Miniature neurotransmission is the trans-synaptic process where single synaptic vesicles spontaneously released from presynaptic neurons induce miniature postsynaptic potentials. Since their discovery over 60 years ago, miniature events have been found at every chemical synapse studied. However, the in vivo necessity for these small-amplitude events has remained enigmatic. In this thesis, I show that miniature neurotransmission is required for the normal structural maturation of Drosophila glutamatergic synapses in a developmental role that is not shared by evoked neurotransmission. Conversely, I find that increasing miniature events is sufficient to induce synaptic terminal growth. I show that miniature neurotransmission acts locally at terminals to regulate synapse maturation via a Trio guanine nucleotide exchange factor (GEF) and Rac1 GTPase molecular signaling pathway. My thesis study establishes that miniature neurotransmission, a universal but often-overlooked feature of synapses, has unique and essential functions in vivo.

For many organisms the sense of smell is critical to survival. Some olfactory stimuli elicit innate responses that are mediated through hardwired circuits that have developed over long periods of evolutionary time. Most olfactory stimuli, however, have no inherent meaning. Instead, meaning must be imposed by learning during the lifetime of an organism. Despite the dominance of olfactory stimuli on animal behavior, the mechanisms by which odorants elicit learned behavioral responses remain poorly understood. All odor-evoked behaviors are initiated by the binding of an odorant to olfactory receptors located on sensory neurons in the nasal epithelium. Olfactory sensory neurons transmit this information to the olfactory bulb via spatially organized axonal projections such that individual odorants evoke a stereotyped map of bulbar activity. A subset of bulbar neurons, the mitral and tufted cells, relay olfactory information to higher brain structures that have been implicated in the generation of innate and learned behavioral responses, including the cortical amygdala and piriform cortex. Anatomical studies have demonstrated that the spatial stereotypy of the olfactory bulb is maintained in projections to the posterolateral cortical amygdala, a structure that is involved in the generation of innate odor-evoked responses. The projections of mitral and tufted cells to piriform cortex however appear to discard the spatial order of the olfactory bulb: each glomerulus sends spatially diffuse, apparently random projections across the entire cortex. This anatomy appears to constrain odor-evoked responses in piriform cortex: electrophysiological and imaging studies demonstrate that individual odorants activate sparse ensembles that are distributed across the extent of cortex, and individual piriform neurons exhibit discontinuous receptive fields such that they respond to structurally and perceptually similar and dissimilar odorants. It is therefore unlikely that olfactory representations in piriform have inherent meaning. Instead, these representations have been proposed to mediate olfactory learning. In accord with this, lesions of posterior piriform cortex prevent the expression of a previously acquired olfactory fear memory and photoactivation of a random ensemble of piriform neurons can become entrained to both appetitive and aversive outcomes. Piriform cortex therefore plays a central role in olfactory fear learning. However, how meaning is imparted on olfactory representations in piriform remains largely unknown. We developed a strategy to manipulate the neural activity of representations of conditioned and unconditioned stimuli in the basolateral amygdala (BLA), a downstream target of piriform cortex that has been implicated in the generation of learned responses. This strategy allowed us to demonstrate that distinct neural ensembles represent an appetitive and an aversive unconditioned stimulus (US) in the BLA. Moreover, the activity of these representations can elicit innate responses as well as direct Pavlovian and instrumental learning. Finally activity of an aversive US representation in the basolateral amygdala is required for learned olfactory and auditory fear responses. These data suggest that both olfactory and auditory stimuli converge on US representations in the BLA to generate learned behavioral responses. Having identified a US representation in the BLA that receives convergent olfactory information to generate learned fear responses, we were then able to step back into the olfactory system and demonstrate that the BLA receives olfactory input via the monosynaptic projection from piriform cortex. These data suggest that aversive meaning is imparted on an olfactory representation in piriform cortex via reinforcement of its projections onto a US representation in the BLA. The work described in this thesis has identified mechanisms by which sensory stimuli generate appropriate behavioral responses. Manipulations of representations of unconditioned stimuli have identified a central role for US representations in the BLA in connecting sensory stimuli to both innate and learned behavioral responses. In addition, these experiments have suggested local mechanisms by which fear learning might be implemented in the BLA. Finally, we have identified a fundamental transformation through which a disordered olfactory representation in piriform cortex acquires meaning. Strikingly this transformation appears to occur within 3 synapses of the periphery. These data, and the techniques we employ, therefore have the potential to significantly impact upon our understanding of the neural origins of motivated behavior.

