Circuit Specificity and Behavioral Role of GABAergic Inputs. onto Midbrain Dopamine Neurons

Transcription

1 Circuit Specificity and Behavioral Role of GABAergic Inputs onto Midbrain Dopamine Neurons by Nicholas J. Edwards B.S. Brigham Young University, 2011 A dissertation submitted in partial fulfillment of the requirements for the Degree of Doctor of Philosophy in the Department of Neuroscience at Brown University Providence, Rhode Island May 2016

2 Copyright 2016 by Nicholas Edwards

3 This dissertation by Nicholas J. Edwards is accepted in its present form by the Department of Neuroscience as satisfying the dissertation requirement for degree of Doctor of Philosophy Date Antonello Bonci, Advisor Recommended to the Graduate Council Date Mark Andermann, Reader Date Yeka Aponte, Reader Date Bruce Hope, Reader Date Christopher Moore, Reader Approved by the Graduate Council Date Jack A. Elias, Dean of Medicine and Biological Sciences ii

4 Nicholas James Edwards 107 N. Glover St. Phone: (801) Baltimore, MD EDUCATION Doctor of Philosophy in Neuroscience, expected April 2016 Brown University and the National Institutes of Health Advisor: Antonello Bonci Bachelor of Science in Neuroscience, 2011 Brigham Young University Minors in Chemistry and English RESEARCH EXPERIENCE Predoctoral Fellow, National Institute on Drug Abuse, 2012-Present Thesis Advisor, Antonello Bonci Examined the role of nucleus accumbens inputs and local connections within the ventral tegmental area (VTA) Performed mouse surgeries to express viral vectors within various brain regions Performed whole cell patch clamp recordings in VTA GABA and dopamine neurons to examine connectivity between brain regions Used in vivo optogenetic experiments to examine the role of nucleus accumbens inputs to the VTA Studied neuromodulation of GABAergic circuits in the VTA Used in vitro optogenetic manipulations to examine the role of presynaptic neuromodulatory receptors on various GABAergic synapses Examined the role of GABA B receptors in cocaine-related behavior Performed cocaine-related behavioral experiments and ex vivo patch clamp recordings to examine plasticity of GABAergic inputs on dopamine neurons Used a viral strategy to genetically delete GABA B receptors from VTA dopamine neurons and determine their role in cocaine-related behavior Research Assistant, Infinity Pharmaceuticals, 2011 Studied reforestation after plant extraction in the Wasatch mountain range iii

5 Research Assistant, Chemistry Department, BYU Supervisor, Dr. Paul B. Savage, Professor, Department of Chemistry and Biochemistry Isolated glycolipid adjuvants from cell extracts to be examined for antigenic properties Research Assistant, Psychology Department, BYU Supervisor, Dr. Dawson Hedges Formulation of demographic questionnaires; recorded event-related potentials from human subjects PUBLICATIONS Edwards NJ, Tejeda HA, Pignatelli M, Zhang S, Liu B, McDevitt RA, Lowell BB, Morales M, Bonci A. Circuit Specificity in the Inhibitory Architecture of the Ventral Tegmental Area (In Submission). INVITED TALKS Brown-NIH Neuroscience Scientific Retreat, November 2011 Marine Biological Laboratory, Woods Hole, MA Talk title: Inhibitory Influence of Accumbal Inputs to the Midbrain. TEACHING/WORK EXPERIENCE Teaching Assistant, Principles of Biology, Fall 2009, Winter & Fall 2010; Supervisor, Dr. John Bell Teaching Assistant, Organic Chemistry; Winter 2010, Fall 2010, Winter 2011, Supervisor, Dr. Paul B. Savage Supervisor, Dialogue-Marketing International, Provo, UT, December August 2009; managed a team of employees GRANTS AND AWARDS Undergraduate Research Award, Department of Chemistry and Biochemistry, BYU, Fall 2009, Winter 2010, Spring/Summer 2010, Fall 2010 Office of Research and Creative Activities (ORCA) Grant, Brigham Young University, 2010 Academic Scholarship, Brigham Young University The Big Ten+ Graduate School Exposition Scholarship, Fall 2010 iv

6 Acknowledgments I would first like to thank my advisor and mentor, Antonello Bonci, for his continuous support of me during the good years and the not-so-good years. Anto has been a huge source of inspiration in his intellectual curiosity, enthusiasm, and amazing leadership. More than anything, I ve learned from him the importance of designing simple, elegant experiments, and to follow where the data takes me. He has been an incredible role model both as a scientist and as a mentor who genuinely cares for the well-being of the people around him. Additionally, Anto has fostered a friendly and collaborative lab environment, where the work is both engaging and fun. I d like to thank the members of the Bonci lab for good friendships, insightful comments, and thoughtful feedback. In particular, Hugo Tejeda and Marco Pignatelli have been outstanding mentors and even better friends. Marco and Hugo not only taught me how to perform patch clamp electrophysiology, but showed me the power of it in answering a broad range of questions. Our conversations sitting at the rig have had an immense impact on me and my approach to science. Ross McDevitt has been an amazing support to me and is one of the best people I know to discuss ideas with. His comments and feedback are always spot on. Wendy Xin has been a great friend and a morale booster as we ve navigated the turbulent waters of science and grad school. I want to express sincere gratitude to Yeka Aponte and Bruce Hope. Bruce has been a good friend and mentor to me. His creativity is inspiring and I m extremely grateful for his support of my career during the many transitions I ve made. Yeka took on a huge responsibility of mentoring me during a difficult and confusing time of my life. v

7 Her attention to detail and precision are incredible. I d like to thank Chris Moore for his uncanny creative insights. Chris probably doesn t know it, but his ability to offer timely encouragement and advice has had a big impact on me. Also, I d like to thank Mark Andermann for taking time out of his schedule to serve as an outside reader on my thesis committee. Most importantly, I want to thank my lovely wife, Kim Edwards, and my beautiful and amazing kids, Colette and Eli. Kim, you are the love of my life. This PhD is as much, and probably more, your accomplishment than it is mine. I honestly could not have done this without you. Your advice and encouragement have carried me through some very difficult times and your enthusiasm and cheerleading have been an inspiration during the victories. Colette and Eli, I feel so lucky and proud to be able to come home (almost) every day to your smiling faces. It really is the best part of my day. Finally, thank you to my parents, Hal and Lynne Edwards, and in-laws, Dennis and Kaye Wright, for instilling in me a strong work ethic and for their unconditional love and support. vi

8 Table of Contents Chapter 1. General Introduction Cellular Composition of the VTA The Behavioral Role of Neurons in the VTA Firing of Dopamine Neurons in vivo Mechanisms Controlling the Firing of Dopamine Neurons Inputs to the VTA from Other Brain Regions Inhibitory Circuits in the VTA Interactions between the VTA and Nucleus Accumbens Dopamine Neurons and Cocaine Summary...19 Chapter 2. Circuit Specificity in the Inhibitory Architecture of the VTA Introduction Methods Results NAc inputs activate GABA B Rs on VTA dopamine neurons Expression of GABA B Rs in the VTA...34 vii

9 2.3.3 Preferential activation of GABA A Rs in GABA neurons and GABA B Rs in dopamine neurons Evidence for synaptic transmission in the activation of GABA B Rs NAc and local VTA GABA afferents activate different receptor pools within VTA dopamine neurons NAc inputs potently inhibit dopamine cell firing through specific activation of GABA B Rs Discussion...41 Chapter 3. Circuit Specific Modulation of GABAergic Synapses in the VTA Introduction Methods Results NAc inputs and local GABAergic inputs are differentially modulated by adenosine A 1 receptors Dopamine D1R activation affects NAc inputs to VTA but does not affect inputs from VTA interneurons Serotonin 5-HT 1B Rs induce long-term depression on NAc inputs but shortterm depression on local GABAergic synapses Cocaine causes long-term depression on NAc inputs to VTA and shortlasting inhibition on VTA GABA to dopamine synapses...72 viii

10 3.4 Discussion...73 Chapter 4. The Role of GABAergic Inputs During and After Cocaine Exposure Introduction Methods Results Repeated cocaine exposure reduces presynaptic GABA release onto dopamine neurons Selective deletion of GABA B Rs from dopamine neurons increases locomotor response to cocaine Discussion...96 Chapter 5. General Discussion and Future Directions References ix

11 List of Figures Chapter 1. General Introduction Figure 1. Reward prediction errors in midbrain dopamine neurons...21 Figure 2. Representation of VTA inputs and outputs...22 Chapter 2. Circuit Specificity in the Inhibitory Architecture of the VTA Figure 1. Optogenetic activation of NAc D1 terminals elicits GABA B activation in VTA dopamine neurons...46 Figure 2. D2 MSNs project directly to the ventral pallidum, but not the VTA...48 Figure 3. GABA B Rs are strongly expressed in dopamine neurons throughout the midbrain...50 Figure 4. Nucleus accumbens inputs preferentially inhibit VTA GABA and dopamine neurons through separate postsynaptic receptors...52 Figure 5. Amplitudes of GABA A oipscs correlate with GABA neuron properties and amplitudes of GABA B oipscs correlate with dopamine neuron properties...54 Figure 6. NAc inputs form dendritic and perisomatic symmetric synapses onto dopamine neurons...56 Figure 7. Evidence for synaptic release of GABA onto GABA B receptors...58 Figure 8. NAc inputs and local GABA interneurons activate separate receptor populations in dopamine neurons...60 x

12 Figure 9. NAc inputs inhibit dopamine neuron firing at multiple frequencies through activation of GABA B Rs...62 Chapter 3. Circuit Specific Modulation of GABAergic Synapses in the VTA Figure 1. Adenosine A 1 Rs have similar effects on NAc synapses to VTA dopamine and GABA neurons, but no effect on local GABAergic inputs...77 Figure 2. Dopamine D1Rs exert differential activity on NAc inputs to VTA and have no effect on local GABAergic inputs to dopamine neurons...79 Figure 3. Local GABAergic inputs to dopamine neurons and NAc inputs are presynaptically inhibited by 5-HT 1B Rs...81 Figure 4. Cocaine inhibits both NAc inputs to the VTA as well as local GABAergic synapses...83 Figure 5. Schematic showing the locations of modulatory receptors on GABAergic inputs in the VTA...85 Chapter 4. The Role of GABAergic Inputs During and After Cocaine Exposure Figure 1. Repeated exposure to cocaine decreases presynaptic GABA release onto GABA B synapses Figure 2. AAV TH-iCre virus allows for selective expression of cre-recombinase in VTA dopamine neurons Figure 3. Conditional deletion of GABA B Rs from dopamine neurons xi

13 Figure 4. Deletion of GABA B Rs from dopamine neurons increases cocaineinduced locomotion xii

14 Chapter 1 General Introduction Obtaining rewards and avoiding harm are essential to the survival of organisms and species. As a consequence, the ability to perform motivated behaviors like reward seeking and behavioral avoidance are under strong evolutionary selective pressure. The midbrain dopamine system is an important component of the brain reward circuitry that appears to be conserved across all vertebrate species (O Connell and Hofmann, 2011). The existence and importance of brain reward circuits was first discovered by Olds and Milner, who showed that animals would learn to press a lever for direct electrical stimulation of various brain regions (Olds and Milner, 1954). In fact, the rewarding effect of electrical stimulation is so strong that animals will suffer punishment or even starve themselves to continue self-stimulating these brain regions (Wise, 2002). Subsequent work has shown that these and other motivated behaviors are dependent on the release of the neurotransmitter dopamine (Fouriezos and Wise, 1976; Wise, 2004). Dopamine neurons within the midbrain are mostly located within the ventral tegmental area (VTA) and substantia nigra pars compacta (SNc). While the SNc is classically thought to control motor behavior, the VTA is thought to be important in rewarding and aversive behavior. However the precise roles of dopamine neurons within these brain regions have been the subject of much debate (Lammel et al., 2014; Wise, 2004). Recent work suggests that the VTA is a highly heterogeneous brain region, containing multiple cell types that play various roles in motivated behaviors. As such, 1

15 understanding the extrinsic inputs and local connectivity of the VTA is necessary in determining how the circuitry of the VTA controls behavior. 1.1 Cellular Composition of the VTA The VTA contains multiple types of neurons that appear to have distinct functions. However, early in vivo and in vitro recordings suggested the existence of only two cell types within the VTA and SNc, dopamine and GABA neurons, which were differentiated by their electrophysiological characteristics (Grace and Onn, 1989; Johnson and North, 1992a). In vivo, dopamine neurons of the lateral VTA and SNc exhibit burst firing, low firing rates, short action potential width and autoreceptor inhibition by dopamine D2 receptor agonists (Grace and Bunney, 1980). In brain slice recordings, dopamine neurons have been identified by the presence of a hyperpolarization-activated cation current (hcurrent), small action potential widths, and low tonic firing rates. GABAergic neurons have been identified by the lack of h-current, short action potential width and high tonic firing rates (Grace and Onn, 1989; Johnson and North, 1992a). Indeed, the use of transgenic mice to identify VTA GABA neurons confirms the validity of these criteria for distinguishing dopamine and GABA neurons within the lateral VTA and SNc (Chieng et al., 2011). Nearly all in vitro studies examining drug-induced synaptic alterations are restricted to specific regions within the lateral VTA and have used these electrophysiological criteria to identify dopamine neurons (Bellone et al., 2011; Chen et al., 2008; Saal et al., 2003; Stuber et al., 2008; Ungless et al., 2001a). However, recent work suggests that other regions of the VTA demonstrate considerable heterogeneity in circuitry and cell types. 2

16 Within the last decade, several labs have concentrated their efforts on elucidating the cellular makeup of the VTA. Although the lateral VTA consists almost entirely of dopamine and GABA neurons, the medial VTA contains a large number of glutamatergic neurons (Yamaguchi et al., 2007, 2011). VTA glutamatergic neurons form excitatory synapses onto dopamine neurons within the VTA and project to other brain regions within the reward circuitry, such as the ventral pallidum and nucleus accumbens (Dobi et al., 2010; Hnasko et al., 2012). Interestingly, some subsets of VTA glutamate neurons corelease dopamine in their projections to the nucleus accumbens, while other glutamate neurons project to the lateral habenula and co-release GABA (Root et al., 2014; Zhang et al., 2015). These findings suggest that VTA glutamate neurons are remarkably promiscuous in their release of neurotransmitters and that they may be important in controlling the local circuitry of the VTA as well as downstream targets. Additionally, investigations of dopamine neurons in the medial VTA suggest that we need to reconsider the idea of a uniform group of dopamine neurons. Medial VTA dopamine neurons display non-canonical electrophysiological characteristics and project to different brain regions than the dopamine neurons contained in the lateral VTA (Lammel et al., 2008; Margolis et al., 2006). Using optogenetic stimulation of excitatory inputs and retrograde labeling of dopamine neurons projecting to specific target regions, Lammel et al. show that distinct populations of dopamine neurons receive separate excitatory inputs and play opposite roles in rewarding and aversive behavior (Lammel et al., 2012). These findings suggest that, within the VTA, there may be parallel streams of information encoded by separate populations of dopamine neurons. 3

17 Local GABA neurons in the VTA make up about 30% of the total population of VTA cells. VTA GABA neurons exert tonic activity in vivo and in vitro and control the firing of dopamine neurons through the activation of GABA A receptors (Steffensen et al., 1998; Tan et al., 2012). Additionally, VTA GABA projections to the nucleus accumbens enhance associative learning by inhibiting cholinergic interneurons (Brown et al., 2012). Many recent studies have examined the behavioral role of local VTA GABA neurons (Cohen et al., 2012; Eshel et al., 2015; Tan et al., 2012; van Zessen et al., 2012). However, as will be discussed later in this chapter, compared to other brain regions, relatively little is known about the synaptic properties and cellular diversity of local inhibitory circuits in the VTA. The advent of genetic targeting and optogenetic manipulations has greatly improved our understanding of the circuit architecture of the VTA within the last decade. The studies highlighted in this section demonstrate the remarkable diversity of neuronal cell types contained in the VTA. In particular, recent studies suggest a role for glutamate transmission in the VTA and highlight a previously unrecognized heterogeneity in VTA dopamine neurons. 1.2 The Behavioral Role of Neurons in the VTA The VTA is a central component of the brain s reward circuitry. As such, the VTA dopamine system is hypothesized to play a role in a dizzying array of behavioral phenomena, including, but certainly not limited to reward prediction, working memory, salience detection, and aversive behavior. Modern advances in genetic targeting of specific cell types combined with the ability to manipulate neuronal activity with 4

