize partial volume effects.

The conn toolbox (Whitfield-Gabrieli & Nieto-Castanon, 2012) was used to linearly de-

trend fMRI time series and to remove confounding effects by nuisance regression; which

included 3 principle components each from subject specific white matter and cerbero-spinal fluid

masks; 6 motion parameters and their first-order temporal derivatives; and point-regressors to

censor time points with mean frame-wise displacement (FD) > .5 mm (one covariate per

Page 59: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

53

censored volume, consisting of a one at the censored time point and zeros for all other time

points). For N-back data, a regressor for each task condition (1-4 back load) was convolved with

a hemodynamic response function and included as an additional nuisance regressor, along with

its first order temporal derivative. Residual time-series were band-pass filtered (0.008 Hz < f <

0.09Hz). Importantly, during both rest and the N-back conditions, the mean FD and number of

time points censored due to motion were not significantly different on guanfacine (all ps > 0.21).

The mean FDs ±SD (range) for all conditions were; rest placebo: 0.13 ± 0.05 (0.07-0.25), N-back

placebo: 0.10 ± 0.05 (0.04-0.22), rest guanfacine: 0.15 ± 0.12 (0.06-0.53), and N-back

guanfacine: 0.10 ± 0.06 (0.05-0.25). The mean censored time points ±SD (range) for all

conditions were; rest placebo: 4.95 ± 7.12 (0-29), N-back placebo: 1.55 ± 2.04 (0-8), rest

guanfacine: 11.65 ± 24.88 (0-97), and N-back guanfacine: 1.70 ± 4.37 (0-19).

Following nuisance regression and band-pass filtering, mean time courses were extracted

from a previously defined set of 333 ROIs (Gordon et al., 2016). These ROIs were created

through a parcellation of resting state data that yielded highly homogenous parcels with a

network structure similar to canonical functional networks. Functional connectivity was

computed as the fisher-transformed pearson’s correlation coefficient between ROI time series.

Univariate Analyses

Univariate functional connectivity differences on drug were detected with 2-way drug x

load repeated measure ANOVAs and paired-t tests (drug vs. placebo) conducted in MATLAB.

Analyses were conducted separately for each functional connection from all 333 ROIs to every

other ROI.

Page 60: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

54

Support Vector Machine Analyses

Linear support vector machine analyses were conducted in MATLAB. We trained and

validated SVM models using leave one out cross validation; for each person, a model was trained

excluding their drug and placebo data, and the trained model was then used to predict their drug

vs. placebo state from connectivity strengths. Before SVM analyses, each feature (i.e. functional

connection) was standardized by subtracting the mean and dividing by the standard deviation

across visits and participants. During LOOCV, to avoid biasing the training model, features were

standardized only on the training data. The same standardizations were then applied to withheld

data (i.e. again using the means and standard deviations only from the training data). When

features were restricted to connections based on univariate paired t-tests, these tests were

conducted only on the training sample to avoid biasing the training model.

The statistical significance of SVM prediction accuracies was assessed through

permutation testing, wherein a null distribution was created by computing LOOCV accuracy

1,000 times with drug and placebo labels randomly switched. A null distribution of the accuracy

differences between rest and working memory conditions was created by computing LOOCV

accuracy 1,000 times with rest and working memory labels randomly switched. Post-hoc

analyses of SVM feature weights were conducted by randomly permuting drug and placebo

labels 15,000 times, and obtaining SVM feature weights for each iteration. The true feature

weight of each connection was then compared against its own null distribution to compute a p-

value.

Page 61: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

55

Graph Theory Analyses

Networks were detected and modularity values were computed using the Louvain algorithm

implemented with the Brain Connectivity Toolbox (Rubinov & Sporns, 2010). This algorithm

was applied to full weighted functional connectivity matrices, with asymmetric weighting of

positive and negative connections (as suggested by Sporns et al. 2011). For weighted and signed

FC networks, modularity can be calculated as:

𝑄∗ = 𝑄+ + 𝑣−

𝑣+ + 𝑣− 𝑄− =

1

𝑣+∑(𝑤𝑖𝑗

+ − 𝑒𝑖𝑗+)𝛿𝑀𝑖𝑀𝑗

𝑖𝑗

1

𝑣+ + 𝑣−∑(𝑤𝑖𝑗

− − 𝑒𝑖𝑗−)𝛿𝑀𝑖𝑀𝑗

𝑖𝑗

Where wij is the weight of connection ij, eij is defined as 𝑠𝑖

±𝑠𝑗±

𝑣± , and 𝛿𝑀𝑖𝑀𝑗= 1 when i and j are in

the same module and 𝛿𝑀𝑖𝑀𝑗= 0 otherwise. A default resolution parameter of 1 was used. As

described in the results section, we used a hierarchical consensus clustering approach to detect

group level networks, and compare network structure between visits. We first applied network

detection 1,000 times within each participant, which produced a network partition for each

participant across all conditions (i.e. drug or placebo visit and rest or working memory task) The

same network detection algorithm was then run on a matrix indicating how often each pair of

nodes was classified in the same network, across the 1,000 times. This provided stable within-

subject network partitions. We then conducted the consensus clustering approach again at the

group level. The network detection algorithm was run on a matrix indicating how often each pair

of nodes was classified in the same network across participants. This yielded one overall set of

functional networks that was consistent across participants and visits (Figure 4A). Network

partitions were compared with the normalized variation of information (VIn), an information

Page 62: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

56

theory metric which quantifies how much information is lost and gained when comparing two

network partitions (Meilă, 2007).

Page 63: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

57

CHAPTER IV: GENERAL CONCLUSION

Conclusion

The study of large-scale brain networks has permitted a deeper understanding of the

neural basis of executive control. A more complete model of network function, however, must

not only link network structure to executive control, but also elucidate the mechanisms

regulating this network structure. The studies presented in this dissertation provide a framework

for understanding one potentially significant regulator of large-scale networks – the LC-NE

system. Study I indicated that, in the resting state, many networks that are important to executive

control couple with the LC-NE system in a behaviorally relevant manner. Hence, networks

contributing to executive control likely do not act in isolation, but coordinate with lower brain-

stem neuromodulatory systems. Furthermore, work in animal models suggests that this

coordination may influence the level of NE signaling, and thereby shape the overall state of

cortical networks (Sara & Hervé-Minvielle, 1995; Sara & Bouret, 2012). Indeed, study II

indicated that NE alpha-2a receptor signaling influenced the overall topology of brain networks

during executive control. Taken together, these studies can be thought of as investigating

different portions of a broader executive circuit; with study I examining the baseline

communication between executive networks and the LC-NE system, and study II examining how

the LC-NE system in turn regulates the communication within these networks. This unified

framework maps well onto anatomical and electrophysiological studies examining LC-NE

function, which suggest that the LC integrates disparate synaptic input from across the cortex,

and broadcasts back out to the entire cortex, regulating overall brain-state (Sara & Bouret, 2012;

Schwarz et al., 2015).The present findings help contextualize past work, and have important

Page 64: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

58

implications for future work. Most directly, Studies I and II have relevance for understanding

how NE acts on brain networks to facilitate executive control. They also have implications for

studies investigating disorders where executive control is impaired. Finally, they have

methodological implications for how psychiatric medications can be better understood using

neuroimaging. Each of these topics is considered below.

Functional networks important for executive control

Despite the advances presented in studies I and II, many questions remain regarding how

NE and other neurmodulators regulate the functional networks underlying executive control.

Firstly, it is an ongoing debate if NE regulates non-specific network and behavioral parameters

such as arousal (Samuels & Szabadi, 2008; Sara & Bouret, 2012), or if NE’s behavioral and

network effects are far more specific (Yu & Dayan, 2002; Clayton et al., 2004). Models of non-

specific NE action point to the fact that—in response to environmental cues—the LC is often

activated in parallel with basic physiological systems, such as the autonomic nervous system.

This is thought to facilitate mental and physical arousal in preparation for an adaptive behavioral

response (Sara & Bouret, 2012). By this view, NE mediates cognition through its influence on

general physiological and behavioral state (i.e. arousal, effort, etc.).

