The Role of Serotonin in Cortical Excitability and Network Dynamics

THE ROLE OF SEROTONIN IN CORTICAL EXCITABILITY

AND NETWORK DYNAMICS

PAVEL ANATOLYEVICH PUZEREY

Submitted in partial fulfillment of the requirements

for the degree of Doctor of Philosophy

Thesis Advisor: Roberto Fernández Gálan

Department of Neurosciences

CASE WESTERN RESERVE UNIVERSITY

January 2015

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the thesis of

Pavel Anatolyevich Puzerey

Candidate for the doctoral degree*.

David Friel

(Chair of thesis committee)

Roberto Galan

Evan Deneris

Christopher Wilson

(date) November 11th, 2014

* We also certify that written approval has been obtained for any proprietary material contained herein.

DEDICATION

This dissertation is dedicated to my loving family. Your unyielding support

and confidence in my resolve kept me afloat in these rough waters.

iii

TABLE OF CONTENTS

List of Tables viii

List of Figures ix

Acknowledgements x

Abstract xi

Chapter 1: Introduction

Overview 1

Anatomical Organization of the Cerebral Cortex 2

Dynamics of Cortical Activity 4

Neuromodulation in the Cortex 9

Serotonin as a Modulator of Cortical Neuronal Excitability 13

Serotonin as a Modulator of Cortical Network Activity 16

Chapter 2: Elevated serotonergic signaling amplifies synaptic noise and facilitates the emergence of epileptiform network oscillations

Summary 20

Introduction 20

Materials and Methods

Thalamocortical Slice Preparation 24

In Vitro Electrophysiology 24

Data Analysis and Statistics 26

Spectral Analysis 28

Computational Model 28

Seizure Induction 31

In Vivo Electrophysiology 32

Biocytin Staining, Histology, and Imaging 33

Results

Spontaneous synaptic activity in the mouse neocortex 34 is partially mediated by 5–HT in vitro

Augmenting endogenous 5–HT signaling transforms 37 cortical network dynamics via 5–HT2 and 5–HT3 receptors

Fluoxetine enhances excitatory synaptic inputs 40 mediating cortical network activity

Enhanced excitatory coupling and synaptic noise 42 are sufficient to simulate fluoxetine–modulated network activity in a model of a cortical microcircuit

In vivo blockade of 5–HT2R activity delays 48 behavioral and electrographic seizure onset and reduces seizure incidence

Discussion 50

Figures 2.1–2.5 63

Table 2.1 73

Chapter 3: Constitutive deletion of Pet–1 leads to altered cell–intrinsic, synaptic, and network excitability in mouse cortex

Summary 75

Introduction 76

Materials and Methods

Thalamocortical Slice Preparation 78

In Vitro Electrophysiology 79

Seizure Induction 81

Data Analysis and Statistics 82

Results

Cell–intrinsic parameters of neuronal excitability 83 are altered in Pet–1 knock–out mice

Cortical pyramidal cells exhibit increased 85 spontaneous synaptic activity in Pet–1 knock–out mice

Cortical network excitability is enhanced in Pet–1 86 knock–out mice

Susceptibility to convulsant–induced seizures is 88 unchanged in Pet–1 knock–out mice

Discussion 90

Figures 3.1–3.4 96

Chapter 4: On How Correlations between Excitatory and Inhibitory

Synaptic Inputs Maximize the Information Rate of Neuronal Firing

Summary 104

Introduction 105

Materials and Methods

Synaptic Inputs 109

Analytical Expression for the Cross–Correlogram 110 of the Synaptic Inputs

Single Compartment Model 111

Determination of Information Rates 112

Results

Magnitude, kinetics, and temporal correlation 115 of synaptic excitation and inhibition

Spiking behavior of a stochastic Hodgkin–Huxley 116

neuron in response to kinetically variant synaptic inputs

Entropy of neural spike trains 118

Information rate of spike trains is insensitive to 119 synaptic kinetics and the relative delay of synaptic inhibition in the balanced conductances regime

Information rate of spike trains exhibits dependence 120 on synaptic kinetics at short delays for inhibition in the balanced currents regime

Discussion 122

Figures 4.1–4.5 130

Chapter 5: General Discussion

Thesis Overview 140

On Noise 144

Serotonin as a neuromodulator of cortical network activity 147

Serotonin and Epilepsy 152

Bibliography 155

vii

LIST OF TABLES

Table 2.1

Parameter values for computational model presented in Fig 2.4 73

viii

LIST OF FIGURES

Figure 1: Circuit diagram of a canonical cortical microcircuit 19 with presently known cellular and subcellular 5-HT receptor expression profiles

Figure 2.1: Spontaneous excitatory transmitter release is 63 mediated in part by 5–HT3 receptors in mouse neocortex

Figure 2.2: Elevated endogenous 5–HT in cortical slices 65 enhances network excitability and transforms network dynamics

Figure 2.3: Enhanced synaptic excitation onto cortical 67 neurons underlies transformation of network dynamics in the presence of fluoxetine

Figure 2.4: Computational model of a cortical network 69 accounts for the emergence of fast runs with increased synaptic noise and excitatory coupling

Figure 2.5: 5–HT2 receptor blockade increases seizure 71 threshold in vivo

Figure 3.1: Altered cell–intrinsic excitability in mice lacking Pet–1 96

Figure 3.2: Increased spontaneous excitatory postsynaptic 98 currents in cortical pyramidal cells from Pet–1 knock–out mice

Figure 3.3: Enhanced cortical network excitability in mice lacking Pet–1 100

Figure 3.4: Seizure susceptibility is unaltered in mice lacking Pet–1 102

Figure 4.1: Modeling excitatory and inhibitory synaptic inputs 130

Figure 4.2: Firing properties of a stochastic Hodgkin–Huxley 132 neuron in different input regimes

Figure 4.3: Entropy of neural spike trains 134

Figure 4.4: Information rates of neural spike trains in the balanced 136 conductances regime

Figure 4.5: Information rates of neural spike trains in the balanced 138 currents regime

ACKNOWLEDGEMENTS

First and foremost, I would like to express my gratitude to my advisor, Roberto

Galán, for the knowledge and skillsets he has bestowed onto me during my time here. His genuine passion for science, his interest in the success of his students,

and his emphasis on scientific independence has been an inspiration to me.

Secondly, I would like thank Christopher Wilson, who throughout my time here

has been a mentor, a thesis committee member, and a friend. Chris taught me

electrophysiological methods in brain slices, a lesson that served as a catalyst for

a lifelong interest in neurophysiology. I would like to also thank members of my thesis committee including David Friel and Evan Deneris. In addition, I would like to extend my gratitude to members of the Case community who have provided material, methodological, and intellectual support. These include but are not limited to Michael Decker, Ted Dick, Lynn Landmesser, David Baekey,

Cathy Mayer, David Nethery, Yee Hsee Hsieh, Gemma Casadesus, George

Dubyak, Isaac Youngstrom, Rishi Dhingra, Becca James, Robert Hyde, Gustav

Karl Steinke, Yenan Zhu, Joanna Pucilowska, Joseph Vithayathil, Kathy Lobur,

Colleen McLaughlin, and Didi Mamaligas. Finally, I would like to sincerely thank

my incredible family and friends who have been my anchor during this

challenging period of my life.

The Role of Serotonin in Cortical Excitability and Network Dynamics

Abstract

PAVEL ANATOLYEVICH PUZEREY

The neocortex is the most recent addition to the vertebrate nervous system, endowing it with a capacity to generate abstract representations of sensory stimuli, mediate complex goal–directed behaviors and store memories of past events as well as numerous other functions. The neuronal activity mediating such functions arises from complex nonlinear interactions between excitatory and inhibitory cell populations within and between cortical regions. These interactions are shaped by the distinct cell–intrinsic excitability of the participating neuronal populations, the properties of the synapses connecting their constituent neurons, and the global network interactions that rise from specific patterns of connectivity. Furthermore, the above properties are all subject to neuromodulation, that is, slow neurochemical control of cell–intrinsic, synaptic, and network excitability. Significant conceptual advances in the study of neuronal networks over the last 40 years have redefined the mantra that “structure begets function,” especially in the context of neuromodulation. Namely, these advances have pointed out that while the anatomy of neuronal networks provides the constraints for their operation, it does not define the specific pattern of activity.

Network activity, it seems, can be reconfigured with neuromodulation to elicit different dynamics within the same anatomical substrate.

Using a combination of patch–clamp recordings in mouse cortical slices, computational modeling, and in vivo acute behavioral seizures I addressed contribution of the monoamine neurotransmitter, serotonin (5– hydroxytryptamine; 5–HT), to the patterning of activity within the neocortex. I first showed that cortical pyramidal cells receive a substantial source of synaptic noise in the form of spontaneous excitatory postsynaptic currents mediated by 5–

HT3 receptors, the only ionotropic 5–HT receptor. Secondly, I showed in a slice exhibiting spontaneous network activity that increasing endogenous 5–HT signaling leads to transformation of cortical activity from sparse and temporally random to clustered and periodic dynamics. Two parallel mechanisms acting through two distinct 5–HT receptors, 5–HT2 and 5–HT3, account for such a transformation of cortical network activity. In collaboration with Dr. Michael J.

Decker, I also induced acute epileptic seizures in awake and behaving mice while performing cortical electroencephalographic recordings to show that blocking 5–

HT2 receptors can substantially delay the onset of cortical epileptiform oscillations. Additionally, working with 5–HT–deficient transgenic mice, I showed that cortical cell–intrinsic, synaptic, and network excitability undergoes dramatic changes; further providing support for the role of 5–HT in controlling the patterning of cortical network activity. Finally, we generated another computational model of a cortical neuron to show that changing cell–intrinsic and network properties (e.g. through neuromodulation) can modulate the rates of

xii information transfer between cortical neurons. Combined together, these results suggest a critical role for 5–HT in shaping the activity patterns of cortical neuronal networks.

xiii

Chapter 1: Introduction

Overview

The operation of the cerebral cortex accounts for many of the complex

sensory and cognitive capacities of vertebrates, yet the underlying principles

governing its function still require elucidation. A significant paradigm shift in thinking about the function of neural networks (Getting 1989) has provided

neuroscientists with the understanding that anatomical connectivity does not

directly translate to functional connectivity (i.e. the correlation in the activity of

distinct neurons/neuronal ensembles/brain regions) under different physiological

contexts. Instead, the anatomy provides the physical substrate on which the

activity unfolds, but its dynamic spatiotemporal patterns depend on a broad set of

parameters of neuronal and synaptic excitability. Moreover, accumulating

evidence from converging lines of investigation shows that most of these

parameters can be tuned by neuromodulation, a process by which

neurotransmitters, acting through intracellular biochemical signaling cascades,

change the excitability of neurons and synapses. In my thesis, I use

complementary in vitro and in vivo electrophysiological techniques, computational modeling, and electroencephalographic (EEG) recordings during acute epileptic seizures in vivo to determine the role of the neuromodulator, serotonin (5–hydroxytryptamine; 5–HT; will be used interchangeably), in the generation of distinct cortical activity patterns. I also use a second computational model to understand how neuromodulation can affect information transfer between cortical neurons.

Anatomical organization of the cerebral cortex

The gross architecture of the cerebral cortex is defined by the arrangement of unique groups of excitatory and inhibitory neurons into six horizontal layers, which are then grouped into vertical columns, or modules, based on their patterns of afferent innervation from the thalamus and their output projections to other cortical and subcortical targets. The early work of Ramon y Cajal revealed what was, at the time, the clearest representation yet of the detailed anatomy of the cerebral cortex (Cajal 1899). His elegant staining preparations demonstrated that the cortex exhibited a roughly consistent six–layered organization across different regions and that the definition of each layer was given by the presence of distinct cell types distributed at varying densities. Subsequently, the German neuroanatomist, Korbinian Brodman, divided up the cortical sheath into a number of areas based on their distinct cytoarchitecture (i.e. the detailed cellular composition), which, in humans and other mammals, had a surprising correspondence to areas responsible for specific sensory, cognitive, or motor functions. Decades later, the seminal work of Mountcastle (Mountcastle 1957) and Hubel and Wiesel (Hubel and Wiesel 1959) confirmed the presence of functional modules in sensory cortical areas, which are vertically built from mini– columns of synaptically connected groups of (~80–300) cortical neurons. Neurons within each mini–column have invariant functional characteristics (e.g. receptive fields) imposed by patterns of thalamic innervation and local intracortical connectivity (Mountcastle 1997). These mini–columns are then arranged adjacently to construct larger columns/modules (300–600 µm diameter) in which

similar response properties are shared among large groups of neurons.

Interactions within a single column are mediated by dense local connectivity

among nearby neurons, while spread of activity between adjacent columns occurs

through patchy long–range excitatory and inhibitory neuronal projections

(Ferezou et al. 2006; Harris et al. 1999; Petersen and Sakmann 2001; Shepherd et

al. 2003). A prime example of modular cortical organization can be seen in the

rodent somatosensory cortex in which vertical chains of cortical neurons, also

known as barrel columns, exhibit an anatomical and functional mapping to

individual whiskers on the rodent’s mystacial pads (Woolsey and Van der Loos

1970). Similar organizational principles hold for the primary visual cortex of

mammals, which exhibit systematic variations in response preference to ocularity

and orientation selectivity of visual stimuli across the horizontal axis of the

cortex, but show mostly invariant selectivity along the vertical axis (Hubel and

Wiesel 1977). Thus, the gross architectural design of the cerebral cortex is predicated on the clustering of functionally related groups of neurons into vertical mini–columns, which then band to form larger columnar units that provide the necessary anatomical substrate to build local cortical circuits.

The operation of local cortical circuits is carried out by the concerted activities of excitatory and inhibitory neuronal populations. Eighty percent of cortical neurons are excitatory, have pyramidal or stellate morphology, and predominately release the neurotransmitter glutamate, while the remainder consists of inhibitory interneurons whose primary action at synapses is to release

γ–aminobutyric acid (GABA) and decrease postsynaptic excitability of their

3 target neurons (Thomson and Lamy 2007). The smallest units of organization between excitatory and inhibitory neurons are circuit motifs, which are five basic ways of connecting these two cell classes. Excitatory neurons can excite other glutamatergic neurons through feed–forward excitation, which can then activate those same neurons that excited them through a feedback loop (i.e. recurrent/feedback excitation). Feed–forward inhibition occurs when an excitatory neuron drives inhibition onto a second excitatory neuron through an intermediate inhibitory neuron, while feedback inhibition results from reciprocal connectivity between excitatory and inhibitory neurons. As a result of such reciprocal connectivity, the cortex is able to instantaneously correlate excitatory and inhibitory neuronal activity, thereby controlling the gain of inhibition proportionally with the amount of excitation and limiting the spread of excitation on a moment–to–moment basis (Fino and Yuste 2011; Holmgren et al. 2003;

Kapfer et al. 2007; Silberberg and Markram 2007; Yoshimura and Callaway

2005). Mutual inhibitory motifs have also been observed in the cortex whereby

GABAergic interneurons inhibit the activity of other inhibitory neurons (Lee et al.

2013; Pfeffer et al. 2013; Pi et al. 2013). The importance of inhibition in controlling cortical excitation is further emphasized by findings that show that synaptic coupling among excitatory cortical neurons is substantially weaker than between excitatory and inhibitory neurons (Cruikshank et al. 2007; Gabernet et al.

2005; Helmstaedter et al. 2008).

Dynamics of cortical activity

Interactions between excitatory and inhibitory neuronal populations within

the cortex govern the spatiotemporal dynamics of network activity. The logic of

cortical anatomy ensures that activation of excitatory neurons will reliably drive

the activation of inhibitory neurons with a short synaptic delay. The near

simultaneity of synaptic excitation of inhibition in the cortex observed in vivo

during spontaneous and sensory–evoked activity (Anderson et al. 2000; Atallah

and Scanziani 2009; Okun and Lampl 2008; Wehr and Zador 2003) and in vitro

(Graupner and Reyes 2013; Shu et al. 2003b) has led to the development of a

conceptual framework of cortical function in which dynamically balanced

excitation and inhibition (E/I) underlie the operation of cortical circuits (Haider et

al. 2006; Haider and McCormick 2009). This balance has been shown to bestow

upon the cortex the ability to modulate the gain of synaptic inputs (Salinas and

Sejnowski 2000), impart specific tuning properties and stimulus selectivity

(Marino et al. 2005; Wehr and Zador 2003; Wu et al. 2008), direct the flow of

activity through cortical circuits (Kremkow et al. 2010a), dictate the variability of neuronal firing (Salinas and Sejnowski 2000), modulate the efficiency of information transfer between neurons (Sengupta et al. 2013), and transform neurons into coincidence detectors by creating short windows of excitation for input integration (Pouille et al. 2009; Pouille and Scanziani 2001). Thus, it appears that the cortical E/I balance is essential for maintaining proper physiological operation of cortical circuits.

The evolution of synaptic excitation and inhibition in time gives rise to a

variety of cortical activation patterns. Given the high degree of reciprocal and

recurrent connectivity in the cortex, the abundance of regenerative ionic

conductances in neurons, and the dependence on inhibition for regulating cortical

activity, it is not surprising that the most prominent patterns of activity to emerge

are oscillations, i.e. periodic cycling of activity in the form of subthreshold or suprathreshold (action potential) voltage deflections. Oscillations in the cortex occur across a broad range of frequencies ranging from 0.05 to 500 Hz with center frequencies (i.e. mean frequency) separated linearly on a logarithmic scale

(Buzsaki and Draguhn 2004). Cortical circuits may exhibit more than one oscillation at a given time owing to the property that oscillations whose center frequencies are integer multiples can co–exist simultaneously (Burns et al. 2011;

Buzsaki and Draguhn 2004; Frohlich et al. 2010; Hansel 1996). They may also

exhibit distinct oscillatory regimes under different operational modes within a

single circuit (Poulet and Petersen 2008; Timofeev et al. 1998). Such modes of cortical circuit function typically correspond to distinct behavioral states and/or performance of specific tasks (Engel et al. 2001; Fries et al. (2002); Steriade et al.

1998; Tallon-Baudry et al. 1997).

Insights into the role of oscillatory dynamics during directed behavior, however, were preceded by the observation that cortical oscillations were most clearly apparent during behavioral states in which cognitive activity was minimal

(e.g. sleep/anesthesia) or during pathological brain states such as epileptic

seizures (Berger 1929; Steriade et al. 1993). Moreover, the development of

epileptiform oscillations within cortical structures has been reliably observed as a

transition from the otherwise physiological slow oscillation (0.1–1 Hz) that

typically occurs during slow wave sleep or anesthesia (Steriade et al. 1993;

Timofeev et al. 1998). The dynamics of this epileptiform regime is typically characterized by the transformation of the cortical slow oscillation to 2–4 Hz spike–wave discharges followed by the onset of 10–15 Hz fast run oscillations in which regenerative network bursts are superimposed on top of a massive depolarizing plateau (Steriade et al. 1998; Timofeev et al. 1998). This is

consistent with a well–recognized phenomenon that most epileptic seizures in

humans (Camfield 2011) and animals occur during sleep (Steriade and Amzica

2003; Timofeev and Steriade 2004). Since the cortex alone is sufficient to generate both the slow oscillation (Amzica and Steriade 1995; Compte et al.

2003; Sanchez-Vives and McCormick 2000; Steriade et al. 1993) and the epileptiform fast run discharges (Castro-Alamancos and Rigas 2002; Lee and

Hablitz 1991; McCormick and Contreras 2001; Timofeev et al. 1998), it is important to understand how such states and the transitions between them are generated and regulated. The cellular and network mechanisms underlying the cortical slow and epileptiform oscillations appear to be similar in that: 1) both originate in cortical layer 5 and subsequently propagate to more superficial cortical layers; 2) both depend on local, recurrent and patchy excitatory long– range projections for synchronization; 3) both are terminated by the “fatigue” of excitation (e.g. depletion of available neurotransmitter pools) and build–up of intrinsic potassium conductances (McCormick and Contreras 2001; Sanchez-

Vives and McCormick 2000). The difference between these activity regimes

arises from the differential contribution of cortical inhibition. The Up and Down

states of the slow oscillations correspond to periods of transient (~0.5–1 s)

depolarization mediated by a simultaneous arrival of excitatory and inhibitory

synaptic barrages that are approximately balanced in conductance magnitude

(Haider et al. 2006). Epileptiform cortical activity, on the other hand, emerges as

a consequence of an imbalance between synaptic excitation and inhibition

(McCormick and Contreras 2001). This is best exemplified by the artificial generation of epileptiform activity in vivo (Timofeev et al. 1998) and in vitro

(Castro-Alamancos and Rigas 2002; Lee and Hablitz 1991) by blocking

components of inhibitory synaptic transmission (e.g. GABAA receptor blockade)

or by using an artificial cerebrospinal fluid (ACSF) lacking in magnesium, which

enhances the activity of excitatory synaptic transmission through N–methyl–D– aspartate (NMDA) receptors (Kohr and Heinemann 1989). Importantly, partial

disinhibition (for example with picrotoxin, GABAA receptor antagonist) of the

cortical network leads to the generation of individual network bursts, also known

as paroxysmal depolarizing shifts (PDS), which are the intracellular correlate of

inter–ictal spikes seen in epileptic seizures and share a remarkable voltage profile

similarity to the plateau depolarizations that occur during spike–wave discharges

(Lee and Hablitz 1991; Timofeev et al. 1998). Providing additional disinhibition,

for example with blockade of both GABAA and GABAB receptors unmasks

recurrent excitatory activity and leads to the generation of epileptiform discharges

that appear indistinguishable from fast run oscillations seen in vivo (Castro-

Alamancos 2000; Castro-Alamancos and Rigas 2002). These findings provide two

important insights: Firstly, that slightly shifting the E/I balance (as with partial

disinhibition) leads to the generation of single epileptiform events that are relatively sparse and have no impact on behavior; secondly, providing additional disinhibition leads to the emergence of a fully epileptiform oscillatory regime that coincides with overt behavioral seizures (Castro-Alamancos 2000). Thus, it seems that across the gradient of the E/I balance, moving in the direction of more excitation leads to the transformation of physiological to pathophysiological oscillations. This suggests that the same cortical circuits can support distinct dynamical regimes that can transition between one another based on the relative involvement between excitatory and inhibitory neuronal activity. Furthermore, synaptic and cell–intrinsic properties of excitability of cortical neurons are subject to modulation by slow neurochemical control, or neuromodulation, through a variety of G–protein coupled neurotransmitter receptors (GPCR), allowing different neuromodulators to affect the evolution of cortical activity based on context such as behavioral state (Kaczmarek and Levitan 1987). In the following section, I discuss the role of neuromodulation in the cortex as it pertains to control of neuronal, synaptic, and network excitability and dynamics.

Neuromodulation in cortex

Biogenic neurochemicals are packaged and released from a variety of central neuromodulatory centers and intra–cortical sources onto cortical neurons to alter their excitability via control of membrane excitability. Some of these

neuromodulators include glutamate, GABA, acetylcholine, norepinephrine,

adenosine, histamine, dopamine, and serotonin (McCormick and Williamson

1989). This list does not take into account a diverse set of modulatory

neuropeptides that can also act to bias neuronal excitability through similar

GPCR–mediated mechanisms. Cognate receptors for different neuromodulators drive distinct biochemical signaling cascades downstream of G–protein activation that results in the modulation of activity of various voltage– and ligand–gated ion channels (McCormick and Williamson 1989; Tanaka and North 1993). Ion channels downstream of neuromodulatory receptors regulate passive or active membrane conductances, which can act as low–pass and high–pass filters, respectively; thus, biasing the neuronal membrane to be resonant at specific oscillatory bands (Buzsaki and Draguhn 2004). By such means, neuromodulation can specify the band–pass filtering properties of cortical neurons. In addition, neuromodulators can affect the non–linear response properties pf cortical neurons.

For instance, the convergence of several neuromodulatory receptors onto potassium channels of different sorts (depending on the receptor being activated) leads the removal of spike frequency adaptation, a mechanism by which the neuron reduces its firing rate in response to a constant input (McCormick and

Williamson 1989). This results from inhibition of potassium currents that typically act to suppress firing. By removing spike frequency adaptation, neuromodulators can control the gain of neuronal responses to synaptic inputs, thereby, controlling the contribution of modulated neurons to network activity. As opposed to inhibiting potassium currents in certain cases as with M–current inhibition with acetylcholine/serotonin or inhibition of the slow

afterhyperpolarizing current with 5–HT (Halliwell 1986; McCormick and

Williamson 1989; Villalobos et al. 2005), neuromodulators can also directly

activate hyperpolarizing potassium currents as with adenosine, GABAB, and serotonin receptors (Araneda and Andrade 1991; McCormick and Williamson

1989; Tanaka and North 1993) and inhibit neuronal excitability. Thus, neuromodulatory influence can bias neuronal excitability bi–directionally by inhibiting or activating cell–intrinsic potassium conductances.

The effects of neuromodulation extend beyond the control of cell–intrinsic excitability to the regulation of synaptic transmission. Presynaptic release of neurotransmitter vesicles has a non–linear relationship to the presynaptic

membrane potential. Once the membrane potential reaches a threshold level of

depolarization, voltage–gated calcium channels open to allow for calcium influx

and release of presynaptic neurotransmitter vesicles. Thus, control of the

presynaptic membrane potential will have a direct influence on the probability of

neurotransmitter release. It is precisely by this mechanism that neuromodulation

regulates synaptic transmission from the presynaptic side. For instance, 5–HT1B

and GABAB receptors negatively regulate thalamocortical synaptic transmission by activating potassium conductances that hyperpolarize the presynaptic membrane (Gothert 1990; Laurent et al. 2002; Porter et al. 2001). Postsynaptic

receptors involved in synaptic transmission within the cortex are also subject to

neuromodulatory regulation. GABAA receptors can be inhibited by activation of

postsynaptic 5–HT2C receptors (Huidobro-Toro et al. 1996) and 5–HT2A

receptors can enhance NMDA receptor–mediated currents (Rahman and Neuman

1993). The above studies represent a few out of many mechanisms ways by which

neuromodulators shape the interactions between synaptically connected neurons.

By virtue of its control of cell–intrinsic and synaptic excitability,

neuromodulation can also lead to changes in the activity of networks of cortical neurons. This has been most intensely studied in the context of the cortical slow oscillation. As mentioned earlier (see Dynamics of cortical activity), the slow oscillation is generated within the cortex and corresponds to alternating periods

(0.5–1 s) of intense synaptic activity among excitatory and inhibitory neurons followed by periods of quiescence and the activation of cell–intrinsic potassium conductances whose primary action is to hyperpolarize the neuronal membrane.

The activity of these hyperpolarizing conductances can be altered with

neuromodulators (McCormick and Williamson 1989). Previous studies have

shown that selective activation of cortical cholinergic and noradrenergic receptors

leads to abolition of the cortical slow oscillation and the onset of a tonic

depolarized state that resembles a continuous Up–state (Castro-Alamancos and

Gulati 2014; Constantinople and Bruno 2011). These findings are consistent with

the known role of cholinergic and noradrenergic neurmodulatory nuclei in the

reticular activating system in bringing about sleep–wake transitions

(Constantinople and Bruno 2011; Steriade et al. 1993). Thus, neuromodulation

can reconfigure cortical network dynamics by specifically acting on cortical

neurons to alter their intrinsic excitability, thereby bringing about a qualitatively

distinct behavioral state (i.e. awake vs. asleep).

The focus of this dissertation is one the role of 5–HT in cortical network

activity. To understand how 5–HT affects the orchestrated activity of neuronal

networks, it is essential to discuss its known effects on neuronal excitability.

Serotonin as a modulator of cortical neuronal excitability

Serotonin is supplied to the cortex by the medial forebrain bundle, which

emanates from the dorsal and median raphe nuclei (DRN and MRN, respectively)

within the midbrain and brainstem (Tork 1990). These parallel fiber systems

deliver two distinct sources of 5–HT to the forebrain with terminals from the

DRN providing diffuse branching with thin varicose axons and no synaptic

contacts and the MRN providing large, non–varicose basket axons that make

synaptic contacts onto their postsynaptic targets (Tork 1990).

Serotonin released within the cortex can exert its action on seven different

classes of 5–HT receptors (5–HT1, 5–HT2, 5–HT3, 5–HT4, 5–HT5, 5–HT6 and

5–HT7 receptors), which also includes a set of receptor subtypes for each class

(Nichols and Nichols 2008). Expression profiles of cortical 5–HT receptors tend to be complex with anterior–posterior gradients in receptor expression as well as laminar, cellular, and subcellular localization (Amargos-Bosch et al. 2004; Blue et al. 1988; Jakab and Goldman-Rakic 2000; Santana et al. 2004; Weber and

Andrade 2010). Despite the rich diversity of cortical 5–HT receptors, the

predominant cortical responses to 5–HT are mediated by 5–HT1, 5–HT2, and 5–

HT3 receptors (Araneda and Andrade 1991; Lee et al. 2010; Tanaka and North

1993). Coupling of 5–HT1 receptors with Gi/o α subunits leads to activation of G–

protein–coupled inwardly rectifying potassium (GIRK) channels that ultimately

lead to the hyperpolarization of the neuronal membrane (Araneda and Andrade

1991; Luscher et al. 1997; Tanaka and North 1993). 5–HT1 receptors are mostly expressed near the axon hillock and are therefore poised to regulate neuronal output (Azmitia et al. 1996). In contrast, expression of 5–HT2 receptors has been

localized mostly to the apical dendrite of pyramidal neurons within the cortex

(Fig. 1) (Aghajanian and Marek 1999; Cornea-Hebert et al. 1999; Jakab and

Goldman-Rakic 2000; Willins et al. 1997). Activation of these receptors leads to

membrane depolarization often superposed with bursts of action potentials

(Araneda and Andrade 1991). Given the localization of 5–HT2 receptors to apical dendrites of pyramidal neurons, 5–HT2 receptors are positioned to modulate the gain of synaptic responses (Zhang and Arsenault 2005). The notion that there is

an anatomical and functional separation of 5–HT1 and 5–HT2 receptors is further

supported by the observation that in some cortical areas (i.e. somatosensory

cortex), some serotonergic afferents from the raphe nuclei align in parallel to the

apical dendrites of 5–HT2 receptor–expressing neurons (Blue et al. 1988; Jansson et al. 2001), while others align in an orthogonal direction that coincides roughly with the axonal expression of 5–HT1 receptors (DeFelipe et al. 2001).

Interestingly, the anatomical separation of serotonergic innervation patterns and expression of 5–HT receptors on cortical pyramidal cells does not preclude a mixed response, especially considering that 80% of 5–HT2 receptor–expression pyramidal neurons also express 5–HT1 receptors (Amargos-Bosch et al. 2004).

Specifically, cells that undergo facilitation of firing in a 5–HT2 receptor– dependent manner can also experience suppression of firing if a different raphe

site is stimulated, suggesting the existence of subcircuit–specific serotonergic

pathways that can effect distinct compartments of cortical pyramidal neurons.

This is consistent with previous findings showing neurons exhibiting 5–HT1

receptor–mediated hyperpolarizations followed by immediate 5–HT2 receptor–

mediated depolarizations (Araneda and Andrade 1991). In addition, inhibition of

firing by 5–HT can be followed with a delay by an enhancement of the firing rate, a property of cortical 5–HT responses that corresponds to specific projection patterns of cortical neurons (Amargos-Bosch et al. 2004). 5–HT3 receptors, on the

other hand, are thought to be expressed exclusively on inhibitory interneurons

within the cortex (Fig.1) (Jakab and Goldman-Rakic 2000; Lee et al. 2010;

Morales and Wang 2002; Puig et al. 2004; Vucurovic et al. 2010). We will

challenge this notion in Chapter 2 (see Chapter 2: Discussion). Thus, responses of

cortical neurons to 5–HT tend to be highly complex, with the predominant effect

at the level of the neuron being inhibition of firing (Jacobs and Azmitia 1992),

likely resulting from a longer time–course of action for 5–HT1 receptors and an

age–dependent decline in the expression of cortical 5–HT2 receptors (Amargos-

Bosch et al. 2004).