18 optogenetics have helped illuminate the roles of VTA neuronal cell types in rewarding and aversive behavior. While dopamine has been implicated in rewarding behaviors and associative conditioning for decades, the causal role of dopamine neurons in these behaviors was difficult to prove until recently. In a landmark study, Deisseroth and colleagues showed for the first time that phasic, but not tonic activation of VTA dopamine neurons was rewarding and sufficient to induce behavioral conditioning (Tsai et al., 2009). Many studies have now confirmed the role of VTA dopamine neurons in encoding rewardrelated behaviors. For example, stimulation of glutamatergic inputs to VTA dopamine neurons is highly rewarding (Lammel et al., 2012; McDevitt et al., 2014) and inhibition of dopamine neurons by direct optogenetic inhibition or by activating local GABA neurons is aversive (Tan et al., 2012). These studies suggest that dopamine neurons bidirectionally control rewarding behaviors and that the local microcircuitry in the VTA is important in the regulation of dopamine neurons. However, the heterogeneity of the VTA suggests that dopamine neurons may play a more nuanced role in behavior. The VTA appears to contain subpopulations of dopamine neurons that receive different inputs and exert opposite behavioral effects (Lammel et al., 2012). Additionally, phasic activation of dopamine neurons has been shown to elicit a depression -like phenotype in mice and dopamine neurons appear to increase their firing rates after chronic social defeat stress (Chaudhury et al., 2013; Friedman et al., 2014). However, another lab has shown the exact opposite result, suggesting that activation of dopamine neurons protects against chronic mild stress (Tye et al., 2013). These 5

19 contrasting results are difficult to reconcile and highlight the idea that there is still much to be discovered about the microcircuitry of the VTA. 1.3 Firing of Dopamine Neurons in vivo The links between dopamine neurons and reward and aversion are evidenced by behavioral experiments as well as direct electrophysiological recordings from dopamine neurons in awake animals. A prominent hypothesis is that dopamine neurons encode reward prediction errors (RPEs), or the difference between expected rewards and obtained rewards. In a series of seminal studies, Schultz and colleagues demonstrated that dopamine neurons fire short bursts of action potentials during the delivery of unexpected rewards (Ljungberg et al., 1992). However, when an expected reward is omitted, dopamine neurons pause their tonic firing. Furthermore, after learning that specific cues predict the receipt of reward, dopamine neurons then begin to fire at the time of the cue, rather than during the receipt of reward (Figure 1; Schultz et al., 1997). These RPE signals have now been identified in monkeys (Schultz et al., 1997), rats (Roesch et al., 2007), and in optogenetically-identified dopamine neurons in mice (Cohen et al., 2012). Furthermore, recent evidence suggests that optogenetic activation and inhibition of dopamine neurons produces the behavioral effects predicted by RPE-based models (Chang et al., 2016; Steinberg et al., 2013). These findings suggest that, across species, the firing of dopamine neurons appears to play an integral role in the expectation of rewards and the motivation to obtain rewards. In addition to encoding reward related information, dopamine neurons change their firing rates in response to aversive events. The majority of dopamine neurons in the VTA 6

20 are inhibited by aversive stimuli such as foot shock (Brischoux et al., 2009; Tsai et al., 1980). However a small proportion of dopamine neurons are phasically excited by aversive stimuli (Cohen et al., 2012; Ungless et al., 2004). These data suggest that dopamine neurons are involved in both reward and aversion and further implicate dopamine neurons in the actions of drugs of abuse and in neuropsychiatric disorders such as depression (Russo and Nestler, 2013). Additionally, the evidence discussed above suggests that changes in the firing of dopamine neurons mediate complex behaviors and highlights the importance of understanding the mechanisms underlying the activity of dopamine neurons. 1.4 Mechanisms Controlling the Firing of Dopamine Neurons In anesthetized and awake animals, dopamine neurons in the VTA and SNc display two different types of activity: tonic and phasic. The tonic background firing rate of dopamine neurons is generally between 1-8 Hz. As discussed above, dopamine neurons also exert phasic bursting and pausing during the receipt and expectation of reward (Figure 1; Schultz et al., 1997). Importantly, the tonic and bursting activities of dopamine neurons are thought to perform different functions through the tonic and phasic release of dopamine in projection targets. Tonic firing releases a baseline level of dopamine in the striatum and activates dopamine D2 receptors (D2Rs), whereas phasic activity triggers a transient surge in dopamine concentration and is thought to be necessary for the activation of dopamine D1 receptors (D1Rs; Goto et al., 2007). D1R- and D2Rexpressing neurons in the striatum exert opposite behavioral effects (Kravitz et al., 2012), suggesting that shifts in tonic and phasic firing of dopamine neurons can rapidly 7

21 alter behavior. The fact that transient alterations in the firing rate of dopamine neurons can affect reward learning (Tsai et al., 2009) illustrates the importance of understanding the mechanisms regulating their firing activity. Although in vivo electrophysiology in awake animals has generated a wealth of data regarding the behavioral events that trigger dopamine neuron activity, most of the mechanistic understanding has been obtained with either anesthetized animals or in vitro patch-clamp electrophysiology. The firing of dopamine neurons is dependent on both intrinsic and synaptic mechanisms. The tonic pacemaker activity of dopamine neurons is controlled by an array of voltage-dependent sodium and potassium channels (Khaliq and Bean, 2010). In slice recordings, dopamine neurons exert tonic activity, but do not exhibit burst firing. In fact, direct current injection into dopamine neurons during patch clamp recordings can increase their firing rate, but does not induce bursting. This suggests that intrinsic mechanisms are not sufficient for burst firing and that tonic and phasic firing are regulated by separate processes. Early recordings in anesthetized animals demonstrated that intracellular Ca 2+ levels control the burst firing of dopamine neurons (Grace and Bunney, 1984). These results suggest that bursting is controlled by either intracellular Ca 2+ stores or potentially through NMDA receptors. Indeed, activation of NMDA receptors by electrical stimulation of afferents, application of NMDA agonists, or even by mimicking NMDA currents with dynamic clamp techniques induces bursting activity in brain slices (Deister et al., 2009; Morikawa et al., 2003). In contrast, AMPA receptors do not appear to be important for burst firing (Johnson and Wu, 2004). Interestingly, selective deletion of NMDA receptors from dopamine neurons abolishes in vivo bursting without affecting tonic firing rates (Zweifel et al., 2009). These findings demonstrate the 8

22 critical role of NMDA receptors and suggest that glutamatergic inputs are important in controlling the phasic firing of dopamine neurons. Afferent inputs are also thought to regulate the inhibition of dopamine neurons, as the oscillatory pacemaker activity of dopamine neurons coupled with excitatory drive from glutamatergic afferents suggest the need for extrinsic inhibitory inputs. Pausing of dopamine neurons is thought to be mediated by a combination of inhibitory receptors. Dopamine neurons express inhibitory ionotropic GABA A and metabotropic GABA B and D 2 receptors (Lacey et al., 1988). While GABA A R inhibition is mediated by a fastactivating (but also fast inactivating) Cl- conductance, GABA B and D2 receptors activate slower g-protein coupled inwardly-rectifying K+ (GIRK) channels (Johnson and North, 1992a; Lüscher et al., 1997). Therefore, the identity of the activated receptor dictates the time course of inhibition. Importantly, the pausing of dopamine neurons during the lack of reward occurs over a time course of hundreds of milliseconds, suggesting that it could be mediated by either ionotropic or metabotropic receptors. However, as opposed to bursting, where NMDA receptors are known to be important (Zweifel et al., 2009), the specific receptor(s) mediating pauses have not yet been determined. Additionally, as will be discussed in subsequent sections, recent advances have furthered our understanding of the inputs that activate GABA A Rs, however the sources of inputs activating GABA B Rs receptors have not been clearly identified. 1.5 Inputs to the VTA from Other Brain Regions By dynamically controlling the firing of dopamine neurons, inputs to the VTA are thought to regulate rewarding and aversive behaviors. However, the roles of specific 9

23 afferent inputs in these behaviors and the circuit specificity of their connections are poorly understood. Recent advances in technology have greatly enhanced our understanding of the inputs to VTA dopamine and GABA neurons. Using a novel and elegant transsynaptic tracing method, Watabe Uchida et al. examined the monosynaptic inputs directly innervating VTA dopamine and GABA neurons (Watabe-Uchida et al., 2012). Their findings highlighted several inputs to dopamine neurons such as the bed nucleus of the stria terminalis, lateral hypothalamus, lateral habenula, dorsal raphe nucleus, laterodorsal tegmental nucleus, and nucleus accumbens (Figure 2). This study also provided quantitative information regarding the relative strength of individual projections to VTA neurons. Interestingly, multiple studies using monosynaptic tracing have now confirmed that VTA dopamine and GABA neurons receive inputs from largely the same brain regions (Beier et al., 2015; Menegas et al., 2015; Watabe-Uchida et al., 2012), suggesting the idea that VTA dopamine and GABA neurons might be regulated by more nuanced forms of synaptic communication. Therefore combinations of other methods, such as optogenetic manipulations, patch clamp electrophysiology, and electron microscopy are important in addressing the functional contributions of excitatory and inhibitory connections and their integration within local circuits. Multiple follow-up studies have examined the roles of various inputs to the VTA. In particular, with monosynaptic tracing, it was shown that the dorsal raphe nucleus (DRN) sends a dense projection to VTA dopamine neurons (Watabe-Uchida et al., 2012). DRN neurons express the neuromodulator serotonin, which is thought to promote aversive behaviors. However, the functional contribution of DRN inputs on synaptic transmission 10

24 and behavior were not known. Recently, McDevitt et al. used an elegant combination of mouse genetics, in vitro and in vivo optogenetics, patch-clamp recordings, and behavior to examine serotonergic and nonserotonergic inputs to the VTA. They found that the DRN inputs to the VTA were largely glutamatergic and evoked intense rewarding behavior (McDevitt et al., 2014). These unexpected findings illustrate the importance of using a multidimensional approach to examine the functional roles of anatomical projections. Additionally, the transsynaptic tracing studies discussed above are not able to adequately address the architecture of local circuits within the VTA. 1.6 Inhibitory Circuits in the VTA As mentioned above, approximately 30% of the neurons in the VTA are GABAergic. Intrinsic GABA neurons and extrinsic GABAergic inputs to the VTA regulate the firing of dopamine neurons. Furthermore, disinhibition, or the removal of tonic inhibition, is a major mechanism by which drugs of abuse hijack the reward circuitry in the VTA (Bonci and Williams, 1996; Cruz et al., 2004; Tan et al., 2010). However, the functions of local and long-range GABAergic circuits in the VTA are poorly understood. The roles of inhibitory GABAergic transmission are perhaps best understood in the cortex and hippocampus (i.e. neocortex). The layered architecture and high degree of connectivity in the neocortex have allowed researchers to thoroughly examine various forms of inhibition. GABAergic neurons within the neocortex are termed interneurons because they generally form local synaptic contacts. Multiple types of interneurons exist within the neocortex, each displaying unique properties and performing different computational roles (Isaacson and Scanziani, 2011; Moore et al., 2010). Neocortical 11

25 interneurons can be classified by their morphology, genetic markers, calcium binding proteins, intrinsic physiological properties, responses to neuromodulators, or their targeting of specific subcellular compartments (Freund and Katona, 2007; McBain and Fisahn, 2001). Moreover, these different properties of interneurons allow them to convey unique forms of inhibition. For example, fast-spiking basket cells in the neocortex target the perisomatic region of nearby pyramidal cells, allowing them to powerfully control the initiation of action potentials (Freund and Katona, 2007). Interestingly, optogenetic activation of fast-spiking cells in the cortex can control cortical oscillations and affect sensory perception (Cardin et al., 2009; Siegle et al., 2014). These findings highlight the fact that GABAergic neurons can powerfully regulate network activity and rapidly control behavior. Compared to cortical and hippocampal networks, the architecture of the VTA is poorly defined. As a consequence, it has been difficult to examine the various forms and sources of inhibition using classical slice physiology experiments. Additionally, most of the work within the midbrain has concentrated on dopamine neurons. However, within the last decade, advances in technology have spurred an interest in defining the behavioral roles of local and extrinsic inhibitory inputs to dopamine neurons. As mentioned above, optogenetic activation of VTA GABA neurons produces aversive behavior (Tan et al., 2012) and can disrupt rewarding behavior (van Zessen et al., 2012), mediated by the local inhibition of dopamine neurons via ionotropic GABA A receptors. Recent work has focused the rostromedial tegmental nucleus (RMTg), or tail of the VTA, a mostly GABAergic nucleus just posterior to the VTA that inhibits dopamine neurons through activation of GABA A Rs (Jhou et al., 2009). The RMTg receives 12

26 excitatory inputs from the lateral habenula, a brain region thought to be involved in the generation of dopamine neuron reward prediction errors (Bromberg-Martin et al., 2010; Tian and Uchida, 2015). The lateral habenula projections to RMTg are currently thought to provide a sort of feedforward inhibition of dopamine neurons. However, the RMTg is difficult to distinguish from the VTA, especially in mice, and it is unknown whether GABA neurons in the RMTg display distinct properties from the VTA GABA neurons. Additionally, a subset of VTA GABA neurons send long-range projections to the nucleus accumbens and can enhance associative learning (Brown et al., 2012). However, it is not known whether the same neurons collateralize to dopamine neurons or whether there are separate subsets of VTA GABA neurons. Therefore, future work examining the heterogeneity of VTA GABA neurons and the various (assumed) forms of inhibition will be necessary. The role of GABAergic inhibition in the VTA is still poorly understood. VTA dopamine and GABA neurons appear to display differential sensitivities to GABA agonists. Interestingly, local infusions of GABA A R agonists paradoxically increase the firing rate of dopamine neurons, augment dopamine release, and produce rewarding behavior (Grace and Bunney, 1984; Kalivas et al., 1990; Xi and Stein, 1998). These findings suggest that GABA A -mediated activation of dopamine neurons might be caused by disinhibition, mediated through inhibition of local GABA neurons. Indeed, there appears to be differential expression of GABA A Rs on dopamine neurons and GABA neurons, making GABA neurons more sensitive to GABA A R agonists (Tan et al., 2010). Taken together, these findings show a role for GABA A Rs in dopamine neurons, but 13

27 suggest the possibility that other inhibitory sources may be important in controlling dopamine neurons. Metabotropic GABA B Rs represent another major source of inhibition in VTA dopamine neurons. The firing rate of dopamine neurons in vivo is dramatically reduced by local infusion of GABA B R agonists (Kalivas et al., 1990; Paladini and Tepper, 1999) and blockade of GABA B Rs can induce burst firing in VTA dopamine neurons (Erhardt et al., 2002). Additionally, activation of GABA B Rs dramatically reduces reward-related behaviors and the effects of drugs of abuse (Kalivas et al., 1990; Slattery et al., 2005). These findings suggest that GABA B Rs may play an integral role in the regulation of dopamine neurons. Although the sources of inhibition converging on VTA dopamine neurons have been difficult to identify, indirect evidence suggests that inputs to GABA A Rs and GABA B Rs come from different afferent inputs. Sugita et al. first suggested this hypothesis when they found that dopamine neurons exhibit spontaneous GABA A -mediated currents, but GABA B currents only occurred with trains of electrical stimulation (Sugita et al., 1992). As has been shown in other brain regions, GABA B Rs are activated by volume release of GABA (Isaacson et al., 1993; Scanziani, 2000), suggesting the possibility that spontaneous release is not strong enough to activate extrasynaptic GABA B Rs. However, Sugita et al. also showed that electrically-evoked GABA B currents were presynaptically inhibited by serotonin (5-HT), whereas GABA A currents were unaffected (Sugita et al., 1992). These results suggest differential expression of presynaptic receptors on inputs mediating GABA A and GABA B currents. In subsequent years, various studies have further shown differential modulation of GABA A and GABA B transmission by dopamine 14

28 D1 receptors, 5-HT 1B receptors, adenosine A 1 receptors, and even cocaine (Bonci and Williams, 1996; Cameron and Williams, 1993, 1994, 1995; Sugita et al., 1992; Wu et al., 1995). From these studies, it appears that GABA release from the inputs activating GABA B Rs comes from different sources than GABA terminals activating GABA A Rs. However, the hypothesis that dopamine neurons receive divergent inhibitory inputs has not been directly tested. Additionally, there is inconclusive evidence regarding which inputs to the VTA activate GABA B Rs. 1.7 Interactions between the VTA and Nucleus Accumbens The nucleus accumbens (NAc) and the VTA each play an integral role in reward seeking behavior and are implicated in drug abuse, depression, and other neuropsychiatric disorders. The NAc is part of the ventral striatum and is the main projection target of VTA dopamine neurons. Importantly, the NAc is thought to control motivated behaviors through the interaction of glutamatergic and dopaminergic inputs. Medium spiny neurons (MSNs) are the main projection neurons of the NAc. MSNs are GABAergic and make up approximately 90% of the neurons in the NAc. There are two largely non-overlapping types of MSNs in the NAc: dopamine D1 receptorcontaining MSNs (D1 MSNs) and dopamine D2 receptor-containing MSNs (D2 MSNs). These MSNs express different neuropeptides, project to distinct brain regions, and exert different roles on motivated behaviors. D1 MSNs express substance P and the endogenous opioid dynorphin and directly innervate the midbrain. On the other hand, D2 MSNs express enkephalin and send projections to the ventral pallidum, which sends GABAergic projections to the midbrain (Gerfen and Surmeier, 2011). The divergent 15