Study II indicated that NE’s network effects may vary depending on the general

behavioral context (i.e. rest vs task). This behavioral distinction is was too uncontrolled,

however, to determine if NE’s network effects vary by general or specific behavioral parameters.

Future studies should also attempt to assess the specificity of NE’s network effects by separately

manipulating two factors: tonic arousal and behavioral domain (i.e. perceptual, executive, etc.).

Previously studies have separately manipulated these factors (Sadaghiani & D’Esposito, 2014),

Page 65: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

59

and a similar approach can be taken while delivering an NE agonist, antagonist or reuptake

inhibitor. If NE’s network effects are similar across different behavioral domains when arousal

levels are comparable, this will validate non-specific accounts of NE’s network effects.

Conversely, if NE’s network effects differ by both behavioral domain and arousal, this will

suggest that NE’s network effects may be more complex than accounted for by models of arousal

alone.

In addition to assessing NE’s network impacts on executive networks, it is also important

to understand how the cortex regulates evoked LC response during executive tasks. A common

conceptual framework in network studies is that the baseline/intrinsic state of a network

constrains its evoked activity (Luczak et al., 2009; Deco et al., 2011). It is interesting to consider

if an analogous concept is also true of the reciprocal networks formed between the cortex and the

LC (Jodo et al., 1998; Schwarz et al., 2008). Study I found that the strength of baseline network

coupling between the LC-NE system (as indexed by pupil diameter) and cingulo-operuclar

regions related to trait-level inattentiveness. Given that optimal attention in everyday life is

thought to rely on appropriate recruitment of the LC, it is conceivable that stronger intrinsic

network coupling between the LC and the cingulo-opercular network may enable adaptive

evoked LC responses. By this view, the intrinsic fidelity of cortical regulatory networks which

maintain resting and evoked LC tone (Jodo et al., 1998) may constrain the ability of the LC to

optimally react to attentionally salient stimuli. We did not directly test this assertion, however, so

future studies should investigate this hypothesis. One way to test this experimentally would be to

assess if individuals with stronger resting-state pupil-brain coordination also demonstrate more

adaptive task-evoked pupillary responses.

Page 66: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

60

Finally, creative multi-modal investigations are needed to delineate other

neuromodulatory impacts on executive networks. NE is only one of numerous neurodmolators

that likely regulate these networks. Furthermore, neuromodulators like NE and acetylcholine can

have highly divergent cortical effects. For instance, acetylcholine shifts neurons into tonic firing

model, while NE suppresses spontaneous firing (Castro-Alamancos & Gulati, 2014). These

divergent neural effects make it important to explore the shared and unique network impacts of

different neuromodulatory systems. Pharmaco-neuromiaging studies such as study II are a

promising avenue to explore this question. Other study designs are also promising. Simultaneous

PET/fMRI studies (Schlemmer et al., 2008; Wehrl et al., 2013) can map endogenous activity at

norepinephrine, dopamine, serotonin or acetylcholine receptors (Wernick & Aarsvold, 2004),

and link it to network structure (Roffman et al., 2016), both at rest and during executive tasks.

Magnetic resonance spectroscopy (MRS) is another promising approach. Although many

neuromodulators cannot be resolved with MRS, cholinergic regulation of the networks

underlying executive control can be studied by measuring localized changes in acetylcholine

with multi-voxel MRS (Mountford et al., 2010) during executive tasks. Network structure can

then be measured with fMRI within the same individual, albeit not simultaneous to MRS

acquisition. While more stable neural features, such as fixed anatomical connections, are an

important basis for functional networks, neuromodulation is also likely an important contributor.

Studies such as those described above are therefore critical for a true mechanistic understanding

of human large-scale brain networks.

Page 67: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

61

Functional connectivity dynamics

In addition to studying static network structure, many recent investigations have focused

on the dynamics of human brain networks. Dynamic adjustments on a core intrinsic network

structure are thought to facilitate adaptive cognition (Cole et al., 2014). The sources of these

dynamics, however, remain unclear. While study II did not directly examine brain network

dynamics, it did reveal that potentiated signaling at NE alpha-2a receptors can modify the overall

architecture of functional brain networks. This finding – taken together with evidence of the

varying levels of LC-NE signaling that occur across general behavioral state (Arnsten, 2009), as

well as in rapid phasic alignment with behaviorally relevant stimuli (Aston-Jones & Cohen,

2005) – suggest that endogenously fluctuating levels of NE signaling continually modify

network structure. For instance, during focused attention or executive control, moderate NE

levels act at high-affinity alpha2-a receptors, which study II indicates drives the brain into a more

modular network configuration. During stress, however, another study found that higher NE

signaling at lower-affinity beta receptors prompts FC changes in the salience network (Hermans

et al., 2011). Hence, the LC-NE system may make context-specific adjustments to the network

structure of the human brain based on general behavioral context and its level of signaling. This

concept aligns with network reset theory of NE function, which posits that NE acts as a rapid

“reset” signaling that reorganizes functional networks to meet the behavioral context at hand

(Bouret & Sara, 2005).

The network reset framework holds important implications for dynamic network studies:

it suggests that NE may facilitate the network dynamics accompanying NE-dependent behavioral

states, including arousal, mood, executive control, and stress. For instance, a recent study found

Page 68: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

62

that dynamic changes in brain network modularity across a stimulus detection task predicted if

participants would correctly complete a trial (Sadaghiani et al., 2015). The mechanisms

regulating ongoing shifts in network modularity, however, were unknown. Study II’s finding that

NE alpha-2a receptors increase modularity, taken together with importance of NE to basic

stimulus detection, (Ikeda et al., 2015), suggests it is conceivable that NE contributes to ongoing

network modularity adjustments. While the results of study II argue in favor of this point, they

cannot directly validate it. Future studies can directly test this assertion by delivering antagonists

rather than agonists, and determining if network dynamics are decreased. This would indicate

that NE signaling at the blocked receptors’ is necessary for network adjustments. NE alpha-2a

antagonists such as yohimbine, however, have much less selective receptor action than

guanfacine (Millan et al., 2000). This type of investigation may thus rely on the future

development of more specific NE alpha-2a antagonists, and thorough safety and proof of concept

testing in animal models. More selective antagonists are available for other NE receptors (Lowe

et al., 2002; Ladage et al., 2013), however, enabling more near-term investigations of their

network effects.

Animal models of noradrenergic modulation of executive control

Animal models of NE’s contribution to executive control have focused on its action on

post-synaptic alpha-2a receptors in the PFC. More specifically, they have focused on NE’s effect

on the sustained delay-related firing of dlPFC networks during working memory (Arnsten,

2011). Several lines of evidence suggest that these models can be expanded. Firstly, NE-

dependent executive control functions, such as working memory, require precise synchronization

and processing across distributed networks beyond the PFC (Palva et al., 2010; Salazar et al.,

Page 69: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

63

2012; Vatansever et al., 2015). Secondly, study II demonstrates that NE also modulates

distributed brain networks during executive control. Thirdly, recent evidence suggests that

discrete oscillatory dynamics, rather than sustained neural activity, underlies working memory

(Lundqvist et al., 2016). Together, this evidence indicates that future animal models of NE’s

contribution to working memory may profit from multi-electrode array recordings in an effort to

delineate NE’s contribution to oscillatory dynamics in local PFC networks, and long-range

neural synchronization across distributed cortical regions.

Attention Deficit Hyperactivity Disorder (ADHD)

While neither study I nor II directly investigated ADHD—a psychiatric disorder defined

by inattentiveness, hyperactivity, and impulsivity—both have important implications for future

studies of the disorder. It is posited that an inability to maintain appropriate levels of arousal may

be an important contributor to the inattentiveness observed in ADHD (Barkley, 1997; Becker,

2013). Study I suggests an interesting testable hypothesis regarding the source of these arousal

deficits: they may result from altered coordination between the cingulo-opercular network and

the LC-NE system. Study I found that inattentiveness, a defining symptom of ADHD, relates to

the coupling between the LC-NE system (as indexed by pupil diameter) and the cingulo-

opercular network. Interestingly, both the cingulo-opercular network and the LC-NE system are

thought to modulate arousal. Hence, optimal attention might rely on appropriate coordination

between the brain networks and neuromodulatory systems underlying arousal. Given evidence of

both cingulate (Bledsoe et al., 2013) and arousal (Becker, 2013) dysfunction in ADHD, is

conceivable that coupling between the cingulo-opercular network and the LC may be

compromised. This hypothesis can be tested by determining if pupil-brain coupling is lessened in

Page 70: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

64

ADHD populations. It would also be interesting to test if pupil-brain coupling is normalized by

pharmacological interventions that aide executive functions, such as with stimulant and non-

stimulant medications (including guanfacine).