The emergence of limbs in vertebrates represents a significant evolutionary innovation. Limbs facilitate diverse motor behaviors, yet require spinal networks that can coordinate the activities of many individual muscles within the limb. Here I describe several efforts to characterize the specification of spinal motor neurons and assembly of spinal circuits in higher vertebrates. I discuss the formation of selective presynaptic sensory inputs to motor pools, a process which has long been thought to occur in an activity-independent manner. I demonstrate an as yet unappreciated role of activity-dependent refinement in patterning the set of sensory-motor connections that link motor pools with synergist function. I also explore the genetic specification of motor pools that project to defined muscle targets. I show that the motor pools that control digits engage distinct developmental genetic programs which reflect underlying differences in Hox and retinoic acid signaling. The divergent mechanisms underlying the specification of digit-innervating motor neurons may reflect the unique status of digit control in the evolution of motor behavior.

This work concerns the use of engineered genetic tools to build maps of the mammalian nervous system. Within the practice of circuit neuroscience, one of the most effective tools to emerge in recent years are the neurotropic viruses. Among these are modified strains of rabies virus which are made safe for laboratory use. We introduce here a novel form of engineered rabies virus with substantially improved utility for exploring the structure and function of neural circuits. Additionally, using this new tool, an investigation of an important motor circuit, the cortico-striatal circuit, is presented.

To understand the emergent properties of neural circuits it would be ideal to record the activity of every neuron in a behaving animal and decode how it relates to behavior. We have achieved this with the cnidarian Hydra vulgaris, using calcium imaging of genetically engineered animals to measure the activity of essentially all of its neurons. While the nervous system of Hydra is traditionally described as a simple nerve net, we surprisingly find instead a series of functional networks that are anatomically non-overlapping and are associated with specific behaviors. Three major functional networks extend through the entire animal and are activated selectively during longitudinal contractions, elongations in response to light and radial contractions, while an additional network is located near the hypostome and is active during nodding. Additionally, we show that the behavior of Hydra is made of regularly occurring radial contractions, which expel the content of the gastric cavity about every 45 minutes. These results demonstrate the functional sophistication of apparently simple nerve nets, and the potential of Hydra and other basal metazoans as a model system for neural circuit studies.