29 circuitry of striatal MSNs suggests that D1 and D2 neurons play opposite roles in behavior. Indeed, optogenetic activation of D1 MSNs is reinforcing, while activation of D2 MSNs is aversive (Kravitz et al., 2012). The firing of NAc MSNs is dependent on inputs from multiple brain regions, including the hippocampus, basolateral amygdala, and cortex (Britt et al., 2012). The effects of dopamine release in the NAc are complicated, given that dopamine can have both pre- and postsynaptic effects on glutamatergic and GABAergic transmission (Tritsch and Sabatini, 2012). Specifically, dopamine induces glutamatergic long-term potentiation and increases the excitability of D1 neurons, while it causes long-term depression of glutamatergic inputs and decreases the excitability of D2 neurons (Gerfen and Surmeier, 2011). Therefore, the net effect appears to be that dopamine increases excitatory drive in D1 neurons, but decreases the excitatory drive of D2 neurons. As discussed above, VTA dopamine neurons display phasic bursts of action potentials during reward predictive behavior and decrease their firing during the omission of reward. Modern advances in fast-scan cyclic voltammetry have greatly increased our understanding of subsecond dopamine release in the NAc (Phillips et al., 2003). As predicted by in vivo recordings of dopamine neurons, the phasic release of dopamine in the nucleus accumbens also transiently increases during the presentation of reward predictive cues (Day et al., 2007). This finding, together with the mechanistic understanding of dopamine in the nucleus accumbens, suggests that phasic firing of dopamine neurons is transiently increases the activity of D1 MSNs and decrease the activity of D2 MSNs. Indeed, a large percentage of neurons in the NAc ramp up their firing rates during the expectation of reward (Schultz et al., 1992). However, the role of 16

30 the increased firing in the NAc is not known. While the expectation is that the changes in MSN firing rates will promote reward behavior, D1 MSNs project directly to the VTA and may serve as a negative feedback mechanism onto dopamine neurons. The role of dopamine inputs to the NAc have been the subject of extensive investigation, however the functions of feedback projections from the NAc to the VTA are poorly understood. Electron microscopy studies have suggested that NAc inputs to the VTA synapse directly onto dopamine neurons (Somogyi et al., 1981). In fact, of all the monosynaptic inputs to VTA dopamine neurons identified by rabies tracing experiments, the strongest input appears to originate in the NAc (Menegas et al., 2015; Watabe-Uchida et al., 2012). However, the functional contribution of these inputs to the VTA has been the subject of intense debate. Multiple labs have now examined the inputs from the NAc to the VTA using optogenetics and patch clamp electrophysiology (Bocklisch et al., 2013; Matsui et al., 2014; Xia et al., 2011). Surprisingly, these studies suggest that NAc D1 MSNs almost exclusively innervate GABA neurons, rather than dopamine neurons. Thus, anatomical studies suggest that NAc input directly inhibit dopamine neurons, while physiological studies suggest that NAc projections disinhibit dopamine neurons through inhibition of VTA GABA neurons. An alternative hypothesis is that NAc inputs to dopamine neurons use a different form of synaptic communication that was not detected in previous physiological experiments. One possibility is that NAc inputs inhibit dopamine neurons exclusively through GABA B Rs. In dopamine neurons, GABA A Rs are activated by single electrical pulses on GABAergic afferents, while GABA B Rs are only activated by trains of stimuli (Johnson and North, 1992a). Thus, it is possible that NAc inputs onto dopamine neurons 17

31 have eluded detection because previous studies had not used the right conditions to detect GABA B responses. Furthermore, the identification of afferents activating dopamine neuron GABA B Rs is an important question because these inputs are powerfully modulated by multiple GPCRs and appear to be involved in drug-related behaviors. 1.8 Dopamine Neurons and Cocaine The dopaminergic system is heavily implicated in the effects of drugs of abuse, as all known drugs of abuse increase extracellular dopamine levels in the NAc (Di Chiara and Imperato, 1988). The acute behavioral effects of cocaine, as well as other psychostimulants, depend on the activation of dopamine receptors (De Wit and Wise, 1977). Cocaine, in particular, increases extracellular dopamine concentrations by inhibiting dopamine reuptake through the dopamine transporter. However, dopamine transporter knockout animals show reduced sensitivity to cocaine, but will still selfadminister cocaine (Rocha et al., 1998). These results suggest that other mechanisms are involved in the behavioral actions of cocaine. Plasticity of the dopamine system is thought to underlie the long-term effects of drugs of abuse. Many studies have now shown that excitatory synapses onto dopamine neurons undergo drug-induced long-term potentiation after even a single injection of cocaine (Borgland et al., 2006; Chen et al., 2008; Saal et al., 2003; Ungless et al., 2001a; Yuan et al., 2013). Moreover, it appears that glutamatergic plasticity is necessary for the sensitizing effects of repeated cocaine injection (Borgland et al., 2006; Ungless et al., 2001a). These studies suggest that increased excitatory drive onto dopamine neurons contributes to maladaptive drug-seeking behavior. However, another possibility is that 18

32 cocaine increases dopaminergic activity by inhibiting GABAergic inputs to dopamine neurons. While the effects of cocaine on excitatory transmission are well documented, the effects on inhibitory transmission are poorly understood. As illustrated above, GABAergic inputs tonically regulate the firing of dopamine neurons in vivo. Disinhibition of the dopamine system by drugs of abuse is well documented and plays an important role in the mechanism of multiple drugs of abuse (Cruz et al., 2004; Johnson and North, 1992b; Matsui et al., 2014; Tan et al., 2012). In slice recordings, cocaine inhibits electrically-evoked GABA B currents, but does not affect GABA A currents (Cameron and Williams, 1994), suggesting that cocaine disinhibits dopamine neurons through circuit-specific modulation of specific GABAergic inputs. The action of cocaine on GABA B is mediated by the release of 5-HT and endocannabinoids and the presynaptic inhibition of GABA-releasing terminals (Cameron and Williams, 1994; Wang et al., 2015). Interestingly, the dose of cocaine required to achieve this effect is lower than the dose required for the blockade of the dopamine transporter, suggesting that the behavioral effects of cocaine may be at least partially mediated by the presynaptic inhibition of GABA release. However, this hypothesis has not been directly tested. Additionally the effects of cocaine exposure on the plasticity of GABA inputs onto dopamine neurons have not been determined. 1.9 Summary The VTA is a highly heterogeneous brain region involved in rewarding and aversive behaviors. As we understand more about the local microcircuitry of the VTA and its inputs from other brain regions, we can begin to understand its roles in drug addiction 19

33 and other neuropsychiatric disorders. While the inputs to the VTA have been extensively studied, the functional roles of these inputs and their effects on VTA microcircuitry are still poorly understood. In the following chapters, I will examine the roles of extrinsic and intrinsic sources of GABAergic inhibition on the physiology of dopamine neurons and their role in behavior. First, I examine the physiological influence of local VTA GABA neurons and long-range nucleus accumbens inputs to the VTA. Then, I will explain the role of various neuromodulators in the circuit-specific regulation of inhibitory synapses within the VTA. Finally, I will demonstrate evidence that inhibitory inputs to the VTA are modified by repeated exposure to cocaine and that these GABAergic inputs dampen the behavioral effects of cocaine. 20

34 Figure 1. Reward prediction errors in midbrain dopamine neurons. Illustration of the iconic dopaminergic prediction error. Dopamine neurons fire to unpredicted outcomes (top) and to cues that predict outcomes (middle); the same neurons will not fire to predicted outcomes (middle) and will suppress firing when predicted outcomes are omitted. This pattern of response is the fingerprint of a prediction error. This figure was adapted from (Schoenbaum et al., 2013). 21

35 Figure 2. Representation of VTA inputs and outputs. Amyg, amygdala; BNST, bed nucleus of stria terminalis; LH, lateral hypothalamus; LC, locus coeruleus; LDT, laterodorsal tegmental nucleus; LHb, lateral habenula; mpfc, medial prefrontal cortex; NAc, nucleus accumbens; OFC, orbitofrontal cortex; PPTg, pedunculopontine tegmental nucleus (aka PPN); RMTg, rostral medial tegmental nucleus; VP, ventral pallidum; vsub, ventral subiculum; VTA, ventral tegmental area. This figure was adapted from (Juarez and Han, 2016). 22

36 Chapter 2 Circuit Specificity in the Inhibitory Architecture of the VTA 2.1 Introduction Ventral tegmental area (VTA) dopamine neurons control motivated behaviors for both natural rewards and drugs of abuse (Di Chiara and Imperato, 1988; Stuber et al., 2008). Dopamine neurons increase their activity during unexpected rewards and pause during the omission of expected rewards (Bromberg-Martin et al., 2010; Schultz et al., 1997). Afferent inputs dynamically regulate the firing of dopamine neurons (Paladini and Roeper, 2014), but the roles of specific excitatory and inhibitory inputs are poorly understood. Interestingly, the vast majority of synapses onto midbrain dopamine neurons are GABAergic (Ribak et al., 1976; Bolam and Smith, 1990; Henny et al., 2012), highlighting the importance of inhibition in the VTA. Recent anatomical studies suggest that the largest input to VTA dopamine neurons arises from the nucleus accumbens (NAc) (Beier et al., 2015; Menegas et al., 2015; Watabe-Uchida et al., 2012). The principal projection neurons of the NAc are GABAergic, implying powerful inhibitory control over dopamine neuron activity. However, the functional role of NAc inputs to the VTA has been the subject of considerable debate. Recent studies combining optogenetic stimulation and slice electrophysiology suggest that NAc inputs preferentially synapse onto GABA neurons of the VTA (Bocklisch et al., 2013; Matsui et al., 2014; Xia et al., 2011). As VTA GABA neurons directly inhibit dopamine neurons (Tan et al., 2012; van Zessen et al., 2012), these findings suggest that NAc projections increase the firing of dopamine neurons 23

37 through disinhibition. Thus, anatomical and electrophysiological studies propose opposite roles for NAc projections in regulating dopamine neurons, highlighting the importance of determining their role in the local microcircuitry of the VTA. Interestingly, indirect evidence suggests that the GABAergic inputs to dopamine neurons activating GABA A receptors (GABA A Rs) and GABA B receptors (GABA B Rs) arise from separate afferent sources (Cameron and Williams, 1993, 1994; Sugita et al., 1992). However, little is known about the circuit organization and functional contribution of local and long-range GABAergic circuits of the VTA. Here we use a multidisciplinary approach to examine the synaptic architecture of NAc projections to the VTA. We show that NAc inputs to the VTA potently inhibit the firing of dopamine neurons through monosynaptic activation of GABA B Rs. Furthermore, GABAergic inputs from the NAc and from local VTA interneurons activate distinct receptor populations within dopamine neurons. Our findings reveal circuit specificity in the local and long-range GABAergic inputs to dopamine neurons and highlight the importance of the NAc as a critical regulator of dopamine neuron activity. 24

38 2.2 Methods Experimental Subjects Pdyn-IRES-Cre (Krashes et al., 2014), VGAT-IRES-Cre (Vong et al., 2011), and A2Acre (Gong et al., 2007) mice on a C57BL/6 background were used for electrophysiological experiments. Mice of either sex were used. Adult (8+ weeks) mice were acclimatized to the animal facility for more than 1 week before undergoing surgery. Animals were housed on a 12 hr:12 hr reverse light:dark cycle with lights off at 7:00 a.m. All experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and with the approval of the National Institute on Drug Abuse animal care and use committee. Surgeries Microinjection needles (29G) were connected to a 2 µl Hamilton syringe and filled with concentrated adeno-associated virus (~10 12 infectious units ml -1 ) encoding DIO-ChR2- eyfp (CaMKII promoter) or DIO-Caspase3-Tevp (CaMKII promoter). Mice were anesthetized with 150 mg kg -1 ketamine and 50 mg kg -1 xylazine and placed in a stereotaxic frame. Microinjection needles were placed unilaterally or bilaterally into the NAc (+1.7 AP, ±1.6 ML, -4.2 DV, 12 angle) and 500 nl virus was injected for 5 min. For VTA injections (-3.3 AP, ±1.9 ML, -4.6 DV, 14 angle), 300 nl virus was injected for 3 min. The needles were left in place for an additional 5 min to allow for diffusion of virus from the injection site. Electrophysiology 25

39 Three to eight weeks after surgery, mice were anesthetized with Euthasol and perfused with ice-cold artificial cerebrospinal fluid (ACSF). To preserve neuronal health, n- methyl-d-glucamine (NMDG) was substituted for sodium in both the cutting and perfusion solutions (Ting et al., 2014). The cutting and perfusion ACSF contained (in mm) 92 NMDG, 20 HEPES, 25 glucose, 30 NaHCO 3, 1.2 Na-phosphate, 2.5 KCl, 5 sodium ascorbate, 3 sodium pyruvate, 2 thiourea, 10 Mg-sulfate, and 0.5 CaCl 2 (ph7.35, ~305 mosm). Brains were sectioned at µm using a VT-1200 vibratome (Leica). The ACSF used for holding slices before recording was identical to the cutting solution, except it contained 1 mm MgCl 2 and 2 mm CaCl 2 and 92 mm NaCl instead of NMDG. The ACSF used to perfuse the slice during recording contained (in mm): 125 NaCl, 2.5 KCl, 1.25 Na-phosphate, 1 MgCl 2, 2.4 CaCl 2, 26 Na-bicarbonate, and 11 glucose. All ACSF solutions were bubbled with 95% O 2 and 5% CO 2. After >1 hour slice incubation, eyfp expression was examined in slices containing the virus injection sites to ensure placement accuracy. If the injection region was adequately infected with virus, sections containing the VTA/SNc were transferred to the recording chamber and superfused with C ACSF. For optogenetic experiments, a 200 µm optical fiber coupled to a diode-pumped solid-state laser was positioned above the slice and aimed at the recorded cell. Whole cell recordings were performed using MΩ patch pipettes backfilled with an internal solution containing (in mm): 115 K-MethaneSulfate, 20 NaCl, 1.5 MgCl 2,10 BAPTA, 10 Na 2 -phosphocreatine, 4 Mg-ATP, and 0.4 Na 2 -GTP (ph 7.35, 285 mosm). Whole-cell voltage clamp recordings were made using a MultiClamp 700B amplifier (10 khz digitization and 1-2 khz low-pass Bessel filter) with pclamp

40 software (Molecular Devices). Loose seal cell-attached recordings were performed in voltage clamp and the voltage was adjusted to keep the holding current around 0 pa. Cell-attached recordings were used to analyze the spike width and tonic firing rate. Throughout the whole-cell recordings, series resistance was monitored (0.1 Hz) with a 5 mv hyperpolarizing step and maintained below 30 MΩ. Cells in which the series resistance changed more than 20% during the recording were excluded from analysis. Dopamine neurons were identified by morphology, spike width, tonic firing rates, the presence of a hyperpolarization-induced h-current, and in a subset of cells, cell filling with biocytin and immunohistochemistry. Optically-evoked GABA B responses were obtained every 30 s with 20 pulses of 473 nm wavelength light (2-5 mw, 20 Hz, 3 ms) with neurons voltage clamped at -55 mv. GABA A responses were obtained every 10 s with 2 pulses of 473 nm wavelength light (2-5 mw, 3 ms) with 50 ms interpulse intervals and neurons voltage clamped at -70 mv. Paired pulse ratios (PPRs) were measured as the amplitude of the second pulse divided by the amplitude of the first pulse (Zucker and Regehr, 2002). Electrically evoked GABA B responses were obtained every 1 min with 10 pulses (70 Hz, 500 µs) using a bipolar stimulating electrode and a constant current stimulation unit with neurons voltage-clamped at -55 mv. To obtain pure GABA B responses with electrical stimulation, the following were added to the ACSF (in µm): 1000 kynurenic acid, 100 picrotoxin, 1 sulpiride, 5 LY , to block AMPA and NMDA, GABA A, dopamine-d2, and mglur1 responses, respectively. For uncaging experiments, ACSF containing picrotoxin (100µM) and Rubi-GABA (10 mm) was recirculated in the bath and responses were obtained every 60 sec with 1 sec pulses of 473 nm light. For iontophoretic experiments, glass pipettes ( 7-10 MΩ) were filled with 27