Another level of complexity to cortical 5–HT signaling is added when one

considers the effects of 5–HT on synaptic transmission among cortical neurons.

Previous work has shown 5–HT reliably enhances spontaneous excitatory

postsynaptic currents (sEPSCs) in layer 5 pyramidal neurons in the medial

prefrontal cortex (mPFC), a phenomenon attributed to the enhancement of

intrinsic cortical circuits in a 5–HT2 receptor–dependent manner (Beique et al.

2007; Lambe et al. 2000). On the other hand, 5–HT can also enhance the activity

of inhibitory interneurons via 5–HT2 and 5–HT3 receptors, leading to either a

long–lasting (5–HT2 receptor–dependent) or a transient (5–HT3 receptor–

dependent) enhancement of spontaneous inhibitory postsynaptic currents (sIPSCs)

onto pyramidal neurons (Puig et al. 2004; Zhou and Hablitz 1999). Presynaptic 5–

HT1B receptors have also been shown to inhibit the release of glutamate from

excitatory cortical terminals (Tanaka and North 1993). Very relevant to the theme

of this dissertation is the finding by Moreau and colleagues that 5–HT can shift

the balance between excitatory and inhibitory synaptic transmission between

cortical neurons in favor of more excitation (Moreau et al. 2010). As discussed

earlier (see Dynamics of Cortical Activity), the E/I balance can govern transitions

between distinct cortical states. Thus, the finding that 5–HT can modulate the E/I

balance suggests that it may itself contribute to transitions between different

dynamical regimes within the cortex. This notion will be the guiding motivation

behind the experiments presented in my work. However, before proceeding to the

main findings of this dissertation, I will discuss work on the role of 5–HT in

cortical network activity that preceded this study.

Serotonin as a modulator of cortical network activity

The dense innervation of cortical neurons by serotonergic afferents (Blue

et al. 1988), the widespread expression of cortical 5–HT receptors (Dougherty and

Aloyo 2011; Nichols and Nichols 2008; Santana et al. 2004), and the

aforementioned effects of 5–HT on cell–intrinsic and synaptic excitability

suggests that 5–HT is likely to affect the collective and synchronized activity of

cortical networks. Surprisingly, little is known about 5–HT’s role in this process.

Early studies using bipolar recording electrodes in the neocortex suggested that 5–

HT may be responsible for generating atropine–resistant “low voltage fast activity,” which in contemporary jargon refers to the desynchronized cortical state observed upon waking and during rapid eye–movement (REM) sleep

(Vanderwolf and Baker 1986). However, it is known today that such transitions are largely mediated by norepinephrine (Castro-Alamancos and Gulati 2014;

Constantinople and Bruno 2011; Steriade et al. 1993). Upon discovering that the mPFC receives the densest innervation of serotonergic afferents and that this coincides with the highest density of 5–HT receptors, serotonergic modulation of

PFC activity became a topic of interest (Celada et al. 2013). Subsequently, Puig and colleagues showed that release of endogenous cortical 5–HT by low–

frequency stimulation of the DRN lead to modulation of distinct bands of

oscillatory activity within the cortex (Puig et al. 2010). Specifically, this study

showed that the cortical slow oscillation was accelerated in response to DRN

stimulation in a 5–HT2 receptor–dependent manner. They also showed that inhibiting 5–HT1 or 5–HT2 receptors led to an increase and decrease, respectively, in the baseline power of gamma (30–60 Hz) oscillations.

Furthermore, high–frequency (100 Hz) stimulation of the DRN in this study led to the desynchronization of cortical slow oscillations and a decrease in the power of gamma oscillations. These findings were the first to directly show the influence of endogenously released 5–HT on cortical network activity. Note that these studies addressed the role of 5–HT as a modulator of cortical network activity. This

17 means that the oscillations had to be ongoing for 5–HT to exert its effect, which was to increase or decrease the spectral power of specific oscillations. The work presented herein will focus on the role of cortical 5–HT, not as a modulator of ongoing network oscillations, but as a facilitator for the emergence of distinct oscillatory regimes.

Figure 1. Circuit diagram of a canonical cortical microcircuit with presently

known cellular and subcellular 5-HT receptor expression profiles. Cortical

pyramidal neurons mainly express 5-HT2 receptors at the apical dendrites and 5-

HT1 receptors at the axon hillock. Activity of fast spiking inhibitory cortical

interneurons is modulated by 5-HT2 receptors, while that of non-fast spiking interneurons is affected by 5-HT3 receptors. Our results suggest that 5-HT3

receptors are also expressed on cortical pyramidal neurons, though the exact

subcellular localization is unknown.

Chapter 2: Elevated serotonergic signaling amplifies synaptic noise and

facilitates the emergence of epileptiform network oscillations

Summary

Serotonin fibers densely innervate the cortical sheath to regulate neuronal

excitability, but serotonin’s role in shaping network dynamics remains

undetermined. We show that serotonin provides an excitatory tone to cortical

neurons in the form of spontaneous synaptic noise through 5–HT3 receptors,

which is persistent and can be augmented using fluoxetine, a selective serotonin

re–uptake inhibitor. Augmented serotonin signaling also increases cortical

network activity by enhancing synaptic excitation through activation of 5–HT2

receptors. This in turn facilitates the emergence of epileptiform network

oscillations (10–16 Hz) known as fast runs. A computational model of cortical

dynamics demonstrates that these two combined mechanisms, increased

background synaptic noise and enhanced synaptic excitation, are sufficient to

replicate the emergence fast runs and their statistics. Consistent with these

findings, we show that blocking 5–HT2 receptors in vivo significantly raises the

threshold for convulsant–induced seizures.

Introduction

Various neurotransmitter systems converge onto the neocortex to regulate

the activity of cortical neurons in a process known as neuromodulation

(Kaczmarek and Levitan 1987). Through the regulation of neuronal excitability,

neuromodulation shapes the patterning of neural activity (Gil et al. 1997; Puig et

al. 2010). Of the gamut of neuromodulators, serotonin (5–hydroxytryptamine; 5–

HT) exerts a powerful influence on cortical neurons owing to an extensive

distribution of 5–HT receptors (Dougherty and Aloyo 2011; Nichols and Nichols

2008; Santana et al. 2004) juxtaposed with dense innervation by serotonergic

axons from the midbrain (Blue et al. 1988). In cortical neurons, metabotropic 5–

HT2 and ionotropic 5–HT3 receptors (5–HT2Rs & 5–HT3Rs) are the primary

mediators of excitatory responses to 5–HT (Nichols and Nichols 2008; Roerig et

al. 1997; Tanaka and North 1993). The effects of these receptors are well

characterized at the level of intrinsic neuronal and synaptic excitability (Araneda

and Andrade 1991; Fink and Gothert 2007; Zhou and Hablitz 1999). However, the

consequences of these effects on cortical network dynamics remain largely

unexplored (Puig and Gulledge 2011).

Cortical network activity emanates from complex interactions between

synaptic inputs (Economo and White 2012; Fernandez et al. 2013) and cell-

intrinsic excitability (Grashow et al. 2010; Kolind et al. 2012)–both are subject to

neuromodulation (Araneda and Andrade 1991; Gil et al. 1997). The temporal

dynamics of such activity vary between irregular (Shadlen and Newsome 1998)

and oscillatory (Buzsaki and Draguhn 2004) patterns of activation, which may

coexist simultaneously (Burns et al. 2011; Frohlich et al. 2010; Hansel 1996) or

under different operational modes within a network (Poulet and Petersen 2008).

Investigation of the invertebrate nervous system has revealed that network

dynamics are reconfigured with neuromodulation to switch between distinct

dynamical regimes, resulting in changes to behavior (Marder 2012).

Complementary findings in vertebrates remain scarce. Despite recent advances in

understanding the role of 5–HT in regulation of cortical network activity

(Andrade 2011; Celada et al. 2013; Nakamura and Wong-Lin 2014; Puig and

Gulledge 2011), comprehensive evidence is still lacking owing to serotonin’s complex and combinatorial effects on neuronal excitability.

Theoretical studies have suggested that network dynamics are sensitive to random fluctuations of synaptic conductances or in other words, synaptic “noise”

(Parga and Abbott 2007; Stacey et al. 2009). Introduction of randomly fluctuating synaptic inputs to simulated neuronal networks facilitates transitions between distinct network states (Destexhe et al. 2001; Frohlich et al. 2010), but to date there is little empirical evidence for this. Within the neocortex, synaptic noise emerges from spontaneous release of synaptic vesicles (Otsu and Murphy 2003;

Wasser and Kavalali 2009). Based on the theoretical considerations, we sought to understand how 5–HT contributes to synaptic noise in the cortex and, more specifically, how this serotonergic tone influences network excitability and network oscillations.

Here we show that synaptic noise can be augmented with the antidepressant fluoxetine (FLX), resulting in enhanced cortical network activity.

Importantly, the increase in noise is accompanied by the emergence of epileptiform cortical dynamics that arise from a 5–HT2R–dependent enhancement

of synaptic excitatory coupling. Using computational simulations, we demonstrate

that enhanced synaptic excitation and noise are sufficient to account for the

observed dynamics. In accord with these findings from in vitro experiments, we

show that blocking 5–HT2Rs in mice significantly delays the onset of

convulsant–induced epileptic seizures in vivo, and may even completely prevent

them.

Materials and Methods

Thalamocortical slice preparation. Thalamocortical slices (350 µm) of

somatosensory cortex were prepared as previously described from juvenile (P13–

21) C57BL/6 wild–type mice (Agmon and Connors 1991; Pucilowska et al.

2012). Animals were anesthetized with vapor isoflurane and decapitated with a

guillotine. The brain was then submerged in ice–cold artificial cerebrospinal fluid

(ACSF) saturated with 95% O2 5% CO2 containing the following: 125 mM NaCl,

2.5 mM KCl, 1 mM MgCl2, 2 mM CaCl2, 25 mM NaHCO3, 1.25 mM NaH2PO4

and 25 mM glucose. Brain slices were cut on a vibratome (Leica VT1200). All

chemical salts and reagents were purchased from Fisher Scientific (Pittsburgh,

PA) and Sigma–Aldrich (St. Louis, MO). After sectioning at the vibratome, slices were transferred to a bath containing room temperature ACSF for 20 minutes to incubate. Subsequently, slices were moved to the recording chamber and perfused with standard ACSF warmed to 31°C with a TC–324B Automatic Temperature

Controller (Warner Instrument Corporation; Hamden, CT) at a rate of 2 mL/minute. Slices were then incubated for one hour before beginning electrophysiological recordings.

In vitro electrophysiology. Pyramidal cells within L2/3 were visually identified at

63× magnification using Kohler illumination with an upright microscope (Zeiss

Axioskop 2 FS+; Germany). Whole–cell patch clamp recordings under the

current–clamp configuration were established in single neurons using borosilicate

glass electrodes (6–10 MΩ) filled with standard internal solution containing the following: 120 mM potassium gluconate, 2 mM KCl, 10 mM HEPES, 10 mM sodium phosphocreatine, 4 mM MgATP, 0.3 mM Na3GTP, 25 mM QX314, and

adjusted to pH 7.4 with KOH. For voltage-clamp recordings, a cesium-based

internal solution was used to improve space clamp and contained the following:

120 mM cesium gluconate, 2 mM CsCl, 10 mM HEPES, 10 mM sodium

phosphocreatine, 4 mM MgATP, 0.3 mM Na3GTP, 20 mM BAPTA, and 25 mM

QX314 to block voltage-gated sodium channels and adjusted to pH 7.4 with

CsOH. Voltage and current clamp experiments were performed using Multiclamp

700B amplifier (Molecular Devices, Foster City, CA) digitized at 10 KHz with

Digidata 1400 data acquisition interface. The data was low–pass filtered online at

1 KHz.

Spontaneous excitatory postsynaptic currents (sEPSCs) were recorded in the

voltage-clamp configuration for 60 seconds. To isolate the excitatory synaptic

currents, the membrane potential of the neuronal membrane was biased to the

reversal potential for inhibitory postsynaptic currents (EIPSC), which was

experimentally determined to be –80 mV, consistent with previous reports

(Chagnac-Amitai and Connors 1989; Hasenstaub et al. 2005). Recordings were

excluded if access resistance exceeded 30 MΩ throughout the duration of the

recordings and/or if the resting membrane potential was more positive than –60

mV. All pharmacological agents were washed into the bath. The slices were then

given one hour to incubate before beginning recordings.

Current-clamp recordings were performed to assess cortical network activity in a

disinhibited slice preparation. To partially disinhibit the cortical network, the

ACSF was modified to contain 1 µM gabazine (GZN), a selective GABAA

receptor antagonist. Under control conditions, spontaneous network activity

appeared as paroxysmal depolarizing shifts (PDS), which had a stereotypical

voltage profile containing an initial plateau depolarization (~60–80 mV relative to

baseline) lasting 400–500 ms succeeded by a decaying tail lasting ~500 ms.

Recordings of cortical network activity lasted 10 minutes and were obtained

without current injection (Ihold = 0 pA). The concentration of GZN was chosen so

to elicit, on average, 1–2 PDS per minute, thus making the recordings amenable to

statistical analysis. At this concentration, the network was not sufficiently

disinhibited to spontaneously exhibit fast run epileptiform oscillations.

Sampling from individual animals for slice experiments involved taking one slice

per animal and recording from 5–7 cells from each slice. No more than one

pharmacological challenge was presented to a given slice. Statistical tests were

carried out on groups of 30–40 cells (corresponding to 5–6 animals/group).

Data analysis and statistics. Spontaneous excitatory postsynaptic currents

(sEPSCs) were detected using a custom algorithm written and implemented in

MATLAB (Mathworks, MA). Detection was based on a threshold for the sEPSC derivative with threshold values for events obtained empirically. sEPSC detection also included criteria for event kinetics. Events with rise time longer than 5 ms

and decay time constant longer than 30 ms were excluded; also the rise time was

not allowed exceed the decay time constant. Events were visually inspected after

being subjected to the exclusion criteria to ensure they exhibited the typical

profile of an alpha function. Network events in current–clamp were detected

manually using a custom interactive program written in MATLAB. Voltage

deflections during PDS network activity typically were between 50–80 mV, depending on the resting potential of the neuron (typically –70 mV), thus could be detected easily by a simple amplitude threshold–based detection. Reverberant afterdischarges during fast runs were detected by setting an absolute amplitude threshold of 30 mV relative to the resting potential and a 10 mV threshold relative to the local minimum (i.e. if the voltage deflection occurred on top of a depolarization plateau). Network–mediated postsynaptic currents (nEPSCs) were detected using an automated custom script in MATLAB that detected currents larger than 100 pA. In the instance of reverberant network currents, the first current in a single network event always exceeded 1 nA and was substantially larger than the succeeding currents. Thus, to separate the first event from the smaller reverberant currents, we used an amplitude threshold of 1 nA. This procedure allowed us to separate the initial network current (i.e. the trigger event) and the sequential afterdischarges (i.e. reverberant currents). We determined statistically significant differences between distributions using the nonparametric

Wilcoxon rank–sum test, which tests the null hypothesis that two independent groups of samples come from distributions with equal medians.

Spectral analysis. To compute the power spectrum of PDS, we first convert the

membrane potential traces into series of discrete events

N st()=∑δ ( t − tp ), (1) p=1

where δ ()x is Dirac’s delta function; t p denotes the time of the p-th PDS onset;

and N denotes the number of PDS in 10 min. The Fourier transform of (1) yields

11∞ N s(ω )= st ( )exp( itdt ωω ) = exp(it ) , ∫−∞ ∑ k 22ππp=1 where i is the imaginary unit and ω is the angular frequency. The power

spectrum is then given by

NN NN 2 11N P()()ωω= s =∑∑exp(ittω( pq−=+)) ∑∑cos(ω(ttpq−)) , (2) 22πp==11 q ππp>= qq 1

Expression (2) is computed for each cell and then averaged for all cells recorded

under the same conditions (more than 30 in each group). Note that at ω = 0 the

2 power is N / (2π ) . Moreover, for stochastic binary processes, as ω →∞ the power converges to the baseline N / (2π ) . The x–axis in Fig. 2f shows the linear

frequency, f = ωπ/2( ) .

Computational model. The model consists of an excitatory neuron coupled to an

inhibitory neuron as depicted in Fig. 4a. The model also includes recurrent

excitation and inhibition. The rate of change of the membrane potential for each

neuron obeys the equation C⋅=− dV/ dt I total , where C is the membrane

capacitance, V is the membrane potential, and I total is the total current, i.e. the

sum of all membrane and synaptic currents. We first note that during a PDS the

current is completely dominated by synaptic currents, in other words, the neurons

are in a high–conductance state. Similarly, in the absence of PDS the neurons are

close to their resting potential whose fluctuations are dominated by synaptic

noise. Consistent with these observations, all voltage–gated membrane currents

can be neglected for the purpose of this study. Thus, the total current consists of

an excitatory and inhibitory component. The excitatory component is mediated by

AMPA and NMDA currents, whereas the inhibitory component is mediated by

+ GABAA current and by a slow outward K current similar to the slow

afterhyperpolarizing current (IsAHP) (Andrade et al. 2012; Villalobos et al. 2005).

In addition, we include a leak current and synaptic noise in the form of mixed excitatory and inhibitory barrages. The total current takes then the form:

I≈ I + I + I + I ++ II total AMPA NMDA GABA A sAHP leak noise . The parameters for all these currents are defined below and their values are listed in Table 1. The functional form of

the AMPA, GABAA and sAHP currents is the same, Ikk= Gt()( V − E k) , where the

subindex k refers to the current type, Gtk ()is the time dependent synaptic

conductance, as explained below, V is the postsynaptic membrane potential and

Ek is the reversal potential for that synapse type. The time dependence of the

synaptic conductances is modeled as a double low–pass filter so that the impulse–

response is the well–known “alpha function”. Specifically, the integration of the

synaptic dynamics is given by

 dFk τθk=−+F kk gHV( pre −)  dt  dG τ k =−+GF  kdt kk

Where Vpre is the membrane potential of the presynaptic neuron, θ is the voltage

threshold for the presynaptic neuron to release neurotransmitter, gk is the

maximal synaptic conductance, and τ k is the synaptic time constant (see Table 1).

The function Hx()− a is the Heaviside function; its value is 1 when xa> and 0

otherwise. The functional form of the NMDA current contains an additional

factor, BV( )=+− 1/( 1 0.7 exp(V /16.13)) , that accounts for the Mg2+–blockade of

NMDA receptors dependent on the postsynaptic membrane potential:

INMDA= G NMDA () t ⋅ BV ( ) ⋅−( V Ek ) . The leak current has the form Ileak = gVLL() − E ,

where the leak conductance, gL , is constant. The synaptic noise has two

independent contributions, both excitatory: 1) small–amplitude noise; and 2)

large–amplitude noise. The former is responsible for fluctuations of the

membrane potential that, at rest, are not sufficient to cause large depolarizations.

The latter is responsible for large but infrequent membrane potential

depolarizations. Both sources of noise are modeled as Poisson processes that are

statistically independent from each other. The small noise impinges independently

on the excitatory and inhibitory neurons, whereas the large noise impinges on the

excitatory neuron only. The events of the Poisson process are processed through a

double low–pass filter to model the kinetics of synaptic noise. The rates of the

Poisson processes, λ ; the peak amplitudes of the events, g ; and their time

constants, τ , are listed on Table 1.

Seizure induction. Juvenile C57BL/6 mice (P18–24) were administered the convulsant, pentetrazol (PTZ; i.p. 80 mg/kg) dissolved in 0.1 M phosphate- buffered saline (PBS). Some animals were treated with ketanserin (KSN; i.p. 5 mg/kg) dissolved in 5% DMSO 0.1 M PBS solution one hour prior to PTZ injection and a control group was pre–treated with DMSO solution alone to exclude the potential effect of the drug vehicle in modulating seizure parameters.

Control animals were not subsequently used in testing the effects of KSN. The starting point of behavioral seizures was considered as the appearance of generalized clonic convulsions with loss of righting reflex (GCC–LOR). The

endpoint for each GCC–LOR seizure was considered as either complete release

from tonus (in animals presenting with tonus) or as cessation of generalized clonic

convulsions and recovery of righting (Loscher et al. 1990). The time elapsed

between PTZ injection and the first GCC–LOR seizure was taken as the seizure

latency. Time elapsed between each subsequent seizure was considered to be the

inter–seizure interval (ISI). In the majority of acute seizure experiments the

animal died after experiencing several seizures. If the seizures persisted for over

one hour after PTZ injection, the animals were euthanized by isoflurane

anesthesia and decapitation. Animals that did not experience seizures within an

hour after PTZ injection were not included in the data analysis. For the

electroencephalography recordings (see next section), the time limit for the

experiment was extended to two hours.

In vivo electrophysiology. We obtained electroencephalographic (EEG)

recordings of PTZ–induced seizures from 13 juvenile (P20–24) C57BL/6 mice.

Using techniques described elsewhere (Berkeley et al. 2002; Papale et al. 2013), we implanted a fine wire EEG recording electrode (Vintage Machine Supplies,

Inc., Medina OH) subdurally over somatosensory cortex and a reference electrode was placed over the ipsilateral sensorimotor cortex. The tip of the electrodes (~ 1 mm) was fed through two small holes in the parietal bone of the calvarium and positioned parallel to the surface of the cortex. Postsurgical pain management was maintained with bupivacaine (2 mg/kg every 12 hours, s.c.) and carprofen (5 mg/kg every 24 hours, s.c.). Animals were allowed to recover 1–2 days before the seizure experiments were performed. EEG recordings were collected using a

CP511 AC amplifier (Grass Technologies, Middleton, WI) using 100× amplification. The raw signal was band–pass filtered on–line between 0.1–100 Hz and digitized at 1 KHz using a Power 1401 mkII data acquisition system

(Cambridge Electronic Design Limited; Cambridge, United Kingdom).

EEG recordings were then processed offline using MATLAB. Raw EEG waveforms were mean–subtracted and passed through a Butterworth filter in the pass band of 8–16 Hz. This pass band was selected based on the relevant physiological frequencies of fast run oscillations in vivo (Steriade et al. 1998).

The onset of electrographic seizures was detected by rectifying and integrating the band–pass filtered EEG waveform with a time constant of τ = 5 s. An amplitude threshold was then set for the filtered and integrated EEG traces from individual animals due to the variable signal–to–noise ratios between different experiments.

The latency to the first electrographic seizure was taken as the time from injection

of convulsant to the first threshold crossing in the processed EEG signal.

Biocytin staining, histology, and imaging. Layer 2/3 PCs were filled with 20 mM

biocytin (Sigma–Aldrich) during patch–clamp recordings. At the end of the

experiments, slices were preserved in 0.1 M PBS solution containing 4%

paraformaldehyde for over 24 hours. Slices were then subjected to 3 washes in 0.1

M PBS solution containing 0.5% triton–X 100 for 15 minutes and left to incubate for 24 hours in the same solution containing avidin–biotin complex (ABC

Staining Kit – Elite Vector Labs). Slices were then washed 3 times in the PBS–

triton solution for 15 minutes and stained with a diaminobenzedine (DAB

Peroxidase Substrate Kit – Elite Vector Labs) solution until desired staining

intensity was achieved. Slices were rinsed 3 times for 10 minutes in 0.1 M PBS

solution followed by series of alcohol and xylene washes to dehydrate the tissue.

Slices were stored in methyl salicylate. Stained neurons were visualized under

brightfield illumination with 20× and 40× magnification using a Leica DM5000 B

upright microscope and imaged with a Leica DFC420 C digital camera and Leica

Application Suite imaging software. Images were further processed and refined

using ImageJ (NIH) graphics editing software.

Results

Spontaneous synaptic activity in mouse neocortex is partially mediated by 5–HT

in vitro

As a first step to investigate the role of serotonin in cortical network

excitability, we measured the contribution of serotonergic signaling to

spontaneous synaptic activity in layer 2/3 pyramidal cells (PCs) in thalamocortical

slices of mouse somatosensory cortex (Agmon and Connors 1991). We chose somatosensory cortex as a target region owing to its distinct organization of neuronal populations into barrel columns, which are specialized for local (< 300

µm) neuronal interactions and have a high degree of recurrent connectivity

(Lefort et al. 2009; Petersen et al. 2003; Thomson and Lamy 2007). Pyramidal cell morphology was determined visually and verified post–hoc with biocytin labeling (Fig. 2.1a). Spontaneous excitatory postsynaptic currents (sEPSCs) were recorded in voltage–clamp using a cesium–based internal solution while blocking glutamatergic synaptic transmission through AMPA and kainate receptors with

2,3–dyhydroxy–6–nitro–7–sulfamoyl–benzo[f]quinoxaline–2,3–dione (NBQX, 10

µM). NMDA receptor contribution was eliminated by holding the cell at a holding potential (–80 mV) at which NMDA receptors are blocked by Mg2+ ions. We

show that bath application of NBQX significantly reduced sEPSC amplitudes (p <

0.01, Wilcoxon rank–sum test; Fig. 2.1b,c) and increased the inter–event intervals

(IEI) (p < 0.01; Wilcoxon rank–sum test; Fig. 1b,d), but failed to eliminate all

sEPSCs, consistent with previous reports implicating other neurotransmitters

involved in spontaneous synaptic release (Pankratov et al. 2007; Roerig et al.

1997). We investigated 5–HT as a potential source of the NBQX–resistant

sEPSCs based on a previous report showing 5–HT as a mediator of fast synaptic

transmission in cortical PCs through 5–HT3 ionotropic receptors (5–

HT3Rs)(Roerig et al. 1997). In agreement with that report, bath application of the

5–HT3R antagonist, granisetron (GSN; 1 µM), with NBQX resulted in a dramatic

34 reduction of sEPSC amplitudes (p < 0.01, Wilcoxon rank–sum test; Fig. 2.1b,c) and a substantial increase in the IEIs (p < 0.01, Wilcoxon rank–sum test; Fig.

2.1b,d). To test whether the serotonergic component depends on neuronal firing, we co–administered tetrodotoxin (TTX; 1 µM), a voltage–gated sodium channel blocker, with NBQX and observed no significant difference in event amplitudes between events recorded in the presence of NBQX alone or with NBQX and TTX in tandem (p > 0.05; Wilcoxon rank–sum test; Fig. 2.1b, c). The event amplitudes in the presence of NBQX and GSN together were significantly smaller than with

NBQX and TTX (p < 0.01, Wilcoxon rank–sum test; Fig. 2.1b, c). While no significant change in IEIs was observed between spontaneous activity recorded with NBQX and NBQX with TTX (p > 0.05, Wilcoxon rank–sum test; Fig. 2.1b, d), there was a significant increase in IEIs recorded with NBQX and GSN as compared to the group recorded with NBQX and TTX (p < 0.01, Wilcoxon rank– sum test; Fig. 2.1b, d). This suggests that the fast ionotropic 5–HT3R signaling observed in cortical PCs arises independently of neuronal firing. Combined, these results indicate that 5–HT signaling through 5–HT3Rs significantly contributes to the NBQX–resistant component of the ongoing spontaneous synaptic activity in the mouse neocortex.

In line with this hypothesis, raising endogenous 5–HT levels should increase spontaneous synaptic activity. Indeed, bath application of the selective serotonin reuptake inhibitor, fluoxetine (FLX, 4 µM), dramatically augmented sEPSC amplitudes (p < 0.01, Wilcoxon rank–sum test; Fig. 2.1e,f) and reduced the IEIs (p < 0.01; Wilcoxon rank–sum test; Fig. 2.1e,g) relative to control. Co–

application of FLX and GSN substantially reduced the FLX–mediated increase in

sEPSC amplitudes (p < 0.01, Wilcoxon rank–sum test; Fig. 2.1e,f), but had no

effect on the FLX–mediated decrease in IEIs (p > 0.05; Wilcoxon rank–sum test;

Fig. 2.1e,g). Note that the FLX and FLX+GSN IEI distributions are overlapping.

To further verify that activating 5–HT3 receptors increases the amplitude and rate

of spontaneous synaptic activity in cortical PCs, we bath applied a 5–HT3

receptor–specific agonist, m–chlorophenylbiguanide (mCPBG) and recorded

sEPSCs. Indeed, administration of mCPBG resulted in a dramatic increase in the

amplitude and frequency of sEPSCs in cortical neurons relative to control

conditions (p < 0.01, Wilcoxon rank–sum test; Fig. 2.1e, f, g). The effect of

mCPBG on sEPSCs amplitude exceeded that of FLX (p < 0.01, Wilcoxon rank–

sum test), but the effect on frequency was comparable (p > 0.05, Wilcoxon rank–

sum test). One potential confound of these results is the possibility that

incomplete voltage–clamp across the somato–dendritic axis (i.e. incomplete

space–clamp) would result in contamination of sEPSCs with inhibitory GABAA

receptor–mediated postsynaptic currents. This scenario is plausible considering

that a large proportion of cortical inhibitory interneurons express 5–HT3Rs and

could be activated by mCPBG to trigger sIPSCs in PCs (Lee et al. 2010; Morales and Wang 2002; Vucurovic et al. 2010). To control for this possible confound, in addition to using a Cs–based internal solution that minimizes space–clamp effects, we recorded the sEPSCs elicited by mCPBG in the presence of gabazine

(GZN; 5 µM), a GABAA receptor–specific antagonist. If the sEPSCs recorded at

the reversal potential for inhibition are contaminated by GABAergic currents due

to incomplete space clamp, then GZN would lead to the abolition of these events.

mCPBG co–administered with GZN did not lead to a reduction in sEPSC

amplitude compared with mCPBG treatment alone but modestly increased the IEI

(data not shown), suggesting that though a small fraction of the recorded events

may arise from GABAergic currents escaping voltage–clamp, the majority of

these events result from 5–HT3 receptor signaling at the level of cortical PCs.

This finding is consistent with the idea that spontaneous synaptic release, or

synaptic noise, in the cortex is mediated in part by 5–HT signaling through 5–

HT3Rs and offers a valuable tool to explore how altered serotonergic signaling

affects network activity in cortical slices.

Augmenting endogenous 5–HT signaling transforms cortical network dynamics

via 5–HT2 and 5–HT3 receptors

What are the consequences of increasing endogenous 5–HT levels on

cortical network activity? To address this question, we employed a disinhibited

slice preparation (Gutnick et al. 1982; Hablitz 1984) using bath application of the

GABAA receptor antagonist, gabazine (GZN; 1 µM). In this paradigm, the

normally quiescent cortical network initiates bursts of activity in the form of

paroxysmal depolarizing shifts (PDS; Fig. 2.2a). The PDS reflects synchronous activation of local cortical networks, thus enabling recordings from single neurons to probe network activity. The voltage profile of a PDS under a current–clamp

recording configuration is characterized by a large and rapid depolarization (ca.

60–80 mV from baseline) lasting ca. 0.5 s followed by a tail that decays to baseline within 0.5–2 s (Fig. 2.2a). The power spectrum density of the voltage

traces (see Methods and Materials) revealed that PDS events occur randomly in

time, represented by a flat spectrum lacking frequency peaks as shown in Fig.

2.2g (Control case). Occurrence of these events can thus be considered a

stochastic process, reflecting a lack of temporal organization in cortical dynamics.

