Geoffrey J. Goodhill - US grants

For the brain to function correctly, its neurons must be connected correctly. Axons often have to grow over long distances to find appropriate targets. Some of the most important guidance cues they use to achieve this feat are gradients of increasing concentration of guidance molecules. The mechanisms by which axons sense and respond to these gradients are largely unknown. Understanding these mechanisms is important for understanding how the brain normally develops. It is also crucial for understanding how neural connections regenerate, and designing effective therapies to enable axons to grow back to appropriate targets and thus restore lost function after injury. Previous work to uncover these mechanisms has been largely qualitative and based on extremely crude technologies. In this project we will develop a uniquely quantitative approach, based on a combination of mathematical modelling and sophisticated technologies borrowed from other domains. We will paint patterns of guidance molecules onto the surface of a collagen gel in which axons are growing and, using quantitative modelling of molecular diffusion, calculate the resulting gradients in the gel which the axons then encounter. By maintaining fine control over the pattern of molecules applied, we will achieve fine control over the parameters of the gradient. This will allow systematic variation of gradient parameters, and thus systematic investigation of the way each of these parameters regulates axon guidance. The principle rewards of developing this technology are that it will provide a unique tool for probing axon guidance mechanisms that will be useful to a wide range of researchers. It will push the field forward by providing a previously unavailable quantitative testbed for studying such issues as receptor-ligand interactions and pharmacological manipulations of gradient sensing processes. An ultimate goal is to speed up the discovery of ways to stimulate recovery from damage to the nervous system.

Two key questions for understanding the visual system are, how is information about the world represented in the cortex, and how are these representations formed during development. The best studied cortical areas in this regard are V1 and V2, where features of the visual scene such as position, orientation, direction and spatial frequency are represented in maps with complex local and global structure. Understanding how these maps develop, and what principles underlie their final structure, is an important step in answering the above questions. The relative contribution of genetic versus epigenetic (particularly activity-dependent) mechanisms in determining visual cortical structure is still a subject of intense debate. This debate impacts directly on the design of effective therapies for treating the effects of early abnormal visual experience, such as strabismus. Visual cortical map structure arises from activity-dependent learning rules that attempt to represent highly correlated input features close together, acting in conjunction with genetically determined constraints such as the shape of the target area. Both the local and global structure of maps in V1 and V2 can be accounted for in terms of the interaction between the correlational structure of afferent activity, patterns of intracortical connections, and the shape of the cortical target region. Changes in map structure due to abnormal rearing paradigms such as monocular deprivation and strabismus follow naturally from the same rules. The hypothesis will be tested using computational models suited to exploring questions of large scale map structure. These models will be applied to low dimensional feature spaces representing position, orientation, ocular dominance, direction, spatial frequency, disparity, and color. Several types of parameter variation will be investigated, including variations in the statistical structure of afferent activity, variations in the pattern of intracortical connections, variations in the way similarity between input features is measured, and variations in the shape of the target region. At each stage, simulation results will be directly compared to experimental data using a range of quantitative techniques. The analysis and simulation of these computational models will lead to experimentally testable predictions concerning the parameters underlying cortical map formation and the effects of visual experience on cortical map structure.

DESCRIPTION (provided by applicant): Correct brain function requires correct brain wiring. An important step in the establishment of appropriate connectivity is the guidance of axons over long distances in the developing brain to find their correct targets. A crucial type of guidance cue axons use is concentration gradients of attractive or repellent factors. Over the past decade many new molecules have been discovered that guide axons in this way. However, as yet we still have very little understanding of the precise mechanisms by which axons detect and respond to gradients. A better understanding of these mechanisms would enable us to understand better (1) how the nervous system is normally constructed, (2) why axons sometimes mistarget during development, (3) the effect of gene deletions and mutations on wiring, and (4) how to encourage axonal regeneration to appropriate targets after injury. The goal of this project is to develop a mechanistic understanding of axonal behavior in gradients by building computational models of gradient detection and directed movement for axons. Two types of models will be investigated. The first type is based on the idea that gradient detection is limited by inevitable stochastic noise in the receptor binding process. These models assume a small, spherical sensing device, and make predictions about the minimum detectable gradient steepness that such sensing devices can detect. The second type of model addresses the unique role that filopodia play in axonal gradient sensing and movement. The model is based on the idea that filopodia act as somewhat independent sensing devices, and it is their combined dynamics that determines the threshold for gradient detection and the trajectories that axons follow. A crucial component of the project is that the computational modeling will be directly tested and constrained using a new, quantitative experimental assay the investigators have recently developed. The assay allows stable molecular gradients of precisely controlled shape to be established in a collagen gel. The system used will be the guidance of dorsal root ganglion axons by gradients of Nerve Growth Factor (NG). Exponentially-shaped gradients of varying steepness of NG will be used to determine the minimum gradient steepness axons can detect as a function of absolute concentration, and the trajectories of axons in these gradients will be quantitatively analyzed.