41 ACSF containing 1mM GABA and M NaCl. A retaining current of -50 na and a 200 na current was used to release GABA. For pharmacological experiments, we recorded baseline responses for at least 10 min and drugs were bath applied for 5-10 min. All drugs used were obtained from either Tocris Bioscience or Sigma. Combination of In situ hybridization and Immunolabeling Male wild-type C57Bl/6 mice (n=3) were anesthetized and perfused transcardially with 4% (w/v) paraformaldehyde (PF) in 0.1 M phosphate buffer (PB) (ph 7.3). Brains were left in 4% PF for 2 h at 4 C, rinsed with PB, and transferred sequentially to 18% sucrose solutions in PB. Coronal sections of the VTA, 12 μm in thickness, were prepared. Cryosections were incubated for 10 min in PB containing 0.5% Triton X-100, rinsed two times for 5 min each with PB, treated with 0.2N HCl for 10 min, rinsed two times for 5 min each with PB and then acetylated in 0.25% acetic anhydride in 0.1 M triethanolamine (ph 8.0) for 10 min. Sections were rinsed two times for 5 min each with PB, and then post-fixed with 4% PF for 10 min. Before hybridization and after a final rinse with PB, the free-floating sections were incubated in hybridization buffer (50% formamide; 10% dextran sulphate; 5 x Denhardt s solution; 0.62 M NaCl; 50 mm DTT; 10 mm EDTA; 20 mm PIPES (ph 6.8); 0.2% SDS; 250 µg/ml salmon sperm DNA; 250 µg/ml trna) for 2 h at 55 C. Sections were hybridized for 16 h at 55 C in hybridization buffer containing [35 S ]- and [33 P ]-labeled single-stranded antisense of GABA B (nucleotides, GenBank accession number AF ) probe at 10 7 cpm/ml. Sections were treated with 4 µg/ml RNase A at 37 C for 1 h, washed with 1X SSC and 50% formamide at 55 C for 1 h, and with 0.1X SSC at 68 C for 1 h. After the last SSC wash, sections were rinsed with PB and incubated for 1 h in PB supplemented with 4% bovine serum albumin and 0.3% 28

42 Triton X-100. This was followed by the overnight incubation at 4 C with an anti-th mouse monoclonal antibody (1:500; MAB 318; Millipore) for which specificity has been documented (Tagliaferro and Morales, 2008). After being rinsed three times for 10 min each in PB, sections were processed with an ABC kit (Vector Laboratories). The material was incubated for 1 h at room temperature in a 1:200 dilution of the biotinylated secondary antibody, rinsed with PB, and incubated with avidin-biotinylated horseradish peroxidase for 1 h. Sections were rinsed and the peroxidase reaction was then developed with 0.05% 3, 3-diaminobenzidine-4 HCl (DAB) and 0.03% hydrogen peroxide (H2O2). Free-floating sections were mounted on coated slides. Slides were dipped in Ilford K.5 nuclear tract emulsion (Polysciences; 1:1 dilution in double distilled water) and exposed in the dark at 4 C for 4 weeks before development. Electron microscopy Methods were described in previous study (Zhang et al., 2015). Briefly, vibratome tissue VTA sections were rinsed and incubated with 1% sodium borohydride to inactivate free aldehyde groups, rinsed and then incubated with blocking solution. Sections were then incubated with primary antibodies [sheep anti-th (1:1,000, EMD Millipore, AB1542), mouse anti-gaba B1 (1:1000, Abcam, ab55051), rabbit anti-yfp (1:2,000, Frontier Institute, GFP-Rb-Af2020) and guinea pig anti-vgat (1:500, Frontier Institute, VGAT- GP-Af1000)]. All primary antibodies were diluted with 1% normal goat serum (NGS), 4% BSA in PB supplemented with 0.02% saponin and incubations were for 24 h at 4 C. Sections were rinsed and incubated overnight at 4 C in the corresponding secondary antibodies. Sections were rinsed in PB, and then in double-distilled water, followed by silver enhancement of the gold particles with the Nanoprobe Silver Kit (2012, 29

43 Nanoprobes) for 7 min at room temperature. Next, sections were incubated in avidinbiotinylated horseradish peroxidase complex in PB for 2 h at room temperature and washed. Peroxidase activity was detected with 0.025% 3,3'-diaminobenzidine (DAB) and 0.003% H 2 O 2 in PB for 5-10 min. Sections were rinsed with PB and fixed with 0.5% osmium tetroxide in PB for 25 min, washed in PB followed by double distilled water and then contrasted in freshly prepared 1% uranyl acetate for 35 min. Sections were dehydrated through a series of graded alcohols and with propylene oxide. Afterwards, they were flat embedded in Durcupan ACM epoxy resin (14040, Electron Microscopy Sciences). Resin-embedded sections were polymerized at 60 C for 2 days. Sections of 65 nm were cut from the outer surface of the tissue with an ultramicrotome UC7 (Leica Microsystems) using a diamond knife (Diatome). The sections were collected on formvar-coated single slot grids and counterstained with Reynolds lead citrate. Sections were examined and photographed using a Tecnai G 2 12 transmission electron microscope (Fei Company) equipped with a digital micrograph 3.4 camera (Gatan). Ultrastructural analysis of brain tissue Serial ultrathin sections of the VTA (bregma mm to mm) from 4 male VGAT-ChR2-eYFP mice. Synaptic contacts were classified according to their morphology and immunolabel, and photographed at a magnification of 6,800-13,000. The morphological criteria used for identification and classification of cellular components or type of synapse observed in these thin sections were as previously described (Zhang et al., 2015). In the serial sections, a terminal or cell body containing greater than 5 immunogold particles were considered as immunopositive terminal or cell body. Pictures were adjusted to match contrast and brightness by using Adobe Photoshop 30

44 (Adobe Systems Incorporated, Seattle, WA). This experiment was successfully repeated three times. Electron microscopy and confocal analysis quantification occurred blindly. Fluorescence microscopy Methods were described in previous study (Zhang et al, 2015). Briefly, free floating coronal sections (40 µm) from VGAT-ChR2-eYFP mice (n = 4) were incubated for 1 h in PB supplemented with 4% BSA and 0.3% Triton X-100. Sections were then incubated with cocktails of primary antibodies: sheep anti-th (1:500) + mouse anti-gaba B1 (1:1000) + rabbit anti-yfp (1:2,000) + guinea pig anti-vgat (1:500) overnight at 4ºC. After rinsing 3 10 min in PB, sections were incubated in a cocktail of the corresponding fluorescence secondary antibodies: Alexa-Fluor-405-anti-sheep + Alexa-Fluor-488-antirabbit + Alexa-Fluor-594-anti-mouse + Alexa-Fluor-647-anti-guinea pig for 2 h at room temperature. After rinsing, sections were mounted on slides. Fluorescent images were collected with an Olympus FV1000 Confocal System (Olympus). Images were taken sequentially with different lasers with 100 oil immersion objectives and Z-axis stacks were collected at 0.2 µm. Imaris microscopy software (Bitplane Inc., South Windsor, CT) was used to analyze Z-stacks of confocal images from 4 VGAT-ChR2-eYFP mice. This experiment was successfully repeated three times. 31

45 2.3 Results NAc inputs activate GABA B Rs on VTA dopamine neurons Recent studies suggest that the NAc sends a dense GABAergic input to the VTA(Beier et al., 2015; Bocklisch et al., 2013; Matsui et al., 2014; Menegas et al., 2015; Watabe- Uchida et al., 2012; Xia et al., 2011). We examined whether NAc projections activate dopamine neuron GABA B Rs. NAc medium spiny neurons (MSNs) express either D1 or D2 dopamine receptors and are the principal projection neurons of the NAc. D1 MSNs of the NAc also express the neuropeptide dynorphin and project directly to the midbrain (Gerfen et al., 1990). To examine the projections from D1 MSNs to midbrain dopamine neurons, we injected adeno-associated virus encoding cre-dependent channelrhodopsin-2 (AAV DIO-ChR2-YFP) into the NAc of preprodynorphin-ires-cre (Dyn-cre) mice (Krashes et al., 2014) (Figure 1a,b). ChR2-YFP fluorescence was prominent in the lateral VTA and substantia nigra (Figure 1c). Whole-cell recordings were performed in horizontal brain slices containing the VTA. We recorded from midbrain dopamine neurons identified by electrophysiological characteristics and, in a subset of cells, the presence of tyrosine hydroxylase (TH) immunofluorescence (Figure 1c). In voltage clamp recordings, stimulation of NAc terminals produced slow optically-evoked IPSCs (oipscs) in dopamine neurons (Figure 1d). The size of oipscs increased with increasing number of light pulses (Figure 1e) and stimulation frequency (Figure 1f). However, oipscs were much less pronounced at 100 Hz (Figure 1f), likely due to the intrinsic kinetics of ChR2 (Zhang et al., 2006). 32

46 We asked whether NAc oipscs in dopamine neurons were mediated through monosynaptic or polysynaptic release of neurotransmitter. We optically stimulated NAc terminals while recording from dopamine neurons and found that application of the voltage-gated sodium channel antagonist tetrodotoxin (TTX, 1 µm) completely blocked oipscs (Figure 1g,h). Co-application of TTX and the potassium channel antagonist, 4- aminopyridine (4-AP, 1mM), which enables ChR2-mediated release of neurotransmitter in the absence of action potentials, restored oipscs (Figure 1g,h). These findings show that light-evoked IPSCs were both action-potential dependent and mediated through monosynaptic release of neurotransmitter from NAc terminals. To determine the ion channel mediating oipscs, we examined the current-voltage relationship and found a reversal potential around -90 mv (Figure 1i), suggesting the activation of a potassium channel. Optically-evoked IPSCs were substantially reduced with application of tertiapin (0.5 µm), a g-protein coupled inwardly-rectifying potassium (GIRK) channel antagonist (Figure 1j,m). D1 MSNs contain both dynorphin, an endogenous agonist of the kappa-opioid receptor (Chavkin et al., 1982), and GABA, both of which can activate GIRKs. NorBNI (0.1 µm), a kappa opioid antagonist, had no effect on oipscs (Figure 1k,m). However the GABA B antagonist CGP (100 µm) blocked oipscs (Figure 1l,m). These findings show that D1 MSNs project directly to the midbrain and monosynaptically activate GABA B Rs in dopamine neurons. We asked whether D2 MSNs also project to the VTA and synapse onto dopamine neurons. We injected AAV DIO-ChR2-YFP into the NAc of A2A-cre mice and performed whole-cell patch clamp electrophysiology in horizontal sections containing the VTA (Figure 2a). Optical stimulation of D2 MSN terminals had no effect on VTA 33

47 GABA or dopamine neurons, but produced GABA A oipscs in ventral pallidal neurons (Figure 2b,c). These findings are consistent with previous reports(bocklisch et al., 2013; Gerfen et al., 1990) which show that D2 MSNs send projections to the ventral pallidum, but do not project directly to the VTA Expression of GABA B Rs in the VTA We next examined the localization of GABA B Rs within the VTA. GABA B Rs are heteromultimeric assemblies containing the GABA B1 and GABA B2 subunits (Kaupmann et al., 1998; Schwenk et al., 2010, 2016). However, the GABA B1 subunit is responsible for binding GABA and is an essential component of functional GABA B Rs (Kaupmann et al., 1998; Schuler et al., 2001). Therefore, to examine colocalization of GABA B Rs with dopamine neurons, we combined immunostaining for TH and in situ hybridization for the GABA B1 R. We observed GABA B1 R expression throughout the VTA and in surrounding regions (Figure 3a). We found GABA B1 R expression in the vast majority of TH+ neurons throughout the VTA and SNc (Figure 3b,c). Additionally, we found GABA B1 R expression in TH- cells (Figure 3d), suggesting that, while most dopamine neurons contain GABA B Rs, non-dopamine cells also express GABA B Rs Preferential activation of GABA A Rs in GABA neurons and GABA B Rs in dopamine neurons We wondered whether NAc inputs to the VTA show selectivity in their connections with GABA or dopamine neurons. In the lateral VTA and substantia nigra pars compacta, GABA and dopamine neurons make up the vast majority of neurons, as glutamatergic neurons are largely confined to the medial VTA (Yamaguchi et al., 2007). GABA and 34

48 dopamine neurons in the lateral VTA can be differentiated by their electrophysiological characteristics and by the absence or presence of TH, respectively. In particular, a recent study using transgenic GAD-GFP mice shows that GABA and dopamine neurons can be completely distinguished by their action potential width (Chieng et al., 2011). We identified dopamine neurons by their morphology, long action potential width, low tonic firing rates, presence of a hyperpolarization-activated cation current (h-current; Figure 4a, Figure 5a-c) and in a subset of cells, TH immunoreactivity. GABA neurons were identified by their short action potential width, high tonic firing rates, lack of h-current (Figure 4a, Figure 5a-c), and in a subset of cells, lack of TH immunoreactivity. Optical stimulation of NAc terminals produced prominent activation of GABA A Rs in VTA GABA neurons, but much smaller GABA A oipscs in dopamine neurons (Figure 4b-d), as reported previously(bocklisch et al., 2013; Matsui et al., 2014; Xia et al., 2011). When we compared the size of GABA A oipscs with distinguishing electrophysiological characteristics of dopamine and GABA neurons, we found that the size of the GABA A oipsc was positively correlated with firing rate (Figure 5e) and negatively correlated with spike width and h-current amplitude (Figure 5d,f). These data further support the idea that NAc D1 inputs more robustly activate GABA A Rs in GABA neurons relative to dopamine neurons. In contrast, light evoked GABA B currents were significantly larger in dopamine neurons than in VTA GABA neurons (Figure 4e-g). The amplitude of the GABA B oipsc was positively correlated with action potential width and h-current (Figure 5g,i) and negatively correlated with firing rate (Figure 5h). These findings indicate that NAc projections preferentially activate GABA A Rs in VTA GABA neurons, while selectively activating GABA B Rs in dopamine neurons (Figure 4h). 35

49 2.3.4 Evidence for synaptic transmission in the activation of GABA B Rs The ability of NAc inputs to activate dopamine neuron GABA B Rs might be due to excess GABA release during optogenetic activation, causing spillover from NAc to GABA neuron synapses. Indeed, electrically-evoked GABA B IPSCs (eipscs) in other brain regions are mediated by spillover of GABA from the synaptic cleft, activating extrasynaptic GABA B Rs (Isaacson et al., 1993; Scanziani, 2000). However, studies with electron microscopy show that dopamine neuron GABA B Rs are located both synaptically and extrasynaptically (Boyes and Bolam, 2003). To examine whether NAc inputs act by synaptic or volume release, we first asked whether NAc inputs to the VTA synapse directly onto dopamine cells. We injected AAV DIO-ChR2-YFP into the NAc of VGATcre mice (Figure 6a). Using confocal microscopy, we detected YFP-VGAT immunofluorescent varicosities in the VTA. With quadruple label immunofluorescence, we found that NAc inputs formed YFP-VGAT varicosities around the cell body of GABA B1 +/TH+ cells (Figure 6b). Using immunoelectron microscopy and serial sectioning, we confirmed that YFP-VGAT varicosities corresponded to axon terminals. These YFP-VGAT axon terminals formed symmetric synapses on GABA B1 + dendrites (Figure 6d). However, we also found that YFP-VGAT axon terminals frequently formed symmetric synapses on GABA B + cell bodies (Figure 6c,f). Furthermore, these terminals synapsed on the cell bodies of TH+ neurons containing GABA B1 Rs (Figure 6e,f), suggesting that NAc inputs to the VTA directly synapse onto cell bodies and dendrites of GABA B R-expressing dopamine neurons. Based on these findings, we hypothesized that NAc inputs activate dopamine neurons via synaptic, rather than volume release. 36

50 To further examine whether activation of GABA B Rs was mediated by volume release of GABA, we used patch clamp electrophysiology to study the temporal dynamics of GABA B R activation. We defined the onset of GABA B currents as the time to 10% maximum current. The onset of electrically evoked GABA B IPSCs was 52 ± 5.0 ms (Figure 7a). When GABA was iontophoretically released as close to the cell as possible, the onset of GABA B IPSCs was 154 ± 5.8 ms (Figure 7a). By iontophoretically applying GABA at different distances from the cell, we estimate that GABA diffused at a rate of approximately 1.09 mm/s. These data indicate that electrically-evoked GABA is acting on postsynaptic GABA B Rs close to GABA release sites. We next asked if GABA B oipscs evoked specifically from NAc inputs are acting by volume or synaptic release. The viscous properties of dextran can be utilized to study neurotransmitter diffusion (Ford et al., 2010; Min et al., 1998). Therefore, if GABA is being volumetrically released, the time to onset of GABA B oipscs should be prolonged in the presence of dextran. We compared the time course of GABA A oipscs on GABA neurons (Figure 7b) with GABA B oipscs on dopamine neurons (Figure 7f) before and after incubation with dextran (10%, 40 kda molecular weight). Neither the time to onset nor the time to peak current of the GABA A oipsc on GABA neurons was different between control and dextran-treated slices (Figure 7c-e). Similarly there were no differences in time to onset or time to peak current for GABA B oipscs on dopamine neurons (Figure 7g-i). These data suggest that optically evoked GABA release acts on GABA B Rs close to NAc terminals. 37

