Understanding how the human brain functions on a molecular and cellular level is nearly impossible with current technology and ethical considerations. Utilizing the small nematode, Caenorhabditis elegans, and its innate behavioral responses to olfactory social cues, we can begin to unravel the mechanisms underlying social behavior. This is made possible given that innate behaviors are crucial for survival, and therefore hardwired into the genome of organisms. This allows for genetic-level analysis of neural circuitries driving behavior. Studying the neuronal mechanisms underlying C. elegans’ behavioral responses to social cues will not only assist in our overall understanding of how the brain perceives stimuli to enact a behavioral response at the cellular and molecular level, but also our understanding as to how the nervous system properly integrates information to enact social behavioral responses: mis-integration and social abnormalities are commonalities seen in many neuropsychiatric disorders, and these studies will provide fruitful insights into the defects observed in these disorders. Lastly, by comparing the perception of several different types of social chemicals, we can further our understanding of neural coding strategies for the various behaviors crucial for survival. Chapter One of this thesis orients the reader to social, innate behavior, and the usefulness of C. elegans as a tool for understanding behavioral coding. Chapter Two explores and establishes the required components of a socially aversive pheromone, providing insight into signaling evolution and co-option of biological machineries. Chapter Three examines how multiple, competing stimuli are integrated to modulate behavioral output, furthering our understanding of molecular and cellular integration and decision making within the nervous system. Chapter Four highlights the importance of predator pressure, and provides insights into circuit strategies of redundant and promiscuous networks of threat detection. Lastly, Chapter Five considers the implications of these findings as a whole, in the perspective of evolutionary strategies leading to neuronal coding of different behavioral outputs. Taken together, this dissertation aimed to fill the void in our understanding of social behavior neural circuitries, and how integration governed at the molecular and cellular level of the nervous system affects those behaviors.

Monoamine Oxidase Inhibitors have sparked great controversy in the treatment of idiopathic Parkinson’s Disease. There is little doubt that Monoamine Oxidase Inhibitors work synergistically with Levodopa to reduce several major debilitating symptoms. Multiple other medications provide a similar symptomatic benefit when combined with Levodopa; thus, a symptomatic benefit alone does little to advance current Parkinson’s treatment. The great controversy in treatment then comes from the possibility that Monoamine Oxidase Inhibitors modify the natural course of Parkinson’s Disease. This class of drug protected nigrostriatal dopaminergic neurons in many cellular and animal studies. Clinical studies involving Monoamine Oxidase Inhibitors are more controversial. Several studies have shown results that suggest a neuroprotective effect while other have not. This may be because the tools used to assess PD progression are inadequate. To see a clear decrease in nigrostriatal dopaminergic death, and thus prove a neuroprotective effect, more advanced techniques to measure the progression of Parkinson’s Disease must be developed. Given the controversy it will be important to revisit the benefits of MAO-B inhibitors once more advanced progression techniques are available.

The monkey lateral intraparietal area (LIP) is involved in visual attention and eye movements. It has traditionally been studied using extracellular recording, where often a single neuron is recorded at a time. Thus we have a wealth of correlational knowledge of what LIP neurons do, but not how or why, i.e. we do not know the circuit mechanisms and functions of the observed LIP activity. In this thesis, we have aimed to uncover the circuit mechanisms underlying LIP activity by building tightly constrained computational models. In Part 1, we found that during two versions of a delayed-saccade task, beneath similar population average firing patterns across time lie radically different network dynamics. When neurons are not influenced by stimuli outside their receptive fields (RFs), dynamics of the high-dimensional LIP network lie predominantly in one multi-neuronal dimension, as predicted by an earlier model. However, when activity is suppressed by stimuli outside the RF, LIP dynamics markedly deviate from a single dimension. The conflicting results can be reconciled if two LIP local networks, each dominated by a single multi-neuronal activity pattern, are suppressively coupled to each other. These results demonstrate the low dimensionality of LIP local dynamics and suggest active involvement of LIP recurrent circuitry in surround suppression and, more generally, in processing attentional and movement priority and in related cognitive functions. In Part 2, we examine the mechanisms of learning in LIP. When monkeys learn to group visual stimuli into arbitrary categories, LIP neurons become category-selective. Surprisingly, the representations of learned categories are overwhelmingly biased: while different categories are behaviorally equivalent, nearly all LIP neurons in a given animal prefer the same category. We propose that Hebbian plasticity, at the synapses to LIP from prefrontal cortex and from lower sensory areas, could lead to the development of biased representations. In our model, LIP category selectivity arises due to competition between inputs encoding different categories, and bias develops due to excitatory lateral interactions among LIP neurons. This model reproduces the different levels of category selectivity and bias observed in multiple experiments. Our results suggest that the connectivity of LIP allows it to learn the behavioral importance of stimuli in order to guide attention.