To determine the effect of enhanced endogenous 5–HT signaling on network

dynamics, we bath–applied FLX (4 µM) for the duration of the recordings and found that FLX dramatically increased the number of network bursts recorded in a

10 min interval relative to control conditions (p < 0.01, Wilcoxon rank–sum test;

Fig. 2.2b,f). Moreover, we show that enhanced serotonergic signaling transforms

the patterning of cortical activity from temporally sparse and random to clustered

and periodic network dynamics as described below (Fig. 2.2b). These clusters of

network activity have a voltage profile characterized by an initial long burst (~500

ms) comparable to the PDS seen in control conditions followed by a sequence of

shorter afterdischarges (duration ~40 ms) superposed on a long depolarizing

plateau. The intra–cluster bursts occurred with a preferred frequency in the low

beta range (10–16 Hz; Fig. 2.2b,g), while inter–cluster intervals showed no temporal patterning. The shift in the baseline of the power spectrum in Figure 2g corresponds to an increase in the total number of events, consistent with findings shown in Figure 2.2f. Note that the afterdischarges are of synaptic origin and persist in the absence of spiking when the recorded neuron is filled with QX–314 to block voltage–gated sodium channels. This pattern of activity has been observed previously in disinhibited rat cortex in vivo and slices in vitro (Castro-

Alamancos 2000; Castro-Alamancos and Rigas 2002), animal models of

spontaneous epileptic seizures (Timofeev et al. 1998), as well as in certain forms of childhood onset epilepsies (Camfield 2011) and has been termed “fast runs”, a designation we will employ here.

FLX increases serotonergic signaling by blocking presynaptic reuptake of

5–HT, thereby increasing transmitter availability for postsynaptic receptors. Thus, if the alteration of cortical network activity observed here is a bona fide serotonergic phenomenon, then some 5–HT receptor(s) must be responsible for the effect. Our observation that 5–HT3Rs generate a significant portion of spontaneous synaptic noise prompted us to ask whether this enhanced noise in the presence of FLX can account for the observed changes in cortical network activity. Co–application of GSN with FLX in disinhibited slices led to a significant reduction of the number of network events (p < 0.01, Wilcoxon rank– sum test; Fig. 2.2f), but did not eliminate the fast runs, the persistence of which is represented by the beta peak in the power spectrum (Fig. 2.2g). This finding suggested the involvement of another 5–HT receptor in the emergence of the oscillatory fast runs. 5–HT2Rs are densely expressed in the cortex (Santana et al.

2004) and mediate excitatory neuronal responses (Araneda and Andrade 1991;

Tanaka and North 1993), making them a suitable candidate. We blocked 5–

HT2Rs with ketanserin (KSN, 10 µM) in the presence of FLX and reduced the total number of network events back to control levels (control vs. KSN; p = 0.13;

Wilcoxon rank–sum test; Fig. 2.2f). Importantly, this led to complete attenuation

of fast runs represented by the elimination of the 10–16 Hz peak in the power

spectrum (Fig. 2.2g). This effect was tested with a gamut of 5–HT2 receptor

39 antagonists, which were all able to eliminate the fast runs albeit with different effects on network excitability. These included ritanserin and glemanserin to selectively antagonize 5–HT2a receptors, and RS–102,221 to antagonize 5–HT2c receptors. Receptor antagonists targeting 5–HT2a receptor subtypes were more effective at reducing network excitability than those targeting the 5–HT2c subtype, with ketanserin being the most effective (data not shown). We also tested the effect of 5–HT1 receptor (5–HT1R) activation on fast run activity owing to their dense expression in the cortex and their inhibitory effects on neuronal excitability via activation of K+ currents (Parks et al. 1998; Tanaka and North

1993). 5–HT1Rs were activated using the specific agonist 8–hydroxy–2–

(dihydropylamino)tetralin (8–OH–DPAT; 4 µm) in the presence of FLX. 8–OH–

DPAT treatment resulted in a pronounced reduction in the number of recorded network events relative to FLX treatment alone (p < 0.01, Wilcoxon rank–sum test; Fig. 2.2e, f), an effect statistically indistinguishable from that seen in the presence of FLX and GSN. It also led to an almost complete abolition of the 10–

15 Hz oscillation (Fig. 2.2e, g), consistent with a slowing of network bursting seen in the voltage profile in Figure 1e. In the presence of 8–OH–DPAT, the network bursts occurred either as single PDS events or as pairs of bursts like those shown in Figure 2e. These results suggest that 1) 5–HT1R activity exerts an inhibitory effect on cortical network activity by precluding the occurrence of fast run depolarizations and reducing network bursting; and that 2) in the presence of

FLX alone, activation of 5–HT1Rs by endogenous 5–HT release is insufficient to prevent the occurrence of fast runs. Thus, our findings show that elevating

endogenous 5–HT leads to an increase in cortical network excitability and the

emergence of epileptiform network oscillations. The former effect is mediated in

part by 5–HT3Rs and 5–HT2Rs while the latter effect is exclusively attributable

to activation of 5–HT2Rs. These observations provide a novel insight into how

altered serotonergic tone in the cortex leads to the transformation cortical network

dynamics.

Fluoxetine enhances excitatory synaptic inputs mediating cortical network

activity

The current–clamp experiments used in the previous section provide a

proxy for the network behavior at the level of neuronal output. Voltage–clamping,

on the other hand, offers insight into the magnitude, structure, and pharmacology

of the inputs arriving at cortical neurons during network activity. We sought to

determine the network mechanisms underlying the emergence of fast runs by

measuring the network–mediated synaptic inputs in L2/3 PCs during

disinhibition–induced activity in voltage–clamp before and after FLX treatment.

With GZN alone, excitatory network inputs manifested as massive currents

(nEPSCs; Fig. 2.3a) that were presumably dominated by AMPA receptor–

mediated currents (Lee and Hablitz 1991), since NMDA receptor contribution was eliminated by holding the cell at a command potential (–80 mV) at which

NMDA receptors are blocked by Mg2+ ions. Addition of FLX (4 µM) altered the

network inputs in the following way. As expected, each network event had two

phases corresponding to the two phases of the voltage profile in Figure 2.2b. The

first phase consisted of a large nEPSC lasting ca. 500 ms followed by the second

phase comprised of substantially smaller nEPSCs which occurred at the same

frequency as the fast runs seen in Fig. 2.2 (Fig. 2.3b). This latter phase has been

previously demonstrated to arise from recurrent glutamatergic synaptic

transmission, which is expressed strongly in the superficial cortical layers and can

be unmasked with sufficient disinhibition (Castro-Alamancos and Rigas 2002).

We found that the charge transferred to each cell per event, taken as the integral of the total synaptic current, was substantially increased (p < 0.05, Wilcoxon rank–sum test; Fig. 2.3c) after FLX treatment. This suggests that each cell received on average more synaptic excitation during a network event when FLX was present. Importantly, we compared the amplitude of the first nEPSC within each event between the two groups by setting a threshold that clearly separated

the first and second phases. Our results show that the first nEPSC within each

network event is significantly larger in the presence of FLX (p < 0.01, Wilcoxon

rank–sum test; Fig. 2.3d). Since this massive event acts as a trigger for the

subsequent 10–16 Hz afterdischarges, these findings suggest that a shift in favor

of more synaptic excitation during the initial burst leads to increased network

activity and the onset of reverberant oscillations within the cortical circuit. We

tested this hypothesis in the following section.

Enhanced excitatory coupling and synaptic noise are sufficient to simulate

fluoxetine–modulated network activity in a model of a cortical microcircuit

Are the increased synaptic noise and enhanced excitatory coupling observed in vitro sufficient to account for the emergence of fast runs, or is it secondary to some unknown biophysical process? This question poses a technical

problem since precise tuning of synaptic excitation and noise in the slice is

impractical. Therefore, we tested this question using a computational model of a

cortical microcircuit.

To this end we adopt a minimal modeling approach and focus on capturing

the essential biophysical properties of neurons leading to realistic cortical

dynamics. The inspiration for our model is taken from a previous computational

study (Parga and Abbott 2007) showing that the key feature of cortical network dynamics is the bistability of the neurons’ membrane potential, as explained below. Bistability can be attained by multiple combinations of the excitatory and inhibitory synaptic conductances (Grashow et al. 2010; Prinz et al. 2003) and hence, does not critically depend on their specific values.

For our purposes, those network models can (Parga and Abbott 2007) be simplified by noting that cortical microcircuits can to a good approximation be considered translation invariant, i.e. the connectivity pattern repeats itself across cortical columns. Fig. 2.4a depicts the connectivity for layer 2/3. Because PDS occur synchronously across PCs within at least 200 µm (data not shown) the study of the dynamics of the microcircuit can be reduced to its simplest structural motif, consisting of a PC and an interneuron (Fig. 2.4a, right). The inputs of the missing neighbors are compensated by proportionately scaling the recurrent excitation and inhibition so that the net synaptic currents are the same as they would be when embedded in the full circuit. When the membrane potential of the neurons fluctuates around the resting (low conductance) state, the dominant current is the leak, as shown in Fig. 2.4b (straight dashed line). During the early phase of the

PDS, the current driving the membrane potential is dominated by the AMPA,

NMDA and to a lesser degree by GABAA components (high–conductance state; early phase). The dependence of the current on the membrane potential is now given by the solid curved line which crosses IC/0= at three points. The two

crossings with a positive slope denote transiently stable points of the membrane

potential, endowing the network with bistable dynamics (i.e. low/high

conductance states). In the late phase of the PDS inhibition becomes more

dominant and the membrane potential decays slowly (high–conductance state; late

phase). The dependence of the current on the membrane potential is now given by

the dashed curved line with only one stable point close to the resting potential,

driving the neurons to repolarize quickly. It is important to note that in the

absence of synaptic noise the neurons would never develop PDS. Noise is

necessary to make the transition to a depolarized state. Once in the depolarized

state, the recruitment of inhibition and decay of excitation bring the membrane

potential back to the resting state. Figure 2.4c shows the connections of the motif

in the control case as well as its dynamics. Figure 2.4d depicts the motif with

increased synaptic excitation, mimicking the effect of FLX shown in Figure 2b.

The activity pattern now switches into a fast run mode in which network

activation occurs repetitively at ∼15 Hz. How does the increase in synaptic

excitation make a qualitative change in the dynamics? Figure 2.4b shows that

while the membrane potential decays, the dependence of the current on the

membrane potential smoothly morphs from the solid curved line to the dashed

curved line, mainly by moving upwards. In this process, the membrane potential

44 switches from a bistable regime to a monostable regime. If the synaptic noise is strong enough on the verge of this transition, it will shift the curve back down, bringing the membrane potential again to a depolarized state. The slower the decay of the membrane potential, the more likely it is that synaptic noise will depolarize it again. By this mechanism, increased synaptic noise creates reverberant fast run oscillations within circuit.

Figure 2.4g shows that the model replicates the experimental results shown in Fig.

2.2f. The first group corresponds to control conditions in which synaptic noise and excitatory coupling are kept at basal conditions. Network events within this group manifest as single PDS–like bursts (Fig. 2.4c). The second group mimics the effects of FLX: an increase in synaptic noise mediated by 5–HT3Rs and an increase in synaptic excitation by 5–HT2Rs. As in vitro, the total number of network events increases significantly relative to control (p < 0.01; Wilcoxon rank–sum test) and so does the variability in network excitability as seen by a substantially larger spread in the distribution. Importantly, under these conditions the network events appear as fast run–like oscillatory bursts with a frequency of

~15 Hz. A third group models the effect of FLX in the presence of GSN, or equivalently the effect of increasing synaptic excitation with only a marginal increase in synaptic noise. Clearly, the reduction in synaptic noise alone reduces the excitability of the network but does not bring it back to the control level nor does it abolish fast run–like oscillations.

To further validate the model’s ability in predicting the effects of pharmacological manipulations presented in Figure 2.2, we modeled the effects of

5–HT1R activation with 8–OH–DPAT, which in the slice had a combined effect of reducing network excitability and abolishing fast runs. This effect is consistent with its effect on neuronal excitability (Araneda and Andrade 1991), which results to a large extent from an increase in potassium conductance through G–protein– coupled inwardly rectifying potassium (GIRK) channels that contribute to the resting leak conductance (Luscher et al. 1997) . We thus simulated the postsynaptic effect of 5–HT1R activation in the presence of FLX by increasing the leak conductance in the excitatory neuron within the model while maintaining high levels of synaptic excitation and noise. This manipulation was able to mimic the effect of 8–OH–DPAT in the slice in two ways: Firstly, the number of network events dropped substantially relative to the group with increased synaptic excitation and noise alone (p < 0.01, Wilcoxon rank–sum test; Fig. 2.4g). This reduction in number of events was statistically indistinguishable from the one observed in the group simulating effects of FLX and GSN (p = 0.14, Wilcoxon rank–sum test), reproducing the results observed in vitro. Secondly, the fast run oscillation was replaced by single bursts or burst pairs, similar to those recorded in the slice (Fig. 2.4e), suggesting that a 5–HT1R–mediated enhancement of leak current is capable of suppressing network activity and epileptiform network dynamics.

5–HT2Rs are also known to inhibit outward potassium current mediating the slow afterhyperpolarization (IsAHP) in cortical neurons (Satake et al. 2008;

Villalobos et al. 2005; Villalobos et al. 2011). To test whether inhibition of IsAHP in the model alone can simulate the effects of FLX observed in the slice, we

reduced this current while keeping control levels of synaptic excitation. Reduction

of the IsAHP generated the fast runs oscillation (Fig. 2.4f) and increased the number

of network bursts to levels observed with increased synaptic excitation and noise

(i.e. FLX group)(p = 0.87 ; Wilcoxon rank–sum test). These results suggest that a

parallel mechanism mediated by 5–HT2R–dependent inhibition of IsAHP could

potentially generate epileptiform fast runs similar to those seen under conditions

of increased excitatory coupling. Whether FLX actually decreases IsAHP in the in vitro preparation remains an open question.

Combined together these results support our hypothesis that increased synaptic noise and excitatory coupling resulting from altered 5–HT signaling are sufficient to account for the increased network excitability and emergence of epileptiform fast run oscillations. The synaptic noise increases the likelihood of transitioning into a depolarized high–conductance state, while the enhanced synaptic excitation facilitates the emergence of oscillatory dynamics.

Furthermore, the model predicts the effects of 5–HT1R agonist 8–OH–DPAT by manipulating one of its known targets, namely the potassium current controlling the resting membrane potential, and provides a parallel mechanism by which fast run oscillations can be generated in the presence of elevated 5–HT tone, namely

5–HT2R–mediated inhibition of the slow afterhyperpolarization current. These results underscore the utility of reduced models of cortical circuits, which in our case is capable of reproducing the in vitro results and making accurate predictions about network behavior under various pharmacological manipulations without fine–tuning the model’s parameters.

In vivo blockade of 5–HT2R activity delays behavioral and electrographic seizure onset and reduces seizure incidence

Fast runs are not an epiphenomenon of the cortical slice preparation, but have been observed during spontaneous epileptic seizures in cats (Steriade et al.

1998; Timofeev et al. 1998) and are a hallmark of EEG recordings from children with Lennox–Gastaut syndrome, a childhood–onset encephalopathy (Camfield

2011). Based on the in vitro efficacy of KSN in abolishing fast runs, we hypothesized that blocking 5–HT2Rs before convulsant–induced seizures in mice may increase the threshold to epileptic seizures or modulate other seizure parameters. To test this hypothesis, we induced seizures in mice with the convulsant, pentetrazol (PTZ; 80 mg/kg, i.p.), and measured the latency to the first seizure characterized by loss of righting and generalized clonic convulsions

(LOR–GCC; Fig. 2.5a) as well as the duration of each seizure. Consistent with our hypothesis, mice that received an injection of KSN (10 mg/kg; i.p.) one hour prior to PTZ injection showed a more than five–fold increase in latency to the first

LOR–GCC seizure (p < 0.01, Wilcoxon rank–sum test; Fig. 2.5b) compared to mice injected with the vehicle control. The duration of seizures was not significantly different (p = 0.59; Wilcoxon rank–sum test; Fig. 2.5c) suggesting that blockade of 5–HT2Rs in vivo raises the threshold to seizure onset but does not affect its termination. Furthermore, KSN treatment reduced seizure incidence in mice (control: 100% seized, n = 17; KSN: 78% seized, n = 18; p = 0.058,

Fisher Exact Test).

The acute behavioral seizures accompanied by loss of righting and

generalized clonic convulsions are an indirect proxy for the oscillatory brain

dynamics that beget them. To directly test whether fast run oscillations occur

during acute seizures in mice treated with PTZ and whether the behaviorally

measured seizure latencies have a correspondence to the onset of such oscillatory

regimes in the cortex we implanted 12 mice with subdural electrodes over the

surface of somatosensory cortex and recorded the cortical electroencephalogram

(EEG) during seizures with and without pre–treatment with KSN. Figure 2.5d shows raw EEG recordings from individual mice pre–treated with the drug vehicle alone (left) or with KSN (right). As visible from both traces, the onset of cortical seizures corresponds to a massive increase in activity in the EEG signal relative to the preceding baseline. The two traces also show that the onset of electrographic seizures in the mouse pre–treated with KSN is substantially delayed (latency = 1765 s) as compared with the control mouse (latency = 71 s).

We filtered the EEG traces in the pass band of 8–16 Hz to capture the fast run oscillations in the beta band as done previously by others (Steriade et al. 1998) and then determined the latency to the emergence of fast runs using a simple threshold (see Materials and Methods). To ensure that the latencies to behavioral seizures are an accurate proxy for the emergence of electrographic fast runs, we compared the latencies to the behavioral seizures with the latency to the onset of

8–16 Hz oscillations in the EEG. Figure 2.5e shows a very close and highly significant pairwise correlation between behaviorally and electrographically measured seizure latencies (R = 0.9967, p < 0.01, Student’s t–test). We thus

49 proceeded to compare the latency to the onset of cortical fast runs between control and KSN–treated animals and observed a significant delay in the onset of fast runs after pre–treatment with KSN (vehicle: n = 5, KSN: n = 7; p < 0.05,

Wilcoxon rank–sum test; Fig. 2.5f), a finding consistent with the those shown in in Fig. 2.5b.

Thus, just as 5–HT2R blockade prevents the emergence of epileptiform fast runs in vitro, it is also able to prevent or delay the onset of epileptic fast runs in vivo, lending support to the idea that serotonergic signaling in a disinhibited cortex facilitates the emergence of epileptiform activity and providing evidence across synaptic, network, and behavioral levels of observation.

Discussion

Despite decades of investigation on the role of 5–HT in neuronal excitability, its effects on dynamics of cortical networks remain unclear (Celada et al. 2013). We sought to further this understanding by determining how changes in endogenous 5–HT affect cortical network dynamics. Our findings demonstrate that cortical neurons are persistently bombarded with spontaneous excitatory synaptic inputs mediated in part by 5–HT3 receptors. This synaptic noise can be augmented with the SSRI, fluoxetine, at concentrations close to those measured in medicated human patients (Bolo et al. 2000). This results in enhanced network excitability and the transformation from temporally random to transiently oscillatory network dynamics, i.e. from PDS to fast runs, a transition which is facilitated by 5–HT2Rs. We show that in cortical slices such a shift is accounted for by increased excitatory coupling between pyramidal cells. A computational

model corroborates that increased synaptic noise and synaptic excitation are

sufficient to explain the increased network excitability and transitions between

both network regimes. Consistent with these results, blocking 5–HT2 receptors in

vivo prior to PTZ–induced convulsions significantly delays the onset of epileptic

seizures and reduces seizure incidence. Together, our findings provide a potential

mechanism for understanding how altered 5–HT tone can bias cortical network activity to transition into an epileptiform state.

The prevailing view of spontaneous excitatory neurotransmitter release onto cortical neurons held glutamate as its chief mediator (Espinosa and Kavalali 2009;

Maffei and Turrigiano 2008; Sutton et al. 2004; Vyleta and Smith 2011), overlooking the contribution of other neurotransmitters (Pankratov et al. 2007;

Roerig et al. 1997). Our findings show that 5–HT3Rs contribute substantially to spontaneous activity in the form of background synaptic noise, providing cortical

PCs with a considerable excitatory tone. What is the source of 5–HT in the slice preparation? While the cortical slice lacks the dorsal raphe nucleus containing the forebrain–projecting serotonergic neurons, it preserves their dense projections to cortical neurons. These remaining axons are the putative source of endogenous 5–

HT in our preparation and, based on results presented in Figure 2.1, support spontaneous vesicle release.

It is important to mention that our results from Figure 2.1 are in conflict with the presently held assumption that 5–HT3 receptors are expressed exclusively on cortical inhibitory interneurons. The evidence for this assumption comes from several studies that definitively show robust 5–HT3 receptor

expression on cortical interneurons (Jakab and Goldman-Rakic 2000; Lee et al.

2010; Morales and Wang 2002; Puig et al. 2004; Vucurovic et al. 2010) but fail to detect expression of these receptors on the principal pyramidal neurons. While it is undeniable that a subclass of cortical interneurons expresses 5–HT3 receptors, we posit that the evidence is inconclusive pertaining to their expression on pyramidal cells. We first draw attention to the study by Puig et al. (Puig et al.

2004) which used in situ hybridization and immunohistochemistry to assess 5–

HT3 receptor expression in GABAergic interneurons of the rat prefrontal cortex.

They report in their results that a minority of 5–HT3 receptor–expressing cells did not express GAD mRNA, a marker of inhibitory neurons and reported unpublished results in which some of these cells expressed vGluT1 mRNA, a marker of glutamatergic neurons. The authors state that they “[…] cannot exclude that a minority of 5–HT3 receptor–positive cells are pyramidal neurons.” These results are consistent with those of Roerig & Katz who were the first to show spontaneous and evoked 5–HT3 receptor–dependent currents on pyramidal cells of the ferret cortex and carried out post–hoc analysis of neuronal morphology to verify that both pyramidal and non–pyramidal neurons functionally expressed these currents (Roerig et al. 1997). A study by Sung et al. used a transgenic mouse line in which 5–HT3 receptor expression was driven in forebrain pyramidal neurons using the CamKII promoter to investigate the modulation of 5–HT3 receptors by alcohol (Sung et al. 2000). Using patch clamp recordings, they first report no 5–HT3 receptor–dependent currents in most principal cells from wild– type animals; however, they mention in the results that 7 out of 48 recorded cells

in wild–type mice showed evoked 5–HT3 receptor–mediated currents. That is,

15% of the cells which presumably should not express 5–HT3 receptors do, thus,

providing supporting evidence for expression of functional 5–HT3 receptors on

cortical pyramidal neurons.

Some of the studies addressing the molecular expression of 5–HT3

receptors in the cortex, such is that of Lee et al., have employed transgenic mouse

lines that express enhanced green fluorescent protein (eGFP) in 5–HT3a receptor–

expressing neurons and were able to demonstrate eGFP expression that was

confined to cortical neurons expressing GAD, but not Satb2, a marker of cortical

pyramidal cells (Lee et al. 2010). A similar finding was reported by Vucurovic

and colleagues (Vucurovic et al. 2010). A critical caveat in these studies pertains

to the subunit composition of 5–HT3 receptors of which there are two types,

5HT3a and 5–HT3b. 5HT3 receptors can exist as homomers composed of only 5–

HT3a subunits (Brown et al. 1998) or as heteromers containing both subunit types

(Davies et al. 1999; Kelley et al. 2003). It has been postulated based on previous work (Morales and Wang 2002; Sudweeks et al. 2002) that the heteromeric form of the receptor was not expressed in the CNS. However, a follow–up study by

Doucet et al. showed cortical expression of the 5–HT3b subunit using a

polyclonal antibody targeted to the intracytoplasmic loop of the 5–HT3 receptor

and demonstrated that this expression differed considerably from localization of

5–HT3a receptors in previous studies (Doucet et al. 2007). The findings of Doucet

and colleauges underscore two important points, namely, that methodological

scope of previous studies investigating 5–HT3 receptor expression was too

limited to detect 5–HT3b–containing serotonin receptors. Secondly, the lack of

the expected anatomical overlap between 5–HT3b and 5–HT3a receptor subunits

in their findings suggests that the previous studies may have inadvertently

overlooked a population of heteromeric 5–HT3 receptor–containing cortical neurons. These findings are consistent with another study that supports the existence of heteromeric 5–HT3 receptors in rodent neurons (Hanna et al. 2000).

Lastly, we note that early studies that confirmed 5–HT3 receptor expression on cortical interneurons (Jakab and Goldman-Rakic 2000; Morales and Wang 2002)

used in situ hybridization and immunohistochemistry to find overlap between

markers of inhibitory interneurons and 5–HT3 receptors, but did not rigorously

investigate overlap with any markers of pyramidal neurons, thus, limiting the

reach of their conclusions that 5–HT3 receptors are expressed exclusively on

interneurons. Given these caveats, we hold that previous studies leave the

possibility open that there is a subset of 5–HT3 receptors, likely to be

heteromeric, expressed on cortical pyramidal cells. Our findings in Figure 2.1

provide another compelling piece of evidence in support of this idea.

An important finding reported herein is that the synaptic noise generated

by 5–HT enhances cortical network activity. That noise can enhance network

activity may seem counterintuitive, yet it has been explored thoroughly in

numerous nonlinear dynamical systems, including neurons (McDonnell and Ward

2011; Stacey and Durand 2001). Importantly, synaptic noise can not only enhance

network activity but also network performance, specifically in the context of

encoding information (Ermentrout et al. 2008; Puzerey and Galán 2014). In this

context, stochastic resonance is probably the clearest example, in which the

detection of subthreshold periodic inputs to a nonlinear system can be enhanced

with the introduction of noise, but it has been expanded to encompass a variety of

dynamic phenomena in which noise facilitates transitions between distinct system

states under a broader concept of stochastic facilitation (McDonnell and Ward

2011). A novel finding of our study is that synaptic noise provided by 5–HT3Rs

promotes transitions in a bistable network moving between quiescent and active

states. This finding is consistent with theoretical work implicating synaptic noise

in the generation of epileptiform oscillations (Stacey et al. 2009). Whether other sources of noise modulate network dynamics remains an open question. We note, for instance, that previous work has shown that 5–HT enhances local glutamatergic transmission in recurrent cortical circuits (Aghajanian and Marek

1999; Beique et al. 2004; Beique et al. 2007). Thus, determining the contribution of other neurotransmitters to synaptic noise and network activity will be useful in determining if stochastic facilitation occurs as a general phenomenon of fast neurotransmission.

The notion that neuromodulation plays an active role in transforming network behavior has been explored thoroughly in the invertebrate central nervous system (CNS), particularly the crustacean stomatogastric system (Marder and Thirumalai 2002). Consequently, several fundamental principles governing network function were brought to light. An indispensable principle to our understanding was that networks, while constrained by the anatomical architecture, exhibit highly dynamic functional connectivity (Getting 1989;

Marder and Thirumalai 2002). In short, the concerted behavior of neuronal networks operating within a defined anatomical circuit can be reconfigured with neuromodulation through changes in cellular, synaptic, or network properties. Our work shows that elevating endogenous cortical 5–HT leads to enhanced signaling through 5–HT2Rs, resulting in increased excitatory synaptic transmission during network activity and the emergence of fast run oscillations. Though previous reports have demonstrated that endogenous 5–HT may modulate network oscillations (Celada et al. 2008; Puig et al. 2010), our work reveals the role of 5–

HT as facilitating the emergence of distinct cortical dynamics. While we provide a network mechanism for the emergence of epileptiform fast runs under elevated

5–HT tone, the cellular cascades downstream of 5–HT2R activation responsible for the enhancement of synaptic excitation require elucidation. 5–HT2Rs have distinct downstream targets in cortical neurons including NMDA receptors

(Rahman and Neuman 1993), GABAA receptors (Huidobro-Toro et al. 1996),

Ca2+–activated K+ channels (Villalobos et al. 2005), and M–channels (Tanaka and

North 1993). Determining the targets responsible for enhanced network responses will be important in understanding how changes in cellular properties can lead to transformation of network activity. One could then use computational models to test the robustness of the network dynamics to changes in these parameters and determine at which parameter boundaries transitions occur between states. It may also be possible that correlated variation in synaptic or membrane conductances may yield similar network behavior (Grashow et al. 2010; Marder 2011; Prinz et al. 2003). Given the prediction made by our computational model, the potassium

currents underlying the slow afterhyperpolarization may be a likely effector

downstream of 5–HT2 receptors underlying generation of fast runs, although the

enhancement of excitatory synaptic coupling in vitro suggests that NMDA

receptors may play a role as well.

Previous reports have suggested that FLX suppresses intrinsic cortical

microcircuits in slices of non–epileptic human prefrontal cortex (PFC) (Komlosi

et al. 2012). While this finding is certainly relevant for understanding how FLX

affects sparsely active cortical microcircuits, it stands in contrast to our findings

that FLX enhances the activity of a partially disinhibited cortical network. This

discrepancy may be attributable to several factors including: 1) the use of a

disinhibited slice in which the release of endogenous 5–HT is likely to be much

higher during active states; 2) using somatosensory instead of PFC slices, which

have different 5–HT receptor densities (Blue et al. 1988; Weber and Andrade

2010); and 3) comparing effects of FLX on sparse versus global and synchronized

cortical activation, the relationship between which is ambiguous. Another

important consideration is that elevating endogenous 5–HT is one of several ways

in which the balance between excitation and inhibition can be shifted to elicit fast

run oscillations. For instance, Castro–Alamancos and Rigas have shown that

partial blockade of GABAA receptors in cortical slices is sufficient to elicit PDS

but not fast run activity, while co–administration with GABAB antagonists generates fast runs (Castro-Alamancos and Rigas 2002), which, as the authors

argued, resulted from the unmasking of recurrent excitatory synaptic activity in

the superficial cortical layers. In support of this idea, we posit that elevated

cortical 5–HT enhances synaptic excitation during network activity, thereby

promoting the expression of fast run oscillations.

Our study focused exclusively on changes in excitatory drive to the

pyramidal cells of the cortex under conditions where intrinsic cortical inhibition is

altered. Therefore, our interpretation of the results above must be confined to the

sphere of cortical function under conditions in which inhibition is diminished,

such as those in which epileptiform activity emanates from such an imbalance.

Nevertheless, it is important to place our results in the context of the present

literature of 5–HT function on cortical circuits in which inhibition is intact.

Though there are several comprehensive reviews on this topic (Andrade 2011;

Celada et al. 2013; Nakamura and Wong-Lin 2014), we will briefly discuss the main principles gathered from past investigations. It has become increasingly clear that making general claims about 5–HT’s role in cortical circuits is far from trivial. This difficulty arises from the heterogeneity in regional, laminar, cellular, and subcellular serotonin receptor expression (Amargos-Bosch et al. 2004; Blue et al. 1988; Jakab and Goldman-Rakic 2000; Santana et al. 2004; Weber and

Andrade 2010). It has been understood, however, that the most frequent responses of cortical neurons to 5–HT are 5–HT1 receptor–mediated hyperpolarizations and

5–HT2 receptor–mediated slow depolarizations (Araneda and Andrade 1991;

Tanaka and North 1993) The former effect is mediated by activation of outward potassium currents mediated by G–protein coupled inwardly–rectifying potassium

(GIRK) channels (Luscher et al. 1997) while the latter is attributed to inhibition of a slow outward M–type potassium current (Tanaka and North 1993). 5–HT1

receptors are typically expressed at the axon hillock where they are poised to

regulate neuronal output, while 5–HT2 receptors are expressed along the apical

dendrite where they can modulate neuronal gain in response to synaptic inputs

(Amargos-Bosch et al. 2004; Zhang and Arsenault 2005). Often neurons

expressing both receptors can have either facilitation or suppression of firing

depending on the site of stimulation in the DRN, the afferent source of

serotonergic fibers innervating the forebrain, suggesting that precise circuits

underlie the direction of the cortical response to 5–HT (Amargos-Bosch et al.