The motivation for this thesis is to devise a simple model of transient dynamics in neural networks. Neural circuits are capable of performing many computations without reaching an equilibrium, but instead through transient changes in activity. Thus, having a good model for transient activity is important. In particular, this thesis focuses on a firing-rate description of neural activity. Firing rates offer a convenient simplification of neural activity, and have been shown experimentally to convey information about stimuli and behavior. This work begins by review the philosophy of modeling firing rates, as well as the problems that go with it. It examines traditional approaches to modeling firing rates, and in particular how common assumptions lead to a model that fails to capture transient dynamics. Chapter 2 applies a traditional model of firing rates in order to gain insight into properties of cortical circuitry. In collaboration with the lab of David Ferster at Northwestern University, we found that surround suppression in cat primary visual cortex is mediated by a withdrawal of excitation in the cortical circuit. In theoretical work, we find that this behavior can only arise if excitatory recurrence alone is strong enough to destabilize visual responses but feedback inhibition maintains stability. Chapter 3 reviews concepts and literature related to the dynamics of large networks of spiking neurons. Population density approaches are common for describing the dynamics of networks of spiking neurons. These approaches allow for a rigorous approach to relate the dynamics of individual neurons to the population firing rate. Chapter 4 explores a method for accurately approximating the firing-rate dynamics of a population of spiking neurons. We describe the population by the probability density of membrane potentials, so the dynamics are governed by a Fokker-Planck equation. Using a spiking model with periodic boundary conditions, we write the Fokker-Planck dynamics in a Fourier basis. We find that the lowest Fourier modes dominate the dynamics. Chapter 5 presents a novel rate model that successfully captures synchronous dynamics. As in the previous chapter, we invoke an approximation to the dynamics of a population of spiking neurons in order to develop a firing-rate model. Our approach derives from an eigenfunction expansion of a Fokker-Planck equation, which is a common approach to solving such problems. We find that a very simple approximation turns out to be surprisingly accurate. This approximation allows us to write a closed-form expression for the firing rate that resembles the equations for a damped harmonic oscillator. Finally, chapter 6 uses the formalism derived in the previous chapter to analyze activity in a large randomly-connected network of neurons. Comparing this large spiking network to a network of two coupled rate units, we find that the firing rate network gives a good approximation to the time-varying activity of a spiking network across a wide range of parameters. Perhaps most surprisingly, we also find that the rate network can approximate the phase diagram of the spiking network, predicting the bifurcation line between synchronous and asynchronous states.

Though the scientific study of stress is relatively new, not even one hundred years old, there has been robust inquiry and discovery in stress research since its instantiation. Yet, many unanswered questions remain on how specific stressors impact the mind, brain and body. Social threat is a pervasive form of stress for species that are organized in social hierarchies, like humans and some animals. Social evaluative threat (SET), occurring when there is potential for negative evaluation or rejection from others, is a pervasive and important form of stress in humans having many links to stress-related physiological outcomes which in turn have important implications for health outcomes. The brain is a critical component in the mind-brain-body-health connection, but less is known about SET at the neural level. Here in this thesis, there are three studies that characterize the neural circuitry that responds to SET. Using a novel imaging technique, arterial spin labeling, Study 1 asks whether SET-related brain circuitry is modulated by a SET-related trait level vulnerability, Fear of Negative Evaluation (FNE). Overall, Study 1 replicated previous work by showing SET-related reactivity in the left pregenual anterior cingulate cortex, right ventromedial prefrontal cortex (vMPFC) and medial periaqueductal gray and extended previous work by showing that changes in the left vMPFC and the right thalamus were predicted by FNE. Using blood oxygenation level-dependent imaging (BOLD), Study 2 asks whether SET influences the brain circuitry on which the formation of relational episodic memory relies. With the use of mediation analysis, it was found that SET impaired relational episodic memory, and that the impairment was a function of activity in the right parahippocampal cortex and bilateral vMPFC. Using BOLD imaging, Study 3 asks whether SET influences the brain circuitry that subserves working memory (WM). With the use of mediation analysis, it was found that SET impaired WM, and that the impairment was a function of activity in bilateral intraparietal sulcus. Links between mind, brain, body and health are discussed throughout this work.

Modern neurotechnologies enable the recording of neural activity at the scale of entire brains and with single-cell resolution. However, the lack of principled approaches to extract structure from these massive data streams prevent us from fully exploiting the potential of these technologies. This thesis, divided in three parts, introduces new statistical machine learning methods to enable the large-scale analysis of some of these complex neural datasets. In the first part, I present a method that leverages Gaussian quadrature to accelerate inference of neural encoding models from a certain type of observed neural point processes --- spike trains --- resulting in substantial improvements over existing methods. The second part focuses on the simultaneous electrical stimulation and recording of neurons using large electrode arrays. There, identification of neural activity is hindered by stimulation artifacts that are much larger than spikes, and overlap temporally with spikes. To surmount this challenge, I develop an algorithm to infer and cancel this artifact, enabling inference of the neural signal of interest. This algorithm is based on a a bayesian generative model for recordings, where a structured gaussian process is used to represent prior knowledge of the artifact. The algorithm achieves near perfect accuracy and enables the analysis of data hundreds of time faster than previous approaches. The third part is motivated by the problem of inference of neural dynamics in the worm C.elegans: when taking a data-driven approach to this question, e.g., when using whole-brain calcium imaging data, one is faced with the need to match neural recordings to canonical neural identities, in practice resolved by tedious human labor. Alternatively, on a bayesian setup this problem may be cast as posterior inference of a latent permutation. I introduce methods that enable gradient-based approximate posterior inference of permutations, overcoming the difficulties imposed by the combinatorial and discrete nature of this object. Results suggest the feasibility of automating neural identification, and demonstrate variational inference in permutations is a sensible alternative to MCMC.