51 2.3.5 NAc and local VTA GABA afferents activate different receptor pools within VTA dopamine neurons Synaptic afferents acting on dopamine neuron GABA A Rs and GABA B Rs are thought to arise from separate inputs (Sugita et al., 1992). Because VTA GABAergic interneurons control dopamine neuron activity through activation of GABA A Rs (Tan et al., 2012; van Zessen et al., 2012), we asked whether they were also capable of activating GABA B Rs. We injected AAV DIO-ChR2-YFP into the VTA of VGAT-cre mice and performed whole cell patch-clamp recordings in VTA dopamine neurons (Figure 8a). Optogenetic activation of VTA GABA interneurons evoked GABA B oipscs in dopamine neurons, but they were significantly smaller than the GABA B oipscs evoked from NAc inputs (Figure 8b). We wondered if NAc inputs or VTA GABA neurons are necessary components of the synaptically-evoked GABA B current in dopamine neurons. AAV containing Credependent caspase (AAV caspase) was injected unilaterally into the VTA of VGAT- IRES-cre mice to ablate local VTA GABA neurons (Figure 8c). We found no differences in electrically-evoked GABA B IPSCs (eipscs) onto dopamine neurons in the AAV caspase-injected hemisphere versus sham controls (Figure 8d). However, GABA A eipscs were reduced in the AAV caspase hemisphere compared to sham controls (Figure 8e). These results show that VTA GABA interneurons contribute to synaptic activation of GABA A Rs on dopamine neurons, but are not a necessary component of synapticallyevoked GABA B activity. To test whether NAc inputs are necessary for synaptically-evoked GABA B activation in dopamine neurons, we injected AAV caspase unilaterally into the NAc of 38

52 Dyn-cre mice to ablate D1 MSNs (Figure 8f). GABA B eipscs were substantially reduced in AAV caspase hemispheres compared to sham controls (Figure 8g). However, the reduction in synaptic GABA B activity could be caused by either decreased presynaptic GABA release or decreased sensitivity of postsynaptic GABA B Rs. To determine whether the effect was presynaptic or postsynaptic, we bath applied Rubi-GABA (10 mm), a caged GABA compound chemically activated in the presence of 473 nm light. Optically uncaged postsynaptic GABA B currents were significantly larger in AAV caspase hemispheres than in sham hemispheres (Figure 8h), suggesting that presynaptic ablation of NAc inputs sensitized postsynaptic GABA B Rs after D1 MSN ablation. Thus, we are likely underestimating the effect of D1 caspase lesions on GABA B eipscs. Taken together, these results suggest that NAc inputs, but not local GABA interneurons, make up a necessary component of the synaptically evoked GABA B current in dopamine neurons NAc inputs potently inhibit dopamine cell firing through specific activation of GABA B Rs As NAc D1 MSN projections inhibit both VTA GABA and dopamine neurons, we asked what the net effect of stimulation would be on the firing of VTA dopamine neurons. We injected AAV containing cre-dependent ChR2-YFP into the NAc of Dyn-cre mice and performed cell-attached recordings to examine the tonic firing rate of VTA dopamine neurons (Figure 9a). Brief trains of optogenetic stimuli (20 Hz, 20 pulses) consistently paused the firing of dopamine cells (Figure 9b). Moreover, the firing of VTA dopamine neurons was inhibited over a broad range of frequencies (Figure 9c-i). Bath application of the GABA A antagonist picrotoxin (100 µm) did not affect the inhibition of dopamine 39

53 neurons induced by optogenetic stimulation. However, bath application of the GABA B antagonist CGP (100 µm) completely blocked the inhibitory effects of NAc terminal stimulation. These results show that NAc inputs to the VTA directly inhibit the firing of dopamine neurons through specific activation of GABA B Rs. 40

54 2.4 Discussion We examined inhibitory inputs to VTA dopamine neurons from the NAc and from VTA GABA interneurons. We show with optogenetic stimulation that NAc inputs synapse onto both GABA and dopamine neurons, but activate two different types of receptors. Whereas NAc projections inhibit VTA GABA neurons through GABA A Rs, these terminals inhibit dopamine neurons through activation of GABA B Rs. Using cell-type specific ablation, we also show that NAc projections provide one of the major inputs activating GABA B Rs in dopamine neurons, while VTA GABA interneurons preferentially inhibit dopamine neurons through GABA A Rs. Although anatomical studies suggest a monosynaptic projection from the NAc to VTA dopamine neurons (Beier et al., 2015; Menegas et al., 2015; Somogyi et al., 1981; Watabe-Uchida et al., 2012), recent optogenetic studies suggest that NAc projections preferentially synapse on VTA GABA neurons, not dopamine neurons (Bocklisch et al., 2013; Matsui et al., 2014; Xia et al., 2011). Here we resolve this apparent conflict by showing that trains of optogenetic stimuli directly inhibit dopamine neurons through monosynaptic activation of GABA B Rs. Interestingly, a previous study also used trains of optogenetic stimuli but did not find evidence for GABA B activation (Xia et al., 2011). Our use of Dyn-cre mice allowed us to directly target striosomal D1 MSNs (Banghart et al., 2015), which preferentially innervate dopaminergic regions of the midbrain (Gerfen, 1984). Additionally, while Xia et al. targeted viral injections to the NAc shell (Xia et al., 2011), our injections infected neurons in both NAc core and shell. Indeed, a recent study shows that NAc inputs to dopamine neurons are clustered in discrete patches within the NAc, one of which is in the NAc core (Menegas et al., 2015). We observed strong 41

55 GABA B R expression in dopamine neurons throughout the various regions of the midbrain. However, we found that NAc inputs preferentially target the lateral VTA and substantia nigra, suggesting that different populations of dopamine neurons might receive inputs from distinct subregions of the striatum or from other brain regions. Therefore, studies examining whether dopamine neurons receive inputs from specific regions of the striatum will be of particular interest. The role of inhibition and the diversity of interneuron subtypes in the VTA are poorly understood compared to those within cortical and hippocampal networks. In the cortex and hippocampus, GABAergic interneurons are composed of multiple different cell types, each exerting distinct forms of inhibition. Recent work in the cortex suggests that, while most interneurons appear to activate GABA A Rs, neurogliaform cells preferentially activate GABA B Rs through volume release of GABA (Craig and McBain, 2014; Oláh et al., 2009). Our results show that within the reward circuitry, long-range NAc inputs to the VTA play a similar role by preferentially activating GABA B Rs on dopamine neurons. However, in contrast to other brain regions where volume release of GABA controls extrasynaptic GABA B R activation (Isaacson et al., 1993; Oláh et al., 2009; Scanziani, 2000), we show with electron microscopy and slice physiology that NAc inputs make bona fide synaptic contacts with dopamine neurons to activate GABA B Rs. To our knowledge, these data provide the first evidence of synaptic activation of GABA B Rs. While NAc terminals project to GABA B R-expressing synapses, we found that VTA GABA interneurons preferentially activate GABA A Rs. These results suggest that NAc inputs and local GABA interneurons activate separate receptor pools within 42

56 dopamine neurons. One potential mechanism for the selectivity of inputs is that various GABAergic afferents target specific subcellular domains within dopamine neurons, similar to neocortical interneurons (Freund and Katona, 2007). A recent study reported that inhibitory synapses in dopamine neurons are generally more abundant in distal dendritic processes (Henny et al., 2012). Using electron microscopy and quadruple-label immunofluorescence, we found that NAc inputs innervate dopaminergic dendrites, but frequently synapse onto dopaminergic cell bodies. Importantly, dendritic inhibition and perisomatic inhibition (i.e. innervation of proximal dendrites and cell bodies) serve different functions. While dendritic inhibition regulates glutamatergic inputs in close proximity to inhibitory synapses, perisomatic inhibition controls action potential generation and the synchrony of neuronal populations (Freund and Katona, 2007). Therefore, further work examining subcellular innervation of various GABAergic inputs will allow us to better understand the integration of excitation and inhibition within dopamine neurons. Furthermore, the presence of perisomatic inhibition by NAc synapses may in part explain their powerful control over dopamine neuron firing, as demonstrated by our optogenetic experiments. Although much work has focused on the role of ionotropic receptors, our data demonstrate an important role for metabotropic GABA B Rs in controlling the activity of dopamine neurons. It may be surprising that a slow receptor can have rapid actions on firing rates. However, recent studies show that GABA B Rs rapidly control oscillatory Up states in pyramidal cells in the entorhinal cortex (Craig et al., 2013). Furthermore, in the somatosensory cortex, in vivo activation of GABA B Rs elicits rapid intrahemispheric inhibition of somatosensory-evoked responses. Although this inhibition is mediated by a 43

57 small hyperpolarization, it strongly inhibits the firing of pyramidal neurons (Palmer et al., 2012). Similarly, we found that even moderate activation of NAc inputs (as little as 2 Hz) quickly and powerfully inhibited the pacemaker firing of dopamine neurons. Our results add to a growing body of literature suggesting that GABA B Rs play a vital role in the rapid control of neuronal networks. While NAc inputs inhibit dopamine neurons via GABA B Rs, they also inhibit VTA GABA neurons through GABA A Rs. As such, we hypothesize that NAc inputs to the VTA might exert bidirectional control over VTA dopamine neurons under different temporal activity patterns. During weak activity, NAc projections would be expected to inhibit VTA GABA neurons via GABA A Rs, resulting in disinhibition of dopamine neurons. Indeed, Bocklisch et al., demonstrate that optogenetic stimulation of NAc inputs can increase the firing rate of dopamine neurons (Bocklisch et al., 2013). However, during synchronous firing or sustained activity, D1 MSNs might provide powerful shunting inhibition via GABA B Rs (Lüscher et al., 1997) that would be expected to override disinhibition, especially considering the perisomatic innervation of NAc inputs to dopamine neurons. Therefore, NAc inputs may act as a feedback system to prevent the over activation of dopamine neurons. In vivo, NAc MSNs ramp up their firing rates to more than 20 Hz during the expectation of reward (Cromwell and Schultz, 2003; Schultz et al., 1992). Within that temporal range, we predict that D1 MSN projections to the VTA would inhibit dopamine neurons via GABA B R activation. Therefore, we hypothesize that GABA B -mediated inhibition by NAc inputs could potentially shape the reward prediction errors observed in dopamine neurons. Indeed, the NAc has been hypothesized to provide information about the timing of reward expectation and the actual delivery of reward 44

58 (Joel et al., 2002; O Reilly et al., 2007). According to these models and our empirical findings, inhibitory feedback from the NAc to the VTA may critically influence responses to both natural rewards and drugs of abuse. In summary, we examined the functional anatomy of local and long range GABAergic circuits and show that the NAc potently inhibits dopamine neurons via GABA B Rs, while local VTA interneurons act primarily through GABA A Rs. Our results highlight a role for feedback inhibition from the NAc and suggest a general framework for understanding the role of GABAergic inhibition in the VTA. 45

59 46

60 Figure 1. Optogenetic activation of NAc D1 terminals elicits GABA B activation in VTA dopamine neurons. (a) Schematic of experimental protocol. Cre-dependent ChR2- YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (b) Example coronal section showing ChR2- YFP fluorescence in the NAc. (c) Horizontal brain section containing the VTA. Recorded cells were filled with biocytin and stained for immunofluorescence (red). TH immunostaining (white) was performed to verify dopamine neurons (scale bar, 60 µm). (d) Representative trace from a voltage clamp recording of a VTA dopamine neuron held at -55 mv, showing outward current produced by optical stimulation of NAc terminals. (e) Input-output curve varying the number of light pulses while keeping frequency constant at 20 Hz, normalized to 20 pulses (n=9 cells, 5 mice). (f) Input-output curve varying the stimulation frequency while keeping the number of light pulses constant at 20 pulses, normalized to 40 Hz (n=9 cells, 6 mice). (g) Representative traces of monosynaptic light-evoked IPSCs before (black) and after tetrodotoxin (TTX; 1 µm, blue) and in the presence of TTX + 4-aminopyridine (4-AP; 1mM, red). (h) Quantification of oipsc amplitudes (normalized to baseline) during TTX and TTX + 4- AP (7 and 6 cells, 4 mice; two-tailed t-test, P < 0.01, t=7.182, df=5). (i) Current-voltage relationship of oipscs recorded in voltage clamp. (j-l) Representative oipsc before (black) and after bath application of (j) the GIRK antagonist tertiapin (purple, 0.5 µm), (k) the kappa-opioid receptor antagonist NorBNI ( red, 0.5 µm), and (l) the GABA B receptor antagonist CGP (blue, 100 µm). (m) Quantification of the effects of bath application of tertiapin, NorBNI, and CGP (n=6 cells each; one-way analysis of variance (ANOVA), P<0.01, F=105.6). All data are shown as mean ± SEM. 47

61 48

62 Figure 2. D2 MSNs project to the ventral pallidum, but not directly to the VTA. (a) Schematic of experimental protocol. Cre-dependent ChR2-YFP was injected into the VTA of A2A-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine and GABA neurons and in ventral pallidal neurons. (b) Summary data of lightevoked GABA A IPSCs in voltage clamp recordings (V m =-70 mv) of VTA dopamine (blue), VTA GABA (red), and ventral pallidal neurons (purple; n=10, 8, and 11 cells; one-way ANOVA, F=8.606, P<0.01). (c) Representative traces of light-evoked GABA A IPSCs in VTA dopamine (blue), VTA GABA (red) and ventral pallidal neurons (purple). Data are shown as mean ± SEM. 49

63 50

64 Figure 3. GABA B Rs are strongly expressed in dopamine neurons throughout the midbrain. (a) Wide-field image showing TH immunoreactivity (brown) and GABA B1 in situ (black). (b) TH neurons co-express GABA B1 mrna (scale bar 25 µm). (c) Summary data showing the ratio of TH neurons co-expressing GABA B1 mrna in the SNc and VTA (substantia nigra pars compacta, SNc; parabrachial pigmented nucleus, PBP; parainterfascicular nucleus, PIF; paranigral nucleus, PN). (d) Summary data showing the ratio of GABA B1 + cells that do not express TH (n=3 mice, 4 sections per mouse). 51

65 52

66 Figure 4. Nucleus accumbens inputs preferentially inhibit VTA GABA and dopamine neurons through separate postsynaptic receptors. (a) Representative traces demonstrating the action potential width (AP width, left), firing rate (middle) and hyperpolarization-activated cation current (h-current, right) of a VTA dopamine neuron (blue) and a GABA neuron (red). (b) Heat map, sorted by AP width, showing the AP width, firing rate, h-current and optically evoked GABA A current (V m =-70 mv) for individual cells (n=45 cells, 6 mice). (c) Representative traces of GABA A currents evoked from a VTA dopamine cell (blue) and a VTA GABA cell (red). (d) Quantification of GABA A oipscs evoked from dopamine and GABA neurons neurons identified by their electrophysiological properties (n=21 and 20 cells, 6 mice, two-tailed t-test, P<0.01, t=6.619, df=39). (e) Heat map, sorted by AP width, showing the AP width, firing rate, h- current and optically evoked GABA B current (V m =-55 mv) for individual cells (n=64 cells, 15 mice). (f) Representative traces of GABA B currents evoked from a VTA dopamine cell (blue) and a VTA GABA cell (red). (g) Quantification of GABA B oipscs evoked from dopamine and GABA neurons identified by their electrophysiological properties (n=46 and 39 cells, 15 mice, two-tailed t-test, P<0.01, t=7.644, df=83). (h) Schematic summarizing the findings presented in this figure. All data are shown as mean ± SEM. 53

67 54

68 Figure 5. Amplitudes of GABA A oipscs correlate with GABA neuron properties and amplitudes of GABA B oipscs correlate with dopamine neuron properties. (a) Representative traces demonstrating the action potential width of a VTA dopamine neuron (blue) and a GABA neuron (red). (b) Representative traces demonstrating the firing rate of a VTA dopamine neuron (blue) and a GABA neuron (red). (c) Representative traces demonstrating the h-current of a VTA dopamine neuron (blue) and a GABA neuron (red). (d) Correlation of GABA A oipscs with AP width (n=45 cells, 6 mice). (e) Correlation of GABA A oipscs with firing rate (n=45 cells, 6 mice). (f) Correlation of GABA A oipscs with h-current (n=45 cells, 6 mice). (g) Correlation of GABA B oipscs with AP width (n=64 cells, 12 mice). (h) Correlation of GABA B oipscs with firing rate (n=64 cells, 12 mice). (i) Correlation of GABA B oipscs with h-current (n=64 cells, 12 mice). 55

69 56

70 Figure 6. Nucleus accumbens inputs form perisomatic and dendritic symmetric synapses onto dopamine neurons. (a) Diagram of fibers from nucleus accumbens neurons expressing YFP (under VGAT-promoter regulation) innervating the VTA. (b) Detection of TH-IR (blue), GABA B1 -IR (red), YFP-IR (green), and VGAT-IR (white). YFP-IR/VGAT-IR varicosity (arrow) is in contact with TH-IR/GABA B1 -IR neurons (Scale bar: 2 µm). (c-d) Electron micrographs showing that axon terminals (ATs) coexpressing YFP-IR (scattered dark material) and VGAT-IR (gold particles; green arrowheads) making symmetric synapses (green arrows) on a cell body (c) and a dendrite (d) expressing GABA B1 -IR (gold particles; red arrowheads). (e) Electron micrograph showing an axon terminal (AT) co-expressing YFP-IR (scattered dark material) and VGAT-IR (gold particles; green arrowhead) making a symmetric synapse (green arrow) on a cell body expressing TH-IR (gold particles; blue arrowheads). (f) Electron micrograph showing an axon terminal (AT) expressing YFP-IR (scattered dark material) making a symmetric synapse (green arrow) on a cell body co-expressing TH-IR (scattered dark material) and GABA B1 -IR (gold particles; red arrowheads; scale bars: 200 nm). 57

