Understanding the neural mechanisms that guide behavior is one of the biggest quests in neuroscience. This question is tackled at different levels of analysis, from studying whole-brain blood oxygenation levels in humans, to minuscule receptor movements in cells. In this dissertation, I describe work aiming to explore the neural basis of behavioral improvement in the larval zebrafish from the network level, across brain areas, to the neuron level, looking at specific neuronal ensembles.In chapter one, I introduce the study of experience dependent changes in behavior through history and describe the advantages of using the ethologically relevant prey capture behavior in larval zebrafish as a model. In chapter two, in collaboration with Claire Oldfield, I studied how experience hunting live prey affects prey capture behavior and the underlying neural activity. I show that previous experience with live prey improves hunting performance compared to larvae that have been fed with inert fish flakes. Consequently, looking at whole-plane neural activity, I observed no differences in the neural representations of prey in the visual areas, however, experienced fish showed increased correlations between output neurons of the tectum and the forebrain, and an increased probability for visual activity to evoke motor action. This led to the hypothesis that experience may lower the threshold for visual information transfer to motor areas, via an increase of activity in the forebrain. To test this hypothesis, I specifically ablated cells in the habenula, one of the forebrain structures, and observed a reduction in eye convergences and prey consumption. These findings show the involvement of the forebrain in experience-dependent improvement of prey capture for the first time. In chapter three, I describe our attempts to study experience-dependent changes in forebrain activity and network dynamics between visual areas and the forebrain at a single-cell resolution, using multi-plane two-photon imaging. This project is still at its beginnings, but I describe the adaptation of the behavioral paradigm to the two-photon microscope, and the potential of recording large populations of neurons at cellular level during a naturalistic behavior.

Synaptic plasticity underlies all learning and memory and is a cornerstone of neurological function. The adaptive and flexible nature of these processes makes them complex and challenging to define with the precision needed to develop quantitative models or effective therapeutics. The main excitatory neurotransmitter in the mammalian nervous system is glutamate which is also critical in setting synaptic strength and plasticity. Glutamate induces synaptic changes through two classes of receptors, ionotropic glutamate receptors and metabotropic glutamate receptors. Ionotropic glutamate receptors are glutamate gated calcium channels which can alter synaptic strength in the time scale of tens of milliseconds to 1 second. Metabotropic glutamate receptors (mGluRs) are Class C G-protein coupled receptors (GPCRs) which induce synaptic changes on the time scale of tens of seconds to minutes, hours and even days. This makes mGluRs critical to understanding the molecular induction of synaptic plasticity and very attractive therapeutic targets. Despite much effort put towards developing therapeutics that target mGluRs none have been approved by the FDA. A deeper understanding of mGluR activation will aid in designing more effective therapeutics as well as provide a more precise role of glutamate on synaptic changes.mGluRs are obligate dimers with large N-terminal domains. Each subunit contains a clamshell like glutamate binding domain (LBD), a rigid cysteine rich domain (CRD) and the canonical 7-pass transmembrane domain that defines all GPCRs. Numerous studies of GPCRs have pointed to the fact that these receptors are highly dynamic and that ligand binding changes the time a receptor spends in several distinct conformational states. The last ten years has seen an explosion in the number of high-resolution GPCRs structures. Unfortunately, these only provide a snapshot of a subset of potential receptor conformations and the amount of knowledge on the conformational dynamics of GPCRs has lagged behind. One very promising technique to address the questions surrounding GPCR conformational dynamics is single molecule fluorescence resonance energy transfer (smFRET). smFRET allows for very precise measurement of receptor motions and compliments the wealth of structural information to create a much more complete understanding of GPCRs. To this end I used smFRET to measure and study the conformational motions associated with activation of mGluR homodimers and heterodimers. First I examined the effects of different regions on mGluR2 homodimer activation and mGluR2/3 heterodimer activation which demonstrated a functional role of a LBD interface and the cysteine loop. Next I examined mGluR7, which has an incredibly low glutamate affinity, and found that while mGluR7 homodimers are virtually insensitive to glutamate, mGluR2/7 heterodimers are present in native tissue and are super sensitive to glutamate due to an apo state rotation demonstrating a potential biologic and modulatory role of heterodimerization. Lastly, I examined regions of mGluR7 and find that the lower dimer interface is responsible for very low glutamate sensitivity and that the cysteine bridge flexibility is responsible for the mGluR2/7 super sensitivity to glutamate.

Many biological systems are composed of heterogeneous elements. This work analyzes the heterogeneities found in signaling systems within the Drosophila nervous system, focusing on the neuromuscular junction and the antennal lobe. The variations found within each system are used to understand some aspect of its organization.