Study II indicated that guanfacine, a medication commonly prescribed for ADHD,

induces widespread neural changes. This has important implications for future studies of the

pharmacology of ADHD, as well as the disorder itself. Based on animal work showing

guanfacine’s effect on the sustained delay-related firing of dlPFC networks, the dlPFC is often

assumed to be the locus of guanfacine’s therapeutic action (Arnsten & Jin, 2012). Several lines

of evidence suggest that the drug may also have important actions outside the PFC. Firstly, study

II revealed widespread action of the drug, but could not definitively ascribe behavioral relevance

to all of these changes. In another recent pharmaco-fMRI study in ADHD participants, however,

the neural changes most predictive of guanfacine’s clinical effects were outside of the dlPFC, in

areas like the mid and posterior cingulate cortices (Bédard et al., 2015). Guanfacine also has non-

cognitive benefits, whose neural sources are unknown, including anxiety reduction and the

amelioration of stress induced drug cravings (Fox & Sinha, 2014). Together, this evidence

suggests that guanfacine may have more distributed neural and behavioral effects than current

animal models would suggest.

In addition to an evolving understanding of the neural locus of guanfacine’s action, a

less-PFC centric model of ADHD itself has also recently been proposed (Castellanos & Proal,

2012). This model suggests that many networks may be implicated in ADHD; including the

fronto-parietal, dorsal attention, motor, and default networks. It is thought that ADHD

pathophysiology, rather than originating solely in the PFC, may stem from aberrant interplay

Page 71: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

65

between default, cognitive control and limbic networks. Given that study II showed that

guanfacine alters the interplay between these networks, this further underscores the importance

of understanding if the drug’s clinical efficacy results from wide-spread or local action. Future

studies in ADHD populations should attempt to link overall network measures which index

large-scale network interaction, such as modularity, to drug-induced behavioral improvements.

Psychiatric translational studies

Large-scale network measurement as a tool in the broader push for personalized medicine

("Precision Medicine Initiative", 2016) is relatively new, but holds great promise. Recently,

baseline brain network architecture has been used to predict drug response in disorders such as

schizophrenia (Sarpal et al., 2016) and social anxiety disorder (Whitfield-Gabrieli et al., 2016). A

new framework also suggests the potential utility of predicting long-term clinical efficacy from

how brain networks reorganize in response to acute drug dosing. Network measures were

recently used to detect whole-brain and immediate effects of selective serotonin reuptake

inhibitors (Schaefer et al., 2014). These acute network changes may be useful in predicting long-

term clinical efficacy for depression, which can often take weeks or longer to manifest (Sugrue,

1983; Harmer et al., 2009). It remains to be seen if acute network changes are related to eventual

clinical efficacy, but this is an intriguing prospect for future studies to investigate.

While promising, many of these studies utilized standard univariate techniques, which

study II suggests may be insufficient to capture the entirety of drug related network change.

Furthermore, the ultimate translational aim of these studies is to provide effective predictions of

which medication, or combinations of medications, a patient will best respond to. It is likely that

this problem will require complex modeling with high sensitivity. Thus, machine learning and

Page 72: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

66

graph theory measures, which can capture meaningful and distributed patterns of network change

with high sensitivity, will likely be useful methodological additions to translational psychiatric

neuroimaging studies.

Conclusions

The discovery of large-scale functional networks has enabled a better understanding of

the circuits underlying processes as diverse as stress, attention, and memory. It has also enabled a

re-conceptualization of many psychiatric disorders as fundamentally disorders of neural

connectivity (Di Martino et al., 2014). A more complete model of both network function and

dysfunction, however, requires a deeper understanding of the neurobiological mechanisms

regulating network structure. The studies presented in this thesis suggest NE as one potentially

significant regulator of large-scale networks. As future studies continue to delineate the

neurobiological mechanisms controlling network structure, this will enable a deeper appreciation

for the brain bases of cognition, as well as expand the translational potential of brain network

science.

Page 73: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

67

APPENDIX

Supplementary Materials from Chapter III

Supplementary Figure 2.1. Guanfacine Did Not Change Blood Pressure or Heart Rate

Blood pressure was measured 30 minutes pre-dosing, 45 minutes post-dosing, and after fMRI

scanning (mean time post-dosing = 150 minutes). Dashed lines represent dosing time.

Page 74: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

68

REFERENCES

Allen, E. A., Damaraju, E., Plis, S. M., Erhardt, E. B., Eichele, T., & Calhoun, V. D. (2014).

Tracking whole-brain connectivity dynamics in the resting state. Cerebral Cortex, 24(3),

663-676.

Andreassi, J. L. (2000). Pupillary response and behavior. In Psychophysiology: Human

Behavior & Physiological Response. (pp. 218–233). Mahwah, N.J: Lawrence Erlbaum

Assoc.

Aoki, C., Venkatesan, C., Go, C. G., Forman, R., & Kurose, H. (1998). Cellular and subcellular

sites for noradrenergic action in the monkey dorsolateral prefrontal cortex as revealed by

the immunocytochemical localization of noradrenergic receptors and axons. Cerebral

Cortex, 8(3), 269–277.

Arnsten, A. F., Cai, J. X., & Goldman-Rakic, P. S. (1988). The alpha-2 adrenergic agonist

guanfacine improves memory in aged monkeys without sedative or hypotensive side

effects: evidence for alpha-2 receptor subtypes. The Journal of Neuroscience, 8(11),

4287–4298.

Arnsten, A. F., & Contant, T. A. (1992). Alpha-2 adrenergic agonists decrease distractibility in

aged monkeys performing the delayed response task. Psychopharmacology, 108(1–2),

159–169.

Arnsten, A. F. T. (1997). Catecholamine regulation of the prefrontal cortex. Journal of

Psychopharmacology, 11(2), 151–162.

Arnsten, A. F. T. (2009). Stress signaling pathways that impair prefrontal cortex structure and

function. Nature Reviews. Neuroscience, 10(6), 410–422.

Page 75: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

69

Arnsten, A. F. T. (2011). Catecholamine influences on dorsolateral prefrontal cortical networks.

Biological Psychiatry, 69(12), e89–e99.

Arnsten, A. F. T., & Jin, L. E. (2012). Guanfacine for the treatment of cognitive disorders: A

century of discoveries at Yale. The Yale Journal of Biology and Medicine, 85(1), 45–58.

Arnsten, A. F. T., Paspalas, C. D., Gamo, N. J., Yang, Y., & Wang, M. (2010). Dynamic

Network connectivity: A new form of neuroplasticity. Trends in Cognitive Sciences,

14(8), 365–375.

Aston-Jones, G., & Cohen, J. D. (2005). An integrative theory of locus coeruleus-norepinephrine

function: adaptive gain and optimal performance. Annual Review of Neuroscience, 28,

403–450.

Baars, B. J. (2002). The conscious access hypothesis: origins and recent evidence. Trends in

Cognitive Sciences, 6(1), 47–52.

Bargmann, C. I., & Marder, E. (2013). From the connectome to brain function. Nature Methods,

10(6), 483–490.

Barkley, R. A. (1997). Behavioral inhibition, sustained attention, and executive functions:

constructing a unifying theory of ADHD. Psychological Bulletin, 121(1), 65–94.

Bassett, D. S., Wymbs, N. F., Porter, M. A., Mucha, P. J., Carlson, J. M., & Grafton, S. T.

(2011). Dynamic reconfiguration of human brain networks during learning. Proceedings

of the National Academy of Sciences, 108(18), 7641–7646.

Beauchaine, T. (2001). Vagal tone, development, and Gray’s motivational theory: toward an

integrated model of autonomic nervous system functioning in psychopathology.