71 58

72 Figure 7. Evidence for synaptic release of GABA onto GABA B receptors. (a) Time to onset (time to 10% max current) of GABA B IPSCs while GABA (1M) was iontophoretically released (open circles) at different distances from the soma compared to the time to onset of electrically-evoked GABA B eipscs (closed circle, n=9 and 10 cells). (b) Experimental schematic for (c-e). GABA A oipscs were evoked in voltage clamp recordings (-70 mv) of VTA GABA neurons. (c) Representative traces of normalized NAc GABA A oipscs in normal artificial cerebrospinal fluid (acsf) versus dextranincubated slices. (d) Summary of time to onset for GABA A oipscs from normal ACSF versus dextran-incubated slices (n= 8 cells, two-tailed t-test, t=0.18, df=14, P=0.86). (e) Summary of time to maximum current for GABA A oipscs from normal ACSF versus dextran-incubated slices (n= 8 cells, two-tailed t-test, t=1.85, df=14, P=0.09). (f) Experimental schematic for (g-i). GABA B oipscs were evoked during voltage clamp recordings (-55 mv) of VTA dopamine neurons. (g) Representative traces of normalized NAc GABA B oipscs onto dopamine neurons in normal acsf versus dextran-incubated slices (h) Summary of time to onset for GABA B oipscs from normal ACSF versus dextran-incubated slices (n= 9 cells, two-tailed t-test, t=0.32, df=16, P=0.75). (i) Summary of time to maximum current for GABA B oipscs from normal ACSF versus dextran-incubated slices (n= 9 cells, two-tailed t-test, t=0.61, df=16, P=0.55). All data are shown as mean ± SEM. 59

73 60

74 Figure 8. NAc inputs and local GABA interneurons activate separate receptor populations in dopamine neurons. (a) Experimental schematic for (b). AAV5 DIO- ChR2-YFP was injected into the VTA of VGAT-cre mice and patch clamp recordings were made from VTA dopamine neurons. (b) Representative trace (left) and quantification (right) showing the effect of 20 Hz, 20 pulse light stimulation on evoked GABA B oipscs from VTA GABA interneurons (red) compared to NAc inputs (blue). (c) Schematic of the experiments for (d,e). Unilateral injections of AAV5 DIO-Caspase3- Tevp (AAV Caspase) were made into the VTA of VGAT-cre mice and electrically evoked GABA B currents were recorded in VTA dopamine neurons. (d) Average GABA B eipscs evoked at different stimulus intensities in sham control versus VTA GABA ablated hemispheres (n=14 cells, 5 mice, two-way ANOVA, Treatment x Intensity, F 3,3 =0.3338, P=0.80) (e) Average GABA A eipscs evoked at different stimulus intensities in sham control versus VTA GABA ablated hemispheres (n=12 and 11 cells, 5 mice, twoway ANOVA, Treatment x Intensity, F 4,4 =4.001, P<0.01). (f) Schematic of the experiments for (g,h). Unilateral injections of AAV5 DIO-Caspase3-Tevp (AAV Caspase) were made into the NAc of Dyn-cre mice and electrically evoked GABA B currents were measured in VTA dopamine neurons. (g) Average GABA B eipscs evoked at different stimulus intensities in sham control versus NAc Dyn-cre ablated hemispheres (n=16 and 18 cells, 5 mice, two-way ANOVA, Treatment x Intensity, F 3,3 =10.73, P<0.01). (h) Summary of postsynaptic GABA B currents in sham controls versus NAc Dyn-cre ablated hemispheres evoked by single-photon uncaging of Rubi-GABA (10 mm; n=9 and 8 cells, 4 mice, unpaired t-test, P<0.05, t=2.579, df=15). All data are shown as mean ± SEM. 61

75 62

76 Figure 9. NAc inputs inhibit dopamine neuron firing at multiple frequencies through activation of GABA B Rs. (a) Experimental schematic for (b-i). Cre-dependent ChR2-YFP was injected into NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (b) Raster plot showing the effect of brief optical stimulation (20 Hz, 20 pulses) of NAc terminals on dopamine cell firing. (c) Representative traces showing the effect of extended (10 s) optogenetic stimulation at multiple frequencies on the firing rate of a dopamine cell. (d-h) Effect of optical stimulation of NAc terminals on the normalized firing rate of dopamine cells at various frequencies. Optical stimulation inhibited dopamine cell firing during control conditions (black, n=10 cells and 4 mice) and during GABA A blockade with picrotoxin (100 µm, red, n=8 cells and 4 mice), but not during GABA B blockade with CGP (100 µm, blue, n=8 cells and 4 mice). (i) Summary data of (d-h), showing the average normalized effect of NAc terminal stimulation on dopamine cell firing during the first 5 sec of optical stimulation. 63

77 Chapter 3 Circuit Specific Modulation of GABAergic Synapses in the VTA 3.1 Introduction The behavioral effects of drugs of abuse are thought to be mediated by increased dopamine release from ventral tegmental area (VTA) dopamine neurons. Drug-induced increases in dopamine neuron activity can be the result of direct perturbations to dopamine neurons or modulation and plasticity of afferent inputs to the dopamine system (Ungless et al., 2001b). Disinhibition, the removal of inhibitory inputs, is a major mechanism by which drugs affect the activity of VTA dopamine neurons. Moreover, the effects of disinhibition are often mediated by neuromodulatory receptors located on GABAergic inputs to VTA dopamine neurons (Cruz et al., 2004; Johnson and North, 1992b; Tan et al., 2010). Neuromodulators influence neuronal function through postsynaptic and presynaptic mechanisms. Many modulators either increase or decrease the probability of neurotransmitter release by activating receptors located on axon terminals (Chesselet, 1984). Multiple modulatory inputs converge in the VTA. For instance, the dorsal raphe nucleus sends serotonin projections to the VTA (McDevitt et al., 2014) and VTA dopamine neurons can somatodendritically release dopamine (Ford et al., 2010). In fact, the first demonstration of presynaptic facilitation of neurotransmitter release in the central nervous system was studied in dopamine cells in the VTA (Cameron and Williams, 1993). Cameron and Williams showed that activation of dopamine D1 receptors presynaptically increases GABA release onto VTA dopamine neurons 64

78 (Cameron and Williams, 1993). Additionally, cocaine causes the release of dopamine, serotonin, and endocannabinoids, all of which presynaptically modulate GABA release onto dopamine neurons in the VTA (Cameron and Williams, 1994, 1995; O Dell and Parsons, 2004; Wang et al., 2015). However, due to the difficulty in isolating specific inputs to the VTA with electrical stimulation, the roles of various neuromodulators on different GABAergic inputs to the VTA are poorly understood. We have previously shown (chapter 2) that VTA dopamine and GABA neurons receive monosynaptic inputs from nucleus accumbens (NAc) projection neurons. Our earlier findings suggest that NAc inputs to dopamine neurons activate different synapses and receptor pools than GABAergic inputs from local interneurons. Using cell-type specific optogenetic activation, we examine the effect of various presynaptic modulators on NAc inputs to the VTA and on local inputs from VTA GABA neurons to dopamine neurons. We show that local and long-range GABAergic inputs to the VTA are differentially modulated. Additionally, our results suggest that cocaine disinhibits dopamine neurons by inhibiting specific GABAergic inputs within the VTA. 65

79 3.2 Methods Experimental Subjects Pdyn-ires-Cre (Krashes et al., 2014) and VGAT-ires-Cre (Vong et al., 2011) mice on a C57BL/6 background were used for electrophysiological experiments. Mice of either sex were used. Adult (8+ weeks) mice were acclimatized to the animal facility for more than 1 week before undergoing surgery. Animals were housed on a 12 hr:12 hr reverse light: dark cycle with lights off at 7:00 a.m. All experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and with the approval of the National Institute on Drug Abuse animal care and use committee. Surgeries Microinjection needles (29G) were connected to a 2 µl Hamilton syringe and filled with concentrated adeno-associated virus (~10 12 infectious units ml -1 ) encoding DIO-ChR2- eyfp (CaMKII promoter). Mice were anesthetized with 150 mg kg -1 ketamine and 50 mg kg -1 xylazine and placed in a stereotaxic frame. Microinjection needles were placed unilaterally or bilaterally into the NAc (+1.7 AP, ±1.6 ML, -4.2 DV, 12 angle) and 500 nl virus was injected for 5 min. For VTA injections (-3.3 AP, ±1.9 ML, -4.6 DV, 14 angle), 300 nl virus was injected for 3 min. The needles were left in place for an additional 5 min to allow for diffusion of virus from the injection site. Electrophysiology 66

80 Eight to ten weeks after surgery, mice were anesthetized with Euthasol and perfused with ice-cold artificial cerebrospinal fluid (ACSF). To preserve neuronal health, n-methyl-dglucamine (NMDG) was substituted for sodium in both the cutting and perfusion solutions (Ting et al., 2014). The cutting and perfusion ACSF contained (in mm) 92 NMDG, 20 HEPES, 25 glucose, 30 NaHCO 3, 1.2 Na-phosphate, 2.5 KCl, 5 sodium ascorbate, 3 sodium pyruvate, 2 thiourea, 10 Mg-sulfate, and 0.5 CaCl 2 (ph7.35, ~305 mosm). Brains were sectioned at µm using a VT-1200 vibratome (Leica). The ACSF used for holding slices before recording was identical to the cutting solution, except it contained 1 mm MgCl 2 and 2 mm CaCl 2 and 92 mm NaCl instead of NMDG. The ACSF used to perfuse the slice during recording contained (in mm): 125 NaCl, 2.5 KCl, 1.25 Na-phosphate, 1 MgCl 2, 2.4 CaCl 2, 26 Na-bicarbonate, and 11 glucose. All ACSF solutions were bubbled with 95% O 2 and 5% CO 2. After >1 hour slice incubation, eyfp expression was examined in slices containing the virus injection sites to ensure placement accuracy. If the injection region was adequately infected with virus, sections containing the VTA were transferred to the recording chamber and superfused with C ACSF. For optogenetic experiments, a 200 µm optical fiber coupled to a diode-pumped solid-state laser was positioned above the slice and aimed at the recorded cell. Whole-cell recordings were performed using MΩ patch pipettes backfilled with an internal solution containing (in mm): 115 K-MethaneSulfate, 20 NaCl, 1.5 MgCl 2,10 BAPTA, 10 Na 2 -phosphocreatine, 4 Mg-ATP, and 0.4 Na 2 -GTP (ph 7.35, 285 mosm). Whole-cell voltage clamp recordings were made using a MultiClamp 700B amplifier (10 khz digitization and 1-2 khz low-pass Bessel filter) with pclamp

81 software (Molecular Devices). Throughout the whole-cell recordings, series resistance was monitored (0.1 Hz) with a 5 mv hyperpolarizing step and maintained below 30 MΩ. Cells in which the series resistance changed more than 20% during the recording were excluded from analysis. Dopamine neurons were identified by morphology, spike width, tonic firing rates, the presence of a hyperpolarization-induced h-current, and in a subset of cells, cell filling with biocytin and immunohistochemistry. Optically-evoked GABA B responses were obtained every 30 s with 20 pulses of 473 nm wavelength light (2-5 mw, 20 Hz, 3 ms) with neurons voltage clamped at -55 mv. GABA A responses were obtained every 10 s with 2 pulses of 473 nm wavelength light (2-5 mw, 3 ms) with 50 ms interpulse intervals and neurons voltage clamped at -70 mv. Paired pulse ratios (PPRs) were measured as the amplitude of the second pulse divided by the amplitude of the first pulse. Therefore, increases in PPR suggest a reduction in the probability of release and decreases in PPR suggest an increase in the probability of neurotransmitter release. For uncaging experiments, ACSF containing picrotoxin (100µM) and Rubi-GABA (10 mm) was recirculated in the bath and responses were obtained every 60 sec with 1 sec pulses of 473 nm light. For pharmacological experiments, we recorded baseline responses for at least 10 min and drugs were bath applied for 5-10 min. All drugs used were obtained from either Tocris Bioscience or Sigma. Statistics All data were presented as the mean ± SEM. Data were analyzed in Clampfit, Ethovision, Excel, and Prism. 68

82 3.3 Results NAc inputs and local GABAergic inputs are differentially modulated by adenosine A 1 receptors Based on evidence from electrical stimulation experiments, the GABAergic inputs activating GABA A Rs and GABA B Rs are differentially affected by various neuromodulators (Cameron and Williams, 1993, 1994; Sugita et al., 1992). We asked whether inputs to dopamine neurons from the NAc and local GABA interneurons are differentially modulated by adenosine A 1 receptor (A 1 R) agonists. We injected Credependent ChR2-YFP into the NAc of Dyn-cre mice and performed whole-cell recordings in VTA dopamine neurons (Figure 1a). Bath application of the A 1 R agonist N 6 -CPA inhibited GABA B oipscs in dopamine neurons (Figure 1b). NAc GABA A oipscs in VTA GABA neurons (Figure 1c) were similarly modulated by N 6 -CPA (Figure 1d). We next asked whether the inhibition by A 1 Rs was due to postsynaptic receptor modulation or a decrease in presynaptic neurotransmitter release. Changes in the presynaptic probability of release are often dependent on calcium and can be measured by paired pulse ratios (Zucker and Regehr, 2002). The paired pulse ratio of GABA A oipscs increased after N 6 -CPA, suggesting a presynaptic mechanism for A 1 Rs (Figure 1d). To compare the local GABAergic synapses within the VTA, we injected Credependent ChR2-YFP into the VTA of VGAT-cre mice and recorded from dopamine cells (Figure 1e). GABA A oipscs evoked from local interneurons were unaffected by bath application of N 6 -CPA (Figure 1e). These data demonstrate that NAc inputs to the VTA are presynaptically modulated by A1Rs, but local VTA GABA to dopamine synapses are unaffected. 69

83 3.3.2 Dopamine D1R activation affects NAc inputs to VTA but does not affect inputs from VTA interneurons Dopamine facilitates electrically-evoked GABA B transmission in the VTA through the activation of dopamine D1 receptors (Bonci and Williams, 1996; Cameron and Williams, 1993). We asked whether D1R agonists differentially affect NAc inputs or local GABA to dopamine neuron synapses. We injected AAV containing cre-dependent ChR2-YFP into the NAc of Dyn-cre mice and performed patch clamp recordings in VTA dopamine neurons (Figure 2a). Bath application of the D1R agonist SKF (10 μm) inhibited GABA B R oipscs (Figure 2b). On the other hand, when we stimulated NAc inputs and recorded from VTA GABA neurons (Figure 2c), we found that bath application of SKF potentiated GABA A -mediated oipscs (Figure 2d,e). The paired pulse ratio decreased after drug application, suggesting a presynaptic facilitation of GABA release from NAc terminals (Figure 2e). To determine the effect of D1R stimulation on local GABAergic inputs, we injected cre-dependent ChR2-YFP into the VTA of VGAT-cre mice and performed patch clamp recordings in VTA dopamine neurons (Figure 2f). Bath application of SKF did not affect GABA A oipscs evoked from local interneurons (Figure 2g). Together, these findings suggest that D1R stimulation has opposite effects on NAc inputs to VTA GABA and dopamine neurons, but does not modulate presynaptic release by local GABAergic interneurons Serotonin 5-HT 1B Rs induce long-term depression on NAc inputs but shortterm depression on local GABAergic synapses Serotonin (5-HT) has been shown to inhibit electrically-evoked GABA B currents in VTA dopamine neurons through activation of 5-HT 1B receptors (Bonci and Williams, 1996; 70