2004). The actions of serotonin receptors extend beyond modulation of cell–

intrinsic properties to the level of synaptic interactions. It is by now well

established that 5–HT2 receptors enhance spontaneous postsynaptic glutamatergic

currents in cortical neurons, a process attributed to the enhancement of activity in

intrinsic cortical circuits by 5–HT (Beique et al. 2007; Lambe et al. 2000). 5–HT

also modulates the activity of inhibitory interneurons via excitation by 5–HT3

receptors and 5–HT2 receptors (Puig et al. 2004; Zhou and Hablitz 1999), though interneurons expressing these two receptors may have distinct molecular, anatomical, and functional profiles. Activation of 5–HT2 receptors on interneurons leads to sustained enhancement of spontaneous inhibitory postsynaptic currents onto pyramidal neurons, while activation of cortical interneurons through 5–HT3 receptors leads to transient increases of inhibitory currents on pyramidal cells (Zhou and Hablitz 1999). The action of 5–HT goes as far as modulating the concerted activities of ensembles of cortical neurons. Puig et al. showed that stimulating the DRN (and recording local field potentials (LFP)

in the PFC of anesthetized rodents) corresponded to acceleration of the cortical slow oscillation (< 2 Hz) in a 5–HT2 receptor–dependent manner (Puig et al.

2010). They also demonstrated, using the same paradigm, that inhibition of 5–

HT1 receptors increased cortical gamma (30–80 Hz) oscillations while 5–HT2 receptor inhibition decreased their power. Our results agree with this work in that our experiments show a facilitating effect of 5–HT2 receptors and a suppressive effect of 5–HT1 receptors on cortical network activity. Nevertheless, it is important to draw the distinction between the role of 5–HT as a modulator of cortical dynamics as shown by Puig et al. and the role of 5–HT as a facilitator to the emergence of said dynamics. Furthermore, given the strict focus on the effects of 5–HT on pyramidal neurons in our study, it will be important to determine the role of inhibitory cortical interneurons in the emergence of oscillatory cortical dynamics and how their activity is shaped by altered 5–HT tone.

The transition between PDS and fast run activity is not an artifact of the disinhibited slice. The PDS is a direct intracellular correlate of the inter–ictal spike that often precedes epileptic seizures and is the most basic unit of epileptiform activity (McCormick and Contreras 2001). Fast runs have been demonstrated in disinhibited rat cortex (Castro-Alamancos 2000), during epileptic seizures in cats (Timofeev et al. 1998), and are an electrographic signature of epileptic seizures in children with Lennox–Gastaut syndrome, a childhood–onset encephalopathy which comprises ~10% of all childhood epilepsies (Camfield

2011). Therefore, taking into consideration that 5–HT2R blockade abolishes fast runs in vitro, increases seizure threshold and delays the onset of electrographic

fast runs in vivo lends support to the idea that excessive 5–HT in the CNS

facilitates the onset of epileptiform network dynamics. These findings are in line

with reports supporting 5–HT’s pro–convulsive effects (Bercovici et al. 2006;

Freitas et al. 2006; O'Dell et al. 2000a; Wada et al. 1992), but stands in opposition to others who claim that central 5–HT is anti–convulsive (Bagdy et al. 2007; Yan et al. 1995). This controversy can be attributed to the lack of consistent seizure paradigms, animal models, or pharmacological agents employed by different investigators (Loscher et al. 1990). Importantly, the studies mentioned above lack a detailed mechanism for how 5–HT may modulate epileptic seizures. Our results clearly support a pro–convulsive role for enhanced 5–HT signaling by providing a mechanistic description for how changes in synaptic noise and excitatory coupling can lead to epileptiform network dynamics. An important subtlety in the interpretation of our in vitro findings is that 5–HT2R activation alone is insufficient to generate fast runs – the cortex must be partially disinhibited. In other words, the cortical network must be close to the border of a transition between two states (i.e. PDS and fast runs) near which a shift in favor of more excitation is sufficient to push the network into an epileptiform regime. This subtlety highlights the role of 5–HT not as an initiator nor as a modulator of epileptiform oscillations, but as a facilitator for their emergence. In the context of human epileptic seizures, our findings imply that 5–HT2R blockade may prove to be therapeutic for epileptic patients in general and, in particular, for those treated with SSRIs, which are often prescribed to patients comorbid for epilepsy and depression (Kanner 2003). Though the association between SSRIs and epileptic

seizures in humans remains anecdotal, largely correlative, and confounded by

factors such as co–administration of SSRIs with anti–epileptic drugs (Braitberg

and Curry 1995; Gross et al. 1998; Pisani et al. 1999; Prasher 1993; Thome-Souza

et al. 2007), there is substantial evidence for a pro–convulsive role of fluoxetine

in animal models of epileptic seizures and epilepsy (Ferrero et al. 2005; Macedo et al. 2004; Morita et al. 2005; Raju et al. 1999; Zienowicz et al. 2005).

Furthermore, our findings may be relevant for other neurological disorders in which 5–HT signaling and network dynamics are altered such as in certain cases

of autism (Veenstra-VanderWeele and Blakely 2012) and schizophrenia

(Aghajanian and Marek 1999).

Figure 2.1. Spontaneous excitatory transmitter release is mediated in part by 5–

HT3 receptors in mouse neocortex

(a) Images of biocytin–labeled pyramidal neurons in L2/3 of mouse somatosensory cortex. Right panel shows neurons under 20× magnification with layer boundaries delineated by opaque white lines. Cortical laminae are labeled accordingly. Left panel shows neurons under 40× magnification with a clear view of the apical dendrites projecting to L1, basal dendrites projecting laterally, and, in some neurons, the axon is visibly projecting towards the deeper layers. (b)

Raw traces of sEPSCs recorded under three conditions: 1) control (no drugs); 2)

10 µM NBQX (AMPAR blockade); and 3) 10 µM NBQX with 1 µM GSN

(AMPAR and 5–HT3R blockade). (c) Cumulative distribution function of sEPSC amplitudes under three experimental conditions (control: n = 28; NBQX: n = 27;

NBQX+GSN: n = 25). (d) Cumulative distribution function of sEPSC inter–event intervals (IEIs) for the data presented in b. (e) Raw traces of sEPSCs recorded under: 1) control conditions; 2) 4 µM FLX; 3) 4 µM FLX with 1 µM GSN; and 5

µM mCPBG. (f) Cumulative distribution function of sEPSC amplitudes under the described experimental conditions (control: n = 28; FLX: n = 24; FLX+GSN: n =

29; mCPBG: n = 22). (g) Cumulative distribution function of sEPSC inter–event intervals for the data presented in e. Note that the FLX and FLX+GSN distributions are overlapping.

Figure 2.2. Elevated endogenous 5–HT in cortical slices enhances network excitability and transforms network dynamics

(a–e) Raw traces of network activity recorded over 10 minutes from L2/3 PC in a disinhibited cortical slice under (a) control conditions (GZN only), (b) with FLX,

DPAT. Top trace represents entire recording and bottom trace zooms in on a single network event. Note that in a and d network activity manifests as a single

PDS, while in b and c it appears as oscillatory fast runs. (f) Box and whisker plots of the number of network bursts observed per recording for the four groups

(control: n = 30; FLX: n = 31; FLX+GSN: n = 31; FLX+KSN: n = 29; FLX+8–

OH–DPAT: n = 27). Asterisks above distributions denote level of significance (*

= p < 0.05; ** = p < 0.01, n.s. = not significant). (g) Power spectrum obtained from recordings of network activity exemplified in a–e. Baseline of power spectrum corresponds to the baseline level of network activity in each group.

Figure 2.3. Enhanced synaptic excitation onto cortical neurons underlies transformation of network dynamics in the presence of fluoxetine.

(a–b) Raw traces of nEPSCs recorded over 10 min interval under (a) control conditions and (b) with FLX. Note, traces in (a) and (b) were obtained from different slices. Top trace represents entire recording and bottom trace zooms in on a single network event. (c) Box plot of excitatory charge transferred to cortical

PCs during each network event under control and FLX conditions (control: n =

28; FLX: n = 26). (d) Cumulative distribution functions of the amplitude of the initial nEPSC during each network event under control and FLX conditions.

Figure. 2.4. Computational model of a cortical network accounts for the

emergence of fast runs with increased synaptic noise and excitatory coupling.

(a) Cortical microcircuit of layer 2/3 and reduction to its core motif. Gray triangles represent cortical pyramidal neurons while gray circles represent cortical interneurons. Excitatory connections are denoted by arrow heads and inhibitory connections by small circles. (b) Dependence of the total current on the membrane potential at the three phases of the PDS. The straight dashed line corresponds to the low conductance state preceding a PDS. During this state, the membrane potential dynamics are dictated by the leak conductance. The curved solid line corresponds to the early phase of the PDS and is equivalent to the high– conductance state during which synaptic conductances dominate the membrane potential dynamics. The curved dashed line corresponds to the late phase of the PDS during which the slow inhibitory conductance competes with fast excitation to repolarize the membrane potential (c) PDS generated by the model. (d) PDS turn into fast runs when synaptic excitation is increased. (e) Increasing the leak current in the presence of enhanced synaptic excitation prevents generation of fast runs, recapitulating in vitro results with FLX + 8–OH–DPAT. (f) Decreasing the slow afterhyperpolarizing current generates fast run oscillations (g) Box plot of the number of network events under control conditions (control; n = 30 trials), under increased synaptic noise and synaptic excitation (FLX; n = 30 trials), under increased synaptic excitation and moderate synaptic noise (FLX+GSN; n = 30 trials), under increased synaptic excitation and enhanced leak current (FLX+DPAT; n = 30 trials), and under decreased slow afterhyperpolarizing current (n = 30 trials). The model replicates the statistics of network bursts from the experiments in Fig. 2.2f.

Figure 2.5. 5–HT2 receptor blockade increases seizure threshold in vivo.

(a) Schematic of experimental design for convulsant–induced seizure protocol.

Control mice are given PTZ injection alone, while the test group is given the KSN

1 hour prior to PTZ injection. Lightning bolts represent appearance of

spontaneous LOR–GCC seizures some time after PTZ injection. (b) Box plot of latency to first seizure in control and KSN conditions (control: n = 17; KSN: n =

14) (c) Box plot of mean seizure duration for each animal in control and KSN groups. (d) Raw traces of electroencephalographic (EEG) recordings of epileptic seizures obtained from mouse treated with PTZ alone (left) and those pre–treated with KSN one hour before PTZ injection (right). t = 0 corresponds to the time of

PTZ injection. (e) Comparison of latency to the onset of behavioral and electrographic seizures. Dashed gray line corresponds to the identity line (x = y) and solid black line corresponds to the linear fit of the two variables (n = 12 mice). (f) Box plot of latency to the onset of 8–16 Hz (beta band) activity in the

EEG during electrographic seizures in control and KSN–treated mice (control: n =

5 mice; KSN = 7 mice).

Table 2.1. List of parameters and their values used in the generation of

computational model described in Fig. 2.4. Subscripted acronyms denote the

receptor or ion–channel type for the conductance (g), time constant (τ), or reversal

potential (E); AMPA, DL–α–amino–3–hydroxy–5–methylisoxazole–propionic

acid; NMDA, N–methyl–D–aspartate; AHP, afterhyperpolarizing; C, membrane capacitance; gL, leak conductance; θ, voltage threshold for the presynaptic neuron to release neurotransmitter; λ, rates of the synaptic noise. Arrows indicate the change of parameters used for Fig. 2.4G

Chapter 3: Constitutive deletion of Pet–1 leads to altered cell–intrinsic, synaptic, and network excitability in mouse cortex

Summary

Neurons originating from the raphe nuclei of the brainstem are the exclusive source of serotonin (5–HT) to the cortex. Their serotonergic phenotype is specified by the transcriptional regulator Pet–1, which is also necessary for maintaining their neurotransmitter identity across development. Transgenic mice in which Pet–1 is genetically ablated show a dramatic reduction (~80%) in forebrain 5–HT levels, yet no investigations have been carried out to assess the impact of such severe 5–HT depletion on the function of target cortical neurons.

Using whole–cell patch clamp methods, neuronal reconstructions, and animal behavior, we investigated the impact of 5–HT depletion on cortical cell–intrinsic and network excitability. We report significant changes in several parameters for cell intrinsic excitability in cortical pyramidal cells (PC). In addition, we show that cortical PCs in mice lacking Pet–1 exhibit increased spontaneous synaptic signaling through 5–HT3 receptors. Furthermore, we demonstrate that changes in synaptic excitability are correlated with an increase in cortical network excitability and oscillatory activity in a 5–HT2 receptor–dependent manner. We suggest that the change in cortical network activity may be due to previously reported hypersensitivity of cortical 5–HT2 receptors in mutant mice. To determine whether the increase in network excitability of Pet–/– mice corresponds

to an increased susceptibility to epileptic seizures, we carried out acute

convulsant–induced seizure experiments. Surprisingly, no significant differences

were observed in seizure parameters between mice lacking Pet–1 and their wild–

type littermates. Our findings provide the first evidence for changes in neuronal

and network excitability in target structures of serotonergic neurons in mice

lacking Pet–1.

Introduction

The monoamine serotonin (5–hydroxytryptamine; 5–HT) is a functionally

diverse molecule that plays a critical role within the central nervous system (CNS)

as a physiological signal in cell–to–cell communication (Nichols and Nichols

2008) and as a developmental signal guiding the patterning of distinct forebrain

structures (Gaspar et al. 2003). A subset of midbrain neurons located within the

dorsal and median raphe nuclei distribute 5–HT–containing axonal terminals

throughout the entire extent of the forebrain (Tork 1990; Waterhouse et al. 1986).

The cortex, in particular, receives dense innervation via the medial forebrain

bundle fibers that originate from the raphe nuclei (Blue et al. 1988; Stone 1990).

5–HT released from their terminals act on six distinct families of G–protein

coupled metabotropic receptors (5–HT1, 5–HT2, 5–HT4, 5–HT5, 5–HT6, and 5–

HT7) and one family of ligand–gated ionotropic 5–HT3 receptors (Dougherty and

Aloyo 2011; Nichols and Nichols 2008; Santana et al. 2004). The complexity of

serotonergic signaling on cortical neurons is further complicated by

heterogeneous expression patterns of specific 5–HT receptors across regions, laminae, and cell types of the cortex (Amargos-Bosch et al. 2004; Blue et al.

1988; Jakab and Goldman-Rakic 2000; Santana et al. 2004; Weber and Andrade

2010) and microcircuit–specific and context–dependent activation of distinct

receptors on neurons expressing multiple 5–HT receptor types. (Amargos-Bosch

et al. 2004). Despite recent advances in understanding the role of 5–HT in cortical

function (Andrade 2011; Celada et al. 2013; Nakamura and Wong-Lin 2014), this

area of inquiry remains in need of further exploration.

Uncovering the fundamental principles of 5–HT system development and

function within the CNS has been aided substantially by the use of transgenic

animals in which different components of the 5–HT system are genetically altered

to augment or diminish systemic and/or region–specific levels of 5–HT (Deneris

and Wyler 2012; Gaspar et al. 2003). The development of a mouse line in which

the ETS domain transcription factor, Pet–1, is genetically ablated has been

particularly useful for understanding the processes underlying specification and

maintenance of the serotonergic neuron phenotype (Hendricks et al. 2003; Liu et

al. 2010). Constitutive deletion of Pet–1 results in a severe reduction (~70% loss)

of 5–HT immunoreactive neurons within the brainstem raphe nuclei which leads

to a dramatic decrease (~80% reduction) in forebrain 5–HT levels (Hendricks et

al. 2003; Liu et al.). These deficits are accompanied by a decrease in the

expression of genes regulating 5–HT synthesis, reuptake, packaging into vesicles,

and storage in the remaining population of 5–HT neurons. The behavioral

consequences of such an extreme depletion in systemic 5–HT levels are

characterized by increased anxiety and aggression (Hendricks et al. ; Liu et al.

2010) as well as altered maternal behavior (i.e. decreased nesting) (Lerch-Haner et al. 2008). In addition to the primary changes in expression of genes regulated

by Pet–1 and the resulting behavioral phenotypes, mice lacking Pet–1 also exhibit

hypersensitive cortical 5–HT2 receptors and an increased expression and

hypersensitivity of cortical 5–HT1 receptors, perhaps as compensation for

decreased 5–HT signaling (Yadav et al. 2011). Whether these perturbations to the

5–HT system correspond to alterations in the function of cortical neurons and

networks remains unknown.

In this study, we address the consequences of systemic 5–HT depletion on

cortical function and neuronal structure in Pet–1 null mice (Pet–1–/–). We employ

single–cell electrophysiology in cortical slices from wild–type (WT) and Pet–1–/–

mice to assess potential changes in cell–intrinsic, synaptic, and network

excitability. We also perform cytochemistry and 3–D neuronal reconstructions to

determine if deletion of Pet–1 leads to abnormal neuronal morphology in the

cortex. Our results show alterations across cell–intrinsic, synaptic, and network levels of cortical excitability in Pet–1–/– mice. In light of the changes in cortical

excitability, we also carried out acute behavioral seizure experiments and

surprisingly found no changes in susceptibility to acute seizures in Pet–1–/– mice.

To our best knowledge, this is the first study to report alterations in cortical

excitability following systemic depletion of 5–HT.

Materials and Methods

Thalamocortical slice preparation. Thalamocortical slices (350 µm) of

somatosensory cortex were prepared as previously described (Agmon and

Connors 1991; Puzerey et al. 2014) from juvenile (P13–21) C57BL/6 wild–type

and Pet–1–/– mice, which were generously donated by Dr. Evan Deneris. Animals

were anesthetized with vapor isoflurane and decapitated with a guillotine. The

brain was then submerged in ice–cold artificial cerebrospinal fluid (ACSF)

saturated with 95% O2 5% CO2 containing the following: 125 mM NaCl, 2.5 mM

KCl, 1 mM MgCl2, 2 mM CaCl2, 25 mM NaHCO3, 1.25 mM NaH2PO4 and 25 mM glucose. Brain slices were cut on a vibratome (Leica VT1200). All chemical salts and reagents were purchased from Fisher Scientific (Pittsburgh, PA) and

Sigma–Aldrich (St. Louis, MO). After the vibratome, slices were transferred to a bath containing room temperature ACSF for 20 minutes to incubate.

Subsequently, slices were moved to the recording chamber and perfused with

standard ACSF warmed to 31°C with a TC–324B Automatic Temperature

Controller (Warner Instrument Corporation; Hamden, CT) at a rate of 2 mL/minute. Slices were then incubated for one hour before beginning electrophysiological recordings.

In vitro electrophysiology. Pyramidal cells within L2/3 were identified by visual inspection at 63× magnification using Kohler illumination with an upright microscope (Zeiss Axioskop 2 FS+; Germany). Patch clamp recordings under the whole cell current–clamp configuration were collected from single neurons using

borosilicate glass electrodes (6–10 MΩ) filled with standard internal solution

containing the following: 120 mM potassium gluconate, 2 mM KCl, 10 mM

HEPES, 10 mM sodium phosphocreatine, 4 mM MgATP, 0.3 mM Na3GTP, 25

mM QX314, and adjusted to pH 7.4 with KOH. For voltage–clamp recordings, a

cesium–based internal solution was used to improve space clamp and contained

79 the following: 120 mM cesium gluconate, 2 mM CsCl, 10 mM HEPES, 10 mM sodium phosphocreatine, 4 mM MgATP, 0.3 mM Na3GTP, 20 mM BAPTA, and

25 mM QX314 to block voltage–gated sodium channels and adjusted to pH 7.4 with CsOH. Electrophysiological recordings were amplified with Multiclamp

700B amplifier (Molecular Devices, Foster City, CA) and digitized at 10 KHz with Digidata 1400 data acquisition interface. Data was low–pass filtered online at 1 KHz.

Spontaneous excitatory postsynaptic currents (sEPSCs) were recorded in the voltage–clamp configuration for 60 seconds. sEPSCs were recorded by holding the membrane potential at the reversal potential for inhibitory postsynaptic currents (EIPSC), which was experimentally determined to be –80 mV, consistent with previous reports (Chagnac-Amitai and Connors 1989;

Hasenstaub et al. 2005). Neurons whose access resistance exceeded 30 MΩ throughout the duration of the recordings were excluded from analysis. Drift of the resting membrane potential to values more positive than –60 mV was also used as an exclusion criterion. All pharmacological manipulations were carried out via bath application of the drug. Slices were then given one hour to incubate before beginning recordings.

Current–clamp recordings were carried out to measure cortical network activity in a disinhibited slice. We added 1 µM gabazine (GZN), a selective

GABAA receptor antagonist, to the bath ACSF to partially disinhibit the slice. In wild–type animals, spontaneous network activity under these conditions manifested as paroxysmal depolarizing shifts (PDS) which have a consistent

voltage profile comprising an initial plateau depolarization (~60–80 mV relative

to baseline) lasting 400–500 ms succeeded by a decaying tail lasting ~500 ms.

Recordings of cortical network activity lasted 10 minutes and were obtained

without current injection (Ihold = 0 pA). The concentration of GZN was chosen so

to elicit on average 1–2 PDS per minute, thus making the recordings amenable to

statistical analysis. At this concentration, the network was not sufficiently

disinhibited to spontaneously exhibit fast run epileptiform oscillations in wild–

type animals, but caused fast runs in Pet–1–/– mice.

One slice was used from a single animal and 5–7 cells were recorded in

each slice. Statistical tests were carried out on groups of 30–40 cells

(corresponding to 5–6 animals/group).

Seizure induction. Juvenile C57BL/6 and Pet–1–/– mice (P18–P24) were injected with the convulsant, pentetrazol (PTZ; i.p. 80 mg/kg) dissolved in 0.1 M phosphate–buffered saline (PBS). The starting point of behavioral seizures was considered as the appearance of generalized clonic convulsions with loss of righting reflex (GCC–LOR) (Loscher et al. 1990). The endpoint for each GCC–

LOR seizure was considered as release from tonus (in animals presenting with tonus) or as cessation of generalized clonic convulsions and recovery of righting.

Time elapsed between PTZ injection and the first GCC–LOR seizure was taken as the seizure latency. In most acute seizure experiments the animal died after experiencing several seizures. If the seizures persisted for over one hour after PTZ injection, the animals were euthanized by isoflurane anesthesia and decapitation.

Animals that did not experience seizures within an hour after PTZ injection were

not included in the data analysis.

Data Analysis and Statistics. Spontaneous excitatory postsynaptic currents

(sEPSCs) were detected using a custom algorithm written in MATLAB

(Mathworks). Detection was based on a threshold for the sEPSC derivative with

threshold values for events obtained empirically. We provided the algorithm with

additional detection criteria based on the event kinetics. Events with rise time exceeding 5 ms and decay time constant longer than 30 ms were excluded; also the rise time was not allowed exceed the decay time constant. For verification of successful detection, events were visually inspected after being passed through the exclusion criteria. Network activity in current–clamp was detected manually

using a custom interactive detection algorithm written in MATLAB. Voltage

deflections during PDS network activity typically were between 50–80 mV,

depending on the resting potential of the neuron (typically –70 mV), thus could be

detected easily by a simple amplitude threshold–based detection. Reverberant

afterdischarges during fast runs were detected by setting an absolute amplitude

threshold of 30 mV relative to the resting potential and a 10 mV threshold relative

to the local minimum (i.e. if the voltage deflection occurred on top of a

depolarization plateau). We determined statistically significant differences

between distributions using the nonparametric Wilcoxon rank–sum test, which

tests the null hypothesis that two independent groups of samples come from

distributions with equal medians. Details regarding the spectral analysis of

82 network activity are described in detail in our previous publication (Puzerey et al.

2014).

Results

Cell–intrinsic parameters of neuronal excitability are altered in Pet–1 knock– out mice

To determine whether the drastic 5–HT depletion in the forebrain of Pet–1 knock–out mice (KO) has any effect on the intrinsic excitability of cortical neurons, we carried out whole–cell patch clamp recordings in layer 2/3 of somatosensory cortex in wild–type (WT) and KO animals from individual pyramidal cells (PCs). We first tested the passive properties of the neuronal membrane, namely the neuronal input resistance (Rinput) and the time constant of the neuronal membrane, τ. Comparison of these parameters between WT and KO mice reveals that both Rinput and τ are increased in KO animals (Rinput: WT mean ± std: 126 ± 73 MΩ; KO mean ± std: 333 ± 134; p < 0.01, Wilcoxon rank–sum test;

τ: WT mean ± std: 9.5 ± 2.6; KO mean ± std: 13.2 ± 1.93; p < 0.01, Wilcoxon rank–sum test; Fig. 3.1a, b). A longer time constant corresponds to slower decay of the membrane potential in response to a transient current and consequently, longer integration time for synaptic inputs. Larger values of τ confer enhanced summation of excitatory inputs in time, thus leading to increased excitability of the neuronal membrane. Rinput dictates the magnitude of the voltage response to an injected current according to Ohm’s law. This means that larger values of Rinput would on average lead to a larger voltage deflection in response to a transient

83 input current of a fixed amplitude. Therefore, the changes observed Rinput in and τ confer enhanced excitability to cortical neurons in KO mice. We also compared the resting membrane potential and report no significant differences between WT

(mean ± std: –67.8 ± 11.3 mV) and KO mice (mean ± std: –64.2 ± 9.3 mV; p =

0.28, Wilcoxon rank–sum test)

We then investigated the active intrinsic membrane properties underlying the generation of action potentials in cortical PCs of both WT and KO mice. We first compared differences in spike threshold by injecting ramp currents in the current clamp configuration and observed no difference between the two groups (WT mean ± std: –31.8 ± 8.4; KO mean ± std: –32.1 ± 6.7; p = 0.33, Wilcoxon rank– sum test). Ramp current injection also allows for measurement of the latency to the onset of the first spike, which serves as a proxy for neuronal excitability.

Figure 3.1c shows that KO mice have substantially delayed spike responses to injection of ramp current compared to their WT littermates (WT mean ± std: 447

± 80 ms; KO mean ± std: 636 ± 233 ms; p < 0.01, Wilcoxon rank–sum test). We also compared frequency–input (F–I) curves which determine neuronal gain and observed no significant differences between the two groups (data not shown).

Despite seeing no differences in neuronal gain, we observed that cortical PCs from KO animals tend to reach a lower maximal firing frequency (MFF) in response to step current injection. Quantifying the MFF for WT and KO groups showed that, indeed, pyramidal cells in KO mice reach a lower MFF on average than those from WTs (WT mean ± std: 47.3 ± 9.3 Hz; KO mean ± std = 32.8 ± 6.9

Hz; p < 0.05, Wilcoxon rank–sum test; Fig. 3.1d). To understand why PCs from

KO mice have lower MFF, we compared the range of values of the input current

that elicit changes in the firing rate (i.e. input dynamic range) and report that this

range is on average lower in KO animals (WT mean ± std: 439 ± 119 pA; KO mean ± std: 273 ± 113 pA; p < 0.01, Wilcoxon rank–sum test; Fig. 3.1e). When considered together these results demonstrate an overall decrease in the active properties of cell–intrinsic excitability, perhaps as compensation for changes in passive membrane properties in the opposite direction.

Cortical pyramidal cells exhibit increased spontaneous synaptic activity in Pet–

1 knock–out mice

One possible outcome of 5–HT depletion in the cortex is a change in synaptic transmission between PCs, which might arise owing to 5–HT’s capacity to modulate or directly mediate synaptic transmission (Aghajanian and Marek

1999; Beique et al. 2007; Foehring et al. 2002; Lambe et al. 2000; Roerig et al.

1997; Zhou and Hablitz 1999). To test this possibility, we recorded spontaneous excitatory postsynaptic currents (sEPSCs) from layer 2/3 cortical pyramidal cells in voltage clamp configuration from WT and KO mice. Our results show a dramatic and significant increase in the amplitude of sEPSC in KO animals as compared with WTs (p < 0.01, Wilcoxon rank–sum test, Fig. 3.2a,b).

Surprisingly, this increase can be normalized near to control levels with an antagonist of 5–HT3 receptors (5–HT3R), granisetron (GSN). This unexpected finding suggests that despite a substantial (~80%) (Hendricks et al. 2003) depletion of forebrain 5–HT levels, there may be a compensatory upregulation of synaptic serotonergic signaling, a finding consistent with other models of

serotonin depletion in which signaling through 5–HT receptors was altered (Liu et

al. 2010; Moya et al. 2011; Yadav et al. 2011). The effect on the frequency of sEPSCs is less obvious. As shown in Figure 2c, shorter (< 100 ms) inter–event intervals (IEI) are decreased, while longer (> 100 ms) are increased in Pet–1 KO animals, a finding that is difficult to interpret. Since the IEI acts as a proxy for presynaptic vesicle release probability, this finding suggests a complex effect of

5–HT on regulation of spontaneous presynaptic release that is perhaps regulated by more than one receptor type (Gothert 1990). GSN treatment decreased in KOs

decreased the IEI relative to both WT and KO groups without pharmacological

treatment (p < 0.01, Wilcoxon rank–sum test, Fig. 3.2a,c).

Cortical network excitability is enhanced in Pet–1 knock–out mice

As a next step to understanding effects of central 5–HT depletion on cortical neurophysiology, we investigated cortical network activity in a partially disinhibited cortical slice (Castro-Alamancos and Rigas 2002; Hablitz 1987;

Puzerey et al. 2014). The slice was disinhibited with the GABAA receptor

antagonist, gabazine (GZN; 1 µM). Under these conditions, cortical networks in

slices from WT animals undergo massive spontaneous depolarization plateaus

known as paroxysmal depolarizing shifts (PDS; Fig. 3.3a) (Hablitz 1987). The voltage profile of a PDS consists of an early plateau phase lasting ~ 500 ms followed by a decay phase that lasts up to 2 s. The PDS events correspond to synchronized activity in local cortical networks (Johnston and Brown 1984;

McCormick and Contreras 2001). Therefore, one can use single–cell recordings from cortical PCs as a read–out of network activity. We thus compared the

voltage profile and statistics of cortical network activity in slices from WT and

KO mice. An unexpected result from these experiments was that unlike the slices

from WT animals, which manifested network activity as sparse (~ 1 PDS/min;

Fig. 3.3a,g) and random (i.e. lacking temporal organization) PDS, the network

activity in KO mice had a voltage profile that was typically characterized by a

massive PDS that was succeeded by a series of afterdischarges with a preferred frequency of 10–15 Hz and a burst duration of ~40 ms (Fig. 3b,g). The afterdischarges were typically superposed on a long depolarizing plateau. The emergence of these afterdischarges resulted in a substantially larger number of

network events recorded over a 10 min period (p < 0.01, Wilcoxon rank–sum test;

Fig. 3.3f). It is important to note that the afterdischarges were not a result of

intrinsic spiking in recorded neurons since they persisted in the presence of a

voltage–gated sodium channel blocker, QX–314, in the intracellular solution.

Thus, the afterdischarges were of synaptic origin, a finding that is verified by

previous studies on such patterns of activity in both cortex and hippocampus (Lee

and Hablitz 1991; Schmitz et al. 1997). This is further verified by dual–cell patch

clamp recordings from pairs of pyramidal cells spaced not more than 100 µm

apart showing that the initial burst and the afterdischarges were almost perfectly

correlated in adjacent neurons (Fig. 3.3d,e). The pattern of network activity in

disinhibited cortical slices from Pet–1 KO mice has been previously reported in

disinhibited rat cortex in vivo (Castro-Alamancos 2000) and is an electrographic

signature of epileptic seizures in certain childhood onset epilepsies (Camfield

2011). We will henceforth refer to this pattern of activity as “fast runs,” a term

87 assigned to this pattern of activity in previous studies (Steriade et al. 1998;

Timofeev et al. 1998).

In our previous study, we showed that elevated endogenous 5–HT signaling in WT animals transforms PDS into fast runs via augmented signaling through 5–HT2 receptors (5–HT2R). Given that Pet–1 KO mice express hypersensitive 5–HT2Rs (Yadav et al. 2011), we hypothesized that fast runs observed in Pet–1 mice could result from excessive 5–HT2R signaling and could, therefore, be abolished by blockade of this receptor. Consistent with our hypothesis, bath application of the selective 5–HT2 receptor antagonist, ketanserin (KSN; 10 µM), reduced the number of network events (p < 0.01,

Wilcoxon rank–sum test; Fig. 3.3f) and abolished fast runs in slices from KO mice. The voltage profile of network activity in KOs treated with KSN transformed to single PDS events like those seen in control WT slices (Fig. 3.3c).