Development and Psychopathology, 13(2), 183–214.

Page 76: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

70

Becker, S. P. (2013). Topical review: sluggish cognitive tempo: research findings and relevance

for pediatric psychology. Journal of Pediatric Psychology, 38(10), 1051–1057.

Beckmann, C. F., DeLuca, M., Devlin, J. T., & Smith, S. M. (2005). Investigations into resting-

state connectivity using independent component analysis. Philosophical Transactions of

the Royal Society of London. Series B, Biological Sciences, 360(1457), 1001–1013.

Bédard, A.-C. V., Schulz, K. P., Krone, B., Pedraza, J., Duhoux, S., Halperin, J. M., & Newcorn,

J. H. (2015). Neural mechanisms underlying the therapeutic actions of guanfacine

treatment in youth with ADHD: a pilot fMRI study. Psychiatry Research, 231(3), 353–

Beissner, F., Meissner, K., Bär, K.-J., & Napadow, V. (2013). The autonomic brain: An

activation likelihood estimation meta-analysis for central processing of autonomic

function. The Journal of Neuroscience, 33(25), 10503–10511.

Bledsoe, J. C., Semrud-Clikeman, M., & Pliszka, S. R. (2013). Anterior cingulate cortex and

symptom severity in attention-deficit/hyperactivity disorder. Journal of Abnormal

Psychology, 122(2), 558–565.

Blondel, V. D., Guillaume, J.-L., Lambiotte, R., & Lefebvre, E. (2008). Fast unfolding of

communities in large networks. Journal of Statistical Mechanics: Theory and Experiment,

2008(10), P10008.

Bouret, S., & Sara, S. J. (2005). Network reset: a simplified overarching theory of locus

coeruleus noradrenaline function. Trends in Neurosciences, 28(11), 574–582.

Braun, U., Schäfer, A., Walter, H., Erk, S., Romanczuk-Seiferth, N., Haddad, L., Bassett, D. S.

(2015). Dynamic reconfiguration of frontal brain networks during executive cognition in

humans. Proceedings of the National Academy of Sciences, 112(37), 11678–11683.

Page 77: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

71

Brozoski, T. J., Brown, R. M., Rosvold, H. E., & Goldman, P. S. (1979). Cognitive deficit

caused by regional depletion of dopamine in prefrontal cortex of rhesus monkey. Science,

205(4409), 929–932.

Buckner, R. L., Andrews-Hanna, J. R., & Schacter, D. L. (2008). The brain’s default network:

anatomy, function, and relevance to disease. Annals of the New York Academy of

Sciences, 1124, 1–38.

Butters, N., & Pandya, D. (1969). Retention of delayed-alternation: effect of selective lesions of

sulcus principalis. Science, 165(3899), 1271–1273.

Carbonell, F., Nagano-Saito, A., Leyton, M., Cisek, P., Benkelfat, C., He, Y., & Dagher, A.

(2014). Dopamine precursor depletion impairs structure and efficiency of resting state

brain functional networks. Neuropharmacology, 84, 90–100.

Casey, B. J. (2002). Neuroscience. Windows into the human brain. Science, 296(5572), 1408–

1409.

Castellanos, F. X., Margulies, D. S., Kelly, C., Uddin, L. Q., Ghaffari, M., Kirsch, A., Milham,

M. P. (2008). Cingulate-precuneus interactions: a new locus of dysfunction in adult

attention-deficit/hyperactivity disorder. Biological Psychiatry, 63(3), 332–337.

Castellanos, F. X., & Proal, E. (2012). Large-scale brain systems in ADHD: Beyond the

prefrontal-striatal model. Trends in Cognitive Sciences, 16(1), 17–26.

Castro-Alamancos, M. A., & Gulati, T. (2014). Neuromodulators produce distinct activated

states in neocortex. The Journal of Neuroscience, 34(37), 12353–12367.

Page 78: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

72

Cechetto, D., & Saper, C. (1990). Role of the cerebral cortex in autonomic function. In A. Loewy

& K. Spyer (Eds.), Central regulation of autonomic functions (pp. 208–223). New York:

Oxford University Press.

Chang, C., & Glover, G. H. (2010). Time-frequency dynamics of resting-state brain connectivity

measured with fMRI. NeuroImage, 50(1), 81–98.

Chen, S., Ross, T. J., Zhan, W., Myers, C. S., Chuang, K.-S., Heishman, S. J., Yang, Y. (2008).

Group Independent Component Analysis Reveals Consistent Resting-State Networks

across Multiple Sessions. Brain Research, 1239, 141–151.

Churchland, P. S., & Sejnowski, T. J. (1988). Perspectives on cognitive neuroscience. Science,

242(4879), 741–745.

Clayton, E. C., Rajkowski, J., Cohen, J. D., & Aston-Jones, G. (2004). Phasic activation of

monkey locus ceruleus neurons by simple decisions in a forced-choice task. The Journal

of Neuroscience, 24(44), 9914–9920.

Clerkin, S. M., Schulz, K. P., Halperin, J. M., Newcorn, J. H., Ivanov, I., Tang, C. Y., & Fan, J.

(2009). Guanfacine potentiates the activation of prefrontal cortex evoked by warning

signals. Biological Psychiatry, 66(4), 307–312.

Cole, M. W., Bassett, D. S., Power, J. D., Braver, T. S., & Petersen, S. E. (2014). Intrinsic and

task-evoked network architectures of the human brain. Neuron, 83(1), 238–251.

Cole, M. W., Reynolds, J. R., Power, J. D., Repovs, G., Anticevic, A., & Braver, T. S. (2013).

Multi-task connectivity reveals flexible hubs for adaptive task control. Nature

Neuroscience, 16(9), 1348–1355.

Page 79: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

73

Cole, M. W., Yarkoni, T., Repovš, G., Anticevic, A., & Braver, T. S. (2012). Global connectivity

of prefrontal cortex predicts cognitive control and intelligence. The Journal of

Neuroscience, 32(26), 8988–8999.

Corbetta, M., Patel, G., & Shulman, G. L. (2008). The reorienting system of the human brain:

from environment to theory of mind. Neuron, 58(3), 306–324.

Coull, J. T., Büchel, C., Friston, K. J., & Frith, C. D. (1999). Noradrenergically mediated

plasticity in a human attentional neuronal network. NeuroImage, 10(6), 705–715.

Critchley, H. D., Nagai, Y., Gray, M. A., & Mathias, C. J. (2011). Dissecting axes of autonomic

control in humans: Insights from neuroimaging. Autonomic Neuroscience, 161(1–2), 34–

42.

Dagli, M. S., Ingeholm, J. E., & Haxby, J. V. (1999). Localization of cardiac-induced signal

change in fMRI. NeuroImage, 9(4), 407–415.

De Luca, M., Beckmann, C. F., De Stefano, N., Matthews, P. M., & Smith, S. M. (2006). fMRI

resting state networks define distinct modes of long-distance interactions in the human

brain. NeuroImage, 29(4), 1359–1367.

Deco, G., Jirsa, V. K., & McIntosh, A. R. (2011). Emerging concepts for the dynamical

organization of resting-state activity in the brain. Nature Reviews. Neuroscience, 12(1),

43–56.

Di Martino, A., Fair, D. A., Kelly, C., Satterthwaite, T. D., Castellanos, F. X., Thomason, M. E.,

Milham, M. P. (2014). Unraveling the miswired connectome: A developmental

perspective. Neuron, 83(6), 1335–1353.

Page 80: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

74

Diaz, B. A., Van Der Sluis, S., Moens, S., Benjamins, J. S., Migliorati, F., Stoffers, D.,

Linkenkaer-Hansen, K. (2013). The Amsterdam resting-state questionnaire reveals

multiple phenotypes of resting-state cognition. Frontiers in Human Neuroscience, 7.

Dosenbach, N. U. F., Fair, D. A., Cohen, A. L., Schlaggar, B. L., & Petersen, S. E. (2008). A

dual-networks architecture of top-down control. Trends in Cognitive Sciences, 12(3), 99–

105.

Dosenbach, N. U. F., Fair, D. A., Miezin, F. M., Cohen, A. L., Wenger, K. K., Dosenbach, R. A.