The cerebral cortex continuously encodes new sensory information and organizes it within an experiential intracortical framework. The cortical integration of internal and external information forms the associations that are the basis for higher order sensory representation, and ultimately, perception. Deciphering the cellular and synaptic principles of sensory-cortical integration requires a system with a simplified interface between the internal and external worlds. The piriform cortex provides a relatively simple substrate for the study of intracortical modulation of sensory coding. Within piriform, primary sensory information from the olfactory bulb converges onto neurons in a single cortical layer, where it directly integrates with intracortical input. The major barrier to studying intracortical influences on sensory representation in piriform has been the inability to isolate single types of intracortical input. Here, we use optogenetic techniques to functionally isolate two important classes of intracortical input to piriform pyramidal neurons, and slice electrophysiology to assess their synaptic properties. We first expressed channelrhodopsin in a small subset of piriform neurons, effectively isolating the recurrent synapses formed onto piriform pyramidal neurons by their peers. Recurrent collaterals form strong excitatory connections that extend throughout piriform without spatial attenuation in strength, linking distant piriform neurons. This extensive recurrent network is constrained by powerful disynaptic inhibition, which can also reduce activation by primary sensory inputs in a timing-dependent manner. Next, we functionally isolated inputs to the piriform from the anterior olfactory nucleus (AON), an early target of olfactory bulb output whose role in olfaction is largely unknown. The AON makes weaker excitatory connections with piriform, but unlike recurrent connections, these inputs do not drive strong disynaptic inhibition. Sequential activation of AON inputs leads to pronounced summation that boosts piriform activation in an NMDA-receptor-dependent manner, and may enhance plasticity of AON-to-piriform synapses. The AON is a potentially powerful modulator of piriform cortex, whose role in odor information processing merits further study. Our results collectively illustrate critical features of intracortical input classes to piriform cortex, and how these inputs may have distinct roles in shaping odor representations and olfactory learning.

Artificial recurrent neural networks (RNNs) are powerful models for understanding and modeling dynamic computation in neural circuits. As such, RNNs that have been constructed to perform tasks analogous to typical behaviors studied in systems neuroscience are useful tools for understanding the biophysical mechanisms that mediate those behaviors. There has been significant progress in recent years developing gradient-based learning methods to construct RNNs. However, the majority of this progress has been restricted to network models that transmit information through continuous state variables since these methods require the input-output function of individual neuronal units to be differentiable. Overwhelmingly, biological neurons transmit information by discrete action potentials. Spiking model neurons are not differentiable and thus gradient-based methods for training neural networks cannot be applied to them. This work focuses on the development of supervised learning methods for RNNs that do not require the computation of derivatives. Because the methods we develop do not rely on the differentiability of the neural units, we can use them to construct realistic RNNs of spiking model neurons that perform a variety of benchmark tasks, and also to build networks trained directly from experimental data. Surprisingly, spiking networks trained with these non-gradient methods do not require significantly more neural units to perform tasks than their continuous-variable model counterparts. The crux of the method draws a direct correspondence between the dynamical variables of more abstract continuous-variable RNNs and spiking network models. The relationship between these two commonly used model classes has historically been unclear and, by resolving many of these issues, we offer a perspective on the appropriate use and interpretation of continuous-variable models as they relate to understanding network computation in biological neural circuits. Although the main advantage of these methods is their ability to construct realistic spiking network models, they can equally well be applied to continuous-variable network models. An example is the construction of continuous-variable RNNs that perform tasks for which they provide performance and computational cost competitive with those of traditional methods that compute derivatives and outperform previous non-gradient-based network training approaches. Collectively, this thesis presents efficient methods for constructing realistic neural network models that can be used to understand computation in biological neural networks and provides a unified perspective on how the dynamic quantities in these models relate to each other and to quantities that can be observed and extracted from experimental recordings of neurons.