DESCRIPTION (provided by applicant): Visual maps are crucial for visual function. Understanding the mechanisms by which visual maps develop and regenerate is therefore crucial for understanding how vision develops, and how vision could be restored after injury. Perhaps the most important model system for visual map development is the topographic projection from the retina to the tectum/superior colliculus. Although the mechanisms of map formation in this system have been studied experimentally for many decades, there is no unifying theoretical framework that can explain more than a small subset of this data. Without such a framework it is impossible to rigorously state what the mechanisms underlying map formation in the retinotectal system really are, and it is difficult to extrapolate beyond current data to predict how novel perturbations to the system will affect map development. Although several theoretical models of retinotectal map formation have been proposed previously, they have two serious limitations. Firstly, they generally address only refinements to map topography by activity-dependent Hebbian learning rules, and do not address how initial topography forms by activity independent processes. Secondly, they generally take no account of any of the data that has emerged since 1995 on the crucial role of ephrin/Eph gradients in retinotectal map formation. To overcome these problems, we propose to develop a comprehensive new model which (1) is based on axon guidance by molecular gradients rather than Hebbian learning rules, and (2) synthesizes recent data on ephrin/Eph gradients with older data regarding surgical manipulations to the system. The PI is in a unique position to be able to do this, since he combines a detailed knowledge of the extensive experimental data in this area with wide-ranging experience in the theoretical modeling of axon guidance and map formation in the visual system. The outcome of this project will be a computational framework that rigorously identifies the relative contributions of mechanisms such as chemotaxis and competition to the formation of retinotectal maps, makes testable predictions, and can guide future experimental work.

ABSTRACT A critical step towards understanding how neural circuits drive behavior is the ability to record the activity of all neurons in an organism while it interacts with its environment in an unconstrained manner. A promising vertebrate model system in this regard is the zebrafish larva, which performs complex visually-driven behaviors such as hunting from an early age, and due to its transparency allows large-scale neural imaging at single- neuron resolution. However so far most investigations into the neural basis of zebrafish hunting behavior have used tethered fish which are unable to move freely. While some new assays have recently begun to overcome this limitation, these still significantly perturb the fish?s natural hunting behavior by requiring both an extremely shallow water depth, and rapid cancellation of lateral movement in order to maintain the fish within the field of view. Here we will develop a new `MesoLFM? imaging technology overcoming both of these limitations, thus permitting the neural basis of truly unconstrained hunting behavior to be investigated for the first time in zebrafish. We will achieve this by combining light field microscopy (LFM), an electrically tunable lens (ETL), and the MesoLens, a giant microscope objective with a 4x/0.47 NA objective, a 10 mm working distance and a 5 mm field of view. LFM allows single-shot capture of 3D volumes, enabling imaging at many Hz. The ETL removes the need for rapid stage or objective movements in z, enabling imaging in relatively deep water. The MesoLens allows long sequences of movements to be captured within a single field of view, enabling imaging of hunting without the need for rapid xy motion cancellation. First, we will perform optical simulations of MesoLFM in order to determine the optimal parameters for components such as the microlens array. Second, we will construct the MesoLFM according to these specifications, and calibrate it using measurements of fluorescent beads and a standard resolution target. Third, we will directly demonstrate MesoLFM imaging of the zebrafish brain. Together this work will establish a new platform for investigating the neural bases of unconstrained behavior. In a future R01 application we will use the MesoLFM to reveal how neural circuits drive natural goal-directed behavior and decision-making.