T., Petersen, S. E. (2007). Distinct brain networks for adaptive and stable task control in

humans. Proceedings of the National Academy of Sciences, 104(26), 11073–11078.

Dwyer, D. B., Harrison, B. J., Yücel, M., Whittle, S., Zalesky, A., Pantelis, C., Fornito, A.

(2014). Large-scale brain network dynamics supporting adolescent cognitive control. The

Journal of Neuroscience, 34(42), 14096–14107.

Eilam-Stock, T., Xu, P., Cao, M., Gu, X., Van Dam, N. T., Anagnostou, E., Fan, J. (2014).

Abnormal autonomic and associated brain activities during rest in autism spectrum

disorder. Brain, 137(Pt 1), 153–171.

Eldar, E., Cohen, J. D., & Niv, Y. (2013). The effects of neural gain on attention and learning.

Nature Neuroscience, 16(8), 1146–1153.

Engberg, G., & Eriksson, E. (1991). Effects of α2-adrenoceptor agonists on locus coeruleus

firing rate and brain noradrenaline turnover in N-ethoxycarbonyl-2-ethoxy-1,2-

dihydroquinoline (EEDQ)-treated rats. Naunyn-Schmiedeberg’s Archives of

Pharmacology, 343(5), 472–477.

Page 81: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

75

Enge, S., Fleischhauer, M., Lesch, K.-P., Reif, A., & Strobel, A. (2014). Variation in key genes

of serotonin and norepinephrine function predicts gamma-band activity during goal-

directed attention. Cerebral Cortex, 24(5), 1195-1205.

Fan, J., Xu, P., Van Dam, N. T., Eilam-Stock, T., Gu, X., Luo, Y., & Hof, P. R. (2012).

Spontaneous brain activity relates to autonomic arousal. The Journal of Neuroscience,

32(33), 11176–11186.

Felleman, D. J., & Essen, D. C. V. (1991). Distributed hierarchical processing in the primate

cerebral cortex. Cerebral Cortex, 1(1), 1–47.

Fox, H., & Sinha, R. (2014). The role of guanfacine as a therapeutic agent to address stress-

related pathophysiology in cocaine-dependent individuals. Advances in Pharmacology,

69, 217–265.

Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Essen, D. C. V., & Raichle, M. E.

(2005a). The human brain is intrinsically organized into dynamic, anticorrelated

functional networks. Proceedings of the National Academy of Sciences, 102(27), 9673–

9678.

Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Essen, D. C. V., & Raichle, M. E.

(2005b). The human brain is intrinsically organized into dynamic, anticorrelated

functional networks. Proceedings of the National Academy of Sciences, 102(27), 9673–

9678.

Frazer, A., & Hensler, J. G. (1999). Serotonin. In Basic neurochemistry: Molecular, cellular and

medical aspects. Baltimore, Maryland: Lippincott, Williams and & Wilkins.

Page 82: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

76

Fries, P. (2005). A mechanism for cognitive dynamics: neuronal communication through

neuronal coherence. Trends in Cognitive Sciences, 9(10), 474–480.

Funahashi, S., Bruce, C. J., & Goldman-Rakic, P. S. (1989). Mnemonic coding of visual space in

the monkey’s dorsolateral prefrontal cortex. Journal of Neurophysiology, 61(2), 331–349.

Gamo, N. J., Wang, M., & Arnsten, A. F. T. (2010). Methylphenidate and atomoxetine enhance

prefrontal function through α2-adrenergic and dopamine D1 receptors. Journal of the

American Academy of Child and Adolescent Psychiatry, 49(10), 1011–1023.

Geva, R., Zivan, M., Warsha, A., & Olchik, D. (2013). Alerting, orienting or executive attention

networks: differential patters of pupil dilations. Frontiers in Behavioral Neuroscience, 7,

145.

Giessing, C., & Thiel, C. M. (2012). Pro-cognitive drug effects modulate functional brain

network organization. Frontiers in Behavioral Neuroscience, 6, 53.

Giessing, C., Thiel, C. M., Alexander-Bloch, A. F., Patel, A. X., & Bullmore, E. T. (2013).

Human brain functional network changes associated with enhanced and impaired

attentional task performance. The Journal of Neuroscience, 33(14), 5903–5914.

Godwin, D., Barry, R. L., & Marois, R. (2015). Breakdown of the brain’s functional network

modularity with awareness. Proceedings of the National Academy of Sciences, 112(12),

3799–3804.

Goldman-Rakic, P. (1995). Cellular basis of working memory. Neuron, 14(3), 477–485.

Gonzalez-Castillo, J., Hoy, C. W., Handwerker, D. A., Robinson, M. E., Buchanan, L. C., Saad,

Z. S., & Bandettini, P. A. (2015). Tracking ongoing cognition in individuals using brief,

Page 83: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

77

whole-brain functional connectivity patterns. Proceedings of the National Academy of

Sciences, 112(28), 8762–8767.

Gordon, E. M., Breeden, A. L., Bean, S. E., & Vaidya, C. J. (2014). Working memory-related

changes in functional connectivity persist beyond task disengagement. Human Brain

Mapping, 35(3), 1004-1017.

Gordon, E. M., Laumann, T. O., Adeyemo, B., Huckins, J. F., Kelley, W. M., & Petersen, S. E.

(2016). Generation and evaluation of a cortical area parcellation from resting-state

correlations. Cerebral Cortex, 26(1), 288–303.

Gordon, E. M., Stollstorff, M., & Vaidya, C. J. (2012). Using spatial multiple regression to

identify intrinsic connectivity networks involved in working memory performance.

Human Brain Mapping, 33(7), 1536–1552.

Grandchamp, R., Braboszcz, C., & Delorme, A. (2014). Oculometric variations during mind

wandering. Frontiers in Psychology, 5, 31.

Gratton, C., Lee, T. G., Nomura, E. M., & D’Esposito, M. (2013). The effect of theta-burst TMS

on cognitive control networks measured with resting state fMRI. Frontiers in Systems

Neuroscience, 7.

Harmelech, T., Preminger, S., Wertman, E., & Malach, R. (2013). The day-after effect: Long

term, hebbian-like restructuring of resting-state fMRI patterns induced by a single epoch

of cortical activation. The Journal of Neuroscience, 33(22), 9488–9497.

Harmer, C. J., Goodwin, G. M., & Cowen, P. J. (2009). Why do antidepressants take so long to

work? A cognitive neuropsychological model of antidepressant drug action. The British

Journal of Psychiatry, 195(2), 102–108.

Page 84: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

78

Haufe, S., Meinecke, F., Görgen, K., Dähne, S., Haynes, J.-D., Blankertz, B., & Bießmann, F.

(2014). On the interpretation of weight vectors of linear models in multivariate

neuroimaging. NeuroImage, 87, 96–110.Hermans, E. J., Marle, H. J. F. van, Ossewaarde,

L., Henckens, M. J. A. G., Qin, S., Kesteren, M. T. R. van, Fernández, G. (2011). Stress-

related noradrenergic activity prompts large-scale neural network reconfiguration.

Science, 334(6059), 1151–1153.

Hnatko, G. (2002). Is guanfacine useful in the treatment of attention deficit hyperactivity

disorder? Journal of Psychiatry and Neuroscience, 27(2), 152.

Honey, C. J., Kötter, R., Breakspear, M., & Sporns, O. (2007). Network structure of cerebral

cortex shapes functional connectivity on multiple time scales. Proceedings of the

National Academy of Sciences, 104(24), 10240–10245.

Hutchison, R. M., Womelsdorf, T., Allen, E. A., Bandettini, P. A., Calhoun, V. D., Corbetta, M.,

… Chang, C. (2013). Dynamic functional connectivity: promise, issues, and

interpretations. NeuroImage, 80, 360–378.

Ikeda, M. Z., Jeon, S. D., Cowell, R. A., & Remage-Healey, L. (2015). Norepinephrine

modulates coding of complex vocalizations in the songbird auditory cortex independent

of local neuroestrogen synthesis. The Journal of Neuroscience, 35(25), 9356–9368.