Organisms possess internal sensory systems to detect changes in physiological state. Despite the importance of these sensory systems for maintaining homeostasis, their development, sensory mechanisms, and circuitry are relatively poorly understood. To help address these gaps in knowledge, I used the tracheal dendrite (td) sensory neurons of Drosophila larvae as a model to gain insights into the cellular and molecular organization, developmental regulators, sensory functions and mechanisms, and downstream neural circuitry of internal sensory systems. In this thesis, I present data to show that td neurons comprise defined classes with distinct gene expression and axon projections to the CNS. The axons of one class project to the subesophageal zone (SEZ) in the brain, whereas the other terminates in the ventral nerve cord (VNC). This work identifies expression and a developmental role of the transcription factor Pdm3 in regulating the axon projections of SEZ-targeting td neurons. I find that ectopic expression of Pdm3 alone is sufficient to switch VNC-targeting td neurons to SEZ targets, and to induce the formation of putative synapses in these ectopic target regions. These results define distinct classes of td neurons and identity a molecular factor that contributes to diversification of central axon targeting. I present data to show that td neurons express chemosensory receptor genes and have chemosensory functions. Specifically, I show that td neurons express gustatory and ionotropic receptors and that overlapping subsets of td neurons are activated by decrease in O2 or increase in CO2 levels. I show that respiratory gas-sensitive td neurons are also activated when animals are submerged for a prolonged duration, demonstrating a natural-like condition in which td neurons are activated. I assessed the roles of chemosensory receptor genes in mediating the response of td neurons to O2 and CO2. As a result, I identify Gr28b as a mediator of td responses to CO2. Deletion of Gr28 genes or RNAi knockdown of Gr28b transcripts reduce the response of td neurons to CO2. Thus, these data identify two stimuli that are detected by td neurons, and establish a putative role for Gr28b in internal chemosensation in Drosophila larvae. Finally, I present data to elucidate the neural circuitry downstream of td sensory neurons. I show that td neurons synapse directly and via relays onto neurohormone populations in the central nervous system, providing neuroanatomical basis for internal sensory neuron regulation of hormonal physiology in Drosophila. These results pave the way for future work to functionally dissect the td circuitry to understand its function in physiology and behavior.