The absence of fast runs after KSN treatment was also apparent by the lack of the

10–15 Hz peak in the power density plot (Fig. 3.3g). Therefore, our results support the idea that hypersensitive 5–HT2 receptors in Pet–1 mice bias cortical network activity to a hyperexcitable regime characterized by epileptiform oscillations.

Susceptibility to convulsant–induced seizures is unchanged in Pet–1 knock–out mice

Our findings in the partially disinhibited slice show a hyperexcitable cortical phenotype and spontaneous emergence of epileptiform oscillations in

Pet–1 KOs that are absent in WT mice, suggesting that the cortex of the mutant

mice is more susceptible generating epileptiform activity. The fast run oscillations

observed in the slice have been previously reported during chemically induced

seizures in rats in vivo (Castro-Alamancos 2000), in spontaneous and electrically

induced seizures in cats (Steriade et al. 1998), and are a hallmark of epileptic

seizures in certain forms of childhood onset epilepsy (Camfield 2011). Findings

from our previous publication (Puzerey et al, 2014) suggest that 5–HT2R activity modulates the latency to the onset of acute epileptic seizures induced by the convulsant pentetrazol (PTZ), leading us to hypothesize that Pet–1 KO mice may be more susceptible to convulsant–induced epileptic seizures (i.e. may have a lower seizure threshold, hence shorter seizure latencies). To test this hypothesis, we induced epileptic seizures in WT and KO mice with PTZ (80 mg/kg; i.p.) and measured the time to the onset of the first epileptic seizure characterized by loss of righting and generalized clonic convulsions (LOR–GCC) as well as the mean duration of seizures in each animal (Fig. 4a). Surprisingly, there were no observed changes in either the seizure latency or seizure duration latency (latency WT mean ± std: 312±139 s ; KO mean ± std: 273±186 s; p = 0.31 , Wilcoxon rank– sum test; duration WT mean ± std: 78±36 s; KO mean ± std: 128±103; p = 0.15,

Wilcoxon rank–sum test; Fig. 3.4 b, c). These results lead us to conclude the

following: 1) despite the sufficiency of 5–HT2R hypersensitivity in generating

epileptiform oscillations in KO mouse slices in vitro, it is not sufficient to alter

the susceptibility to epileptic seizures in vivo; and 2) perhaps compensatory

molecular mechanisms involving up–regulation of other 5–HT receptors (e.g. 5–

HT1 receptors) or structural changes in neuronal morphology, cortical lamination,

or bundling neurites are sufficient to maintain overall normal levels of global

excitability in the CNS as a whole.

Discussion

The use of transgenic animals has been an invaluable tool to understanding the function of specific genes; however, the lack of a complete phenotypic profile of a transgenic animal may conceal physiologically relevant biological compensations.

In this study, we demonstrate that mice with a constitutive deletion of the ETS domain transcription factor, Pet–1, which normally confers the serotonergic phenotype to midbrain raphe neurons, exhibit unexpected changes in the cell– intrinsic and network excitability of the cortex. Firstly, our results point to altered intrinsic neuronal excitability in pyramidal neurons of layer 2/3 in the somatosensory cortex. We observed that some parameters of intrinsic excitability were biased towards a hyperexcitable phenotype while others showed a change towards decreased excitability. The direction of these changes could be predicted based on whether they related to passive or active membrane properties. More specifically, measures of passive membrane properties, the membrane time constant (τ) and the input resistance in Pet–1–/– are increased, thus leading to

enhanced responsiveness to transient synaptic inputs. On the other hand, measures

of active membrane excitability like maximal firing frequency, input dynamic

range, and spike latency to ramp current injection all exhibited a significant

tendency towards decreased excitability. These findings show that systemic

depletion of forebrain 5–HT levels produces a complex phenotype in cell–

intrinsic excitability of cortical pyramidal cells. We also demonstrate that these

changes are accompanied by an alteration in synaptic excitability characterized by

an increased rate and amplitude of sEPSCs. The normalization of the sEPSC

amplitudes to control levels with the 5–HT3 receptor antagonist, granisetron,

suggests that Pet-1–/– mice exhibit increased synaptic signaling through these

receptors on cortical pyramidal neurons. Furthermore, we report dramatic

differences in cortical network excitability and dynamics characterized by the

transformation of single network bursts to oscillatory epileptiform discharges. The

emergence of these network oscillations appears to depend on signaling through

5–HT2 receptors, given that they were abolished by a receptor–specific

antagonist. Surprisingly, the appearance of epileptiform activity in cortical slices

did not correlate with changes in susceptibility to convulsant-induced seizures in

vivo. Together, our results show significant changes in cell–intrinsic, synaptic, and network excitability in the cortex of mice lacking the Pet–1 transcription factor and suggest the emergence of a complex set of biological adaptations in response to systemic 5–HT depletion. To our best knowledge, this is the first study to report neurophysiological changes in a target structure of the 5–HT system after genetic ablation of the Pet-1 gene.

The specification, maturation, and maintenance of the central 5–HT

neuronal phenotype is a complex process characterized by distinct developmental

phases that proceed under guidance of specific transcriptional programs. We refer

the reader to a comprehensive review on the development of the 5–HT system and

its relevance to neuropsychiatric disorders (Deneris and Wyler 2012) and will

only focus here on the known functions of the Pet-1 gene. The importance of this

gene for various physiological functions is underscored by several studies

pointing to dysregulation of circadian rhythms (Ciarleglio et al. 2014; Paulus and

Mintz 2012), respiration and thermoregulation (Cummings et al. 2011; Cummings et al. 2010; Erickson et al. 2007; Hodges et al. 2011), maternal behavior (Lerch-

Haner et al. 2008), adult neurogenesis (Diaz et al. 2013), and neonatal growth

(Narboux-Neme et al. 2013) following its constitutive genetic deletion. Such a broad range of physiological defects resulting from systemic 5–HT depletion highlights the importance of the serotonergic system in controlling developmental and autonomic processes. Despite the breadth of systems controlled by 5–HT investigated in the aforementioned studies, not a single study has yet investigated the consequences of Pet-1 deletion on the function of the cortex, a forebrain structure densely innervated by serotonergic afferents. Our study provides a compelling piece of evidence for the importance of the 5–HT system, specified and maintained by the Pet-1 gene, for the normal physiology of the cortical neurons, synapses and networks. To the best of our knowledge, this is the first study to find direct changes in neuronal function in a region secondary to the site of Pet-1 deletion. That is, deletion of Pet-1 in the raphe nuclei leads to functional alterations in the cortex, a site innervated by these structures.

To understand how global alterations in cortical 5–HT levels may affect cell–intrinsic, synaptic, and network function in Pet-1–/– mice, we speculate on some potential pathways by which these changes may be actualized through the known functions of 5–HT receptors. We address first the changes in cell–intrinsic

parameters of neuronal excitability. Our findings show that the membrane time

constant τ and the input resistance are increased in Pet-1–/– mice. These two

parameters of passive membrane excitability are directly related to the membrane

capacitance (via τ = RCin m ), which is typically determined by the size of the neuronal membrane. One possible explanation for altered passive membrane properties reported here could be an alteration in cell morphology. I am presently testing this hypothesis by filling pyramidal cells in L2/3 of somatosensory cortex from Pet-1–/– mice with biocytin and carrying out 3D morphological reconstructions.

The finding that Pet-1–/– mice exhibit an increase in the amplitude of 5–

HT3 receptor–mediated sEPSCs was unexpected given that these animals have a

severe reduction (~80%) in forebrain 5–HT levels. However, Pet-1–/– mice also

show decreased levels in the expression of the 5–HT transporter Sert, which

mediates reuptake of 5–HT into presynaptic axon terminals, and a parallel

decrease in the expression of 5–HT1 receptors on raphe neurons, which provide

an auto–inhibitory signal to regulate 5–HT release (Hendricks et al. 2003; Liu et

al. 2010). The overall effect of diminished expression of these proteins would

result in increased synaptic concentrations of the remaining 5–HT that persists in

spite of Pet-1–/– deletion, which could potentially account for the increased

amplitude of 5–HT3 receptor–dependent sEPSCs onto cortical neurons. We

speculate that an alternative explanation for the increased sEPSCs amplitude

would be elevated expression of postsynaptic 5–HT3 receptors on cortical

neurons. Support for this idea remains to be found though transgenic mice lacking

Sert have been shown to exhibit adaptive changes in 5–HT3 receptor expression,

suggesting that similar adaptations could be possible in Pet-1–/– mice (Mossner et

al. 2004). We note also that 5–HT3 receptor expression on cortical pyramidal

neurons has been previously challenged given the prevailing notion that 5–HT3

receptors are expressed exclusively on cortical inhibitory interneurons. We refer

the reader to our previous publication in which we address the controversy

regarding 5–HT3 receptor expression on cortical pyramidal cells (Puzerey et al.

2014).

Another surprising finding of this study was the difference in cortical

network dynamics in the disinhibited slice between wild–type and Pet-1–/– mice.

While the disinhibition–induced network activity in wild–types manifested as

individual PDS, the 5–HT–deficient mice exhibited fast run epileptiform

oscillations that could be abolished via blockade of 5–HT2 receptors. This is

unexpected since our intuition would lead us to think that an 80% drop in

forebrain 5–HT levels would result in decreased signaling through 5–HT2

receptors. However, a possible explanation for this finding comes from a previous

study that utilized the Pet-1–/– transgenics to study mechanisms of clozapine

action. In this study, Yadav and colleagues showed a significant increase in

intrinsic signaling of 5–HT2 receptors in mice lacking Pet–1 (Yadav et al. 2011).

Such changes in sensitivity of 5–HT2 receptors have been reported in other

models of 5–HT depletion (Narboux-Neme et al. 2013). Combining this hypersensitivity of 5–HT2 receptors with the aforementioned decreases in the expression of Sert and 5–HT1 receptors in raphe neurons, both of which would

94 act to increase synaptic concentrations of the remaining 5–HT (Hendricks et al.

2003; Liu et al. 2010), we hypothesize that the emergence of fast runs in Pet-1–/– mice results from these compensatory increases in postsynaptic 5–HT signaling.

This hypothesis is consistent with our previous study in which elevation of endogenous cortical 5–HT levels with the serotonin reuptake inhibitor, fluoxetine, led to the emergence of fast runs in a 5–HT2 receptor–dependent manner in cortical slices from wild–type mice (Puzerey et al. 2014).

The study of transgenic animal models for the purpose of understanding the function of biological systems is undoubtedly useful but may be confounded by secondary homeostatic compensation that is often unexpected. These compensations may arise as a result of inherent correlations in the expression of certain genes that result from their embedding into highly interconnected transcriptional or protein networks (Marder 2011) or from intrinsic protein properties. Such compensations are certainly familiar in the study of the 5–HT system since several transgenic mouse lines in which distinct components of this system are genetically altered exhibit correlated compensations in expression or sensitivity of related proteins (Fabre et al. 2000; Goodfellow et al. 2012;

Hendricks et al. 2003; Liu et al. 2010; Mossner et al. 2004; Moya et al. 2011;

Narboux-Neme et al. 2011; Veenstra-VanderWeele et al. 2012; Yadav et al.

2011). Our study builds upon this idea and shows new evidence for a functional increase in signaling through 5–HT2 and 5–HT3 receptors in Pet-1–/– mice that correspond to altered cortical network activity and intrinsic cell excitability.

Figure 3.1. Altered cell–intrinsic excitability in mice lacking Pet–1.

(a) Box plots representing input resistance measurements from layer 2/3 cortical pyramidal neurons in WT (n = 26) and Pet–1–/– mice (n = 30). (b) Box plots

representing the membrane time constants from pyramidal cells, τ, in WT (n = 47)

and Pet–1–/– mice (n = 24). (c) Box plots of spike latency times in response to

ramp current injection from WT (n = 22) and Pet–1–/– mice (n = 30). (d) Box plots

of maximal firing frequency in Hz in response to step current from WT (n = 22)

and Pet–1–/– mice (n = 30). (e) Box plots of input dynamic range, taken as the

range of current values that elicit changes in the neuronal firing rate, from cortical

pyramidal cells in WT (n = 22) and Pet–1–/– mice (n = 30). Asterisks denote level

of significance (*: p < 0.05; **: p < 0.01).

Figure 3.2. Increased spontaneous excitatory postsynaptic currents in cortical

pyramidal cells from Pet–1 knock–out mice.

(a) Raw traces of spontaneous excitatory postsynaptic currents (sEPSCs) in layer

2/3 pyramidal neurons from WT mice (left), Pet–1–/– mice (middle), and Pet–1–/–

mouse brain slices treated with the 5–HT3 receptor antagonist, granisetron (right).

(b) Cumulative distribution of sEPSC amplitudes recorded in cortical slices from

WT control slices (n = 48 cells), Pet–1–/– slices (n = 20 cells), and Pet–1–/– slices

treated with granisetron (n = 42 cells). (c) Cumulative distribution of sEPSC

inter–event intervals (IEI) recorded in cortical slices from the same cells as in (b).

Figure 3.3. Enhanced cortical network excitability in mice lacking Pet–1.

(a–c) Raw traces of disinhibition–induced cortical network activity recorded from

L2/3 pyramidal cells under (a) control conditions (1 µM gabazine) in slices from

WT mice (n = 30 cells), (b) control conditions from Pet–1–/– mice (n = 32 cells), and (c) in the presence of the 5–HT2 receptor antagonist, ketanserin, in slices from Pet–1–/– mice (n = 19 cells). Top trace in a–c corresponds to the entire 10 minute recording period and bottom trace inside the rectangle shows a zoomed trace of a single network event. (c) Box plots showing quantification of network activity as the number of network bursts recorded over 10 minutes. Asterisks above correspond to the level of significance of the difference between adjacent distributions. (d) Power spectral density plot as measured from the three groups mentioned above. Note that the presence of the oscillatory fast runs corresponds to a peak in the 10–15 Hz band, while the baseline of the function denotes baseline activity levels in each group.

100

101

Figure 3.4. Seizure susceptibility is unaltered in mice lacking Pet–1.

(a) Experimental paradigm for inducing acute epileptic seizures with the convulsant, pentetrazol. Mice were administered a vehicle control solution 1 hour before PTZ injection. Electric bolts correspond to onset of spontaneous epileptic seizures. (b) Box plots of latency to the first LOR–GCC seizure in WT (n = 17 mice) and Pet–1–/– mice (n = 13 mice). (c) Box plots of mean seizure duration from the same mice as in (b).

102

103

Chapter 4: On How Correlations between Excitatory and Inhibitory

Synaptic Inputs Maximize the Information Rate of Neuronal Firing

Summary

Cortical neurons receive barrages of excitatory and inhibitory inputs

which are not independent, as network structure and synaptic kinetics impose

statistical correlations. Experiments in vitro and in vivo have demonstrated

correlations between inhibitory and excitatory synaptic inputs in which inhibition

lags behind excitation in cortical neurons. This delay arises in feed–forward inhibition circuits and ensures that coincident excitation and inhibition do not preclude neuronal firing. Conversely, inhibition that is too delayed broadens neuronal integration times, thereby diminishing spike–time precision and

increasing the firing frequency. This led us to hypothesize that the correlation

between excitatory and inhibitory synaptic inputs modulates the encoding of

information of neural spike trains. We tested this hypothesis by investigating the

effect of such correlations on the information rate (IR) of spike trains using the

Hodgkin–Huxley model in which both synaptic and membrane conductances are

stochastic. We investigated two different synaptic input regimes: balanced

synaptic conductances and balanced currents. Our results show that correlations arising from the synaptic kinetics, τ, and millisecond lags, δ, of inhibition relative to excitation strongly affect the IR of spike trains. In the regime of balanced synaptic currents, for short time lags (δ ∼ 1 ms) there is an optimal τ that

maximizes the IR of the postsynaptic spike train. Given the short time scales for

104

monosynaptic inhibitory lags and synaptic decay kinetics reported in cortical

neurons under physiological contexts, we propose that feed–forward inhibition in

cortical circuits is poised to maximize the rate of information transfer between

cortical neurons. Our results also provide a possible explanation for how certain

drugs and genetic mutations affecting synaptic kinetics can deteriorate

information processing in the brain.

Introduction

The rate and timing of firing in cortical neurons is strongly affected by the

interaction between synaptic excitation and inhibition (Salinas and Sejnowski

2000). The architecture of cortical circuits ensures that the magnitude of

excitatory and inhibitory synaptic inputs is approximately balanced on average

and temporally correlated (Haider et al. 2006; Shu et al. 2003b), albeit with a

small time delay for inhibition of ~1–10 ms (Okun and Lampl 2008; Wehr and

Zador 2003; Wu et al. 2008). This correlation in amplitude and timing presumably

arises in feed–forward inhibition (FFI) circuits, an anatomical motif present

ubiquitously throughout the cortex which drives monosynaptic excitation and

disynaptic inhibition onto target neurons (Cruikshank et al. 2007; Porter et al.

2001; Sun et al. 2006). The functional consequences of the correlations imposed by such a layout are far–reaching, encompassing a range of functions such as gain modulation for rapidly fluctuating synaptic inputs (Chance et al. 2002; Pouille et al. 2009; Salinas and Sejnowski 2000; Shu et al. 2003a), shaping of neuronal tuning properties and stimulus selectivity (Marino et al. 2005; Wehr and Zador

2003; Wu et al. 2006), directing the propagation of activity by selectively gating

105 firing in neuronal ensembles (Kremkow et al. 2010a; Kremkow et al. 2010b), and creating “windows of integration” during which excitatory inputs can temporally summate to promote spike generation before being rapidly suppressed by inhibition (Pouille et al. 2009; Pouille and Scanziani 2001). Furthermore,

Marsalek and colleagues demonstrated that small differences in the timing between presynaptic excitatory and inhibitory inputs (i.e. input correlation) is directly correlated with temporal jitter in postsynaptic spikes, i.e. output reliability

(Marsalek et al. 1997). Since neurons may represent information through the precise timing of spikes (Dan et al. 1998; deCharms and Merzenich 1996; Liu et al. 2001; Nemenman et al. 2008; Strong et al. 1998), it stands to reason that the control of spike timing by correlated excitation and inhibition is likely to govern the transfer of information between cortical neurons. Previous investigations using realistic simulations of cortical neurons have shown that, indeed, balanced excitatory and inhibitory synaptic currents maximize both coding and metabolic efficiency of neuronal spikes (Sengupta et al. 2013). In that study, however, the excitatory and inhibitory synaptic conductances were uncorrelated and as a result did not exhibit the correlation characteristic of cortical dynamics under experimental contexts (Okun and Lampl 2008; Wehr and Zador 2003; Wu et al.

2008). Furthermore, Kawaguchi and colleagues have shown that the relative balance between excitation and inhibition of a random synaptic input to simulated pyramidal neurons controls the maximal information content of spike trains in the presence of background synaptic noise (Kawaguchi et al. 2011), yet the timing of excitation and inhibition with respect to each other were not considered. To our

106

best knowledge, the relevance of statistical correlations between balanced

excitatory and inhibitory synaptic inputs for the information rates of neural spike

trains have not been investigated. This may in part be due to the high

computational cost of realistic simulations of stochastic neuronal dynamics. Here,

we have overcome this limitation by using the stochastic–shielding approximation, which was recently introduced by our lab, speeding up stochastic simulations by up to two orders of magnitude while preserving accuracy

(Schmandt and Galán 2012).

A critical factor that influences the correlation between synaptic conductances and their effect on firing of cortical neurons is the time–course of the conductance change (Svirskis and Rinzel 2000). This time–varying conductance shapes the trajectory of the membrane potential towards spike threshold and, consequently, alters the probability of firing an action potential. In support of this notion, previous findings have shown that the precision of spike– timing in pyramidal neurons has an inverse relationship with the decay kinetics of excitatory postsynaptic currents (Rodriguez-Molina et al. 2007), that is, spike precision decrements as the postsynaptic currents slow down. The kinetics of postsynaptic synaptic responses may be modified by electrotonic filtering of the inputs across the dendritic arbor (Kleppe and Robinson 1999), activation of distinct afferents (Walker et al. 2002), changes in the driving force (Salin and

Prince 1996), developmental changes in postsynaptic receptor (or receptor subunit) expression (Bannister et al. 2005; Cohen et al. 2000; Kirson and Yaari

1996), interaction with intrinsic conductances (Miller et al. 1985; Wilson 1995),

107 and the presence of receptor–specific drugs (Cohen et al. 2000; Orser et al. 1994;

Poncer et al. 1996). To our knowledge, the relationship between the kinetics of synaptic conductances and the information rates of neural spike trains has not been investigated.

In this study, we test the hypothesis that the rate of information transfer in cortical neurons depends on the correlation between concurrent excitatory and inhibitory synaptic inputs. We predict that an optimal time lag (δ) between excitation and inhibition would maximize information transfer between cortical neurons, since lags that are too short would preclude neuronal firing while long lags will likely decrease the precision of neuronal firing by prolonging the window of integration of presynaptic inputs. This prediction is consistent with a previous study showing an optimal time scale of rapidly fluctuating inputs for spike time reliability (Galán et al. 2008). Extending this hypothesis further, we predict that the rate of information transfer will depend on the kinetics of the synaptic conductance change, which inarguably affects the temporal correlation between synaptic excitation and inhibition (Svirskis and Rinzel 2000). To test this hypothesis, we employ a biologically inspired Hodgkin–Huxley–type simulated neuron with stochastic ion channel gating (Schmandt and Galán 2012) and inject it with Poisson trains of matched excitatory and inhibitory synaptic inputs. We test the impact of relative lag times between excitation and inhibition on the information rates of our model neuron across a range of lags and decay kinetics of synaptic conductances. Moreover, we compare the dependency of the information rates on the lags and kinetics in two synaptic regimes of 1) balanced

108

conductances; and 2) balanced currents; this distinction is functionally important

since the driving force can directly control the ratio between excitatory and

inhibitory synaptic currents. Our findings reveal that the information rate of the neural spike train is indeed dependent on the synaptic kinetics as well as the relative delay times between excitation and inhibition. We show that the dependence of the information rate on the synaptic kinetics shows an optimum at

short and physiologically relevant monosynaptic delay times and that this

dependence is present in the balanced currents, but not in the balanced

conductances regime.

Materials and Methods

Synaptic inputs

All simulations were carried out using the MATLAB R2013b software

package (Mathworks). We modeled the synaptic events as Poisson trains with a

rate of λ = 5 ms–1, which was fixed across all simulations. The synaptic train was then convolved with an “alpha function” to yield the time–dependent

conductance, g(t), of the following form,

− τ − τ g(t ) = Ge( t/ r − et/ ) (3) where the constant G is set at 300 pS for all simulations trials in the balanced

conductances regime, e is the base of the natural logarithm, and τ r and τ are the

rise and decay time–constants, respectively. τ r is set at a fixed value of 0.2 ms

while τ is varied across the range of 1 ≤ τ ≤ 10 ms. The excitatory and inhibitory synaptic conductances (gexc and ginh, respectively) were created as identical

109

realizations of a Poisson process in the balanced conductances input regime. For

the balanced currents input regime, the inhibitory conductance was multiplied by

a factor of 8, yielding a maximal conductance amplitude of 2400 pS. To generate

conductance traces in which inhibition lagged behind excitation, we offset the two

waveforms by a lag time, δ, which for a given simulation was taken from a range

of lags, (1 ≤ δ ≤ 10 ms).

Analytical expression for the cross–correlogram of the synaptic inputs To calculate an analytical expression for the cross–correlogram of the excitatory and inhibitory synaptic inputs, we first note that the inhibitory input is identical with the excitatory input but delayed with a lag, δ, so that the cross–correlogram C()δ is actually equivalent to the auto–correlogram of the excitatory input. We also note that the excitatory input is the convolution of a Poisson process with the kinetics of a single synaptic event given by (1) and recall the following two theorems of time–series analysis:

1) The Wiener–Khinchin theorem, stating that for a given signal the auto–correlogram is the Fourier transform of its power spectrum; and 2) the convolution theorem, stating that the power spectrum of the convolution of two signals is the product of their power spectra. Therefore, since the power spectrum of a Poisson process is a constant, the cross–correlogram is determined by the Fourier transform of the power spectrum of a

single synaptic event. Defining ατ=1/ r and βτ=1/ , the power spectrum of a single

synaptic event is given by:

2 ∞ (αβ− )2 ke−−αt−= e βω t e i dt k , (4) ∫( ) 2222 0 (αωβω++)( )

110

2 where k is a constant and ... denotes the square of the modulus of a complex number.

The un–normalized cross–correlogram of the excitatory and inhibitory inputs is then given by the Fourier transform of (2)

2 k ∞ (αβ− ) C()δω= ediωδ . (5) ∫ 2222 2π −∞ (αωβω++)( )

To solve the integral in (3) we apply the residues theorem to a closed integration path containing two poles on the upper–half of the complex plane,ω= ii αβ, , whose respective residues are

(αβ−−)ee−−αδ (αβ) βδ ii and − . 22ααβ( ++) βαβ( )

Thus, expression (3) yields

k (αβ− ) ee−−βδ αδ C (δ ) = − . 2(αβ+ )  β α

Finally, the cross–correlogram, normalized so that C (01) = reads

−δτ −δτ ττee/ r − / C (δ ) = r . ττr −

This analytical expression accurately describes the cross–correlogram obtained from the numerical simulations, as shown in Figure 1C.

Single compartment model

Neuronal dynamics were simulated in a single compartment model of a

Hodgkin–Huxley (Hodgkin and Huxley 1952) neuron with stochastic voltage– gated fast Na+ and delayed rectifier voltage–gated K+ channels, and a

111

deterministic leak conductance as detailed in (Schmandt and Galán 2012).

Modeling of stochastic ion channel gating was made more computationally

efficient by applying the stochastic shielding approximation of Markov chains,

which reduces the number of ion channel state transitions requiring stochastic

simulations and only considers the stochastic transitions between observable

states (Schmandt and Galán 2012). The fluctuations in the membrane voltage were described by the following current balance equation:

membrane currents dV  Cm = gNa() t( E Na −+ Vt ()) gKK () t( E −+ Vt ()) gleak ( E leak −+ V ) dt +g() t( E −+ Vt ()) g () t( E − Vt ()) exc exc inh inh synaptic currents

where Cm corresponds to the membrane capacitance, gNa, gK, and gleak are the

+ + Na , K , and leak conductances with their respective reversal potentials, ENa, EK,

and Eleak. The excitatory and inhibitory synaptic conductances (gexc and ginh) along

with their respective reversal potentials, Eexc (0 mV) and Einh (–80 mV), dictate

the extent to which synaptic currents affect the membrane potential fluctuations.

The membrane potential was simulated with a time resolution of dt = 0.01 ms.

Determination of information rates

Spike train entropy was determined using the “direct method” (Nemenman

et al. 2008). This approach quantifies the entropy of the spike trains without

making assumptions about the nature of the stimulus. Spike trains were binned in

small time windows (Δt = 5 ms) and spikes were counted for each bin. A value of

0 was assigned to each bin containing no spikes and a value of 1 for those

containing one spike. With the maximal firing rate of the model peaking at 80 Hz

112

(inter–spike interval = 12.5 ms), our choice of Δt ensures that at most one spike can occur within a given time bin, therefore, providing information rates for timing of action potentials with millisecond precision. The resultant binary strings of 0’s and 1’s were used to generate words of length n where n = 2, 4, 6, 8, 10, yielding words that spanned time windows of T = nΔt. Probability distributions were then generated to quantify the occurrence probability of a given word, P(W), within a response pattern. Noise entropy, which measures the reproducibility of spike trains in response a fixed input stimulus across trials (56 trials), was measured as with respect to the conditional probability of a word occurring at time t and calculated with the following equation:

Hnoise = −∑ PW( | t) log2 PW( | t) W t where the operator 〈…〉t denotes averaging over time. The total entropy, which quantifies the possible permutations of output patterns with respect to a broad set of inputs was determined by presenting the model neuron with a different input pattern across 56 trials measuring the occurrence probability of a given word. The total entropy was calculated as:

Htotal = −∑ PW()log()2 PW . W

By definition, the information encoded by the spike train is the difference between the total and noise entropies. We thus computed the information as:

IH=total − H noise . (6)

Both the noise and the total entropies were normalized by T to yield entropy rates

(bits/s). Since entropy is sensitive to the word length, we extrapolated the entropy

113

rates in the limit of T→ ∞, yielding the true rates, and using (4) the information

rate.

Results

To investigate the role of temporal and cross–correlations in synaptic

inputs on information transmission in cortical neurons we modeled excitatory and

inhibitory synaptic conductances as two separate input channels injected into a

single–compartment conductance–based Hodgkin–Huxley model neuron with

stochastic biophysics. Ion channel stochasticity is essential for this model to

recreate biologically faithful spike behavior (Fitzhugh 1965; Schneidman et al.

1998; Skaugen 1979; Strassberg and Defelice 1993). To carry out this

computationally expensive task, we applied the stochastic shielding

approximation to ion channel gating, which has been shown to recreate the

behavior of stochastic Hodgkin–Huxley models using substantially less

computational power while preserving accuracy (Schmandt and Galán 2012).

Central to our method was our ability to generate trains of synaptic inputs with

Poisson statistics and precisely controlled temporal and cross–correlations. The temporal correlations of the barrages arise from the synaptic kinetics whereas cross–correlations are created by shifting two identical barrages relative to each other.

114

Magnitude, kinetics, and temporal correlation of synaptic excitation and

inhibition

We modeled excitatory and inhibitory conductances as two separate

channels of Poisson–distributed events whose rate was set by the fixed parameter

λ (5 ms–1), which is inversely proportional to the average inter–event interval, and whose kinetics were varied across a range of τ values (1–10 ms), representing the time–constant of the synaptic conductance decay. This time constant introduces a temporal correlation (auto–correlation time) in the synaptic barrage (see

Methods). Excitatory and inhibitory conductance amplitudes were either matched

2 2 (gexc = 3 pS/µm , ginh = 3 pS/µm ) to simulate the balanced conductances regime

or the inhibitory conductance was multiplied by a factor of 8, which in our model

generated approximately balanced excitatory and inhibitory synaptic currents (gexc

2 2 = 3 pS/µm , ginh = 24 pS/µm ) on average. Figure 4.1A represents the raw conductance traces in both the balanced conductance (Fig. 4.1A, left) and balanced currents regimes (Fig. 4.1A, right). For determining the effect of synaptic kinetics on information rates in cortical neurons, we generated synaptic input trains with different decay kinetics. The τ value was identical for a given pair of excitatory and inhibitory conductances, but was varied across different simulations. Figure 4.1B depicts a unitary synaptic conductance across a range of

τ values. This value visibly sets the width of the time–varying conductance without affecting the rise time or the time at peak amplitude (Fig. 4.1B). To understand how the kinetics shape the correlation structure between synaptic inputs, we analyzed the cross–correlogram between excitatory and inhibitory

115

synaptic conductances as a function of the synaptic kinetics (Fig. 4.1C). Clearly,

the cross–correlation between excitation and inhibition is affected by the kinetics

of the synaptic conductances, which are identical for both channels. As expected,

the decay of the cross–correlation decays proportionally with the decay of the

conductance waveform itself. Indeed, the numerically determined cross–

correlation of the synaptic input trains (circles) accurately fits the analytically

derived cross–correlation values (lines). Central to our goal is the ability to also

manipulate the cross–correlation between synaptic excitation and inhibition. To

this end, we introduce a relative lag, δ, which delays the inhibitory conductance with respect to the excitatory. Figure 4.1d shows the raw conductance traces with inhibition lagging excitation by δ = 5 ms. In the context of cortical networks, such lagged correlations can arise through the FFI circuit (Fig. 4.1D; inset schematic).

This motif enables disynaptic inhibition generated by local interneurons (black circle) to lag behind monosynaptic excitation (black triangle) with delays ranging from 1–10 ms (Okun and Lampl 2008; Wehr and Zador 2003; Wu et al. 2008) thus, providing cortical neurons with windows of integration whose width is determined by the relative lag between excitatory and inhibitory conductances and their decay kinetics.