Jäkälä, P., Riekkinen, M., Sirviö, J., Koivisto, E., Kejonen, K., Matti, & Riekkinen, P.

(1999). Guanfacine, but not clonidine, improves planning and working memory

performance in humans. Neuropsychopharmacology, 20(5), 460–470.

Jodo, E., Chiang, C., & Aston-Jones, G. (1998). Potent excitatory influence of prefrontal cortex

activity on noradrenergic locus coeruleus neurons. Neuroscience, 83(1), 63–79.

Page 85: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

79

Joshi, S., Li, Y., Kalwani, R. M., & Gold, J. I. (2016). Relationships between pupil diameter and

neuronal activity in the locus coeruleus, colliculi, and cingulate cortex. Neuron, 89(1),

221-234.

Kitzbichler, M. G., Henson, R. N. A., Smith, M. L., Nathan, P. J., & Bullmore, E. T. (2011).

Cognitive effort drives workspace configuration of human brain functional networks. The

Journal of Neuroscience, 31(22), 8259–8270. Kristjansson, S. D., Stern, J. A., Brown, T.

B., & Rohrbaugh, J. W. (2009). Detecting phasic lapses in alertness using pupillometric

measures. Applied Ergonomics, 40(6), 978–986.

Ladage, D., Schwinger, R. H. G., & Brixius, K. (2013). Cardio-selective beta-blocker:

pharmacological evidence and their influence on exercise capacity. Cardiovascular

Therapeutics, 31(2), 76–83.

Lancichinetti, A., & Fortunato, S. (2012). Consensus clustering in complex networks. Scientific

Reports, 2.

Li, B. M., & Mei, Z. T. (1994). Delayed-response deficit induced by local injection of the alpha

2-adrenergic antagonist yohimbine into the dorsolateral prefrontal cortex in young adult

monkeys. Behavioral and Neural Biology, 62(2), 134–139.

Liang, X., Zou, Q., He, Y., & Yang, Y. (2016). Topologically reorganized connectivity

architecture of default-mode, executive-control, and salience networks across working

memory task loads. Cerebral Cortex, 26(4), 1501–1511.

Loewenfeld, I. (1999). The Pupil: Anatomy, Physiology, and Clinical Applications. Oxford:

Butterworth-Heinemann.

Page 86: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

80

Lowe, M. D., Lynham, J. A., Grace, A. A., & Kaumann, A. J. (2002). Comparison of the affinity

of beta-blockers for two states of the beta 1-adrenoceptor in ferret ventricular

myocardium. British Journal of Pharmacology, 135(2), 451–461.

Lowenstein, O. (1955). Pupillary reflex shapes and topical clinical diagnosis. Neurology, 5(9),

631–644.

Luczak, A., Barthó, P., & Harris, K. D. (2009). Spontaneous events outline the realm of possible

sensory responses in neocortical populations. Neuron, 62(3), 413.

Lüdtke, H., Wilhelm, B., Adler, M., Schaeffel, F., & Wilhelm, H. (1998). Mathematical

procedures in data recording and processing of pupillary fatigue waves. Vision Research,

38(19), 2889–2896.

Lundqvist, M., Rose, J., Herman, P., Brincat, S. L., Buschman, T. J., & Miller, E. K. (2016).

Gamma and beta bursts underlie working memory. Neuron, 90(1), 152–164.

Ma, L., Narayana, S., Robin, D. A., Fox, P. T., & Xiong, J. (2011). Changes occur in resting

state network of motor system during 4 weeks of motor skill learning. NeuroImage,

58(1), 226–233.

Marder, E. (2012). Neuromodulation of neuronal circuits: back to the future. Neuron, 76(1), 1–

11.

McAllister, T. W., McDonald, B. C., Flashman, L. A., Ferrell, R. B., Tosteson, T. D., Yanofsky,

N. N., … Saykin, A. J. (2011). Alpha-2 adrenergic challenge with guanfacine one month

after mild traumatic brain injury: altered working memory and BOLD response.

International Journal of Psychophysiology, 82(1), 107–114.

Page 87: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

81

Meilă, M. (2007). Comparing clusterings—an information based distance. Journal of

Multivariate Analysis, 98(5), 873–895.

Menon, V., & Uddin, L. Q. (2010). Saliency, switching, attention and control: a network model

of insula function. Brain Structure & Function, 214(5–6), 655–667.

Milazzo, A.-C., Ng, B., Jiang, H., Shirer, W., Varoquaux, G., Poline, J. B., … Greicius, M. D.

(2016). Identification of mood-relevant brain connections using a continuous, subject-

driven rumination paradigm. Cerebral Cortex, 26(3), 933–942.

Millan, M. J., Newman-Tancredi, A., Audinot, V., Cussac, D., Lejeune, F., Nicolas, J. P., …

Gobert, A. (2000). Agonist and antagonist actions of yohimbine as compared to

fluparoxan at alpha(2)-adrenergic receptors (AR)s, serotonin (5-HT)(1A), 5-HT(1B), 5-

HT(1D) and dopamine D(2) and D(3) receptors. Significance for the modulation of

frontocortical monoaminergic transmission and depressive states. Synapse, 35(2), 79–95.

Mountford, C. E., Stanwell, P., Lin, A., Ramadan, S., & Ross, B. (2010). Neurospectroscopy: the

past, present and future. Chemical Reviews, 110(5), 3060–3086.

Müller, U., Clark, L., Lam, M. L., Moore, R. M., Murphy, C. L., Richmond, N. K., … Robbins,

T. W. (2005). Lack of effects of guanfacine on executive and memory functions in

healthy male volunteers. Psychopharmacology, 182(2), 205–213.

Munakata, Y., Herd, S. A., Chatham, C. H., Depue, B. E., Banich, M. T., & O’Reilly, R. C.

(2011). A unified framework for inhibitory control. Trends in Cognitive Sciences, 15(10),

453–459.

Page 88: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

82

Murphy, P. R., O’Connell, R. G., O’Sullivan, M., Robertson, I. H., & Balsters, J. H. (2014).

Pupil diameter covaries with BOLD activity in human locus coeruleus. Human Brain

Mapping, 35(8), 4140–4154.

Negrao, B. L., Bipath, P., van der Westhuizen, D., & Viljoen, M. (2011). Autonomic correlates

at rest and during evoked attention in children with attention-deficit/hyperactivity

disorder and effects of methylphenidate. Neuropsychobiology, 63(2), 82–91.

Nieuwenhuis, S., De Geus, E. J., & Aston-Jones, G. (2011). The anatomical and functional

relationship between the P3 and autonomic components of the orienting response.

Psychophysiology, 48(2), 162–175.

Owen, A. M. (1997). The functional organization of working memory processes within human

lateral frontal cortex: the contribution of functional neuroimaging. The European Journal

of Neuroscience, 9(7), 1329–1339.

Palva, J. M., Monto, S., Kulashekhar, S., & Palva, S. (2010). Neuronal synchrony reveals

working memory networks and predicts individual memory capacity. Proceedings of the

National Academy of Sciences, 107(16), 7580–7585.

Park, H.-D., & Tallon-Baudry, C. (2014). The neural subjective frame: from bodily signals to

perceptual consciousness. Philosophical Transactions of the Royal Society of London.

Series B, Biological Sciences, 369(1641), 20130208.

Power, J. D., Barnes, K. A., Snyder, A. Z., Schlaggar, B. L., & Petersen, S. E. (2012). Spurious

but systematic correlations in functional connectivity MRI networks arise from subject

motion. NeuroImage, 59(3), 2142–2154.

Page 89: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

83

Power, J. D., Cohen, A. L., Nelson, S. M., Wig, G. S., Barnes, K. A., Church, J. A., … Petersen,

S. E. (2011). Functional network organization of the human brain. Neuron, 72(4), 665–

Precision Medicine Initiative. (2016). https://www.nih.gov/precision-medicine-initiative-cohort-

program

Pugliese, C. E., Anthony, L. G., Strang, J. F., Dudley, K., Wallace, G. L., Naiman, D. Q., &

Kenworthy, L. (2016). Longitudinal examination of adaptive behavior in autism spectrum

disorders: influence of executive function. Journal of Autism and Developmental

Disorders, 46(2), 467–477.

