The long-term goal of my research is to understand the role of sensory experience in shaping the connectivity and functional properties of developing neural circuits. We focus on the relatively simple and experimentally tractable visual system of Xenopus laevis tadpoles; a preparation which is amenable to a wide variety of experimental approaches, ranging from molecular biology, in vivo and in vitro whole-cell electrophysiology, live cell imaging, and behavior. This also provides us a unique platform in which to model and understand various neurodevelopmental disorders such as autism, schizophrenia and epilepsy.
One commonly held view in developmental neurobiology is that the initial connectivity of sensory systems is mediated by molecular cues which provide a rough wiring diagram, and then connectivity is further refined by neural activity [1]. This suggests that neural activity is necessary in order for mature forms of sensory processing to emerge. However, there is accumulating evidence from various species that several aspects of sensory processing emerge long before sensory circuits in the brain are fully mature. For example, in the visual system of human infants, peak contrast sensitivity reaches near-adult levels by 9 weeks of age, visual acuity improves by 8 months, and orientation tuning by 3 months (see [2] for review). Despite this precocious emergence of visual function, the human brain continues to grow, change and be altered by neural activity during the first 6 years of life. Remarkably, the visual system is able to maintain this high level of functionality even as the size of the eye and retina, along with the distance between the eyes, increases during this period, forcing a remapping of the visual space. This leads to one of the principal questions driving my research:
How are developing neural circuits able to remain sufficiently stable to function robustly, yet at the same time remain sufficiently flexible to accommodate the massive changes in neuronal architecture and synaptic organization that occur during development?
One approach for addressing this question is to start with a neural circuit with known behavioral function and to study its response properties at different points in development to stimuli known to elicit behavior. The next step is to characterize how the different neuronal elements that form the circuit work together to generate these behaviorally relevant responses, and characterize how these neural correlates change developmentally. By understanding how behaviorally relevant stimuli are encoded at different times in development, we can deduce general principles by which neural circuits can self organize to produce a consistent behavioral output despite variations in the maturational state of the individual cells, neuronal architecture, synaptic transmission and ion channel types comprising the circuit.
My research takes this approach by focusing on the optic tectum of Xenopus laevis tadpoles. For years, the frog visual system has been an incredibly fruitful preparation in which to study sensory-motor integration as well as visual system development of vertebrates [3-5], and this work has played an important role in shaping our view of how the CNS develops. Since the tectum receives direct multisensory input and its output is directly related with behavior, it is the ideal preparation to study the development of neural circuits involved in behavior. Experimentally, the developing tectal circuit of Xenopus laevis tadpoles provides tremendous advantages over other preparations: First, we have developed robust behavioral assays that are sensitive to changes in tectal circuit development. Second, the ability to easily perform in vivo whole-cell recordings and Ca++ imaging of populations of tectal neurons poses a unique opportunity to study the single-cell and network-level mechanisms by which sensory stimuli are processed. Third, the isolated whole tadpole brain preparation allows us to study electrophysiological responses in developing neurons in an intact brain with most of its circuitry preserved, providing a naturalistic yet reduced preparation [6, 7]. Fourth, the optic tectum receives direct visual, auditory and mechanosensory input, allowing for in vivo manipulations of neural activity by altering the sensory environment [8]. Fifth, because it has a relatively permeable blood brain barrier, it is possible to do in vivo pharmacological manipulations by introducing drugs directly into the rearing solution [9, 10]. Sixth, there are several established techniques for altering gene expression in the in vivo tectum and retina, including viral gene transfer, electroporation of DNA plasmids, RNA interference, morpholino antisense-oligonucleotides as well as various novel transgenesis techniques such as pTransgenesis [6, 11, 12].
Thus, the Xenopus tadpole visual system is one of the few preparations in which one can take a unique, integrative approach where one can work at the interface between the molecular, cellular, circuit and behavioral levels in an intact and highly-plastic vertebrate animal as it goes through development. Furthermore, understanding how small perturbations in this developmental process can cascade into large scale defects in neural circuit function will provide important insights into the mechanisms underlying a variety of neurodevelopmental disorders such as autism, schizophrenia and childhood epilepsies.
Current Research Activities
The optic tectum, and its mammalian homologue the superior colliculus, are central for transforming visual input into orienting behavior and for performing multisensory integration [13-16]. The optic tectum receives a topographically-organized projection from the retina. Initial establishment of this map depends on molecular cues, but further refinement, maturation and maintenance requires neural activity [17, 18]. Patterned visual activity plays a central role in this process, regulating both dendritic growth and formation of new synapses [5, 19, 20].
Most of our research focuses on a period in early development (between developmental stages 42 and 49) when the retinotectal map is being established and further refined, and innervation from other sensory modalities is also occurring. By focusing on key developmental stages (see Fig. 1) during this period we can track the functional development of tectal circuits over time, and work over the last several years has documented this process in extensive detail. Our work focuses on various levels of analysis, from single cells, to behavior and multisensory integration. Some of our major research efforts are described below.