The response of a neuron encoding information about a sensory stimulus is influenced by the context in which that information is presented. In the primary visual cortex (area V1), neurons respond selectively to stimuli presented to a relatively constrained region of visual space known as the classical receptive field (CRF). These responses are influenced by stimuli in a much larger region of visual space known as the extra-classical receptive field (eCRF). In that they cannot directly evoke a response from the neuron, surround stimuli in the eCRF provide the context for the input to the CRF. Though the past few decades of research have revealed many details of the complex and nuanced interactions between the CRF and eCRF, the circuit mechanisms underlying these interactions are still unknown. In this thesis, we present a simple, novel cortical circuit model that can account for a surprisingly diverse array of eCRF properties. This model relies on extensive recurrent interactions between excitatory and inhibitory neurons, connectivity that is strongest between neurons with similar stimu- lus preferences, and an expansive input-output neuronal nonlinearity. There is substantial evidence for all of these features in V1. Through analytical and computational modeling techniques, we demonstrate how and why this circuit is able to account for such a comprehensive array of contextual modulations. In a linear network model, we demonstrate how surround suppression of both excitatory and inhibitory neurons is achieved through the selective amplification of spatially-periodic pat- terns of activity. This amplification relies on the network operating as an inhibition-stabilized network, a dynamic regime previously shown to account for the paradoxical decrease in in- hibition during surround suppression (Ozeki et al., 2009). With the addition of nonlinearity, effective connectivity strength scales with firing rate, and the network can transition be- tween different dynamic regimes as a function of input strength. By moving into and out of the inhibition-stabilized state, the model can reproduce a number of contrast-dependent changes in the eCRF without requiring any asymmetry in the intrinsic contrast-response properties of the cells. This same model also provides a biologically plausible mechanism for cortical normalization, an operation that has been shown to be ubiquitous in V1. Through a winner-take-all population response, we demonstrate how this network undergoes a strong reduction in trial-to-trial variability at stimulus onset. We also propose a novel mechanism for attentional modulation in visual cortex. We then go on to test several of the critical pre- dictions of the model using single unit electrophysiology. From these experiments, we find ample evidence for the spatially-periodic patterns of activity predicted by the model. Lastly, we show how this same circuit motif may underlie behavior in a higher cortical region, the lateral intraparietal area.

Despite their small brains, insects are capable of incredible navigational feats. Even Drosophila melanogaster (the common fruit fly) uses visual cues to remember locations in the environment. Investigating sophisticated navigation behaviors, like visual place learning, in a genetic model organism enables targeted studies of the neural circuits that give rise to these behaviors. Recent work has shown that the ellipsoid body, a midline structure deep within the fly brain, is critical for certain navigation behaviors. However, nearly all aspects of visual place learning remain mysterious. What visual features are used to encode place? What is the site of learning? How do the learned actions integrate with the core navigation circuits? To begin to address these questions I have established an experimental platform where I can measure neural activity using a genetically encoded calcium indicator in head-fixed behaving Drosophila. I further developed a virtual reality paradigm where flies are conditioned to prefer certain orientations within a virtual environment. In dendrites of ellipsoid body neurons, I observe a range of specific visual responses that are modified by this training. Remarkably, I find that distinct calcium responses are observed during presentation of preferred visual features. These studies reveal learning-associated neural activity changes in the inputs to a navigation center of the insect brain.

Early life stress (ELS) leads to alterations in anatomy and function of the adult hippocampal dentate gyrus (DG), but the mechanisms by which these lasting changes occur have not been fully elucidated. We tested the hypothesis that the immediate decrease in cell proliferation and neurogenesis induced by stress is the key mediator of the negative long-term outcomes of ELS. First, we tested whether inhibition of cell proliferation during early life is sufficient to reproduce the ELS-induced reduction in adult DG neurogenesis. We demonstrate that targeting dividing stem cells for elimination during the first or third postnatal weeks leads to diminished adult neurogenesis and reduction of the stem cell pool. Also, we hypothesized that ELS leads to more persistent effects on DG function than stress later in life because of the stress-induced elimination of specific birth cohorts of DG granule cells (GCs) that have distinct functions. We tested whether different birth cohorts of DG GCs differ in function by assessing behavioral and stress response outcomes of pharmacogenetic elimination or optogenetic activation of adult GCs born during the first or third postnatal week. We demonstrate that dorsal GCs born during the first or third postnatal week may be involved in modulating exploratory and anxiety behavior, but that only third postnatal week born GCs stimulate HPA activity. These results suggest that mature DG GCs may differ in specific functions with birth date determining their functional role. Third, we directly assessed the effect of ELS on DG development to better understand the immediate effects of ELS on the DG and to identify other potential mediators of the long-term effects. We demonstrate that ELS using the limited bedding/nesting paradigm leads to developmental delay of the DG. The work presented in this dissertation contributes to our understanding of the mechanisms by which ELS produces lasting impairments in DG function and also to our knowledge of how DG GC function is specified.