84 Cameron and Williams, 1994, 1995). We asked whether 5-HT 1B Rs modulate GABAergic synapses from NAc inputs and from local VTA interneurons. We injected cre-dependent ChR2-YFP into the NAc of Dyn-cre mice and recorded from VTA dopamine neurons (Figure 3a). GABA B -mediated oipscs in dopamine neurons were potently inhibited by the 5-HT 1B agonist CP (2 μm, Figure 3b,c). We asked whether the effect of CP on GABA B oipscs was mediated by a presynaptic or postsynaptic mechanism. We therefore bath applied Rubi-GABA (10 mm), a caged form of GABA that is chemically activated by 473 nm light, in the presence of the GABA A blocker picrotoxin (100 μm). Postsynaptic uncaged GABA B currents were unaffected by bath application of CP 93129, suggesting a presynaptic mechanism of 5-HT 1B Rs on NAc inputs to dopamine neurons. We then compared NAc inputs to VTA GABA neurons by optically evoking ChR2- mediated GABA A oipscs (Figure 3d). CP strongly inhibited GABA A oipscs in VTA GABA neurons (Figure 3e,f). Bath application of CP increased the pairedpulse ratio (Figure 3f), suggesting that activation of 5-HT 1B Rs reduced presynaptic release of GABA from NAc terminals. We then examined the effect of CP on local GABA transmission to VTA dopamine neurons. We injected AAV containing credependent ChR2-YFP into the VTA of VGAT-cre mice and recorded from dopamine neurons (Figure 3g). Bath application of CP produced a short-term inhibition of GABA A oipscs evoked from VTA GABA terminals (Figure 3h,i). We found an increase in the paired-pulse ratio (Figure 3i), suggesting that activation of 5-HT 1B Rs presynaptically decreased the probability of release. These data suggest that activation of 5-HT 1B Rs produces long-term depression on NAc inputs to the VTA and a weaker shortterm depression on the local connections from VTA GABA to dopamine neurons. 71

85 3.3.4 Cocaine causes long-term depression on NAc inputs to VTA and short-lasting inhibition on VTA GABA to dopamine synapses Cocaine inhibits electrically-evoked GABA B currents in dopamine neurons through the endogenous release of serotonin and endocannabinoids (Cameron and Williams, 1994, 1995; Wang et al., 2015). We asked whether cocaine inhibits NAc and local GABA inputs within the VTA. We injected AAV containing cre-dependent ChR2-YFP into the NAc of Dyn-cre mice and performed whole-cell recordings in VTA dopamine neurons (Figure 4a). Bath application of cocaine (10 μm) inhibited GABA B oipscs evoked from NAc terminals (Figure 4b). We also found a long-lasting inhibition of GABA A oipscs in VTA GABA neurons (Figure 4 c,d). We examined the effect of cocaine on GABA A oipscs on VTA GABA to dopamine neuron synapses. Cre-dependent ChR2-YFP was injected into the VTA of VGAT-cre mice and we performed patch clamp recordings in dopamine neurons (Figure 4e). Cocaine induced a short-lasting inhibition on the local GABA inputs within the VTA (Figure 4f). These findings indicate that cocaine inhibits multiple GABAergic inputs to the VTA, but exerts a stronger effect on NAc inputs than on local GABAergic connections. 72

86 3.4 Discussion We demonstrate that NAc inputs to the VTA and local GABAergic connections onto dopamine neurons are differentially modulated by adenosine A 1 R, dopamine D 1 R, and serotonin 5-HT 1B agonists (Figure 5). Furthermore, we show that cocaine inhibits multiple sources of GABAergic inputs to the VTA. Our results show that activation of adenosine A 1 Rs inhibits GABA release from NAc inputs to the VTA, but has no effect on VTA GABA to dopamine neuron synapses. These findings reveal circuit-specific modulation of GABAergic inputs in the VTA. Interestingly, whereas activation of adenosine A 1 Rs selectively inhibits excitatory inputs in other brain regions (Mitchell et al., 1993; Prince and Stevens, 1992; Thompson et al., 1992), it does not affect excitatory synaptic transmission in the VTA (Wu et al., 1995). Additionally, there are no postsynaptic effects of A 1 R agonists in dopamine neurons (Wu et al., 1995), suggesting that the actions of adenosine A 1 R agonists in the VTA are completely mediated by their effects on GABAergic transmission. Caffeine is an antagonist of adenosine A 1 Rs and is known to cause rewarding and locomotor effects (Brockwell et al., 1991; Griffiths and Woodson, 1988; Snyder et al., 1981). Additionally, dopamine neurons in the VTA and substantia nigra are crucial for locomotor activity and rewarding behaviors. However, the rewarding and locomotor stimulating effects of caffeine are incompletely understood. Our findings that adenosine A 1 agonists inhibit transmission from NAc inputs to VTA might implicate these terminals in the behavioral actions of caffeine. 73

87 We found that D1R agonists did not affect local GABA to dopamine neuron synapses, but had opposite effects on the NAc inputs to VTA GABA and dopamine neurons. While D1R agonists facilitated NAc transmission onto GABA neurons, NAc inputs to dopamine neurons were inhibited. These findings are in stark contrast with other results which show that dopamine facilitates electrically-evoked GABA B transmission onto dopamine neurons (Cameron and Williams, 1993). One explanation for our findings is that the NAc inputs are not the same inputs as those recruited during electrical stimulation of GABA B currents. However, this is unlikely as other modulators and cocaine have the same effects on NAc GABA B oipscs as have been shown for electrically-evoked GABA B IPSCs (Cameron and Williams, 1994; Wu et al., 1995). Additionally, we have previously shown (chapter 2) that NAc inputs represent a majority of the GABAergic inputs activated by electrically-evoked GABA B currents in dopamine neurons. Another possible explanation is that D1R activation initially facilitates GABA release from NAc terminals, but evokes a feedback mechanism that reduces postsynaptic GABA B R activity. Indeed, we found that almost all of the neurons we recorded from showed an initial facilitation of GABA B oipscs during D1R activation. One possibility is that increased GABA release would activate presynaptic GABA B Rs on dopamine neurons and inhibit further GABA release from NAc terminals. Additionally, increases in presynaptic release would also cause an increase in other co-transmitters found in NAc terminals. Substance P, a neuropeptide transmitter also located in NAc terminals in the VTA (Bolam and Smith, 1990; Lu et al., 1997), postsynaptically inhibits GABA B R currents but does not affect GABA A R activity (Xia et al., 2010). Therefore, D1R stimulation might increase the release of both GABA and substance P, selectively 74

88 inducing feedback of NAc inputs onto dopamine neurons, but not GABA neurons. This hypothesis remains to be tested. We show that 5-HT 1B receptors potently produce long-lasting inhibition on NAc inputs to VTA dopamine and GABA neurons and induce short-term inhibition on local GABA to dopamine synapses. Previous studies suggest that 5-HT 1B Rs presynaptically inhibit GABAergic inputs onto dopamine neuron GABA B Rs, but do not affect the inputs onto GABA A Rs (Johnson et al., 1992). Although we found that local VTA GABA inputs to dopamine neurons were inhibited by 5-HT 1B Rs, the inhibition was much less pronounced and of shorter duration than the effect on NAc inputs. We also found differential effects of cocaine on the various GABAergic responses within the VTA. Cocaine decreased GABAergic responses from NAc and from VTA GABA to dopamine synapses, but had a stronger and longer-lasting effect on NAc responses, similar to the effect of 5-HT 1B R agonists. The cocaine-induced inhibition of electrically-evoked GABA B responses in dopamine neurons is mediated by endogenous release of 5-HT and endocannabinoids acting on presynaptic GABA terminals (Cameron and Williams, 1994, 1995; Wang et al., 2015). We suggest that the differential sensitivity of NAc inputs and local GABA inputs to cocaine is due to their differential response to 5-HT. The behavioral effects of cocaine are thought to be mediated through increased dopamine release in the NAc (Di Chiara and Imperato, 1988). One mechanism for the increase in dopamine release is by direct inhibition of the dopamine transporter at dopaminergic terminals. However cocaine has a stronger affinity for other monoamine transporters, like 5-HT, than it does for the dopamine transporter (Ritz et al., 1987). Therefore, cocaine might act to increase dopamine neuron activity by reducing the 75

89 inhibition from local interneurons and from the NAc. The role of GABAergic inputs to the VTA in behavioral response to cocaine is addressed in the next chapter. Here we elucidate multiple sources of neuromodulation on GABAergic terminals that could prove to be effective pharmacological targets for the treatment of addiction. Taken together, our findings suggest that NAc inputs to dopamine neurons are tightly regulated by multiple sources of neuromodulation, whereas local GABAergic synapses are likely regulated by separate modulators. 76

90 77

91 Figure 1. Adenosine A 1 Rs have similar effects on NAc synapses to VTA dopamine and GABA neurons, but no effect on local GABAergic inputs. (a) Experimental schematic for (b). Cre-dependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (b) Group (left) and representative (right) effects of the adenosine A 1 receptor agonist N 6 - CPA (1 µm) on GABA B currents from NAc inputs to dopamine neurons (n=8 cells and 5 mice). (c) Experimental schematic for (d). Cre-dependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA GABA neurons. (d) Group (left) and representative (right) effects of N 6 -CPA on GABA A currents from NAc inputs to VTA GABA neurons (n=7 cells and 5 mice). Paired pulse ratios before and after N 6 -CPA application shown in inset (n=7 cells and 5 mice, paired t- test, p<0.05, t=2.774, df=6). (e) Experimental schematic for (f). Cre-dependent ChR2- YFP was injected into VTA of VGAT-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (f) Group (left) and representative (right) effects of N 6 -CPA on GABA A currents from local VTA GABA inputs to dopamine neurons (n=12 cells and 6 mice). 78

92 79

93 Figure 2. Dopamine D1Rs exert differential activity on NAc inputs to VTA and have no effect on local GABAergic inputs to dopamine neurons. (a) Experimental schematic for (b). Cre-dependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (b) Group (left) and representative (right) effects of the dopamine D1 receptor agonist SKF (10 µm) on GABA B currents from NAc inputs to dopamine neurons (n=8 cells and 4 mice). (c) Experimental schematic for (d). Cre-dependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA GABA neurons. (d) Group (left) effect of SKF on GABA A currents from NAc inputs to the VTA GABA neurons (n=7 cells and 5 mice). (e) Representative trace (above) showing the effect of SKF on GABA A oipscs. Paired pulse ratios before and after N 6 -CPA application (below, n=7 cells and 5 mice, paired t-test, p<0.01, t=4.221, df=6). (f) Experimental schematic for (g). Cre-dependent ChR2-YFP was injected into VTA of VGAT-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (g) Group (left) and representative (right) effects of SKF on GABA A currents from local VTA GABA inputs to dopamine neurons (n=8 cells and 4 mice). 80

94 81

95 Figure 3. Local GABAergic inputs to dopamine neurons and NAc inputs are presynaptically inhibited by 5-HT 1B Rs. (a) Experimental schematic for (b-c). Credependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (b) Group effect of the selective 5-HT 1B agonist CP (2 µm) on GABA B currents from NAc inputs to dopamine neurons (n=9 cells and 4 mice). (c) Representative effect of CP on the GABA B current from NAc inputs to a dopamine neuron (above). Example of the effect of CP on uncaged GABA B currents evoked with Rubi-GABA (10 mm, below). (d) Experimental schematic for (e-f). Cre-dependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA GABA neurons. (e) Group effect of CP on GABA A currents from NAc inputs to VTA GABA neurons (n=9 cells and 5 mice). (f) Representative effect of CP on the GABA A current from NAc inputs to a VTA GABA neuron (above). Effect of CP on paired pulse ratios on the GABA A current from NAc inputs to VTA GABA neurons (below, n=9 cells and 5 mice; paired t-test, p<0.05,t=2.865, df=8) (g) Experimental schematic for (h-i). Cre-dependent ChR2-YFP was injected into the VTA of VGAT-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (h) Group effect of CP on GABA A currents from VTA GABA inputs to VTA dopamine neurons (n=6 cells and 3 mice). (i) Representative effect of CP on the GABA A current from VTA GABA inputs to a VTA dopamine neuron (above). Effect of CP on paired pulse ratios on the GABA A current from VTA GABA inputs to VTA dopamine neurons (below, n=6 cells and 3 mice; paired t-test, p<0.05, t=3.344, df =5). 82

96 83

97 Figure 4. Cocaine inhibits both NAc inputs to the VTA as well as local GABAergic synapses. (a) Experimental schematic for (b). Cre-dependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (b) Group (left) and representative (right) effects of cocaine (10 µm) on GABA B currents from NAc inputs to dopamine neurons (n=6 cells and 3 mice). (c) Experimental schematic for (d). Cre-dependent ChR2-YFP was injected into the NAc of Dyn-cre mice and whole-cell patch clamp recordings were performed in VTA GABA neurons (n=7 cells and 3 mice). (d) Group (left) and representative (right) effects of cocaine on GABA A currents from NAc inputs to VTA GABA neurons. (e) Experimental schematic for (f). Cre-dependent ChR2-YFP was injected into the VTA of VGAT-cre mice and whole-cell patch clamp recordings were performed in VTA dopamine neurons. (f) Group (left) and representative (right) effects of cocaine on GABA A currents from local VTA GABA inputs to dopamine neurons (n=9 cells and 3 mice). 84

98 85

99 Chapter 4 The Role of GABAergic Inputs During and After Cocaine Exposure 4.1 Introduction Ventral tegmental area (VTA) dopamine neurons control motivated behaviors for both natural rewards and drugs of abuse (Di Chiara and Imperato, 1988; Stuber et al., 2008). Afferent inputs to dopamine neurons dynamically regulate their firing activity and are important for the control of dopamine-dependent behaviors (Paladini and Roeper, 2014). The vast majority of inputs to midbrain dopamine neurons are GABAergic (Ribak et al., 1976; Bolam and Smith, 1990; Watabe-Uchida et al., 2012; Henny et al., 2012; Lerner et al., 2015; Menegas et al., 2015; Beier et al., 2015), highlighting the importance of inhibition in the VTA. GABAergic inhibition in the VTA is mediated by fast GABA A receptors (GABA A Rs) and metabotropic GABA B receptors (GABA B Rs). While recent studies have examined GABA A -mediated inhibition of dopamine neurons by local VTA interneurons (Eshel et al., 2015; Tan et al., 2012; van Zessen et al., 2012), little is known about the behavioral role of GABA B transmission in dopamine neurons. GABA B Rs powerfully control dopamine neuron activity both in vitro and in vivo (Erhardt et al., 2002; Klitenick et al., 1992; Lacey et al., 1988; Paladini and Tepper, 1999), suggesting that these receptors may play a vital role in regulating dopaminedependent behaviors. GABA B R agonists reduce craving for cocaine in humans (Haney et al., 2006) and attenuate the rewarding effects of cocaine in animal models (Slattery et al., 2005). However, the location of GABA B Rs responsible for these effects is unclear. As dopamine neurons are crucially involved in cocaine-related behaviors, we hypothesized 86

100 that GABA B Rs located specifically in dopamine neurons might exert endogenous activity that counteracts the behavioral effects of cocaine. Furthermore, in acute brain slices, cocaine inhibits presynaptic GABA inputs acting on GABA B Rs of dopamine neurons, but has no effect on the inputs that activate GABA A Rs (Cameron and Williams, 1994). These results suggest that 1) cocaine may cause acute behavioral effects through disinhibition of dopamine neurons and 2) GABAergic inputs to dopamine neurons synapsing on GABA A Rs and GABA B Rs arise from separate sources (Cameron and Williams, 1994; Sugita et al., 1992; Cameron and Williams, 1993). We have previously shown that, in dopamine neurons, NAc inputs selectively activate GABA B Rs, while VTA GABA neurons preferentially inhibit dopamine neurons via GABA A Rs. Moreover, we have shown that specific GABAergic inputs to the VTA are inhibited by cocaine. Here we examine the effect of repeated cocaine exposure on electrically-evoked GABA B inputs and test the role of GABA B receptors in behavioral response to cocaine. Using cell-type specific genetic deletion of GABA B Rs from dopamine neurons, we show that endogenous activation of GABA B Rs dampens cocaine-induced locomotion. 87

101 4.2 Methods Experimental Subjects Wild-type C57/BL/6 mice purchased from Jackson Laboratories were used for sensitization and electrophysiology experiments presented in figure 1. Ai9 ROSAtdTomato (Madisen et al., 2010) mice on a C57BL/6 background were used for histological experiments. GABA B1 -loxp mice (Haller et al., 2004) were maintained on a Balb/c background and used for electrophysiological and behavioral experiments. Mice of either sex were used. Adult (8+ weeks) mice were acclimatized to the animal facility for more than 1 week before undergoing surgery. Animals were housed on a 12 hr:12 hr reverse light:dark cycle with lights off at 7:00 a.m. All experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and with the approval of the National Institute on Drug Abuse animal care and use committee. Surgeries Microinjection needles (29G) were connected to a 2 µl Hamilton syringe and filled with concentrated adeno-associated virus (~10 12 infectious units ml -1 ) encoding icre (TH promoter), or eyfp (synapsin promoter). Mice were anesthetized with 150 mg kg -1 ketamine and 50 mg kg -1 xylazine and placed in a stereotaxic frame. Microinjection needles were placed bilaterally into the VTA (-3.3 AP, ±1.9 ML, -4.6 DV, 14 angle) and 300 nl of virus was injected for 3 min. The needles were left in place for an additional 5 min to allow for diffusion of virus from the injection site. 88