Animals have evolved sensory systems that afford innate and adaptive responses to stimuli in the environment. Innate behaviors are likely to be mediated by hardwired circuits that respond to invariant predictive cues over long periods of evolutionary time. However, most stimuli do not have innate value. Over the lifetime of an animal, learning provides a mechanism for animals to update the predictive value of cues through experience. Sensory systems must therefore generate neuronal representations that are able to acquire value through learning. A fundamental challenge in neuroscience is to understand how and where value is imposed in brain during learning. The olfactory system is an attractive sensory modality to study learning because the anatomical organization is concise in that there are relatively few synapses separating the sense organ from brain areas implicated in learning. Thus, the circuits for learned olfactory behaviors appear to be relatively shallow and therefore more experimentally accessible than other sensory systems. The goal of this thesis is to characterize the representation and function of neural circuits involved in olfactory associative learning. Odor perception is initiated by the binding of odors onto olfactory receptors expressed in the sensory epithelium. Each olfactory receptor neuron (ORN) expresses one of 1500 different receptor genes, the expression of which pushes the ORN to project with spatial specificity onto a defined loci within the olfactory bulb, the olfactory glomeruli. Therefore, each and every odor evokes a stereotyped map of glomerular activity in the bulb. The projection neurons of the olfactory bulb, mitral and tufted (M/T) cells, send axons to higher brain areas, including a significant input to the primary olfactory cortex, the piriform cortex. Axons from M/T cells project diffusely to the piriform without apparent spatial preference; as a consequence, the spatial order of the bulb is discarded in the piriform. In agreement with anatomical data, electrophysiological and optical imaging studies also demonstrate that individual odorants activate sparse subsets of neurons across the piriform without any spatial order. Moreover, individual piriform neurons exhibit discontinuous receptive fields that defy chemical or perceptual categorization. These observations suggests that piriform neurons receive random subsets of glomerular input. Therefore, odor representations in piriform are unlikely to be hardwired to drive specific behaviors. Rather, this model suggests that value must be imposed upon the piriform through learning. Indeed, the piriform has been shown to be both sufficient and necessary for aversive olfactory learning without affecting innate odor responses. However, how value is imposed on odor representations in the piriform and downstream associational areas remain largely unknown. We first developed a strategy to track neural activity in a population of neurons across multiple days in deep brain areas using 2-photon endoscopic imaging. This allowed us to assay changes in neural responses to odors during learning in piriform and in downstream associative areas. Using this technique, we first observe that piriform odor responses are unaffected by learning, so learning must therefore impose discernable changes in neural activity downstream of piriform. Piriform projects to multiple downstream areas that are implicated in appetitive associative learning, such as the orbitofrontal cortex (OFC). Imaging of neural activity in the OFC reveal that OFC neurons acquire strong responses to conditioned odors (CS+) during learning. Moreover, multiple and distinct CS+ odors activatethe same population of OFC neurons, and these responses are gated by context and internal state. Together, our imaging data shows that an external and sensory representation in the piriform is transformed into an internal and cognitive representation of value in the OFC. Moreover, we found that optogenetic silencing of the OFC impaired the ability of mice to acquire learned associations. Therefore, the robust representation of expected value of the odor cues is necessary for the formation of appetitive associations. We made an important observation: once the task has been learned with a set of odors, the OFC representation decays after learning has plateaued and remains silent even when mice encounter novel odors they haven’t previously experienced. Moreover, silencing the OFC when it was not actively engaged during the subsequent learning of new odors had no effect on learning. These sets of imaging and silencing experiments reveal that the OFC is only important during initial learning; once task structure has been acquired, it is no longer needed. Task performance after initial task acquisition must therefore be accommodated by other brain regions that can store the learned association for long durations. We therefore searched for other brain regions that held learned associations long-term. In the medial prefrontal cortex (mPFC), we observe that the learned representation persists throughout the entire course of training. Unlike the OFC, not only does this representation encode the positive expected value of CS+ odors, it also encodes the negative expected value of CS- odors in a non-overlapping ensemble of neurons. We further show through optogenetic silencing that this representation is necessary for task performance after the task structure has already been acquired. Therefore, while the OFC representation is required for initial task acquisition, the mPFC representation is required for subsequent appetitive learning and performance. Why would a learned representation vanish in the OFC and betransfered elsewhere? We hypothesize that the brain may allocate a portion of its real estate to be a cognitive playground where experimentation and hypothesis testing takes place. Once this area solves a task, it may unload what it has learned to storage units located elsewhere to free up space to learn new tasks. We further imaged another associative area, the basolateral amygdala (BLA), and found a representation of positive value that appears to be generated from a Hebbian learning mechanism. However, the silencing of this representation during learning had no effect. This suggests that while multiple and distributed brain areas encode cues that predict the reward, not all may be necessary for the learning process or for task performance. In summary, we have described a series of experiments that map the representation and function of different associational areas that underlie learning. The data and the techniques employed have the potential to significantly advance the understanding of learned behavior.