Raffard, S., & Bayard, S. (2012). Understanding the executive functioning heterogeneity in

schizophrenia. Brain and Cognition, 79(1), 60–69.

Roffman, J. L., Tanner, A. S., Eryilmaz, H., Rodriguez-Thompson, A., Silverstein, N. J., Ho, N.

F., … Catana, C. (2016). Dopamine D1 signaling organizes network dynamics underlying

working memory. Science Advances, 2(6), e1501672.

Rosenberg, M. D., Finn, E. S., Scheinost, D., Papademetris, X., Shen, X., Constable, R. T., &

Chun, M. M. (2016). A neuromarker of sustained attention from whole-brain functional

connectivity. Nature Neuroscience, 19(1), 165–171.

Rosenthal, M., Wallace, G. L., Lawson, R., Wills, M. C., Dixon, E., Yerys, B. E., & Kenworthy,

L. (2013). Impairments in real-world executive function increase from childhood to

adolescence in autism spectrum disorders. Neuropsychology, 27(1), 13–18.

Rubinov, M., & Sporns, O. (2010). Complex network measures of brain connectivity: Uses and

interpretations. NeuroImage, 52(3), 1059–1069.

Page 90: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

84

Rubinov, M., & Sporns, O. (2011). Weight-conserving characterization of complex functional

brain networks. NeuroImage, 56(4), 2068–2079.

Sadaghiani, S., & D’Esposito, M. (2015). Functional characterization of the cingulo-opercular

network in the maintenance of tonic alertness. Cerebral Cortex, 25(9), 2763-2773.

Sadaghiani, S., Poline, J.-B., Kleinschmidt, A., & D’Esposito, M. (2015). Ongoing dynamics in

large-scale functional connectivity predict perception. Proceedings of the National

Academy of Sciences, 112(27), 8463-8468.

Salazar, R. F., Dotson, N. M., Bressler, S. L., & Gray, C. M. (2012). Content-specific fronto-

parietal synchronization during visual working memory. Science, 338(6110), 1097–1100.

Sami, S., Robertson, E. M., & Miall, R. C. (2014). The Time course of task-specific memory

consolidation effects in resting state networks. The Journal of Neuroscience, 34(11),

3982–3992.

Samuels, E. R., & Szabadi, E. (2008). Functional Neuroanatomy of the noradrenergic locus

coeruleus: its roles in the regulation of arousal and autonomic function part i: principles

of functional organisation. Current Neuropharmacology, 6(3), 235–253.

Sara, S. J. (2009). The locus coeruleus and noradrenergic modulation of cognition. Nature

Reviews. Neuroscience, 10(3), 211–223.

Sara, S. J., & Bouret, S. (2012). Orienting and reorienting: the locus coeruleus mediates

cognition through arousal. Neuron, 76(1), 130–141.

Sara, S. J., & Hervé-Minvielle, A. (1995). Inhibitory influence of frontal cortex on locus

coeruleus neurons. Proceedings of the National Academy of Sciences of the United States

of America, 92(13), 6032.

Page 91: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

85

Sarpal, D. K., Argyelan, M., Robinson, D. G., Szeszko, P. R., Karlsgodt, K. H., John, M., …

Malhotra, A. K. (2016). Baseline striatal functional connectivity as a predictor of

response to antipsychotic drug treatment. The American Journal of Psychiatry, 173(1),

69–77.

Scahill, L., Chappell, P. B., Kim, Y. S., Schultz, R. T., Katsovich, L., Shepherd, E., … Leckman,

J. F. (2001). A placebo-controlled study of guanfacine in the treatment of children with

tic disorders and attention deficit hyperactivity disorder. The American Journal of

Psychiatry, 158(7), 1067–1074. Schaefer, A., Burmann, I., Regenthal, R., Arélin, K.,

Barth, C., Pampel, A., … Sacher, J. (2014). Serotonergic modulation of intrinsic

functional connectivity. Current Biology, 24(19), 2314–2318.

Scheinin, M., Lomasney, J. W., Hayden-Hixson, D. M., Schambra, U. B., Caron, M. G.,

Lefkowitz, R. J., & Fremeau, R. T. (1994). Distribution of alpha 2-adrenergic receptor

subtype gene expression in rat brain. Brain Research., 21(1–2), 133–149.

Schlemmer, H.-P. W., Pichler, B. J., Schmand, M., Burbar, Z., Michel, C., Ladebeck, R., …

Claussen, C. D. (2008). Simultaneous MR/PET imaging of the human brain: feasibility

study1. Radiology, 248(3), 1028-1035.

Schölvinck, M. L., Maier, A., Ye, F. Q., Duyn, J. H., & Leopold, D. A. (2010). Neural basis of

global resting-state fMRI activity. Proceedings of the National Academy of Sciences of

the United States of America, 107(22), 10238–10243.

Schulz, K. P., Clerkin, S. M., Fan, J., Halperin, J. M., & Newcorn, J. H. (2013). Guanfacine

modulates the influence of emotional cues on prefrontal cortex activation for cognitive

control. Psychopharmacology, 226(2), 261–271.

Page 92: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

86

Schulz, K. P., Clerkin, S. M., Newcorn, J. H., Halperin, J. M., & Fan, J. (2014). Guanfacine

modulates the emotional biasing of amygdala-prefrontal connectivity for cognitive

control. European Neuropsychopharmacology, 24(9), 1444–1453.

Schwarz, A. J., Gozzi, A., & Bifone, A. (2008). Community structure and modularity in

networks of correlated brain activity. Magnetic Resonance Imaging, 26(7), 914–920.

Schwarz, L. A., Miyamichi, K., Gao, X. J., Beier, K. T., Weissbourd, B., DeLoach, K. E., …

Luo, L. (2015). Viral-genetic tracing of the input-output organization of a central

noradrenaline circuit. Nature, 524(7563), 88–92.

Seeley, W. W., Menon, V., Schatzberg, A. F., Keller, J., Glover, G. H., Kenna, H., … Greicius,

M. D. (2007). Dissociable Intrinsic connectivity networks for salience processing and

executive control. The Journal of Neuroscience, 27(9), 2349–2356.

Seli, P., Smallwood, J., Cheyne, J. A., & Smilek, D. (2015). On the relation of mind wandering

and ADHD symptomatology. Psychonomic Bulletin & Review, 22(3), 629–636.

Shanmugan, S., Wolf, D. H., Calkins, M. E., Moore, T. M., Ruparel, K., Hopson, R. D., …

Satterthwaite, T. D. (2016). Common and dissociable mechanisms of executive system

dysfunction across psychiatric disorders in youth. The American Journal of Psychiatry,

173(5), 517–526.

Shen, K., Hutchison, R. M., Bezgin, G., Everling, S., & McIntosh, A. R. (2015). Network

structure shapes spontaneous functional connectivity dynamics. The Journal of

Neuroscience, 35(14), 5579–5588.

Page 93: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

87

Shirer, W. R., Ryali, S., Rykhlevskaia, E., Menon, V., & Greicius, M. D. (2012). Decoding

subject-driven cognitive states with whole-brain connectivity patterns. Cerebral Cortex,

22(1), 158–165.

Siegel, M., Buschman, T. J., & Miller, E. K. (2015). Cortical information flow during flexible

sensorimotor decisions. Science, 348(6241), 1352–1355.

Siegle, G. J., Ichikawa, N., & Steinhauer, S. (2008). Blink before and after you think: Blinks

occur prior to and following cognitive load indexed by pupillary responses.

Psychophysiology, 45(5), 679–687. Siegle, G. J., Steinhauer, S. R., Stenger, V. A.,

Konecky, R., & Carter, C. S. (2003). Use of concurrent pupil dilation assessment to

inform interpretation and analysis of fMRI data. NeuroImage, 20(1), 114–124.

Smith, S. M., Fox, P. T., Miller, K. L., Glahn, D. C., Fox, P. M., Mackay, C. E., … Beckmann,

C. F. (2009). Correspondence of the brain’s functional architecture during activation and

rest. Proceedings of the National Academy of Sciences, 106(31), 13040–13045.