Spiking behavior of a stochastic Hodgkin–Huxley neuron in response to kinetically variant synaptic inputs

The synaptic conductances described in the section above were injected into a single compartment Hodgkin–Huxley model of a neuronal membrane (100

µm2) with stochastic voltage–gated Na+ and K+ conductances, and a deterministic

116 leak conductance. When the magnitude of the synaptic conductances was set to zero, the neuron fired spontaneously at ~30 Hz. This spontaneous firing resulted from the stochastic flickering of voltage–gated ion channels, as did the subthreshold oscillations of the membrane voltage seen during periods of quiescence (Fig. 4.2A; top row). Depicted in Figure 4.2A are also the spike traces of the model neuron injected with matched excitatory and inhibitory conductances

(middle row) and matched currents (bottom row). In both regimes, neurons were presented with trains of synaptic events with either fast (τ = 1 ms; left column) or slow kinetics (τ = 10 ms; right column). In the presence of excitatory and inhibitory balanced conductances with fast kinetics, the firing rate increased to

~60 Hz (Fig. 4.2A; left column, middle row) and further increased to ~80 Hz when the kinetics were slow (Fig. 4.2A; right column, middle row). In the synaptic input regime of balanced currents, the spike rate increased to ~60 Hz when the synaptic kinetics were fast, but dropped to ~30 Hz when the conductance decay was slow (Fig. 4.2A; right column, bottom row). Thus, the decay time constant of the synaptic conductance impacts the firing rate differentially in the presence of balanced conductances versus balanced currents.

This becomes apparent when the firing rate is determined across the full range of synaptic kinetics and lags in both the balanced conductance (Fig. 4.2B) and balanced currents (Fig. 4.2C) regimes. Longer decay kinetics effectively increase the firing rate when the synaptic conductances are balanced and reduce the firing rate when the currents are balanced. Moreover, when the synaptic conductances are balanced, the firing rate shows no dependence on the lag between excitation

117

and inhibition, but when the currents are balanced the firing rate is highly

sensitive to short lags (0 > δ > 3 ms; Fig. 4.2C).

Entropy of neural spike trains

Information content of a sequence of action potentials is by definition related to the variability of spike timing in response to an input signal. Repeated presentation of the same input conductance to the model neuron, therefore, enables us to measure the reproducibility of the resulting spike pattern. The top panel of Figure 4.3A shows the spike trains in response to the same input conductance across ten trials (frozen input). Applying entropy measures as a proxy for spike variability, we obtain the “noise entropy” of the response across repeated presentations of the input signal. Noise entropy, however, informs us only about the spike variability to a single input pattern. To account for the full spectrum of potential spike responses of the model neuron, we presented a different set of input conductances across trials (unfrozen input; Fig. 4.3A, bottom), this time yielding the “total entropy” of the spike train. Entropy measurements were carried out using the “direct method” (see Methods), which involved converting the output signal into a binary string of 0’s and 1’s by binning the spike trace with small time windows (∆t = 5 ms) and counting spikes within each bin. A bin containing no spikes corresponds to a 0, while a bin containing one or more spikes corresponds to a 1 (Figure 4.3B). We then

generated sequences of words of various lengths (T = ∆t × number of bins) which

were then used to calculate the entropy based on the probability of occurrence of

each possible word. Entropy, being an extensive property, scales with the length

118

of the signal being measured and is sensitive to the temporal resolution of binning

(Strong et al. 1998). Thus, to estimate the maximal entropy of the spike trains we

extrapolated the entropy for words of length T → ∞ (Fig. 4.3C). Furthermore, to

deal with the issue of extensivity, entropies were normalized by time to give

entropy rates per time unit (bits/s).

We calculated the entropy rates for sets of paired excitatory and inhibitory

conductances across a range of lag times for inhibition. Figure 4.3D shows the

entropy rates as a function of δ for the sample traces shown in Figure 4.3A. These rates are exemplary of only a single value of the synaptic kinetics and are presented strictly heuristically. Our results show that the total and noise entropies are initially very low when excitatory and inhibitory conductances occur simultaneously but rise rapidly across a short range of δ values until they plateau around δ = 2 ms. The following sections will deal with the use of these entropy rates for the determination of spike train information rates.

Information rate of spike trains is insensitive to synaptic kinetics and the relative delay of synaptic inhibition in the balanced conductances regime

Measuring the information rates of neural spike trains requires that we take the difference between the total and noise entropy rates. This difference quantifies the information rate (IR) without necessitating assumptions about the nature of the signal being represented. We applied this measure to spike trains generated across

a range of τ and δ values to assess the dependence of the IR on the temporal

correlation of excitatory and inhibitory synaptic inputs. This was first done for the

balanced conductances regime. The top panel of Figure 4.4A shows the

119 dependence of the information across a range of τ and δ values. The IR remains high and constant across different values of δ when the synaptic kinetics are fast

(τ < 4 ms) and increases by no more than 40% with increasing lags when the kinetics are slow (τ > 5 ms). The bottom left panel of Figure 4.4A shows three slices taken from the surface plot corresponding to the IR for three values of δ as a function of the kinetics. The selection of these three points will be clearly explained in the next section. Visible from this panel is that, for the three different

δ values, the change in IR follows a similar trajectory: for fast kinetics the IR increases until reaching a maximum at ~ 3ms and then decreases for higher values of τ. The relationship between the IR and τ cannot be accounted for by changes in firing rate, which increases monotonically with increasing τ for all values of δ

(Fig. 4.4A, bottom right; Fig. 4.2B). To quantify the range of information rates of the spike train across the full range of synaptic kinetics, we took the difference between the maximal and minimal IR values along the τ dimension for different δ values and saw that the IR range was highest for short lags (35 bits/s) and decreased steadily with increasing lags. This decrease in the IR range of the spike train corresponds to the flattening of the IR curve across the τ dimension with increasing values of δ.

Information rate of spike trains exhibits dependence on synaptic kinetics at short delays for inhibition in the balanced currents regime

We next determined how the IR changes with τ and δ in the balanced synaptic currents regime. The surface plot in top panel of Figure 4.5A shows an entirely different dependency of the IR on synaptic kinetics and relative lags times. For δ

120

> 2 ms, the IR remains high and relatively constant across different values of τ; however, as the synaptic lags decrease below 2 ms, the IR begins to show an optimal dependence to the synaptic kinetics. The bottom left panel of Figure 4.5A shows the IR as a function of τ for three values of δ. These results show that for instantaneous lags (δ = 0 ms) the IR is relatively low (IR < 20 bits/s) and decreases slowly across the τ dimension; for δ = 4 ms, the IR does not undergo dramatic changes and remains relatively high (IR > 95 bits/s); for δ = 0.8 ms, however, the IR begins at an intermediate value (70 bits/s) and increases until it reaches an optimum at τ = 4 ms, then drops 55% relative to the maximal value

(Fig. 4.5A, bottom left). Again, this dependency of the IR on the kinetics cannot be explained by changes in firing rate which decrease approximately monotonically with increasing τ values (Fig. 4.5A, bottom right). Applying the same analysis used in the previous section, we compute the IR range across the τ dimension for different values of δ and observe that the IR exhibits the most dramatic dependence of synaptic kinetics at an optimal value of the relative lag (δ

= 0.8 ms). Thus, the selection of the three δ values shown in the bottom panels of

Figures 4.4A and 4.5A and circled in red in Figures 4.4B and 4.5B are based on the range of δ values within which the optimum occurs (δ = 0 ms: δ = 5 ms) and the δ value at the optimum (δ = 0.8 ms). This optimal dependence of the IR on synaptic kinetics is only present when the synaptic currents are balanced, but not the synaptic conductances (Fig. 4.4B; Fig. 4.5B). Interestingly, when the synaptic input trains are normalized by the integral of their conductance, which eliminates the scaling of the time–averaged mean conductance by τ, we see that the IR

121

decreases slowly and monotonically with τ in the balanced conductance regime

and the observed peak of the IR as a function of τ disappears in the balanced

currents regime (data not shown). This regime, however, is not physiologically

relevant considering that changes in synaptic kinetics in biological neurons are

not compensated for by changes in the amplitude of the synaptic inputs on an

event–by–event basis.

Discussion

In this study we set out to investigate how the encoding of information in

neurons depends on the temporal and cross–correlation of balanced synaptic inputs. We manipulated the correlation between identical trains of excitatory and inhibitory inputs by directly controlling the decay kinetics (τ) of the synaptic conductance and/or the relative time delay between excitation and inhibition (δ),

with inhibition always lagging behind. Our results show that the encoding of

information in neural spike trains exhibits a dependence on the correlation

between balanced excitatory and inhibitory synaptic currents and that this

dependence is absent in the input regime of balanced synaptic conductances.

Specifically, our findings demonstrate that the synaptic kinetics modulate the IR

range at which the spike train maximally encodes information, but do so only

when synaptic inhibition lags behind excitation with very short monosynaptic

delays (δ < 2 ms). Furthermore, our model exhibits an optimal delay (δ = 0.8 ms)

for inhibition at which the modulation of the IR by the synaptic kinetics is

highest. Such delays between excitation and inhibition are within the

physiological range of monosynaptic lags obtained from in vitro and in vivo

122 recordings of synaptic barrages in cortical neurons (Okun and Lampl 2008; Wehr and Zador 2003; Wu et al. 2008). The optimum of the IR as a function of τ emerges as the result of the following: As stated in the Methods, the IR is determined as a difference between the total and noise entropies which represent the variability of the spike patterns in response to unfrozen and frozen input trains, respectively. Intuitively and empirically, the total entropy is substantially larger across different values of τ and δ and changes most drastically at values of

δ < 2ms (data not shown), at which the neuronal spiking is subject to a dramatic modulation by the inhibitory inputs. It is at this exact range of δ that the noise entropy is the highest across the τ (for τ < 5 ms) and δ dimensions. Why is the noise entropy highest when synaptic kinetics are fast? To answer this question we consider the relationship between synaptic input–driven spiking and spontaneous firing from stochastic fluctuations of intrinsic regenerative conductances. During synaptic bombardment, the synaptic conductance is the dominant driver of neuronal firing as it overwhelms the intrinsic conductances in both magnitude and duration. However, with increasing synaptic kinetics (smaller τ), the integral of the synaptic conductance decreases and the dominance of the synaptic conductance abates, so that stochastic fluctuations of the intrinsic conductances allow for spontaneous firing. As a result, the spike patterns become more variable, thereby increasing the noise entropy. Thus, the peak of the IR emerges as a result of this increase in the noise entropy at small values of τ and δ and endows the neuronal membrane with the observed dependence of the IR on synaptic kinetics.

These results are consistent with a previous report showing that spike–time

123

reliability, an analytic measure related to the information capacity of a spike train,

shows an optimal value at specific auto–correlation times of their uncorrelated

synaptic inputs (Galán et al. 2008). Therefore, we conclude that synaptic kinetics

as well as relative delays between synaptic excitation and inhibition may be tuned

to optimize information transfer between neurons.

The time–course of the postsynaptic response may be subject to modulation by various factors including electrotonic distance of inputs from sites

of integration (Kleppe and Robinson 1999), pattern of afferent activation (Walker

et al. 2002), driving force (Salin and Prince 1996), postsynaptic receptor (and

subunit) type (Bannister et al. 2005; Cohen et al. 2000; Kirson and Yaari 1996),

intrinsic conductances (Miller et al. 1985; Wilson 1995), and the presence of

receptor–specific drugs (Cohen et al. 2000; Orser et al. 1994; Poncer et al. 1996).

The many ways in which the kinetics of the postsynaptic response to incoming

inputs can be altered provides cortical neurons with a myriad of mechanisms to

tune the correlation structure of incoming synaptic inputs. In particular, drugs,

neuromodulators, etc. may change the synaptic kinetics to the point that the IR is

outside its range, thereby deteriorating the processing of information in the brain

and altering the state of awareness and consciousness.

We have shown here that the IR is also sensitive to the precise arrival

times of inhibition with respect to excitation. Precisely controlling monosynaptic

delay times for inhibition may be less trivial than tuning the synaptic kinetics, but

is still possible. In the context of a feed–forward inhibitory circuit, one potential

mechanism to tune inhibitory lags may be to alter the integration times of the

124 feed–forward interneuron. Experiments in rats have shown that integration time in layer 4 stellate cells of somatosensory cortex is tightly regulated by thalamocortical feed–forward inhibition, thus controlling the precise spike timing of those neurons (Gabernet et al. 2005). Cortical interneurons also receive reciprocal inhibition (Lee et al. 2013; Pfeffer et al. 2013; Pi et al. 2013) and, as a consequence, are likely to have their integration windows regulated by inhibitory circuits. The size of the integration window of the feed–forward interneuron would control its precise spike timing and, thus, the lag of the inhibition in the excitatory neuron. Another way in which delays in inhibition can be modulated in a feed–forward circuit is through recruitment of distinct inhibitory networks

(Beierlein et al. 2003). These networks are comprised of molecularly and physiologically distinct interneuron populations that exhibit differential responsiveness to temporally patterned inputs and distinct synaptic dynamics.

Pivotal to the simulations carried out in this study was our ability to efficiently simulate the spike behavior of the model neuron repeatedly across numerous trials (sampling rate = 10 KHz; 5 s/trial; 56 trials for each δ and τ; which yields ~2,2 Gigabytes per data point in Figs. 4a and 5a). Stochastic simulations of ion channels are notoriously expensive computationally and often create the bottleneck for generating sufficient data across a large enough parameter range. We used the stochastic shielding approximation (SSA) for simulating stochastic ion channel gating dynamics (Schmandt and Galán 2012) to avoid this problem. The SSA reduces the number of ion channel states requiring stochastic simulation, and therefore, dramatically reduces the computational load.

125

Modeling of the synaptic inputs required that several assumptions be made

about the nature of cortical excitation and inhibition. First, the model assumes that

excitatory and inhibitory synaptic inputs are correlated. This assumption has been validated by in vitro (Graupner and Reyes 2013) and in vivo (Okun and Lampl

2008; Wehr and Zador 2003; Wu et al. 2008) recordings of synaptic barrages from cortical neurons showing that, indeed, excitation and inhibition are correlated in magnitude and timing, with inhibition tracking excitation by a few milliseconds. Secondly, the rate of synaptic events in time was assumed to be fast

(5 ms inter–event interval), corresponding to high levels of correlated activity in presynaptic neurons. Recordings from cortical neurons in awake behaving mice during sensory stimulation (Crochet and Petersen 2006), in anaesthetized ferrets during spontaneous active states (Haider et al. 2006), and in spontaneous active cortical slices (Compte et al. 2008) have confirmed high rates of synaptic

bombardment, therefore, lending validation to the use of high rates of synaptic

activity in our model. It is important to note, however, that synaptic inputs onto

cortical neurons have also been shown to occur as sparse and synchronous

population events (DeWeese and Zador 2006; Wehr and Zador 2003). Our study

focused exclusively on synaptic regimes with high levels of activity, thus, it will

be important to understand how temporal correlations between excitation and

inhibition in sparse regimes affect information encoding. Previous findings by

Miura and colleaugues suggest that balanced excitation and inhibition in cortical

neurons may in fact decouple irregularity of the spike train from rate modulations

in firing, which may arise from changes in the synaptic input rate (Miura et al.

126

2007). Thus, the information rate of the spike trains, being dependent on

irregularity of the spike times, may be insensitive to changes in synaptic input rate

if excitation and inhibition are balanced. On a related note, the magnitude of

synaptic inhibition has been shown to have an inverse relationship with the

overall rate of synaptic activity (Taub et al. 2013). This dependence shifts the relative balance between excitation and inhibition and may have a profound effect on encoding of information in cortical neurons. Future studies will need to address this problem to better understand the role of synaptic dynamics in neural coding.

The bulk of our study focused on the role of balanced synaptic inputs in encoding of information. A previous study by Sengupta and colleagues demonstrated that uncorrelated and balanced synaptic currents maximize the coding and metabolic efficiency of neuronal spikes by reducing the spike rate without substantially affecting the information rates (Sengupta et al. 2013). Our work applies the information theoretic approach using a similar model of a stochastic Hodgkin–Huxley neuron to address a different question: Are the correlations between synaptic inputs relevant for information processing? Our results indeed show a dependency of information encoding on the correlation between balanced synaptic currents. Moreover, the dependence of the information rates on the synaptic kinetics cannot be accounted for by changes in firing rate.

Though balanced synaptic currents effectively decrease the firing rate as the kinetics slow down, this relationship is monotonic and does not exhibit the optimum dependence upon the kinetics seen for the information rate.

127

Using a conductance–based single–compartment Hodgkin–Huxley model

offers insight into the interaction between synaptic inputs and an active neuronal

membrane, but it ignores the complex shape and electrotonic geometry of cortical

neurons. These features are important for spatiotemporal integration of synaptic

inputs (Bernander et al. 1991) since distal dendritic inputs may be processed

differently due to interactions with active dendritic conductances (Miller et al.

1985), differences in electrotonic properties (Kleppe and Robinson 1999) or

longer integration times caused by differences in feed–forward inhibition (Pouille et al. 2009; Pouille and Scanziani 2001). Future studies should consider the complex geometry of cortical neurons and how it may impact information processing.

An important aspect of the work presented herein is its focus on information encoding at the level of individual neurons. Though single cells certainly have the capacity to encode and represent information (Nemenman et al.

2008), distributed networks of anatomically and functionally connected neurons

(i.e. neural ensembles) also carry out this task (Ince et al. 2013; Nicolelis et al.

1995; Rothschild et al. 2010). The role of balanced synaptic inputs on information transfer in cortical networks has been addressed in previous studies. For instance, using multi–site recordings of local field potentials (LFP) in rats and monkeys,

Shew et al. showed that cortical networks with balanced excitation and inhibition maximize information capacity and transfer (Shew et al. 2011). Our results are in agreement with these findings, showing optimal information rates of neural spike trains when synaptic currents are balanced. It is important to note, however, that

128

LFP recordings of cortical networks capture coordinated activity of large

ensembles of neurons operating at substantially slower time–scales than that of

single neurons. Thus, the nature of the computations performed and information

encoded at the level of single cells versus that of neural ensembles is likely to

have marginal correspondence.

In conclusion, we provide a biologically realistic model of cortical neurons with stochastic ion channel biophysics and synaptic inputs and apply information theoretic approaches to show that information rates of neural spike trains is dependent on the temporal correlation of balanced synaptic currents. Our

findings emphasize the importance of these correlations for information encoding

and suggest that cortical neurons may optimize this process through precise

tuning of synaptic kinetics and synchrony of excitatory and inhibitory inputs.

129

Figure 4.1. Modeling excitatory and inhibitory synaptic inputs

(A, left) Raw traces of identical and balanced excitatory (gexc; red trace) and inhibitory (ginh; blue trace) synaptic conductances. (A, right) Raw traces of excitatory and inhibitory synaptic conductances necessary to generate approximately balanced postsynaptic currents. Note that the vertical scale bars differ in magnitude between (A, left) and (A, right). (B) Raw traces of a single synaptic event plotted as the time course of the synaptic conductance change across different values of τ, which corresponds to the decay time constant. As expected, only the decay phase of the conductance is affected by varying τ. (C)

Normalized cross–correlation plot of balanced excitatory and inhibitory synaptic conductances shown in (A, left) calculated across a range of τ values. Colored circles correspond to the experimentally determined cross–correlation of Poisson synaptic input trains, while solid colored lines correspond to the analytically derived cross–correlation (see Methods). Note that the width of the cross– correlation function broadens with increasing τ. (D) Raw traces of identical balanced excitatory and inhibitory synaptic conductances offset with respect to each other by δ = 5 ms. The inset shows a feed–forward inhibition circuit configuration that can generate such lags between identical trains of excitation and inhibition. The left triangle corresponds to the afferent input that activates excitatory target neurons (right triangle) monosynaptically and inhibitory interneurons (circle) disynaptically.

130

131

Figure 4.2. Firing properties of a stochastic Hodgkin–Huxley neuron in different input regimes.

(A) Raw traces of membrane potential dynamics in three different input regimes: top row corresponds to spontaneous firing in the absence of synaptic inputs; middle row corresponds to firing in response to balanced synaptic conductances; bottom row corresponds to firing in response to balanced synaptic currents. The left column represents neuronal firing in response to synaptic events with very fast kinetics (τ = 1 ms) and the right column represents firing in response to synaptic events with slow kinetics (τ = 10 ms). Note that the offset between excitation and inhibition in all traces is set at δ = 5 ms. (B) Surface plot of firing rates of the model neuron in response to balanced synaptic conductances with varying synaptic kinetics (τ) and relative lags between excitation and inhibition

(δ). (C) Surface plot of firing rates of the model neuron in response to balanced synaptic currents with varying synaptic kinetics and relative lags between excitation and inhibition.

132

133

Figure 4.3. Entropy of neural spike trains.

(A, top) Sample raster plots of neuronal firing in response to the presentation of a

fixed stimulus (i.e. frozen input) across 10 trials. (A, bottom) Raster plots of

neuronal firing in response to the presentation of different stimuli (i.e. unfrozen

input) across 10 trials. (B) Schematic showing how spike trains (represented by spike raster) were converted to binary strings of 0s and 1s by binning the voltage trace into time bins of size Δt = 5 ms. From these strings, words of various lengths were generated and the probability of their occurrence was calculated to yield entropy rates. (C) Entropy rates for spike responses in response to frozen input

(noise entropy) and in response to unfrozen input (total entropy) calculated across different word durations. The true entropy rates were extrapolated by taking the entropy rate in the limit of T→∞ (or 1/T → 0). (D) True entropy rates of neural

spike trains in response to synaptic inputs with different lag times (δ) between

excitation and inhibition.

134

135

Figure 4.4. Information rates of neural spike trains in the balanced conductances

regime.

(A, top) Surface plot of the information rate of neural spike trains as a function of

the synaptic kinetics (τ) and delays in inhibition relative to excitation (δ). (A,

bottom left) Information rate as a function of synaptic kinetics for three different

values of δ. This plot corresponds to three different slices taken from (A, top). (A, bottom right) Plot of firing rate as a function of synaptic kinetics for the same three δ values presented in (A, bottom left) shows that the dependency of the information rate on the synaptic kinetics is not accounted for by similar changes in firing rate. (B) Plot of information range as a function of relative delay between excitation and inhibition. The red dots correspond to the three values of δ shown in (A, bottom left) and (A, bottom right) (see Results for explanation).

136

137

Figure 4.5. Information rates of neural spike trains in the balanced currents

regime.

(A, top) Surface plot of the information rate of neural spike trains as a function of

the synaptic kinetics (τ) and delays in inhibition relative to excitation (δ). (A,

bottom left) Information rate as a function of synaptic kinetics for three different

138

139

Chapter 5: General Discussion

Thesis Overview

The central question of this thesis relates to the primary role of

neuromodulation in tuning cortical excitability and shaping the temporal

patterning of activation of cortical networks. For the larger part of my thesis I

focused on the action of the monoamine, serotonin (5–HT), on cortical neurons

and networks. First, I used single–cell patch clamp electrophysiological

techniques, pharmacological manipulations, computational modeling, and acute

behavioral seizures coupled with in vivo electroencephalographic (EEG)

recordings to address the role of 5–HT in facilitating transitions between distinct

epileptiform states within the cortex. From these experiments (described in detail

in Chapter 2), I found that cortical neurons receive a persistent and ongoing

source of synaptic noise in the form of sEPSCs mediated by ionotropic 5–HT3

receptors that can be augmented with the selective serotonin reuptake inhibitor

(SSRI), fluoxetine. I then used a partially disinhibited cortical slice preparation

that allowed me to probe synchronized network activity with recordings from single cortical neurons. In this paradigm, I showed that increasing endogenous

cortical 5–HT levels with fluoxetine leads to an increase in network excitability

and a transformation of cortical network dynamics from temporally sparse and

random network bursts (PDS) to clustered and periodic epileptiform oscillations

(fast runs). Importantly, I showed that augmenting synaptic noise with fluoxetine

can account for part of the increase in cortical network excitability, but that the

transformation of the dynamics depends on an enhancement in excitatory

140

coupling between cortical pyramidal cells in a 5–HT2 receptor–dependent manner. Furthermore, using convulsant–induced acute behavioral seizures I demonstrated that blocking 5–HT2 receptors prior to convulsant administration leads to an increase in seizure threshold and a delay in the emergence of cortical epileptiform oscillations in vivo. These findings provide a mechanistic account for how increases in endogenous cortical 5–HT can cause a change in cortical network dynamics. Importantly, these results find agreement with the notion that control of the E/I balance through neuromodulation effectively changes the qualitative behavior of an anatomically unchanging (given the short time scale of the experiments) neuronal network.

Chapter 3 of this dissertation dealt with the consequences of systemic forebrain 5–HT depletion on cortical neuron and network excitability. For this project I used a transgenic mouse lacking the ETS domain transcription factor

Pet–1 (Hendricks et al. 2003) generously provided by Dr. Evan Deneris. Mice lacking Pet–1 show a 70% loss of serotonergic neurons within the raphe nuclei and an 80% reduction in forebrain 5–HT levels. The motivation behind this line of questioning was to understand the effects of systemic forebrain 5–HT depletion on the excitability and activity of cortical neurons and networks. I showed that cortical neurons within the somatosensory cortex of Pet–1 null mice exhibit altered cell–intrinsic excitability. Specifically, I showed that measures of passive membrane properties of cortical pyramidal neurons are biased to be more excitable than those of wild–type C57BL/6 mice, while active membrane properties are less excitable. Furthermore, I reported that synaptic excitability is

141

altered in Pet–1 null mice as well. This was measured as an increase in the

amplitude of sEPSCs impinging onto cortical neurons. Synaptic hyperexcitability

was associated with altered network activity in mice lacking Pet–1. Using a

disinhibited slice preparation, I showed that cortical networks in Pet–1 null mice

were also hyperexcitable, exhibiting epileptiform fast runs at concentrations that

only elicited single network bursts (PDS) in slices from wild–type mice. Such

hyperexcitability was normalized to control levels with the 5–HT2 receptor

antagonist, a finding that was consistent with a previous report showing hyper–

sensitive 5–HT2 receptors in mice lacking Pet–1. I also tested seizure susceptibility in Pet–1 null mice using the same protocol applied to wild–type mice in Chapter 2, but observed no obvious changes in parameters of seizure susceptibility, suggesting potential compensatory mechanisms in extra–cortical circuits involved in seizure generalization. The main conclusion from these findings was that a structure downstream of serotonergic raphe neurons (i.e. the cortex) was dramatically affected by deletion of the Pet–1 gene. This resulted in compensatory changes in the cell intrinsic, synaptic, and network excitability within the cortex of Pet–1 null mice. To my best knowledge, this is the first report of altered cortical excitability in these mice.

In Chapter 4, I use a computational modeling approach to understand how the statistical correlations of excitatory and inhibitory synaptic inputs contribute to information processing in cortical neurons embedded in feed–forward inhibition circuit motifs. Using a Hodgkin–Huxley model of a cortical neuron endowed with stochastic ionic conductances and synaptic inputs, I assessed how

142

the delay between synaptic excitation and inhibition, as well as their kinetics,

affects the information rate of neural spike trains. I tested how the information rate changes as a function of both the synaptic decay kinetics and excitatory– inhibitory delays in a balanced synaptic conductance and balanced synaptic currents regimes. These two cases correspond to distinct input regimes in which the E/I balance is biased to more excitation (balanced conductances) or equal excitation and inhibition (balanced currents). Since this work and that of others

(Moreau et al. 2010) show that 5–HT can effectively shift the E/I balance, this study indirectly addresses the contribution of neuromodulation to information processing among cortical neurons. I show that the information rate is critically dependent on the statistical correlations between excitatory and inhibitory inputs, but only when their synaptic currents are balanced. Moreover, there exists an optimal sensitivity of the information rate of neural spikes on the kinetics of synaptic inputs and such sensitivity emerges at short and physiologically relevant delays between excitatory and inhibitory synaptic inputs. These results provide a previously unexplored and important understanding that the statistical structure of synaptic excitation and inhibition is critical for information processing in the cortex. Furthermore, our findings suggest that neuromodulation, which can modify the E/I balance as well as the relative delays between excitation and inhibition arriving to cortical neurons, is also important for transmitting information among neurons.

143

On noise

A surprising finding reported in Chapter 2 of this thesis is that synaptic noise, in the form of 5–HT3 receptor–mediated sEPSCs, can enhance the

excitability of cortical networks. To be more precise, I demonstrated that blocking

5–HT3 receptors with granisetron, which reduces the synaptic noise onto

pyramidal neurons in vitro, leads to a reduction of neuronal network excitability.

One could argue, however, that the reduction in excitability was a result of

changes in activity of 5–HT3 receptor–expressing cortical inhibitory interneurons

(Puig et al. 2004). This is unlikely since 5–HT3 receptors on interneurons are excitatory, therefore, their blockade would lead to a decrease in interneuron activity and a resultant disinhibition, which should, in theory, increase network excitability. Instead, I demonstrate that blocking 5–HT3 receptors in the cortex leads to a decrease in network excitability in a disinhibited slice. Furthermore, using a simulated cortical network which is able to recreate the bistable activation dynamics observed in the disinhibited slice, I showed that augmenting synaptic noise by increasing the amplitude of sEPSCs is sufficient to increase the number of transitions to activated network states (simulated PDS). Combined, these results provide support for the notion that noise can promote transitions between distinct network states (in this case PDS or quiescence).

Noise in nervous system function has largely been treated by neurophysiologists as a nuisance that often obscures some other more interesting or informative signal. However, the work of Mainen and Sejnowski (Mainen and

Sejnowski 1995) substantially changed this point of view and offered a more

144

constructive role for noise in the nervous system. Using simple but elegant in

vitro electrophysiology from cortical neurons in slices, they demonstrated that injection of a constant current leads to variable spike trains while injection of noisy current (i.e. current with stochastic fluctuations) led to highly reliable patterns of spiking. This finding revealed that input noise can increase the fidelity of neuronal responses, a finding that has critical implications for neuronal coding.

Since cortical neurons are persistently bombarded by barrages of synaptic inputs in vivo (Haider and McCormick 2009), noise may serve a physiological

computational role on cortical networks (McDonnell and Ward 2011). Subsequent

studies have revealed that, in fact, noise can modulate the gain of neuronal

responses in the cortex (Destexhe et al. 2001; Haider and McCormick 2009; Shu

et al. 2003a). In the wake of these and other studies, a few important concepts, initially developed by physicists, were applied to the role of noise in the nervous system. Such concepts include stochastic resonance, coherence resonance, and stochastic facilitation. Stochastic resonance refers to a phenomenon in which detection of a subthreshold periodic input (e.g. subthreshold theta oscillations in entorhinal cortical cells) can be enhanced by introduction of an optimal amount of random input fluctuations (McDonnell and Ward 2011). In this context, large

enough random fluctuations of the input at the peak of depolarization can bring

the neuron above firing threshold, thus facilitating the detection of the periodic

input signal. Coherence resonance is a related phenomenon except it does not

require weak periodic forcing (i.e. periodic subthreshold inputs) to elicit

detectable periodic activity. Thus, within the context of coherence resonance,

145

input noise without weak periodic forcing can enhance the intrinsic periodicity of

a bistable system and induce oscillations. Another way to describe coherence

resonance then is “noise–induced oscillations.” Since oscillations in the activity of

cortical neurons are critical for the computational tasks mediated by the cortex,

coherence resonance may bestow onto noise yet another physiologically relevant

role. Stochastic facilitation is a more recent and general concept that encompasses

the two preceding. Stochastic facilitation refers to any role of noise that enhances

the performance of a system with distinct states (McDonnell and Ward 2011).