The mushroom body (MB) is the site of odor association learning in Drosophila.  In the canonical model, there are two types of reinforcing dopamine neurons (DANs): one set for rewarding unconditioned stimuli (US), and one responding to aversive US.  When DANs are activated together with an odor (the conditioned stimulus, or CS), plasticity is induced in the downstream output neurons (MBONs).  We have identified a DAN (V1) that surprisingly responds preferentially to odors, and responds weakly or not at all to various classical US.  In order to explore the relationship between V1 odor responses and the established roles of the MB, I characterized the responses of DAN V1, and probed its relationship to odor-driven behavior, associative conditioning, and activity in other MB compartments. These data show that V1 receives recurrent input from identified MBONs, contributes to the activity of an MBON that enhances alerting behavior, and that its odor responses are modulated by conditioning. We therefore present the study of the alpha2 compartment, which V1 innervates, as the dissection of an atypical compartment of the MB, one that acts as a hub by which various information from other compartments and brain areas is integrated in order to alter a behavioral response to odor. This work furthers our understanding of the MB not simply as an engine of classical learning, but as a system of diverse interconnected modules that allow coordinated fine control of behavior.

Locomotion, including running, walking, and swimming, is a complex behavior enabling animals to interact with the environment. Vertebrate locomotion depends upon sets of interneurons in the spinal cord, known as the central pattern generator (CPG). The CPG performs multiple roles: pattern formation (left-right alternation and flexor-extensor alternation) and rhythm generation (the onset and frequency of locomotion). Many studies have begun to unravel the organization of the neuronal circuits underlying left-right and flexor-extensor alternation. However, despite pharmacologic, lesion, and optogenetic studies suggesting that the rhythm generating neurons are ispilaterally-projecting glutamatergic neurons, the precise cellular identification of rhythm generating neurons remains largely unknown. Traditionally, CPG networks (both pattern formation and rhythm generation) are thought to reside upstream of motor neurons, which serve as the output of the spinal cord. Recently however, it has been discovered that direct stimulation of lumbar motor neurons using the intact ex vivo neonate mouse spinal cord preparation can activate CPG networks to produce locomotor-like behavior. Furthermore, depressing motor neuron discharge decreases locomotor frequency, whereas increasing motor neuron discharge accelerates locomotor frequency, suggesting that motor neurons provide ongoing feedback to the CPG. However, the circuit mechanisms through which motor neurons can influence activity in the CPG in mammals remain unknown. Here, I used motor neurons as a means of accessing CPG interneurons by asking how motor neuron activation might induce locomotor-like activity. Through intracellular recording and morphological assays, I discovered that ventral spinocerebellar tract (VSCT) neurons are activated monosynaptically following motor neuron axon stimulation through chemical and electrical synapses. A subset of VSCT neurons were located close to or within the motor neuron nucleus. VSCT neurons were found to be excitatory, have descending spinal axon collaterals, and influence motor neuron output, suggesting that VSCT neurons are positioned advantageously to initiate and maintain locomotor-like rhythmogenesis. Intracellular recording from VSCT neurons revealed that they exhibit rhythmic activity during locomotor-like activity. VSCT neurons were found to contain the rhythmogenic pacemaker Ih current and to be connected to other VSCT neurons, at least through gap junctions. Optogenetic and chemogenetic manipulation of VSCT neuron activity provided evidence that VSCT neurons are both necessary and sufficient for the production of locomotor-like activity. Silencing VSCT neurons prevented the induction of such activity, whereas activation of VSCT neurons was capable of inducing locomotor-like activity. The production of locomotor-like activity by VSCT neuron photoactivation was dependent upon both electrical communication through gap junctions as well as the pacemaker Ih current. The evidence presented in this thesis suggests that VSCT neurons are critical components for rhythm generation in the mammalian CPG and are key mediators of locomotor activity.