At the Drosophila larval neuromuscular junction (NMJ), a motor neuron releases glutamate from many boutons onto the muscle it innervates. A post-synaptically localizing calcium sensor was developed to visualize the distribution of synaptic strength across boutons. While the Ca²⁺ signals were uniform within a given connection (i.e., bouton and postsynapse pair), they differed considerably among connections of an NMJ. A gradient of transmission strength was observed along axonal branches, from weak proximal connections to strong distal ones. Presynaptic imaging revealed a matching axonal gradient, with higher Ca²⁺ influx and exocytosis at distal boutons. The results suggest that transmission strength is mainly determined presynaptically at the level of individual boutons, possibly by one or more factors existing in a gradient.

The antennal lobe (AL) is a structure that processes odors. It is partitioned into glomeruli, each of whose inputs detect unique chemical features. How specialized is the circuitry of each glomerulus to operate on its inputs? We utilize a methodology that allows a simultaneous, unbiased comparison of the functional organization of many glomeruli. We find that the glomerulus DA1, which is selectively activated by a single pheromone, is organized differently than glomeruli that are activated by many different odorants. In contrast to what is seen in other glomeruli, ipsilateral and contralateral stimuli elicit distinct spatial patterns of activity within DA1. DA1 experiences greater and more rapid inhibition than other glomeruli, allowing it to report slight inter-antennal delays in stimulus onset in a "winner-take-all" manner. We propose that DA1's specializations help the fly localize and orient with respect to pheromone sources. Our results show how homologous circuits reflect the unique contingencies of their inputs.

Together, these works show how the heterogeneity found within diverse systems can provide a basis for understanding their organization.

Synaptic scaling is a mechanism that neurons use to maintain homeostatic levels of excitability, allowing them to increase their sensitivity when excitatory inputs are weak and to decrease sensitivity at times of elevated excitation. In this way individual neurons and whole neural circuits are thought to maintain a balance between excitation and inhibition, and to operate in a part of their dynamic range that allows for their output to increase and decrease in response to altered inputs. So far, scaling has been studied as cell-wide responses to long-term genetic and pharamcological manipulations of excitability in an entire cell, group of cells, or in whole brain regions. However, under physiological conditions scaling will be elicited not by these kind of global changes in input or excitability, but rather by changes in synaptic drive at a subset of inputs. This thesis sets out to determine whether local excitation elicits scaling responses and, if so, whether the changes would be confined to the site of excitation or would trigger a global response throughout the cell.

Our lab has developed LiGluR, a light-controlled glutamate receptor that rapidly and reliably depolarizes neurons, resulting in calcium influx and inducing action potential firing. In chapter 2 of this thesis I describe the optimization of subcellular LiGluR activation in cultured hippocampal neurons, discussing critical parameters for spatially restricting the illumination. Establishing a targeted illumination system and verifying the spatial resolution of light activation is essential for the experiments later described in chapter 3. The conditions for long-term light activation are also set to take advantage of the noninvasive nature of LiGluR, thus enabling the study of processes that require prolonged, chronic manipulation. The light-activation protocol considers both the light for activation and the light for measuring output, determining maximum power levels for robust photostimuluation while maintaining good spatially resolution and cell health. Light activation is carried out on a high resolution scanning confocal microscope so that small features (such as synaptic receptors) can be properly visualized. I also discuss the expression pattern of LiGluR in neurons and illustrate a co-expression system for driving channel localization to postsynaptic sites.

In chapter 3, I describe LiGluR activation in combination with time-lapse imaging of AMPA receptors (AMPARs), a well established expression locus of different modes of synaptic plasticity. LiGluR is an ideal tool to build upon previous studies for its spatial and temporal properties. With targeted illumination I am able to activate LiGluR within a single neuron with sub-cellular resolution. By combining this approach with bath application of TTX, which blocks action potentials without affecting light-induced depolarization, the effect of LiGluR activation is further restricted. Synaptic scaling occurs on the time scale of hours; LiGluR is able to provide the reliable, long-term stimulation necessary to evoke these changes. The non-invasive nature of LiGluR is also ideal for coupling with time-lapse imaging of other fluorescently-tagged proteins that can be expressed within the same cell.

To measure synaptic scaling I monitored the surface expression of AMPA-type glutamate receptors using an ecliptic-tagged GluR2 subunit, a highly pH-sensitive reporter that is fluorescent when the receptor is surface expressed and quenched when it is exposed to the lumen of a vesicle. Expressing this receptor in cultured hippocampal neurons results in punctate surface expression concentrated at spines along the dendritic shaft. The fluorescence intensity increases in hippocampal neurons when culture-wide activity is blocked with TTX, verifying that the system is a faithful reporter of AMPAR accumulation under synaptic scaling conditions. We found that local excitation in a distal sub-region of the dendritic tree induced a global reduction in the surface population of AMPARs at dendritic spines. The loss of AMPARs began within 20-30 minutes and leveled off within 1 hour. This scaling response was blocked by chelating intracellular calcium with BAPTA-AM, blocking voltage dependent calcium channels with cadmium, or inhibiting the activity of calcium/calmodulin-dependent protein kinases with KN-93. Imaging of internal calcium with a fluorescent calcium indicator dye showed that the optical excitation of LiGluR in the distal dendrite does not lead to a calcium rise outside of the illuminated area, indicating that the spread of the homeostatic effect propagates via another means. We consider the possibility that phosphorylated CaMKII, which has been shown to be highly mobile, spreads throughout the dendrite to give a global homeostatic reduction in AMPA receptor surface levels.