Enhancement, however, is a term that bears with it the subjective connotation of

better performance. In the context of our findings, I discard this connotation and

simply posit that synaptic noise mediated by 5–HT3 receptors may be a biological

substrate for stochastic facilitation in cortical networks owing to its ability to

increase the probability of transitioning into an active network state. By such

means, stochastic facilitation via synaptic noise may enhance cortical network

excitability. Such noise–dependent regulation of cortical network activity

provides a new perspective through which to understand neuromodulation. In the context of our results, persistent noise mediated by fast ionotropic 5–HT3 receptors provides an excitatory tone that can be set by global neuromodulatory states (i.e. high/low 5–HT) that impacts cortical network excitability.

Another interesting role for noise gathered from this thesis is the one discussed in Chapter 4 (see Discussion). As stated earlier, I showed that the modulation of the information rate of neural spiking depends on the statistical correlation of excitatory and inhibitory synaptic inputs, which is dictated by the

146

relative lag between excitation and inhibition as well as the decay kinetics of their

conductances. This optimal modulation of the information rate was present at

shorter time constants of the synaptic conductances and at short time lags between

excitation and inhibition. The shorter the time constants of the synaptic inputs, the

shorter the duration of the resultant postsynaptic current. Thus, for a constant

input rate, inputs of shorter durations will produce shorter periods in which the

neuronal firing is governed by synaptic drive. As a result, the stochastic

fluctuations of cell–intrinsic conductances underlying action potential generation

are able to elicit random firing when synaptic inputs are absent, thus augmenting

the variability of the neural spike train and increasing its information rate. In the

context of these findings, I show that the noise intrinsic to the neuron’s ionic

conductances interacts with the stochastic synaptic inputs to modulate the

encoding of information.

Serotonin as a neuromodulator of cortical network activity

Both in experimental animals and in humans, 5–HT affects higher level cognitive functions known to depend on the activity of cortical networks

(Brigman et al. 2010; Clarke et al. 2004; Harmer et al. 2003; Schmitt et al. 2006), suggesting that 5–HT does in fact influence cortical network activity. Detailed neurophysiological experiments to assess the role of 5–HT on network activity in the cortex has been sparse; however, a few important studies have made efforts in this direction. For instance, populations of cortical neurons in nonhuman primates have been shown to exhibit persistent firing activity during the delay period of an oculomotor delayed response task, a form of activity thought to be

147

the neural substrate of spatial working memory (SWM) (Fuster 1973). Work from

Albert Compte’s lab showed that antagonizing 5–HT2 receptors locally during

this task leads to a reduction in the persistent firing during the delay period in cortical pyramidal neurons and inhibitory interneurons (Williams et al. 2002).

Cano–Calino and colleagues then employed an elegant computational model of

SWM using a cortical network endowed with intrinsic responses to 5–HT to show

that 5–HT can modulate the network mechanisms underlying the generation of

persistent active states. While this line of investigation certainly yielded valuable

insights into the contribution of cortical serotonergic signaling during a behavioral

task and provided a potential network mechanism using a computational model, it

account for a highly specific dynamical regime of cortical networks. Thus, the

ability to generalize 5–HT’s role in the cortex to other patterns of activity is fairly

limited based on the findings of these studies alone.

Other studies, such as that of Puig and colleagues (Puig et al. 2010), have

addressed more general roles for 5–HT in cortical network activity. In this study,

the investigators inquired about the effect of endogenous 5–HT on synchronous

activity in the cortex. As discussed in the Chapter 1 (see Dynamics of Cortical

Activity), synchronous activity in the cortex can manifest as a periodic oscillations

in subthreshold membrane potential fluctuations or neuronal firing (Buzsaki and

Draguhn 2004). Puig and colleagues looked at one instance of such synchronous

activity – the cortical slow oscillation (< 2 Hz). While recording LFPs in the

mPFC and simultaneously stimulating the DRN with low–frequency bipolar

stimulation, they observed that 5–HT could increase the frequency of the cortical

148

slow oscillation. High–frequency stimulation, however, led to desynchronization

of both slow and gamma oscillations in their experimental paradigm. From this

work, the investigators concluded that serotonergic 5–HT1 and 5–HT2 receptors

had dual effects on synchronous cortical network activity. On the one hand, 5–

HT2 receptors acted to synchronize cortical pyramidal neurons involved in the

slow oscillation with no substantial contribution from 5–HT1 receptors. In

contrast, 5–HT1 receptors acted to desynchronize cortical gamma oscillations by

modulating the synchrony of the pacemaking fast–spiking interneurons. These

results are consistent with findings that 5–HT2 receptors and 5–HT1 receptors,

while both exhibiting a high degree of overlap on pyramidal neurons (Amargos-

Bosch et al. 2004), are differentially expressed on populations of inhibitory

interneurons (Jakab and Goldman-Rakic 2000; Puig et al. 2004). The study of

Puig et al. has been inarguably the most comprehensive analysis of serotonergic

modulation of ongoing cortical oscillations. A very intriguing aspect of their work

is the finding that DRN stimulation (increased cortical [5–HT]) can accelerate the

slow oscillation, since such acceleration is observed spontaneously in anesthetized

cats shortly before the cortical network enters into an epileptiform dynamical

regime characterized by fast run oscillations (10–15 Hz) (Steriade et al. 1993;

Timofeev et al. 1998). Perhaps one mechanism by which the cortical slow

oscillations can transform into epileptiform fast runs is through a transient

increase in cortical 5–HT levels. In favor of this idea is the finding that during slow–wave sleep (i.e. when the slow oscillation is prominent) ~25% of DRN

neurons increase their firing rates (Sakai 2011). Such changes in DRN activity

149 could presumably lead to elevations in cortical 5–HT concentrations and increased synchrony cortical pyramidal neurons. An excess in the synchronized activity of cortical neurons could potentially lead to the emergence of pathophysiological oscillatory regimes (McCormick and Contreras 2001).

The findings reported in this thesis also provides support for this idea, albeit, with a more subtle interpretation. As opposed to demonstrating modulation of ongoing oscillations like those shown by Puig et al. in vivo (Puig et al. 2010), I show that 5–HT can facilitate the emergence of fast run oscillations given that the preceding pattern of activity is a temporally random non–oscillatory network regime (i.e. PDS). Mechanistically the emergence of fast run oscillations in my work is accounted for by a 5–HT2 receptor–dependent increase in excitatory synaptic coupling between cortical neurons, though the contribution of 5–HT– modulated cell–intrinsic potassium conductances, such as those mediating the slow afterhyperpolarization (sAHP), cannot be excluded (Andrade et al. 2012;

Villalobos et al. 2005). It is therefore possible that such mechanisms could account for the spontaneous emergence of epileptiform oscillations in the cortex in vivo (Timofeev et al. 1998). Since the degree of recurrent excitatory activity among cortical neurons dictates the occurrence of epileptiform fast runs (Castro-

Alamancos and Rigas 2002), my work suggests that increasing endogenous 5–HT at the border of a transition between states (i.e. PDS to fast run) suffices to render the network “epileptic”. The computational model presented in Chapter 2 suggests that the transition is facilitated by a shift between synaptic excitation and inhibition. Increase in serotonergic signaling biases the E/I ratio in favor of more

150

excitation and allows for reverberant network discharges. As suggested by the

study of Castro–Alamancos and Rigas, the E/I ratio shift occurs through

unmasking of recurrent excitatory synaptic activity in superficial cortical layers

that mediates the oscillatory fast run discharges (Castro-Alamancos and Rigas

2002). The findings I present in this thesis are in agreement with this idea and

uncover a novel role for 5–HT in transforming cortical network dynamics.

One shortcoming of the disinhibited slice paradigm employed in Chapter

2 is that prior to manipulating components of the serotonergic system, the slice is already biased to an epileptiform state via partial disinhibition (i.e. shifting E/I balance). This makes interpretations about transitions between physiological and pathophysiological cortical regimes difficult. Part of this difficulty arises from the difficult of recreating physiological activity regimes (e.g. slow oscillations) in slices of the mouse cortex. Though previous reports have claimed to elicit periodic transitions between Up and Down states in cortical slices from the same mouse species used here by modifying the ACSF recipe (Cossart et al. 2003), we were unable to recreate their results with the same recipe and similar ones used by other groups in different animal models (Favero and Castro-Alamancos 2013; Shu et al. 2003b). Perhaps one way to circumvent this problem would be to use a different animal model in which Up and Down state transitions can be reliably reproduced with the same ACSF recipe and then manipulate extracellular 5–HT

levels to determine whether the slow oscillation can be transformed into an

epileptiform activity regime. This approach could encounter the problem of

151 species–specific differences in expression of 5–HT receptors (Nichols and

Nichols 2008).

Serotonin and epilepsy

In Chapter 2 of this dissertation I showed that systemic blockade of 5–

HT2 receptors prior to pentetrazol (convulsant) administration in vivo delays the onset of behavioral seizures and epileptiform fast run oscillations in the neocortex, a finding that is consistent with the in vitro results in the disinhibited slice. These results argue that 5–HT acts as a pro–convulsant by biasing the cortical E/I balance in favor of more excitation, thus predisposing the cortical network to elicit epileptic activity states characterized by repetitive discharges.

This finding stands in agreement with previous reports showing pro–convulsive effects of 5–HT. For instance, Bercovici et al. showed that experimentally induced absence seizures in rats were shortened in duration if 5–HT synthesis was inhibited prior to the seizure with para–chlorophenylalanine (PCPA) (Bercovici et al. 2006). Another study carried out by Freitas et al. used a pilocarpine model of status epilepticus to assess the effects of different pharmacological agents on seizure parameters (Freitas et al. 2006). Their results showed that fluoxetine pre– treatment resulted in pro–convulsive changes in seizure threshold, severity, and seizure–related mortality. In line with these studies, experiments carried out by

Ritz’s group reported attenuation of cocaine–induced convulsive seizures with antagonism of 5–HT2 receptors (O'Dell et al. 2000b). Consistent with these findings were those of Wada and colleagues in which fully kindled hippocampal seizures were modulated by serotonergic pharmacological agents (Wada et al.

152

1992). Specifically, they showed that 5–HT2 receptor agonists shortened the seizure latencies in kindled animals (Wada et al. 1992). Elevation of 5–HT levels with SSRIs in rats was also associated with an increase in seizure severity and increased immediate early gene expression in the hippocampus (Zienowicz et al.

2005). This list of studies frames the role of 5–HT, acting through 5–HT2 receptors, as a pro–convulsant. The studies above agree with the results I present in this thesis; namely, that 5–HT2 receptors act to facilitate the onset of seizures and epileptiform activity patterns within the neocortex. Antagonism of these receptors in mice treated with the convulsant PTZ increased seizure threshold and reduced seizure incidence. Thus, my results are in accord with the studies above that show a pro–convulsive role for 5–HT.

In contrast to the above studies, others have shown that 5–HT may exhibit anti–convulsive properties (Bagdy et al. 2007; Pisani et al. 1999). In a study carried out by Borowicz et al., fluoxetine was shown to increase anticonvulsive activities of commonly prescribed anti–epileptics in mice (Borowicz et al. 2007) probably through increasing serotonergic signaling within the CNS. In a different study, Hernandez and colleagues showed that increasing systemic 5–HT levels with fluoxetine in rats with pilocarpine–induced epilepsy resulted in a reduced spontaneous seizure rate (Hernandez et al. 2002). Work carried out in baboons with photosensitive epilepsy showed that administration of 5–HT prior to administration of tricyclic antidepressants (TCA) that normally augment epileptic response ended up in abolishing their pro–epileptic properties, suggesting that 5–

153

HT acted as an anti–convulsant and that the TCAs acted through a non–

serotonergic mechanism to exacerbate epileptic seizures (Trimble et al. 1977).

The discrepancies in results between different studies can be attributed to a broad variety of factors. These include but are not limited to the animal species and/or strain used for the experiments, whether a chronic or an acute epileptic model was being used, the type of convulsant used to generate seizures, the type of pharmacological agent used to target the serotonergic system and the concentration in which it was applied, as well as other differences in experimental approach (Loscher et al. 1990). Despite the controversy, the results presented here

clearly show the action of endogenous 5–HT to be pro–epileptic and blockade of

5–HT2 receptors to be protective from epileptic seizures in vivo. Further work

will be necessary to determine whether 5–HT can act both as a pro–convulsant

and an anti–convulsant based on its concentration in the circuits responsible for

generating epileptic seizures. This question is grounded on legitimate premises

considering that previously tested TCA, known to alter serotonergic concentration

in the brain, can have either effect based on their administered concentration

(Jobe et al. 1984; Meldrum et al. 1982; Woods et al. 1983).

154

Bibliography

Aghajanian GK, and Marek GJ. Serotonin, via 5-HT2A receptors, increases EPSCs in layer V pyramidal cells of prefrontal cortex by an asynchronous mode of glutamate release. Brain Res 825: 161-171, 1999. Agmon A, and Connors BW. Thalamocortical responses of mouse somatosensory (barrel) cortex in vitro. Neuroscience 41: 365-379, 1991. Amargos-Bosch M, Bortolozzi A, Puig MV, Serrats J, Adell A, Celada P, Toth M, Mengod G, and Artigas F. Co-expression and in vivo interaction of serotonin1A and serotonin2A receptors in pyramidal neurons of prefrontal cortex. Cereb Cortex 14: 281-299, 2004. Amzica F, and Steriade M. Disconnection of intracortical synaptic linkages disrupts synchronization of a slow oscillation. J Neurosci 15: 4658-4677, 1995. Anderson JS, Carandini M, and Ferster D. Orientation tuning of input conductance, excitation, and inhibition in cat primary visual cortex. J Neurophysiol 84: 909-926, 2000. Andrade R. Serotonergic regulation of neuronal excitability in the prefrontal cortex. Neuropharmacology 61: 382-386, 2011. Andrade R, Foehring RC, and Tzingounis AV. The calcium-activated slow AHP: cutting through the Gordian knot. Front Cell Neurosci 6: 47, 2012. Araneda R, and Andrade R. 5-Hydroxytryptamine2 and 5-hydroxytryptamine 1A receptors mediate opposing responses on membrane excitability in rat association cortex. Neuroscience 40: 399-412, 1991. Atallah BV, and Scanziani M. Instantaneous modulation of gamma oscillation frequency by balancing excitation with inhibition. Neuron 62: 566-577, 2009. Azmitia EC, Gannon PJ, Kheck NM, and Whitaker-Azmitia PM. Cellular localization of the 5-HT1A receptor in primate brain neurons and glial cells. Neuropsychopharmacology 14: 35-46, 1996. Bagdy G, Kecskemeti V, Riba P, and Jakus R. Serotonin and epilepsy. J Neurochem 100: 857-873, 2007. Bannister NJ, Benke TA, Mellor J, Scott H, Gurdal E, Crabtree JW, and Isaac JT. Developmental changes in AMPA and kainate receptor-mediated quantal transmission at thalamocortical synapses in the barrel cortex. J Neurosci 25: 5259-5271, 2005. Beierlein M, Gibson JR, and Connors BW. Two dynamically distinct inhibitory networks in layer 4 of the neocortex. J Neurophysiol 90: 2987-3000, 2003. Beique JC, Chapin-Penick EM, Mladenovic L, and Andrade R. Serotonergic facilitation of synaptic activity in the developing rat prefrontal cortex. J Physiol 556: 739-754, 2004. Beique JC, Imad M, Mladenovic L, Gingrich JA, and Andrade R. Mechanism of the 5- hydroxytryptamine 2A receptor-mediated facilitation of synaptic activity in prefrontal cortex. Proc Natl Acad Sci U S A 104: 9870-9875, 2007. Bercovici E, Cortez MA, Wang X, and Snead OC, 3rd. Serotonin depletion attenuates AY- 9944-mediated atypical absence seizures. Epilepsia 47: 240-246, 2006. Berger H. Electroencephalogram in humans. Archiv fur psychiatrie und nervenkrankheiten 87: 527-570, 1929. Berkeley JL, Decker MJ, and Levey AI. The role of muscarinic acetylcholine receptor- mediated activation of extracellular signal-regulated kinase 1/2 in pilocarpine-induced seizures. J Neurochem 82: 192-201, 2002.

155

Bernander O, Douglas RJ, Martin KA, and Koch C. Synaptic background activity influences spatiotemporal integration in single pyramidal cells. Proc Natl Acad Sci U S A 88: 11569-11573, 1991. Blue ME, Yagaloff KA, Mamounas LA, Hartig PR, and Molliver ME. Correspondence between 5-HT2 receptors and serotonergic axons in rat neocortex. Brain Res 453: 315- 328, 1988. Bolo NR, Hode Y, Nedelec JF, Laine E, Wagner G, and Macher JP. Brain pharmacokinetics and tissue distribution in vivo of fluvoxamine and fluoxetine by fluorine magnetic resonance spectroscopy. Neuropsychopharmacology 23: 428-438, 2000. Borowicz KK, Furmanek-Karwowska K, Sawicka K, Luszczki JJ, and Czuczwar SJ. Chronically administered fluoxetine enhances the anticonvulsant activity of conventional antiepileptic drugs in the mouse maximal electroshock model. Eur J Pharmacol 567: 77-82, 2007. Braitberg G, and Curry SC. Seizure after isolated fluoxetine overdose. Ann Emerg Med 26: 234-237, 1995. Brigman JL, Mathur P, Harvey-White J, Izquierdo A, Saksida LM, Bussey TJ, Fox S, Deneris E, Murphy DL, and Holmes A. Pharmacological or genetic inactivation of the serotonin transporter improves reversal learning in mice. Cereb Cortex 20: 1955-1963, 2010. Brown AM, Hope AG, Lambert JJ, and Peters JA. Ion permeation and conduction in a human recombinant 5-HT3 receptor subunit (h5-HT3A). J Physiol 507 ( Pt 3): 653-665, 1998. Burns SP, Xing D, and Shapley RM. Is gamma-band activity in the local field potential of V1 cortex a "clock" or filtered noise? J Neurosci 31: 9658-9664, 2011. Buzsaki G, and Draguhn A. Neuronal oscillations in cortical networks. Science 304: 1926- 1929, 2004. Cajal SRy. Comparative study of the sensory areas of the human cortex. 1899. Camfield PR. Definition and natural history of Lennox-Gastaut syndrome. Epilepsia 52 Suppl 5: 3-9, 2011. Castro-Alamancos MA. Origin of synchronized oscillations induced by neocortical disinhibition in vivo. J Neurosci 20: 9195-9206, 2000. Castro-Alamancos MA, and Gulati T. Neuromodulators produce distinct activated States in neocortex. J Neurosci 34: 12353-12367, 2014. Castro-Alamancos MA, and Rigas P. Synchronized oscillations caused by disinhibition in rodent neocortex are generated by recurrent synaptic activity mediated by AMPA receptors. J Physiol 542: 567-581, 2002. Celada P, Puig MV, and Artigas F. Serotonin modulation of cortical neurons and networks. Front Integr Neurosci 7: 25, 2013. Celada P, Puig MV, Diaz-Mataix L, and Artigas F. The hallucinogen DOI reduces low- frequency oscillations in rat prefrontal cortex: reversal by antipsychotic drugs. Biol Psychiatry 64: 392-400, 2008. Chagnac-Amitai Y, and Connors BW. Synchronized excitation and inhibition driven by intrinsically bursting neurons in neocortex. J Neurophysiol 62: 1149-1162, 1989. Chance FS, Abbott LF, and Reyes AD. Gain modulation from background synaptic input. Neuron 35: 773-782, 2002.

156

Ciarleglio CM, Resuehr HE, Axley JC, Deneris ES, and McMahon DG. Pet-1 deficiency alters the circadian clock and its temporal organization of behavior. PloS one 9: e97412, 2014. Clarke HF, Dalley JW, Crofts HS, Robbins TW, and Roberts AC. Cognitive inflexibility after prefrontal serotonin depletion. Science 304: 878-880, 2004. Cohen AS, Lin DD, and Coulter DA. Protracted postnatal development of inhibitory synaptic transmission in rat hippocampal area CA1 neurons. J Neurophysiol 84: 2465- 2476, 2000. Compte A, Reig R, Descalzo VF, Harvey MA, Puccini GD, and Sanchez-Vives MV. Spontaneous high-frequency (10-80 Hz) oscillations during up states in the cerebral cortex in vitro. J Neurosci 28: 13828-13844, 2008. Compte A, Sanchez-Vives MV, McCormick DA, and Wang XJ. Cellular and network mechanisms of slow oscillatory activity (<1 Hz) and wave propagations in a cortical network model. J Neurophysiol 89: 2707-2725, 2003. Constantinople CM, and Bruno RM. Effects and mechanisms of wakefulness on local cortical networks. Neuron 69: 1061-1068, 2011. Cornea-Hebert V, Riad M, Wu C, Singh SK, and Descarries L. Cellular and subcellular distribution of the serotonin 5-HT2A receptor in the central nervous system of adult rat. J Comp Neurol 409: 187-209, 1999. Cossart R, Aronov D, and Yuste R. Attractor dynamics of network UP states in the neocortex. Nature 423: 283-288, 2003. Crochet S, and Petersen CC. Correlating whisker behavior with membrane potential in barrel cortex of awake mice. Nat Neurosci 9: 608-610, 2006. Cruikshank SJ, Lewis TJ, and Connors BW. Synaptic basis for intense thalamocortical activation of feedforward inhibitory cells in neocortex. Nat Neurosci 10: 462-468, 2007. Cummings KJ, Commons KG, Hewitt JC, Daubenspeck JA, Li A, Kinney HC, and Nattie EE. Failed heart rate recovery at a critical age in 5-HT-deficient mice exposed to episodic anoxia: implications for SIDS. Journal of applied physiology 111: 825-833, 2011. Cummings KJ, Li A, Deneris ES, and Nattie EE. Bradycardia in serotonin-deficient Pet-1- /- mice: influence of respiratory dysfunction and hyperthermia over the first 2 postnatal weeks. American journal of physiology Regulatory, integrative and comparative physiology 298: R1333-1342, 2010. Dan Y, Alonso JM, Usrey WM, and Reid RC. Coding of visual information by precisely correlated spikes in the lateral geniculate nucleus. Nat Neurosci 1: 501-507, 1998. Davies PA, Pistis M, Hanna MC, Peters JA, Lambert JJ, Hales TG, and Kirkness EF. The 5- HT3B subunit is a major determinant of serotonin-receptor function. Nature 397: 359- 363, 1999. deCharms RC, and Merzenich MM. Primary cortical representation of sounds by the coordination of action-potential timing. Nature 381: 610-613, 1996. DeFelipe J, Arellano JI, Gomez A, Azmitia EC, and Munoz A. Pyramidal cell axons show a local specialization for GABA and 5-HT inputs in monkey and human cerebral cortex. J Comp Neurol 433: 148-155, 2001. Deneris ES, and Wyler SC. Serotonergic transcriptional networks and potential importance to mental health. Nat Neurosci 15: 519-527, 2012. Destexhe A, Rudolph M, Fellous JM, and Sejnowski TJ. Fluctuating synaptic conductances recreate in vivo-like activity in neocortical neurons. Neuroscience 107: 13- 24, 2001.

157

DeWeese MR, and Zador AM. Non-Gaussian membrane potential dynamics imply sparse, synchronous activity in auditory cortex. J Neurosci 26: 12206-12218, 2006. Diaz SL, Narboux-Neme N, Trowbridge S, Scotto-Lomassese S, Kleine Borgmann FB, Jessberger S, Giros B, Maroteaux L, Deneris E, and Gaspar P. Paradoxical increase in survival of newborn neurons in the dentate gyrus of mice with constitutive depletion of serotonin. Eur J Neurosci 38: 2650-2658, 2013. Doucet E, Latremoliere A, Darmon M, Hamon M, and Emerit MB. Immunolabelling of the 5-HT 3B receptor subunit in the central and peripheral nervous systems in rodents. Eur J Neurosci 26: 355-366, 2007. Dougherty JP, and Aloyo VJ. Pharmacological and behavioral characterization of the 5- HT2A receptor in C57BL/6N mice. Psychopharmacology (Berl) 215: 581-593, 2011. Economo MN, and White JA. Membrane properties and the balance between excitation and inhibition control gamma-frequency oscillations arising from feedback inhibition. PLoS Comput Biol 8: e1002354, 2012. Engel AK, Fries P, and Singer W. Dynamic predictions: oscillations and synchrony in top- down processing. Nat Rev Neurosci 2: 704-716, 2001. Erickson JT, Shafer G, Rossetti MD, Wilson CG, and Deneris ES. Arrest of 5HT neuron differentiation delays respiratory maturation and impairs neonatal homeostatic responses to environmental challenges. Respiratory physiology & neurobiology 159: 85- 101, 2007. Ermentrout GB, Galán RF, and Urban NN. Reliability, synchrony and noise. Trends in neurosciences 31: 428-434, 2008. Espinosa F, and Kavalali ET. NMDA receptor activation by spontaneous glutamatergic neurotransmission. J Neurophysiol 101: 2290-2296, 2009. Fabre V, Beaufour C, Evrard A, Rioux A, Hanoun N, Lesch KP, Murphy DL, Lanfumey L, Hamon M, and Martres MP. Altered expression and functions of serotonin 5-HT1A and 5-HT1B receptors in knock-out mice lacking the 5-HT transporter. Eur J Neurosci 12: 2299-2310, 2000. Favero M, and Castro-Alamancos MA. Synaptic cooperativity regulates persistent network activity in neocortex. J Neurosci 33: 3151-3163, 2013. Ferezou I, Bolea S, and Petersen CC. Visualizing the cortical representation of whisker touch: voltage-sensitive dye imaging in freely moving mice. Neuron 50: 617-629, 2006. Fernandez FR, Malerba P, Bressloff PC, and White JA. Entorhinal stellate cells show preferred spike phase-locking to theta inputs that is enhanced by correlations in synaptic activity. J Neurosci 33: 6027-6040, 2013. Ferrero AJ, Cereseto M, Reines A, Bonavita CD, Sifonios LL, Rubio MC, and Wikinski SI. Chronic treatment with fluoxetine decreases seizure threshold in naive but not in rats exposed to the learned helplessness paradigm: Correlation with the hippocampal glutamate release. Prog Neuropsychopharmacol Biol Psychiatry 29: 678-686, 2005. Fink KB, and Gothert M. 5-HT receptor regulation of neurotransmitter release. Pharmacol Rev 59: 360-417, 2007. Fino E, and Yuste R. Dense inhibitory connectivity in neocortex. Neuron 69: 1188-1203, 2011. Fitzhugh R. A kinetic model of the conductance changes in nerve membrane. J Cell Comp Physiol 66: 111-118, 1965. Foehring RC, van Brederode JF, Kinney GA, and Spain WJ. Serotonergic modulation of supragranular neurons in rat sensorimotor cortex. J Neurosci 22: 8238-8250, 2002.

158

Freitas RM, Sousa FC, Viana GS, and Fonteles MM. Effect of gabaergic, glutamatergic, antipsychotic and antidepressant drugs on pilocarpine-induced seizures and status epilepticus. Neurosci Lett 408: 79-83, 2006. Fries P, Schroder JH, Roelfsema PR, Singer W, and Engel AK. Oscillatory neuronal synchronization in primary visual cortex as a correlate of stimulus selection. J Neurosci 22: 3739-3754, 2002. Frohlich F, Sejnowski TJ, and Bazhenov M. Network bistability mediates spontaneous transitions between normal and pathological brain states. J Neurosci 30: 10734-10743, 2010. Fuster JM. Unit activity in prefrontal cortex during delayed-response performance: neuronal correlates of transient memory. J Neurophysiol 36: 61-78, 1973. Gabernet L, Jadhav SP, Feldman DE, Carandini M, and Scanziani M. Somatosensory integration controlled by dynamic thalamocortical feed-forward inhibition. Neuron 48: 315-327, 2005. Galán RF, Ermentrout GB, and Urban NN. Optimal time scale for spike-time reliability: theory, simulations, and experiments. J Neurophysiol 99: 277-283, 2008. Gaspar P, Cases O, and Maroteaux L. The developmental role of serotonin: news from mouse molecular genetics. Nat Rev Neurosci 4: 1002-1012, 2003. Getting PA. Emerging principles governing the operation of neural networks. Annu Rev Neurosci 12: 185-204, 1989. Gil Z, Connors BW, and Amitai Y. Differential regulation of neocortical synapses by neuromodulators and activity. Neuron 19: 679-686, 1997. Goodfellow NM, Bailey CD, and Lambe EK. The native serotonin 5-HT(5A) receptor: electrophysiological characterization in rodent cortex and 5-HT(1A)-mediated compensatory plasticity in the knock-out mouse. J Neurosci 32: 5804-5809, 2012. Gothert M. Presynaptic serotonin receptors in the central nervous system. Ann N Y Acad Sci 604: 102-112, 1990. Grashow R, Brookings T, and Marder E. Compensation for variable intrinsic neuronal excitability by circuit-synaptic interactions. J Neurosci 30: 9145-9156, 2010. Graupner M, and Reyes AD. Synaptic input correlations leading to membrane potential decorrelation of spontaneous activity in cortex. J Neurosci 33: 15075-15085, 2013. Gross R, Dannon PN, Lepkifker E, Zohar J, and Kotler M. Generalized seizures caused by fluoxetine overdose. Am J Emerg Med 16: 328-329, 1998. Gutnick MJ, Connors BW, and Prince DA. Mechanisms of neocortical epileptogenesis in vitro. J Neurophysiol 48: 1321-1335, 1982. Hablitz JJ. Picrotoxin-induced epileptiform activity in hippocampus: role of endogenous versus synaptic factors. J Neurophysiol 51: 1011-1027, 1984. Hablitz JJ. Spontaneous ictal-like discharges and sustained potential shifts in the developing rat neocortex. J Neurophysiol 58: 1052-1065, 1987. Haider B, Duque A, Hasenstaub AR, and McCormick DA. Neocortical network activity in vivo is generated through a dynamic balance of excitation and inhibition. J Neurosci 26: 4535-4545, 2006. Haider B, and McCormick DA. Rapid neocortical dynamics: cellular and network mechanisms. Neuron 62: 171-189, 2009. Halliwell JV. M-current in human neocortical neurones. Neurosci Lett 67: 1-6, 1986. Hanna MC, Davies PA, Hales TG, and Kirkness EF. Evidence for expression of heteromeric serotonin 5-HT(3) receptors in rodents. J Neurochem 75: 240-247, 2000. Hansel D. Synchronized chaos in local cortical circuits. Int J Neural Syst 7: 403-415, 1996.