102 Electrophysiology Three to eight weeks after surgery, mice were anesthetized with Euthasol and perfused with ice-cold artificial cerebrospinal fluid (ACSF). To preserve neuronal health, n- methyl-d-glucamine (NMDG) was substituted for sodium in both the cutting and perfusion solutions (Ting et al., 2014). The cutting and perfusion ACSF contained (in mm) 92 NMDG, 20 HEPES, 25 glucose, 30 NaHCO 3, 1.2 Na-phosphate, 2.5 KCl, 5 sodium ascorbate, 3 sodium pyruvate, 2 thiourea, 10 Mg-sulfate, and 0.5 CaCl 2 (ph7.35, ~305 mosm). Brains were sectioned at µm using a VT-1200 vibratome (Leica). The ACSF used for holding slices before recording was identical to the cutting solution, except it contained 1 mm MgCl 2 and 2 mm CaCl 2 and 92 mm NaCl instead of NMDG. The ACSF used to perfuse the slice during recording contained (in mm): 125 NaCl, 2.5 KCl, 1.25 Na-phosphate, 1 MgCl 2, 2.4 CaCl 2, 26 Na-bicarbonate, and 11 glucose. All ACSF solutions were bubbled with 95% O 2 and 5% CO 2. After >1 hour slice incubation, sections containing the VTA were transferred to the recording chamber and superfused with C ACSF. For optogenetic experiments, a 200 µm optical fiber coupled to a diode-pumped solid-state laser was positioned above the slice and aimed at the recorded cell. Whole cell recordings were performed using MΩ patch pipettes backfilled with an internal solution containing (in mm): 115 K-MethaneSulfate, 20 NaCl, 1.5 MgCl 2,10 BAPTA, 10 Na 2 -phosphocreatine, 4 Mg-ATP, and 0.4 Na 2 -GTP (ph 7.35, 285 mosm). Whole-cell voltage clamp recordings were made using a MultiClamp 700B amplifier (10 khz digitization and 1-2 khz low-pass Bessel filter) with pclamp

103 software (Molecular Devices). Cell-attached recordings were performed in voltage clamp at -70 mv and were used to analyze the spike width and tonic firing rate. Throughout the whole-cell recordings, series resistance was monitored (0.1 Hz) with a 5 mv hyperpolarizing step and maintained below 30 MΩ. Cells in which the series resistance changed more than 20% during the recording were excluded from analysis. Dopamine neurons were identified by morphology, spike width, tonic firing rates, the presence of a hyperpolarization-induced h-current, and in a subset of cells, cell filling with biocytin and immunohistochemistry. Electrically evoked GABA B responses were obtained every 1 min with 10 pulses (70 Hz, 500 µs) using a bipolar stimulating electrode and a constant current stimulation unit with neurons voltage-clamped at -55 mv. To obtain pure GABA B responses with electrical stimulation, the following were added to the ACSF (in µm): 1000 kynurenic acid, 100 picrotoxin, 1 sulpiride, 5 LY , to block AMPA and NMDA, GABA A, dopamine-d2, and mglur1 responses, respectively. For pharmacological experiments, we recorded baseline responses for at least 10 min and drugs were bath applied for 5-10 min. All drugs used were obtained from either Tocris Bioscience or Sigma. Behavior For cocaine-induced locomotion experiments, mice were allowed to recover from surgery for at least 4 weeks prior to behavioral testing. Prior to injections, mice were handled and habituated for 3 days. Animals were given intraperitoneal injections of saline (0.9% NaCl) for 2 days and cocaine (15 mg kg -1 ) or saline for 5 consecutive days. On injection days, mice were placed in a 20 cm square box with grated flooring for 15 min prior to 90

104 injection. Following injection, mice were immediately returned to the locomotor box and were monitored for locomotor activity for 15 min. For open-field tests, mice were placed in a 40 cm diameter circular arena for 30 min. Immunohistochemistry and microscopy After behavioral testing, all mice were deeply anesthetized and perfused with 4% paraformaldehyde. Coronal sections of the VTA (60 μm) were obtained using a VT-1200 vibratome (Leica). Free-floating coronal VTA sections from GABA B1 -loxp mice were washed before being incubated in blocking solution. Sections were then incubated with primary antibody cocktail (mouse anti-gaba B1 1: chicken anti-th 1:500) overnight at 4 C. Sections were rinsed before incubation at room temperature for 2 h in secondary antibody cocktail (anti-mouse 594, 1:500 + anti-chicken 647, 1:500). Sections were then rinsed, mounted on slides, and covered with antifade fluorescent mounting medium. Fluorescence images were taken with an Olympus Fluoview FV1000 confocal microscope. Statistics All data were presented as the mean ± SEM. Data were analyzed in Clampfit, Ethovision, Excel, and Prism. 91

105 4.3 Results Repeated cocaine exposure reduces presynaptic GABA release onto dopamine neurons Acute cocaine exposure in brain slices presynaptically reduces GABA B -mediated inhibitory postsynaptic currents (IPSCs) in dopamine neurons (Cameron and Williams, 1994). We hypothesized that repeated in vivo cocaine exposure exerts long-lasting effects on GABA B transmission in the VTA. Mice were injected for 5 consecutive days with either saline or cocaine (15 mg kg -1, intraperitoneal) and placed in locomotor activity boxes for 15 min (Figure 1a,b). Mice injected with cocaine showed increased locomotor activity compared to saline controls (Figure 1b). One day after the last cocaine injection, we prepared horizontal midbrain slices and performed patch clamp recordings in VTA dopamine neurons. Bath application of cocaine inhibited GABA B eipscs to a similar extent in saline- and cocaine-treated animals (Figure 1c), demonstrating that prior in vivo cocaine experience does not alter the sensitivity to cocaine on this pathway. We sought to determine whether repeated cocaine exposure has long-term effects on GABA B transmission in dopamine neurons. When GABA B eipscs were evoked at various stimulus intensities, there was a significant reduction in the GABA B current from cocaine-treated animals compared to saline controls (Figure 1d). To test whether the reduction in synaptically evoked GABA B eipscs was due to a decrease in presynaptic GABA release or a decrease in the postsynaptic sensitivity of GABA B Rs, we recorded from dopamine neurons and bath applied baclofen, a GABA B agonist (10 µm). Baclofen produced a similar effect on saline- and cocaine-treated slices (Figure 1e), indicating that 92

106 there was no change in postsynaptic GABA B receptor expression or sensitivity. These results suggest that, in addition to its acute actions on GABA transmission, repeated exposure to cocaine produces a long-lasting inhibition of presynaptic GABA release on the inputs to GABA B Rs Selective deletion of GABA B Rs from dopamine neurons increases locomotor response to cocaine Cocaine has been proposed to exert its acute behavioral effects through the disinhibition of dopamine neurons (Cameron and Williams, 1994; O Dell and Parsons, 2004). Therefore, we sought to determine the role of GABA B Rs on the behavioral actions of cocaine. We used an adeno-associated virus (AAV) to express cre-recombinase under the control of the tyrosine hydroxylase (TH) promoter (AAV TH-iCre) to specifically target dopamine neurons (Gompf et al., 2015). We verified the selectivity of this virus by injecting it into the VTA of Rosa-TdTomato reporter mice (Figure 2a). TdTomato in the VTA co-expressed with TH immunofluorescence (Figure 2b), indicating that the AAV TH-iCre virus causes cre-mediated recombination specifically in dopamine neurons. GABA B Rs are heteromultimeric assemblies of the GABA B1 and GABA B2 subunits, KCTD proteins and G-protein effectors (Kaupmann et al., 1998; Schwenk et al., 2010, 2016). However, the GABA B1 subunit is responsible for binding GABA and is an essential component of functional GABA B Rs (Kaupmann et al., 1998; Schuler et al., 2001). To conditionally delete GABA B Rs from dopamine neurons, we injected AAV THiCre or control AAV expressing YFP (AAV Control) into the VTA of GABA B1 -loxp mice (Haller et al., 2004) (Figure 3a). To examine postsynaptic GABA B responses, we used patch clamp electrophysiology to record from dopamine neurons. Bath application 93

107 of baclofen (10 µm), a GABA B agonist, produced GABA B responses that were significantly reduced in dopamine neurons from AAV TH-iCre slices compared to AAV Controls (Figure 3b). However, there was no difference in the tonic firing rate (Figure 3c) or input resistance (Figure 3d) of dopamine neurons between groups. To further verify the efficacy of the GABA B1 deletion, we examined co-expression of GABA B1 Rs and TH in AAV Control and AAV TH-iCre animals. In AAV Control animals, GABA B1 immunofluorescence was observed in 91 ± 1.6% of TH+ immunofluorescent cells, whereas co-labeling was observed in only 37 ± 2.5% of TH+ cells in AAV TH-iCre mice (Figure 3e,f). Importantly, there was no difference in the total number of TH+ cells between groups (Figure 3g). Taken together, these results show that we were effectively able to delete GABA B1 Rs from VTA dopamine neurons without affecting the intrinsic properties or the overall number of dopamine cells. We tested the behavioral role of GABA B Rs in dopamine neurons by examining locomotor activity in response to repeated cocaine injections in AAV TH-iCre- and AAV Control-injected GABA B1 loxp mice (Figure 4a). AAV TH-iCre and AAV Control mice showed similar levels of basal locomotor activity when placed in an open field test (Figure 4b) and similar time spent in the center of the open field (Figure 4c). Additionally, there was no difference between AAV Control and AAV TH-iCre mice on locomotor activity during habituation sessions in which mice were injected with saline (Figure 4d). However, compared to AAV Controls, AAV TH-iCre mice significantly increased their locomotor activity on cocaine (15 mg kg -1 ) injection days (Figure 4d). These data indicate that reduction of GABA B Rs from dopamine neurons increases 94

108 locomotor activity in response to cocaine, but has no effect on general locomotion or anxiety-related behavior. 95

109 4.4 Discussion We found that repeated cocaine exposure reduces presynaptic GABA release onto VTA dopamine neurons. Moreover, we show that selective genetic deletion of GABA B Rs from dopamine neurons in adult mice increases locomotor responses to cocaine. The effect of acute and chronic exposure to cocaine on excitatory glutamatergic transmission has been well characterized (Chen et al., 2008; Ungless et al., 2001b). In particular, cocaine induces long-term potentiation (LTP) of excitatory transmission by postsynaptically altering the expression of glutamatergic receptors within dopamine neurons (Schilström et al., 2006; Yuan et al., 2013). However, recent evidence implicates inhibitory GABAergic transmission mediated by GABA A Rs in the acute and chronic effects of cocaine (Bocklisch et al., 2013; Graziane et al., 2013). Acute application of cocaine in slice recordings has been shown to selectively inhibit GABAergic inputs acting on dopamine neuron GABA B receptors, but does not affect the inputs onto GABA A Rs (Bonci and Williams, 1996; Cameron and Williams, 1994). Here we show that repeated in vivo cocaine administration induces long-term changes in GABA B transmission onto dopamine neurons. Specifically, we show that cocaine decreases presynaptic release of GABA without affecting postsynaptic GABA B R expression or sensitivity. These findings suggest that, in addition to increasing excitatory transmission, cocaine causes both acute and long-lasting disinhibition of dopamine neurons by inhibiting the GABAergic inputs activating GABA B Rs. The mechanism underlying presynaptic inhibition of GABA release after repeated cocaine exposure has not yet been demonstrated. Cocaine has been shown to acutely decrease presynaptic GABA release onto dopamine neurons through the endogenous 96

110 release of serotonin and endocannabinoids (Cameron and Williams, 1994; Wang et al., 2015). Endocannabinoids are released by postsynaptic neurons and are known to cause long-lasting inhibition of presynaptic neurotransmitter release (Kano et al., 2009). If serotonin or endocannabinoids were exerting tonic or long-lasting activity in our experiments, we would expect at least a partial occlusion on the effects of bath application of cocaine. However, we found that repeated exposure to cocaine did not impact the effect of acute bath application GABA B R responses, suggesting that another mechanism might be responsible for the long-term depression of GABA release. Another possibility is that repeated cocaine increased the extracellular concentration of adenosine. Adenosine A 1 Rs presynaptically inhibit GABA release in the VTA, as we and others (Wu et al., 1995) have shown. Additionally, the metabolism of adenosine in the VTA is dramatically altered after withdrawal from cocaine and morphine (Bonci and Williams, 1996). Therefore, future experiments will address the role of adenosine, serotonin, and endocannabinoids in the presynaptic inhibition of GABA release after cocaine exposure. Intrinsic feedback mechanisms onto dopamine neurons blunt the behavioral responses to drugs of abuse. For example, during cocaine exposure, the somatodendritic release of dopamine acts as a feedback mechanism that prevents continued dopamine release (Bello et al., 2011; Holroyd et al., 2015). As evidenced by our optogenetic and lesion studies in chapter 2, our results suggest that NAc inputs to dopamine neurons could act as a feedback mechanism to regulate cocaine-induced behavior. Indeed, lesions of the NAc or severing the connections from the forebrain to the VTA reduce the inhibitory effects of cocaine on the firing rate of dopamine neurons (Einhorn et al., 1988). We have shown in chapter 3 that NAc D1 MSNs provide a major GABAergic input to 97

111 dopamine neurons and that these inputs are potently inhibited by cocaine. Future work will therefore address the specific role of NAc inputs to the VTA in cocaine responses. We asked whether activation of GABA B Rs regulates cocaine-induced behavior. Although selective deletion of GABA B Rs from midbrain dopamine neurons had no effect on general locomotion, it significantly increased cocaine-induced locomotor activity. Our findings suggest that GABAergic inputs to midbrain dopamine neurons, acting on GABA B Rs, are active during cocaine exposure and serve to dampen behavioral responses to cocaine. Interestingly, we found no effect of cocaine on the locomotor activity of mice injected with control virus. GABA B1 loxp mice are bred on a BALB/c background because of strain differences in the viability of GABA B1 constitutive knockout animals (Haller et al., 2004; Schuler et al., 2001). In various addiction models, BALB/c mice display little to no behavioral sensitivity to cocaine (Eisener-Dorman et al., 2011; Ruth et al., 1988; Thomsen and Caine, 2011; Zombeck et al., 2010). However, we found that deletion of GABA B Rs from dopamine neurons in the same line of mice unmasked a strong locomotor response to cocaine. Therefore, one intriguing possibility is that different strains of mice express varying levels of GABA B Rs. If this hypothesis is correct, it could have important clinical implications for patients who may also have different levels of GABA B R expression. Our results suggest that, although cocaine appears to induce feedback mechanisms by activating GABA B Rs on dopamine neurons, it can also hijack that feedback by inhibiting the presynaptic release of GABA. Thus, our findings highlight the importance of endogenous GABAergic inhibition of VTA dopamine neurons in dampening the effects of cocaine. Interestingly, positive allosteric modulators of 98

112 GABA B Rs, which augment their endogenous activation, reduce the rewarding effects of cocaine in preclinical animal models (Slattery et al., 2005). Our results suggest that the efficacy of GABA B positive allosteric modulators might be mediated by their ability to augment GABAergic feedback onto dopamine neurons. Additionally, recent work shows that specific GABA B R subtypes have differential roles in cocaine-mediated behavior (Jacobson et al., 2016), suggesting that the development of targeted agonists for specific subtypes of GABA B Rs on dopamine neurons might provide novel therapeutics for the treatment of cocaine addiction. 99

113 100

114 Figure 1. Repeated exposure to cocaine decreases presynaptic GABA release onto GABA B synapses. (a) Experimental schematic for (b-e). Wild-type C57/Bl6 mice were habituated with saline injections for 2 days, then injected with cocaine (10 mg/kg) for 5 consecutive days. On the 8 th day, we obtained brain sections from mice for patch-clamp recordings. (b) Locomotor activity of mice after injections of saline and cocaine (n=7 mice, two-way ANOVA, Group x Treatment, F 6,6 =7.05, P<0.01). (c) The effect of bath application of cocaine (10 µm) on electrically-evoked GABA B IPSCs from saline and cocaine treated mice (n=10 cells, 6 mice each). (d) Average GABA B eipscs evoked at different stimulus intensities in saline-treated versus cocaine-treated mice (n=13 and 14 cells, 7 mice each group, two-way ANOVA, Treatment x Intensity, p<0.05, F 4,4 =2.858). (e) Summary of voltage clamp recordings (V m =-55 mv) from dopamine neurons. The GABA-B agonist baclofen (10 µm) followed by the GABA-B antagonist CGP (100 µm) were bath applied to examine postsynaptic GABA-B currents in saline-treated versus cocaine-treated mice (n=11 and 12 cells, 7 mice each, two-way ANOVA, Treatment x Time, F 101,101 =0.2867, P=1.00). 101

115 102

116 Figure 2. AAV TH-iCre virus allows for selective expression of cre-recombinase in VTA dopamine neurons. (a) Experimental schematic. AAV2,10 TH-iCre (AAV THiCre) was injected bilaterally into the VTA of Rosa-TdTomato reporter mice. (b) Representative wide-field (above) and 20x (below) fluorescent images of TdTomato expression (red), immunofluorescent TH-IR expression (green), and overlap (scale bars, 100 µm). 103

117 104