Our method employed here for spatially-controlled neuronal stimulation coupled with time-lapse imaging can further be extended to examine other important questions underlying neuronal circuit function.

Information processing in the brain involves communication between neurons at structures called synapses. Billions of neurons can have trillions of synaptic connections, each with varying properties of communication with nearby neurons. My dissertation research studies neurotransmitter release at synapses of the fruit fly (Drosophila) larval neuromuscular junction (NMJ), the sites of contact between a motor neuron and a muscle fiber. These glutamatergic synapses closely resemble the excitatory synapses of the vertebrate central nervous system. Each fly NMJ consists of an axon of a single neuron which forms hundreds of synapses onto a single postsynaptic cell, so that pre-post molecular cues are expected to be the same. Nevertheless, basal presynaptic strength (the probability that an action potential will evoke transmitter release, Pr) differs greatly (by as much as 100-fold) between synapses. My analysis shows that synapses with differing Pr are distributed randomly so that synapses situated within a micrometer of each other within the same bouton can differ greatly in Pr. Despite a general tendency toward facilitation of synapses with low basal Pr and depression of synapses with high basal Pr, I observe a remarkable diversity in facilitation/depression across synapses with similar basal Pr. The molecular underpinnings and functional relevance of these forms of synapse diversity remain to be elucidated. I find that the combination of diverse basal transmitter release and plasticity between synapses of type Ib axons, and between type Ib and type Is axons, generate broad distributions of weak and strong synapses across firing frequency, a robustness that maintains a stable average output. I examine the role in this process of the presynaptic glutamate-activated inhibitory auto-receptor, the G protein coupled receptor DmGluR. In Chapter 1, I review fundamental concepts of probabilistic synaptic release and propose applications of these models to the Drosophila larval neuromuscular junction. In Chapter 2, I report on input-specific differences between type Ib and Is axons in basal transmission properties using a ground-breaking in vivo imaging and analysis technique, Quantal Synaptic Optical Reconstruction (“QuaSOR”), which makes it possible to measure evoked and spontaneous transmission at hundreds of identified synapses simultaneously. I characterize differences in transmission, plasticity and homeostatic plasticity between two convergent inputs--from Ib and Is motor neurons-- onto the same postsynaptic muscle fiber. In Chapter 3, I formalize an analytical framework to examine the spatiotemporal nature of evoked and spontaneous transmission at low and high frequency of motor neuron action potential firing in relation to the spatial location and the distribution of presynaptic molecules. I characterize a fundamental result from my research demonstrating how the NMJ as a whole exhibits stable levels of neurotransmitter release despite large variance at individual synapses. Next, I examine spatiotemporal patterns of activity and conclude that active zones are not spatially arranged by synaptic strength but rather are distributed across the entirety of the neuromuscular junction in a more random fashion.

Experience influences how we perceive the world, how we interact with our environment, and how we develop. In fact, almost all animals can modify their behavior as a result of experience. Psychologists have long distinguished between different forms of learning and memory, and later determined that they are encoded in distinct brain areas. Neuroscientists dating back to Santiago Ramón y Cajal suggested that learning and memory might be encoded as changes in synaptic connections between neurons, but it wasn’t until the second half of the 20th century that experimental evidence corroborated this idea. Specific firing patterns of neurons during learning results in strengthening or weakening of synapses, and morphological modifications that can lead to long lasting changes in neural circuits. The molecular mechanisms that drive these changes are remarkably conserved across vertebrates. However, understanding how synaptic plasticity is integrated at a network level remains a big challenge in neuroscience. Relatively few studies have focused on how neural circuits encode changes in behavior in a natural context.

In this dissertation I present two such examples using calcium imaging and optogenetic techniques in larval zebrafish. First, in collaboration with others, I studied neural activity that drives the onset of the very first behavior of the fish, tail coiling. We characterized the transition of spontaneous activity in the spinal cord from a sporadic, uncorrelated state, to a regular and synchronized state. We determined that this transition occurred through coalescence of local microcircuits, and that inhibiting the early sporadic activity impaired normal development of the central pattern generator network. Second, I studied how experience of prey affects hunting behavior and underlying neural activity. I developed a paradigm to assess the effect of prior hunting on prey capture behavior and showed that experienced fish initiate more captures than their naïve counterparts. I next established that experience does not affect the ability of the fish to see prey. Rather, experience increases the probability that activity in visual areas will evoke a capture initiation. This is accomplished by increasing the impact of information transfer from visual to motor areas, possibly due to increased activity in forebrain areas for experienced fish.