159

Harmer CJ, Bhagwagar Z, Perrett DI, Vollm BA, Cowen PJ, and Goodwin GM. Acute SSRI administration affects the processing of social cues in healthy volunteers. Neuropsychopharmacology 28: 148-152, 2003. Harris JA, Petersen RS, and Diamond ME. Distribution of tactile learning and its neural basis. Proc Natl Acad Sci U S A 96: 7587-7591, 1999. Hasenstaub A, Shu Y, Haider B, Kraushaar U, Duque A, and McCormick DA. Inhibitory postsynaptic potentials carry synchronized frequency information in active cortical networks. Neuron 47: 423-435, 2005. Helmstaedter M, Staiger JF, Sakmann B, and Feldmeyer D. Efficient recruitment of layer 2/3 interneurons by layer 4 input in single columns of rat somatosensory cortex. J Neurosci 28: 8273-8284, 2008. Hendricks TJ, Fyodorov DV, Wegman LJ, Lelutiu NB, Pehek EA, Yamamoto B, Silver J, Weeber EJ, Sweatt JD, and Deneris ES. Pet-1 ETS gene plays a critical role in 5-HT neuron development and is required for normal anxiety-like and aggressive behavior. Neuron 37: 233-247, 2003. Hernandez EJ, Williams PA, and Dudek FE. Effects of fluoxetine and TFMPP on spontaneous seizures in rats with pilocarpine-induced epilepsy. Epilepsia 43: 1337-1345, 2002. Hodges MR, Best S, and Richerson GB. Altered ventilatory and thermoregulatory control in male and female adult Pet-1 null mice. Respiratory physiology & neurobiology 177: 133-140, 2011. Hodgkin AL, and Huxley AF. A quantitative description of membrane current and its application to conduction and excitation in nerve. J Physiol 117: 500-544, 1952. Holmgren C, Harkany T, Svennenfors B, and Zilberter Y. Pyramidal cell communication within local networks in layer 2/3 of rat neocortex. J Physiol 551: 139-153, 2003. Hubel DH, and Wiesel TN. Ferrier lecture. Functional architecture of macaque monkey visual cortex. Proceedings of the Royal Society of London Series B, Containing papers of a Biological character Royal Society 198: 1-59, 1977. Hubel DH, and Wiesel TN. Receptive fields of single neurones in the cat's striate cortex. J Physiol 148: 574-591, 1959. Huidobro-Toro JP, Valenzuela CF, and Harris RA. Modulation of GABAA receptor function by G protein-coupled 5-HT2C receptors. Neuropharmacology 35: 1355-1363, 1996. Ince RA, Panzeri S, and Kayser C. Neural codes formed by small and temporally precise populations in auditory cortex. J Neurosci 33: 18277-18287, 2013. Jacobs BL, and Azmitia EC. Structure and function of the brain serotonin system. Physiological reviews 72: 165-229, 1992. Jakab RL, and Goldman-Rakic PS. Segregation of serotonin 5-HT2A and 5-HT3 receptors in inhibitory circuits of the primate cerebral cortex. J Comp Neurol 417: 337-348, 2000. Jansson A, Tinner B, Bancila M, Verge D, Steinbusch HW, Agnati LF, and Fuxe K. Relationships of 5-hydroxytryptamine immunoreactive terminal-like varicosities to 5- hydroxytryptamine-2A receptor-immunoreactive neuronal processes in the rat forebrain. Journal of chemical neuroanatomy 22: 185-203, 2001. Jobe PC, Woods TW, Mcnulty M, and Dailey JW. Anticonvulsant and Proconvulsant Effects of Tricyclic Antidepressants. Fed Proc 43: 569-569, 1984. Johnston D, and Brown TH. The synaptic nature of the paroxysmal depolarizing shift in hippocampal neurons. Annals of neurology 16 Suppl: S65-71, 1984.

160

Kaczmarek LK, and Levitan IB. Neuromodulation : the biochemical control of neuronal excitability. New York: Oxford University Press, 1987, p. ix, 286 p. Kanner AM. Depression in epilepsy: prevalence, clinical semiology, pathogenic mechanisms, and treatment. Biol Psychiatry 54: 388-398, 2003. Kapfer C, Glickfeld LL, Atallah BV, and Scanziani M. Supralinear increase of recurrent inhibition during sparse activity in the somatosensory cortex. Nat Neurosci 10: 743-753, 2007. Kawaguchi M, Mino H, and Durand DM. Stochastic resonance can enhance information transmission in neural networks. IEEE Trans Biomed Eng 58: 1950-1958, 2011. Kelley SP, Dunlop JI, Kirkness EF, Lambert JJ, and Peters JA. A cytoplasmic region determines single-channel conductance in 5-HT3 receptors. Nature 424: 321-324, 2003. Kirson ED, and Yaari Y. Synaptic NMDA receptors in developing mouse hippocampal neurones: functional properties and sensitivity to ifenprodil. J Physiol 497 ( Pt 2): 437- 455, 1996. Kleppe IC, and Robinson HP. Determining the activation time course of synaptic AMPA receptors from openings of colocalized NMDA receptors. Biophys J 77: 1418-1427, 1999. Kohr G, and Heinemann U. Effects of NMDA antagonists on picrotoxin-, low Mg2+- and low Ca2+-induced epileptogenesis and on evoked changes in extracellular Na+ and Ca2+ concentrations in rat hippocampal slices. Epilepsy Res 4: 187-200, 1989. Kolind J, Hounsgaard J, and Berg RW. Opposing Effects of Intrinsic Conductance and Correlated Synaptic Input on V-Fluctuations during Network Activity. Front Comput Neurosci 6: 40, 2012. Komlosi G, Molnar G, Rozsa M, Olah S, Barzo P, and Tamas G. Fluoxetine (prozac) and serotonin act on excitatory synaptic transmission to suppress single layer 2/3 pyramidal neuron-triggered cell assemblies in the human prefrontal cortex. J Neurosci 32: 16369- 16378, 2012. Kremkow J, Aertsen A, and Kumar A. Gating of signal propagation in spiking neural networks by balanced and correlated excitation and inhibition. J Neurosci 30: 15760- 15768, 2010a. Kremkow J, Perrinet LU, Masson GS, and Aertsen A. Functional consequences of correlated excitatory and inhibitory conductances in cortical networks. J Comput Neurosci 28: 579-594, 2010b. Lambe EK, Goldman-Rakic PS, and Aghajanian GK. Serotonin induces EPSCs preferentially in layer V pyramidal neurons of the frontal cortex in the rat. Cereb Cortex 10: 974-980, 2000. Laurent A, Goaillard JM, Cases O, Lebrand C, Gaspar P, and Ropert N. Activity- dependent presynaptic effect of serotonin 1B receptors on the somatosensory thalamocortical transmission in neonatal mice. J Neurosci 22: 886-900, 2002. Lee S, Hjerling-Leffler J, Zagha E, Fishell G, and Rudy B. The largest group of superficial neocortical GABAergic interneurons expresses ionotropic serotonin receptors. J Neurosci 30: 16796-16808, 2010. Lee S, Kruglikov I, Huang ZJ, Fishell G, and Rudy B. A disinhibitory circuit mediates motor integration in the somatosensory cortex. Nat Neurosci 16: 1662-1670, 2013. Lee WL, and Hablitz JJ. Initiation of epileptiform activity by excitatory amino acid receptors in the disinhibited rat neocortex. J Neurophysiol 65: 87-95, 1991. Lefort S, Tomm C, Floyd Sarria JC, and Petersen CC. The excitatory neuronal network of the C2 barrel column in mouse primary somatosensory cortex. Neuron 61: 301-316, 2009.

161

Lerch-Haner JK, Frierson D, Crawford LK, Beck SG, and Deneris ES. Serotonergic transcriptional programming determines maternal behavior and offspring survival. Nat Neurosci 11: 1001-1003, 2008. Liu C, Maejima T, Wyler SC, Casadesus G, Herlitze S, and Deneris ES. Pet-1 is required across different stages of life to regulate serotonergic function. Nat Neurosci 13: 1190- 1198, 2010. Liu RC, Tzonev S, Rebrik S, and Miller KD. Variability and information in a neural code of the cat lateral geniculate nucleus. Journal of Neurophysiology 86: 2789-2806, 2001. Loscher W, Nolting B, and Fassbender CP. The role of technical, biological and pharmacological factors in the laboratory evaluation of anticonvulsant drugs. I. The influence of administration vehicles. Epilepsy Res 7: 173-181, 1990. Luscher C, Jan LY, Stoffel M, Malenka RC, and Nicoll RA. G protein-coupled inwardly rectifying K+ channels (GIRKs) mediate postsynaptic but not presynaptic transmitter actions in hippocampal neurons. Neuron 19: 687-695, 1997. Macedo DS, Santos RS, Belchior LD, Neto MA, Vasconcelos SM, Lima VT, Fonteles MM, Viana GS, and de Sousa FC. Effect of anxiolytic, antidepressant, and antipsychotic drugs on cocaine-induced seizures and mortality. Epilepsy Behav 5: 852-856, 2004. Maffei A, and Turrigiano GG. Multiple modes of network homeostasis in visual cortical layer 2/3. J Neurosci 28: 4377-4384, 2008. Mainen ZF, and Sejnowski TJ. Reliability of spike timing in neocortical neurons. Science 268: 1503-1506, 1995. Marder E. Neuromodulation of neuronal circuits: back to the future. Neuron 76: 1-11, 2012. Marder E. Variability, compensation, and modulation in neurons and circuits. Proc Natl Acad Sci U S A 108 Suppl 3: 15542-15548, 2011. Marder E, and Thirumalai V. Cellular, synaptic and network effects of neuromodulation. Neural networks : the official journal of the International Neural Network Society 15: 479-493, 2002. Marino J, Schummers J, Lyon DC, Schwabe L, Beck O, Wiesing P, Obermayer K, and Sur M. Invariant computations in local cortical networks with balanced excitation and inhibition. Nat Neurosci 8: 194-201, 2005. Marsalek P, Koch C, and Maunsell J. On the relationship between synaptic input and spike output jitter in individual neurons. Proc Natl Acad Sci U S A 94: 735-740, 1997. McCormick DA, and Contreras D. On the cellular and network bases of epileptic seizures. Annu Rev Physiol 63: 815-846, 2001. McCormick DA, and Williamson A. Convergence and divergence of neurotransmitter action in human cerebral cortex. Proc Natl Acad Sci U S A 86: 8098-8102, 1989. McDonnell MD, and Ward LM. The benefits of noise in neural systems: bridging theory and experiment. Nat Rev Neurosci 12: 415-426, 2011. Meldrum BS, Anlezark GM, Adam HK, and Greenwood DT. Anticonvulsant and Proconvulsant Properties of Viloxazine Hydrochloride - Pharmacological and Pharmacokinetic Studies in Rodents and the Epileptic Baboon. Psychopharmacology 76: 212-217, 1982. Miller JP, Rall W, and Rinzel J. Synaptic amplification by active membrane in dendritic spines. Brain Res 325: 325-330, 1985. Miura K, Tsubo Y, Okada M, and Fukai T. Balanced excitatory and inhibitory inputs to cortical neurons decouple firing irregularity from rate modulations. J Neurosci 27: 13802-13812, 2007.

162

Morales M, and Wang SD. Differential composition of 5-hydroxytryptamine3 receptors synthesized in the rat CNS and peripheral nervous system. J Neurosci 22: 6732-6741, 2002. Moreau AW, Amar M, Le Roux N, Morel N, and Fossier P. Serotoninergic fine-tuning of the excitation-inhibition balance in rat visual cortical networks. Cereb Cortex 20: 456- 467, 2010. Morita K, Hamamoto M, Arai S, Kitayama S, Irifune M, Kawahara M, Kihira K, and Dohi T. Inhibition of serotonin transporters by cocaine and meprylcaine through 5-TH2C receptor stimulation facilitates their seizure activities. Brain Res 1057: 153-160, 2005. Mossner R, Schmitt A, Hennig T, Benninghoff J, Gerlach M, Riederer P, Deckert J, and Lesch KP. Quantitation of 5HT3 receptors in forebrain of serotonin transporter deficient mice. Journal of neural transmission 111: 27-35, 2004. Mountcastle VB. The columnar organization of the neocortex. Brain : a journal of neurology 120 ( Pt 4): 701-722, 1997. Mountcastle VB. Modality and topographic properties of single neurons of cat's somatic sensory cortex. J Neurophysiol 20: 408-434, 1957. Moya PR, Fox MA, Jensen CL, Laporte JL, French HT, Wendland JR, and Murphy DL. Altered 5-HT2C receptor agonist-induced responses and 5-HT2C receptor RNA editing in the amygdala of serotonin transporter knockout mice. BMC pharmacology 11: 3, 2011. Nakamura K, and Wong-Lin K. Functions and computational principles of serotonergic and related systems at multiple scales. Front Integr Neurosci 8: 23, 2014. Narboux-Neme N, Angenard G, Mosienko V, Klempin F, Pitychoutis PM, Deneris E, Bader M, Giros B, Alenina N, and Gaspar P. Postnatal growth defects in mice with constitutive depletion of central serotonin. ACS chemical neuroscience 4: 171-181, 2013. Narboux-Neme N, Sagne C, Doly S, Diaz SL, Martin CB, Angenard G, Martres MP, Giros B, Hamon M, Lanfumey L, Gaspar P, and Mongeau R. Severe serotonin depletion after conditional deletion of the vesicular monoamine transporter 2 gene in serotonin neurons: neural and behavioral consequences. Neuropsychopharmacology 36: 2538- 2550, 2011. Nemenman I, Lewen GD, Bialek W, and van Steveninck RRD. Neural coding of natural stimuli: Information at sub-millisecond resolution. Plos Computational Biology 4: 2008. Nichols DE, and Nichols CD. Serotonin receptors. Chem Rev 108: 1614-1641, 2008. Nicolelis MA, Baccala LA, Lin RC, and Chapin JK. Sensorimotor encoding by synchronous neural ensemble activity at multiple levels of the somatosensory system. Science 268: 1353-1358, 1995. O'Dell LE, Kreifeldt MJ, George FR, and Ritz MC. The role of serotonin(2) receptors in mediating cocaine-induced convulsions. Pharmacol Biochem Behav 65: 677-681, 2000a. O'Dell LE, Li R, George FR, and Ritz MC. Molecular serotonergic mechanisms appear to mediate genetic sensitivity to cocaine-induced convulsions. Brain Res 863: 213-224, 2000b. Okun M, and Lampl I. Instantaneous correlation of excitation and inhibition during ongoing and sensory-evoked activities. Nat Neurosci 11: 535-537, 2008. Orser BA, Wang LY, Pennefather PS, and MacDonald JF. Propofol modulates activation and desensitization of GABAA receptors in cultured murine hippocampal neurons. J Neurosci 14: 7747-7760, 1994. Otsu Y, and Murphy TH. Miniature transmitter release: accident of nature or careful design? Sci STKE 2003: pe54, 2003.

163

Pankratov Y, Lalo U, Verkhratsky A, and North RA. Quantal release of ATP in mouse cortex. J Gen Physiol 129: 257-265, 2007. Papale LA, Makinson CD, Christopher Ehlen J, Tufik S, Decker MJ, Paul KN, and Escayg A. Altered sleep regulation in a mouse model of SCN1A-derived genetic epilepsy with febrile seizures plus (GEFS+). Epilepsia 54: 625-634, 2013. Parga N, and Abbott LF. Network model of spontaneous activity exhibiting synchronous transitions between up and down States. Front Neurosci 1: 57-66, 2007. Parks CL, Robinson PS, Sibille E, Shenk T, and Toth M. Increased anxiety of mice lacking the serotonin1A receptor. Proc Natl Acad Sci U S A 95: 10734-10739, 1998. Paulus EV, and Mintz EM. Developmental disruption of the serotonin system alters circadian rhythms. Physiology & behavior 105: 257-263, 2012. Petersen CC, Grinvald A, and Sakmann B. Spatiotemporal dynamics of sensory responses in layer 2/3 of rat barrel cortex measured in vivo by voltage-sensitive dye imaging combined with whole-cell voltage recordings and neuron reconstructions. J Neurosci 23: 1298-1309, 2003. Petersen CC, and Sakmann B. Functionally independent columns of rat somatosensory barrel cortex revealed with voltage-sensitive dye imaging. J Neurosci 21: 8435-8446, 2001. Pfeffer CK, Xue M, He M, Huang ZJ, and Scanziani M. Inhibition of inhibition in visual cortex: the logic of connections between molecularly distinct interneurons. Nat Neurosci 16: 1068-1076, 2013. Pi HJ, Hangya B, Kvitsiani D, Sanders JI, Huang ZJ, and Kepecs A. Cortical interneurons that specialize in disinhibitory control. Nature 503: 521-524, 2013. Pisani F, Spina E, and Oteri G. Antidepressant drugs and seizure susceptibility: from in vitro data to clinical practice. Epilepsia 40 Suppl 10: S48-56, 1999. Poncer JC, Durr R, Gahwiler BH, and Thompson SM. Modulation of synaptic GABAA receptor function by benzodiazepines in area CA3 of rat hippocampal slice cultures. Neuropharmacology 35: 1169-1179, 1996. Porter JT, Johnson CK, and Agmon A. Diverse types of interneurons generate thalamus- evoked feedforward inhibition in the mouse barrel cortex. J Neurosci 21: 2699-2710, 2001. Pouille F, Marin-Burgin A, Adesnik H, Atallah BV, and Scanziani M. Input normalization by global feedforward inhibition expands cortical dynamic range. Nat Neurosci 12: 1577- 1585, 2009. Pouille F, and Scanziani M. Enforcement of temporal fidelity in pyramidal cells by somatic feed-forward inhibition. Science 293: 1159-1163, 2001. Poulet JF, and Petersen CC. Internal brain state regulates membrane potential synchrony in barrel cortex of behaving mice. Nature 454: 881-885, 2008. Prasher VP. Seizures associated with fluoxetine therapy. Seizure 2: 315-317, 1993. Prinz AA, Billimoria CP, and Marder E. Alternative to hand-tuning conductance-based models: construction and analysis of databases of model neurons. J Neurophysiol 90: 3998-4015, 2003. Pucilowska J, Puzerey PA, Karlo JC, Galán RF, and Landreth GE. Disrupted ERK signaling during cortical development leads to abnormal progenitor proliferation, neuronal and network excitability and behavior, modeling human neuro-cardio-facial-cutaneous and related syndromes. J Neurosci 32: 8663-8677, 2012. Puig MV, and Gulledge AT. Serotonin and prefrontal cortex function: neurons, networks, and circuits. Mol Neurobiol 44: 449-464, 2011.

164

Puig MV, Santana N, Celada P, Mengod G, and Artigas F. In vivo excitation of GABA interneurons in the medial prefrontal cortex through 5-HT3 receptors. Cereb Cortex 14: 1365-1375, 2004. Puig MV, Watakabe A, Ushimaru M, Yamamori T, and Kawaguchi Y. Serotonin modulates fast-spiking interneuron and synchronous activity in the rat prefrontal cortex through 5-HT1A and 5-HT2A receptors. J Neurosci 30: 2211-2222, 2010. Puzerey PA, Decker MJ, and Galan RF. Elevated serotonergic signaling amplifies synaptic noise and facilitates the emergence of epileptiform network oscillations. J Neurophysiol 2014. Puzerey PA, and Galán RF. On how correlations between excitatory and inhibitory synaptic inputs maximize the information rate of neuronal firing. Front Comput Neurosci 8: 59, 2014. Rahman S, and Neuman RS. Activation of 5-HT2 receptors facilitates depolarization of neocortical neurons by N-methyl-D-aspartate. Eur J Pharmacol 231: 347-354, 1993. Raju SS, Noor AR, Gurthu S, Giriyappanavar CR, Acharya SB, Low HC, and Quah SH. Effect of fluoxetine on maximal electroshock seizures in mice: acute vs chronic administration. Pharmacol Res 39: 451-454, 1999. Rodriguez-Molina VM, Aertsen A, and Heck DH. Spike timing and reliability in cortical pyramidal neurons: effects of EPSC kinetics, input synchronization and background noise on spike timing. PloS one 2: e319, 2007. Roerig B, Nelson DA, and Katz LC. Fast synaptic signaling by nicotinic acetylcholine and serotonin 5-HT3 receptors in developing visual cortex. J Neurosci 17: 8353-8362, 1997. Rothschild G, Nelken I, and Mizrahi A. Functional organization and population dynamics in the mouse primary auditory cortex. Nat Neurosci 13: 353-360, 2010. Sakai K. Sleep-waking discharge profiles of dorsal raphe nucleus neurons in mice. Neuroscience 197: 200-224, 2011. Salin PA, and Prince DA. Spontaneous GABAA receptor-mediated inhibitory currents in adult rat somatosensory cortex. J Neurophysiol 75: 1573-1588, 1996. Salinas E, and Sejnowski TJ. Impact of correlated synaptic input on output firing rate and variability in simple neuronal models. J Neurosci 20: 6193-6209, 2000. Sanchez-Vives MV, and McCormick DA. Cellular and network mechanisms of rhythmic recurrent activity in neocortex. Nat Neurosci 3: 1027-1034, 2000. Santana N, Bortolozzi A, Serrats J, Mengod G, and Artigas F. Expression of serotonin1A and serotonin2A receptors in pyramidal and GABAergic neurons of the rat prefrontal cortex. Cereb Cortex 14: 1100-1109, 2004. Satake T, Mitani H, Nakagome K, and Kaneko K. Individual and additive effects of neuromodulators on the slow components of afterhyperpolarization currents in layer V pyramidal cells of the rat medial prefrontal cortex. Brain Res 1229: 47-60, 2008. Schmandt NT, and Galán RF. Stochastic-shielding approximation of Markov chains and its application to efficiently simulate random ion-channel gating. Physical Review Letters 109: 118101, 2012. Schmitt JA, Wingen M, Ramaekers JG, Evers EA, and Riedel WJ. Serotonin and human cognitive performance. Current pharmaceutical design 12: 2473-2486, 2006. Schmitz D, Empson RM, Gloveli T, and Heinemann U. Serotonin blocks different patterns of low Mg2+-induced epileptiform activity in rat entorhinal cortex, but not hippocampus. Neuroscience 76: 449-458, 1997.

165

Schneidman E, Freedman B, and Segev I. Ion channel stochasticity may be critical in determining the reliability and precision of spike timing. Neural Computation 10: 1679- 1703, 1998. Sengupta B, Laughlin SB, and Niven JE. Balanced excitatory and inhibitory synaptic currents promote efficient coding and metabolic efficiency. PLoS Comput Biol 9: e1003263, 2013. Shadlen MN, and Newsome WT. The variable discharge of cortical neurons: implications for connectivity, computation, and information coding. J Neurosci 18: 3870-3896, 1998. Shepherd GM, Pologruto TA, and Svoboda K. Circuit analysis of experience-dependent plasticity in the developing rat barrel cortex. Neuron 38: 277-289, 2003. Shew WL, Yang H, Yu S, Roy R, and Plenz D. Information capacity and transmission are maximized in balanced cortical networks with neuronal avalanches. J Neurosci 31: 55- 63, 2011. Shu Y, Hasenstaub A, Badoual M, Bal T, and McCormick DA. Barrages of synaptic activity control the gain and sensitivity of cortical neurons. J Neurosci 23: 10388-10401, 2003a. Shu Y, Hasenstaub A, and McCormick DA. Turning on and off recurrent balanced cortical activity. Nature 423: 288-293, 2003b. Silberberg G, and Markram H. Disynaptic inhibition between neocortical pyramidal cells mediated by Martinotti cells. Neuron 53: 735-746, 2007. Skaugen E, Walloe, L. Firing behavior in a stochastic nerve membrane model based upon the Hodgkin-Huxley equations. Acta Physiol Scand 107: 343-363, 1979. Stacey WC, and Durand DM. Synaptic noise improves detection of subthreshold signals in hippocampal CA1 neurons. J Neurophysiol 86: 1104-1112, 2001. Stacey WC, Lazarewicz MT, and Litt B. Synaptic noise and physiological coupling generate high-frequency oscillations in a hippocampal computational model. J Neurophysiol 102: 2342-2357, 2009. Steriade M, and Amzica F. Sleep oscillations developing into seizures in corticothalamic systems. Epilepsia 44 Suppl 12: 9-20, 2003. Steriade M, Amzica F, Neckelmann D, and Timofeev I. Spike-wave complexes and fast components of cortically generated seizures. II. Extra- and intracellular patterns. J Neurophysiol 80: 1456-1479, 1998. Steriade M, Nunez A, and Amzica F. Intracellular analysis of relations between the slow (< 1 Hz) neocortical oscillation and other sleep rhythms of the electroencephalogram. J Neurosci 13: 3266-3283, 1993. Stone JBD, & I. Tork. The Neuroanatomist's Colouring Book. Sydney: Maitland Publications, 1990. Strassberg AF, and Defelice LJ. Limitations of the Hodgkin-Huxley Formalism - Effects of Single-Channel Kinetics on Transmembrane Voltage Dynamics. Neural Computation 5: 843-855, 1993. Strong SP, Koberle R, van Steveninck RRD, and Bialek W. Entropy and information in neural spike trains. Physical Review Letters 80: 197-200, 1998. Sudweeks SN, Hooft JA, and Yakel JL. Serotonin 5-HT(3) receptors in rat CA1 hippocampal interneurons: functional and molecular characterization. J Physiol 544: 715-726, 2002. Sun QQ, Huguenard JR, and Prince DA. Barrel cortex microcircuits: thalamocortical feedforward inhibition in spiny stellate cells is mediated by a small number of fast- spiking interneurons. J Neurosci 26: 1219-1230, 2006.

166

Sung KW, Engel SR, Allan AM, and Lovinger DM. 5-HT(3) receptor function and potentiation by alcohols in frontal cortex neurons from transgenic mice overexpressing the receptor. Neuropharmacology 39: 2346-2351, 2000. Sutton MA, Wall NR, Aakalu GN, and Schuman EM. Regulation of dendritic protein synthesis by miniature synaptic events. Science 304: 1979-1983, 2004. Svirskis G, and Rinzel J. Influence of temporal correlation of synaptic input on the rate and variability of firing in neurons. Biophys J 79: 629-637, 2000. Tallon-Baudry C, Bertrand O, Delpuech C, and Permier J. Oscillatory gamma-band (30- 70 Hz) activity induced by a visual search task in humans. J Neurosci 17: 722-734, 1997. Tanaka E, and North RA. Actions of 5-hydroxytryptamine on neurons of the rat cingulate cortex. J Neurophysiol 69: 1749-1757, 1993. Taub AH, Katz Y, and Lampl I. Cortical balance of excitation and inhibition is regulated by the rate of synaptic activity. J Neurosci 33: 14359-14368, 2013. Thome-Souza MS, Kuczynski E, and Valente KD. Sertraline and fluoxetine: safe treatments for children and adolescents with epilepsy and depression. Epilepsy Behav 10: 417-425, 2007. Thomson AM, and Lamy C. Functional maps of neocortical local circuitry. Front Neurosci 1: 19-42, 2007. Timofeev I, Grenier F, and Steriade M. Spike-wave complexes and fast components of cortically generated seizures. IV. Paroxysmal fast runs in cortical and thalamic neurons. J Neurophysiol 80: 1495-1513, 1998. Timofeev I, and Steriade M. Neocortical seizures: initiation, development and cessation. Neuroscience 123: 299-336, 2004. Tork I. Anatomy of the serotonergic system. Ann N Y Acad Sci 600: 9-34; discussion 34- 35, 1990. Trimble M, Anlezark G, and Meldrum B. Seizure activity in photosensitive baboons following antidepressant drugs and the role of serotoninergic mechanisms. Psychopharmacology (Berl) 51: 159-164, 1977. Vanderwolf CH, and Baker GB. Evidence that serotonin mediates non-cholinergic neocortical low voltage fast activity, non-cholinergic hippocampal rhythmical slow activity and contributes to intelligent behavior. Brain Res 374: 342-356, 1986. Veenstra-VanderWeele J, and Blakely RD. Networking in autism: leveraging genetic, biomarker and model system findings in the search for new treatments. Neuropsychopharmacology 37: 196-212, 2012. Veenstra-VanderWeele J, Muller CL, Iwamoto H, Sauer JE, Owens WA, Shah CR, Cohen J, Mannangatti P, Jessen T, Thompson BJ, Ye R, Kerr TM, Carneiro AM, Crawley JN, Sanders-Bush E, McMahon DG, Ramamoorthy S, Daws LC, Sutcliffe JS, and Blakely RD. Autism gene variant causes hyperserotonemia, serotonin receptor hypersensitivity, social impairment and repetitive behavior. Proc Natl Acad Sci U S A 109: 5469-5474, 2012. Villalobos C, Beique JC, Gingrich JA, and Andrade R. Serotonergic regulation of calcium- activated potassium currents in rodent prefrontal cortex. Eur J Neurosci 22: 1120-1126, 2005. Villalobos C, Foehring RC, Lee JC, and Andrade R. Essential role for phosphatidylinositol 4,5-bisphosphate in the expression, regulation, and gating of the slow afterhyperpolarization current in the cerebral cortex. J Neurosci 31: 18303-18312, 2011. Vucurovic K, Gallopin T, Ferezou I, Rancillac A, Chameau P, van Hooft JA, Geoffroy H, Monyer H, Rossier J, and Vitalis T. Serotonin 3A receptor subtype as an early and

167

protracted marker of cortical interneuron subpopulations. Cereb Cortex 20: 2333-2347, 2010. Vyleta NP, and Smith SM. Spontaneous glutamate release is independent of calcium influx and tonically activated by the calcium-sensing receptor. J Neurosci 31: 4593-4606, 2011. Wada Y, Nakamura M, Hasegawa H, and Yamaguchi N. Role of serotonin receptor subtype in seizures kindled from the feline hippocampus. Neurosci Lett 141: 21-24, 1992. Walker HC, Lawrence JJ, and McBain CJ. Activation of kinetically distinct synaptic conductances on inhibitory interneurons by electrotonically overlapping afferents. Neuron 35: 161-171, 2002. Wasser CR, and Kavalali ET. Leaky synapses: regulation of spontaneous neurotransmission in central synapses. Neuroscience 158: 177-188, 2009. Waterhouse BD, Mihailoff GA, Baack JC, and Woodward DJ. Topographical distribution of dorsal and median raphe neurons projecting to motor, sensorimotor, and visual cortical areas in the rat. J Comp Neurol 249: 460-476, 478-481, 1986. Weber ET, and Andrade R. Htr2a Gene and 5-HT(2A) Receptor Expression in the Cerebral Cortex Studied Using Genetically Modified Mice. Front Neurosci 4: 2010. Wehr M, and Zador AM. Balanced inhibition underlies tuning and sharpens spike timing in auditory cortex. Nature 426: 442-446, 2003. Williams GV, Rao SG, and Goldman-Rakic PS. The physiological role of 5-HT2A receptors in working memory. J Neurosci 22: 2843-2854, 2002. Willins DL, Deutch AY, and Roth BL. Serotonin 5-HT2A receptors are expressed on pyramidal cells and interneurons in the rat cortex. Synapse 27: 79-82, 1997. Wilson CJ. Dynamic modification of dendritic cable properties and synaptic transmission by voltage-gated potassium channels. J Comput Neurosci 2: 91-115, 1995. Woods TW, Jobe PC, and Dailey JW. Anti-Convulsant and Proconvulsant Doses of Desipramine in the Genetically Epilepsy-Prone Rat. Fed Proc 42: 363-363, 1983. Woolsey TA, and Van der Loos H. The structural organization of layer IV in the somatosensory region (SI) of mouse cerebral cortex. The description of a cortical field composed of discrete cytoarchitectonic units. Brain Res 17: 205-242, 1970. Wu GK, Arbuckle R, Liu BH, Tao HW, and Zhang LI. Lateral sharpening of cortical frequency tuning by approximately balanced inhibition. Neuron 58: 132-143, 2008. Wu GK, Li P, Tao HW, and Zhang LI. Nonmonotonic synaptic excitation and imbalanced inhibition underlying cortical intensity tuning. Neuron 52: 705-715, 2006. Yadav PN, Abbas AI, Farrell MS, Setola V, Sciaky N, Huang XP, Kroeze WK, Crawford LK, Piel DA, Keiser MJ, Irwin JJ, Shoichet BK, Deneris ES, Gingrich J, Beck SG, and Roth BL. The presynaptic component of the serotonergic system is required for clozapine's efficacy. Neuropsychopharmacology 36: 638-651, 2011. Yan QS, Jobe PC, and Dailey JW. Further evidence of anticonvulsant role for 5- hydroxytryptamine in genetically epilepsy-prone rats. Br J Pharmacol 115: 1314-1318, 1995. Yoshimura Y, and Callaway EM. Fine-scale specificity of cortical networks depends on inhibitory cell type and connectivity. Nat Neurosci 8: 1552-1559, 2005. Zhang ZW, and Arsenault D. Gain modulation by serotonin in pyramidal neurones of the rat prefrontal cortex. J Physiol 566: 379-394, 2005. Zhou FM, and Hablitz JJ. Activation of serotonin receptors modulates synaptic transmission in rat cerebral cortex. J Neurophysiol 82: 2989-2999, 1999.

168

Zienowicz M, Wislowska A, Lehner M, Taracha E, Skorzewska A, Maciejak P, and Plaznik A. The effect of fluoxetine in a model of chemically induced seizures--behavioral and immunocytochemical study. Neurosci Lett 373: 226-231, 2005.

169