Sridharan, D., Levitin, D. J., & Menon, V. (2008). A critical role for the right fronto-insular

cortex in switching between central-executive and default-mode networks. Proceedings

of the National Academy of Sciences, 105(34), 12569–12574.

Stanley, M. L., Simpson, S. L., Dagenbach, D., Lyday, R. G., Burdette, J. H., & Laurienti, P. J.

(2015). Changes in brain network efficiency and working memory performance in aging.

PLoS ONE, 10(4).

Stark, L., Campbell, F. W., & Atwood, J. (1958). Pupil Unrest: An Example of Noise in a

Biological Servomechanism. Nature, 182(4639), 857–858.

Page 94: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

88

Steiger, J. (1980). Tests for comparing elements of a correlation matrix. Psychological Bulletin,

97(2), 245–251.

Stevens, A. A., Tappon, S. C., Garg, A., & Fair, D. A. (2012). Functional brain network

modularity captures inter- and intra-individual variation in working memory capacity.

PLoS ONE, 7(1), e30468.

Stevens, W. D., Buckner, R. L., & Schacter, D. L. (2010). Correlated low-frequency BOLD

fluctuations in the resting human brain are modulated by recent experience in category-

preferential visual regions. Cerebral Cortex, 20(8), 1997–2006.

Sugrue, M. F. (1983). Chronic antidepressant therapy and associated changes in central

monoaminergic receptor functioning. Pharmacology & Therapeutics, 21(1), 1–33.

Szatmari, P., Bartolucci, G., & Bremner, R. (1989). Asperger’s syndrome and autism:

Comparison of early history and outcome. Developmental Medicine & Child Neurology,

31(6), 709–720. h

Tavares, A., Handy, D. E., Bogdanova, N. N., Rosene, D. L., & Gavras, H. (1996). Localization

of alpha 2A- and alpha 2B-adrenergic receptor subtypes in brain. Hypertension, 27(Pt 1),

449–455.

Troyb, E., Rosenthal, M., Eigsti, I.-M., Kelley, E., Tyson, K., Orinstein, A., … Fein, D. (2014).

Executive functioning in individuals with a history of ASDs who have achieved optimal

outcomes. Child Neuropsychology: A Journal on Normal and Abnormal Development in

Childhood and Adolescence, 20(4), 378–397.

Page 95: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

89

Uhlén, S., & Wikberg, J. E. S. (1991). Delineation of rat kidney α2A- and α2B-adrenoeeptors

with [3H]RX821002 radiollgand binding: computer modelling reveals that guanfacine is

an α2A-selective compound. European Journal of Pharmacology, 202(2), 235–243.

Unsworth, N., & Robison, M. K. (2015). Individual differences in the allocation of attention to

items in working memory: Evidence from pupillometry. Psychonomic Bulletin &

Review, 22(3), 757-765.

Vaidya, C. J., & Gordon, E. M. (2013). Phenotypic variability in resting-state functional

connectivity: current status. Brain Connectivity, 3(2), 99–120.

Van Dijk, K. R. A., Hedden, T., Venkataraman, A., Evans, K. C., Lazar, S. W., & Buckner, R. L.

(2010). Intrinsic functional connectivity as a tool for human connectomics: theory,

properties, and optimization. Journal of Neurophysiology, 103(1), 297–321.

Varazzani, C., San-Galli, A., Gilardeau, S., & Bouret, S. (2015). Noradrenaline and dopamine

neurons in the reward/effort trade-off: a direct electrophysiological comparison in

behaving monkeys. The Journal of Neuroscience, 35(20), 7866–7877.

Vatansever, D., Menon, D. K., Manktelow, A. E., Sahakian, B. J., & Stamatakis, E. A. (2015).

Default mode dynamics for global functional integration. The Journal of Neuroscience,

35(46), 15254–15262.

Vinck, M., Batista-Brito, R., Knoblich, U., & Cardin, J. A. (2015). Arousal and locomotion make

distinct contributions to cortical activity patterns and visual encoding. Neuron, 86(3),

740–754.

Page 96: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

90

Wang, L. E., Fink, G. R., Diekhoff, S., Rehme, A. K., Eickhoff, S. B., & Grefkes, C. (2011).

Noradrenergic enhancement improves motor network connectivity in stroke patients.

Annals of Neurology, 69(2), 375–388.

Wang, L., Saalmann, Y. B., Pinsk, M. A., Arcaro, M. J., & Kastner, S. (2012).

Electrophysiological low-frequency coherence and cross-frequency coupling contribute

to bold connectivity. Neuron, 76(5), 1010–1020.

Wang, M., Ramos, B. P., Paspalas, C. D., Shu, Y., Simen, A., Duque, A., Arnsten, A. F. T.

(2007). Alpha2A-adrenoceptors strengthen working memory networks by inhibiting

cAMP-HCN channel signaling in prefrontal cortex. Cell, 129(2), 397–410.

Wang, M., Yang, Y., Wang, C.-J., Gamo, N. J., Jin, L. E., Mazer, J. A., … Arnsten, A. F. T.

(2013). NMDA receptors subserve persistent neuronal firing during working memory in

dorsolateral prefrontal cortex. Neuron, 77(4), 736–749.

Wang, Z., Chen, L. M., Négyessy, L., Friedman, R. M., Mishra, A., Gore, J. C., & Roe, A. W.

(2013). The relationship of anatomical and functional connectivity to resting-state

connectivity in primate somatosensory cortex. Neuron, 78(6), 1116–1126.

Ward, A. A., & Reed, H. L. (1946). Mechanism of pupillary dilatation elicited by cortical

stimulation. Journal of Neurophysiology, 9, 329–335.

Wehrl, H. F., Hossain, M., Lankes, K., Liu, C.-C., Bezrukov, I., Martirosian, P., … Pichler, B. J.

(2013). Simultaneous PET-MRI reveals brain function in activated and resting state on

metabolic, hemodynamic and multiple temporal scales. Nature Medicine, 19(9), 1184–

1189.

Page 97: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

91

Wernick, M. N., & Aarsvold, J. N. (2004). Emission Tomography: The Fundamentals of PET

and SPECT. Academic Press.

Whitfield-Gabrieli, S., Ghosh, S. S., Nieto-Castanon, A., Saygin, Z., Doehrmann, O., Chai, X. J.,

Gabrieli, J. D. E. (2016). Brain connectomics predict response to treatment in social

anxiety disorder. Molecular Psychiatry, 21(5), 680–685.

Whitfield-Gabrieli, S., & Nieto-Castanon, A. (2012). Conn: a functional connectivity toolbox for

correlated and anticorrelated brain networks. Brain Connectivity, 2(3), 125–141.

Wilhelm, B., Wilhelm, H., Lüdtke, H., Streicher, P., & Adler, M. (1998). Pupillographic

assessment of sleepiness in sleep-deprived healthy subjects. Sleep, 21(3), 258–265.

Williams, L. (Lea) M., Das, P., Liddell, B. J., Olivieri, G., Peduto, A. S., David, A. S., Harris, A.

W. F. (2007). Fronto-limbic and autonomic disjunctions to negative emotion distinguish

schizophrenia subtypes. Psychiatry Research: Neuroimaging, 155(1), 29–44.

Williams, L. M., Das, P., Harris, A. W. F., Liddell, B. B., Brammer, M. J., Olivieri, G., …

Gordon, E. (2004). Dysregulation of arousal and amygdala-prefrontal systems in

paranoid schizophrenia. The American Journal of Psychiatry, 161(3), 480–489.

Yamashita, M., Kawato, M., & Imamizu, H. (2015). Predicting learning plateau of working

memory from whole-brain intrinsic network connectivity patterns. Scientific Reports, 5,

7622.

Yellin, D., Berkovich-Ohana, A., & Malach, R. (2015). Coupling between pupil fluctuations and

resting-state fMRI uncovers a slow build-up of antagonistic responses in the human

cortex. NeuroImage, 106, 414–427.

Page 98: NORADRENERGIC MODULATION OF FUNCTIONAL BRAIN …

92

Yu, A. J., & Dayan, P. (2002). Expected and unexpected uncertainty: ACh and NE in the

neocortex. Advances in Neural Information Processing Systems, 15, 157–164.