Homeostatic Control of Neural Excitability and Synaptic Transmission
In order for a neural circuit to process and relay sensory information optimally, it needs to adapt its response properties to best match the type of sensory input it receives [21]. We found that developing tectal neurons are known to adapt their intrinsic excitability in response to changing levels of synaptic input. Between developmental stages 42-49 the developing tectum undergoes temporally coordinated changes in synaptic strength and intrinsic neuronal excitability [6]. For example tectal neuron intrinsic excitability peaks during stages 45/46, but by stage 49, when the tectal cell growth begins to stabilize, neurons become significantly less excitable (Fig. 2A). In parallel, as dendrites grow the total amount of excitatory synaptic drive (sEPSC frequency x amplitude) increases steadily during this time (Fig. 2B). Changes in intrinsic excitability are mediated, at least in part, by changes in the amplitude of voltage-gated Na+ currents. Because changes in synaptic and intrinsic properties are reciprocal between stages 45 and 49 (intrinsic excitability wanes as total synaptic drive waxes), this suggests that a homeostatic mechanism may work to balance neuronal activity at an optimal level. This is manifested as stable neuronal input/output functions during this developmental period (Fig. 2C,D). Genetic manipulations which experimentally alter synaptic drive in single tectal neurons cause compensatory changes in their intrinsic excitability, suggesting that regulation of synaptic and intrinsic properties during development is linked by a common mechanism.
We are now testing whether the intrinsic properties of neurons are tuned to respond in an optimal manner to the temporal input patterns that they receive from the retina at different stages in development. This would indicate that neurons are continuously adapting their membrane properties to generate a stable output even as the amount and spatiotemporal pattern of synaptic input changes over development. In the future we will work towards determining the mechanisms by which a neuron maintains a specific set-point for its output, and in determining the pathways and molecules involved in maintaining this homeostatic behavior. We are also aiming to use gene expression profiling to understand how various ion channel genes and trafficking proteins vary as neurons undergo homeostatic adaptation.
Development and Plasticity of Recurrent Neural Circuits in the Optic Tectum
The process by which neurons adapt to developmental changes in sensory input, is not limited to properties of single neurons, but can also function at the network level. In the retinotectal pathway of Xenopus tadpoles, a tectal neuron's receptive field (RF) size is determined by both the degree of convergence from different retinal ganglion cells (RGC) as well as by the pattern of excitatory and inhibitory inputs received from neighboring tectal cells. This local neural circuitry within the tectum refines over development, such that recurrent excitation becomes more temporally compact and less variable across trials as the circuit matures. This can be seen both a the level of spike output and subthreshold synaptic currents, indicating a developmental change in the dynamics of the whole network [22]. These local circuits are also sculpted by visual experience, such that the temporal properties of recurrent intratectal activity are tuned to reflect the temporal properties of visual input and show a large degree of plasticity. For example, we have found that the timing of visually-evoked spike output can be sculpted by repeatedly activating the retinal inputs with a specific temporal pattern. As a result, this input pattern is reflected in the pattern of recurrent spiking activity following this conditioning (Fig. 3), and this change can be explained by spike-timing dependent plasticity of recurrent intratectal inputs [22].
We have also described changes in network dynamics using high-speed imaging of Ca++ signals [10]. This allows us to simultaneously measure the neuronal activity of large populations of tectal neurons in vivo in response to visual stimuli with a 10 msec temporal resolution. We find that, in response to a visual stimulus, tectal responses become more synchronous across neurons and show less trial-to trial variability over development. This change can be disrupted if animals are reared in the dark or with NMDA receptor blockade. In sum, the retinotectal circuit shows a remarkable degree of flexibility, allowing both individual tectal neurons and local tectal networks to adapt to developmental changes in sensory input.
Visually guided behavior and the tectum
What are some of the functional consequences of this adaptability? One possibility is that adaptability allows tectal circuits to alter their response properties in order fine-tune how they respond to meaningful sensory stimuli, such as those that would elicit important behaviors. This would result in more precise visually guided behavioral output.
We have described a robust visual avoidance behavior in Xenopus tadpoles [23] in which swimming tadpoles avoid collisions with moving visual stimuli of different sizes. We found that tadpoles more reliably avoid spots of a specific size (2 mm radius). Tuning of this behavioral response develops between stages 45 and 49, and is correlated with refinement of visual receptive fields. Disrupting normal development of receptive fields also disrupts the tuning of the avoidance behavior. This behavior, unlike other visuomotor behaviors such as the optomotor response, requires an intact optic tectum. The open question now is how does the tectal circuitry encode visual stimuli such that it triggers an avoidance response, and how does the developing visual circuitry enable the spatial tuning preferences of the behavior?
To answer this question we are taking a multi-level approach in which we present behaviorally relevant stimuli to immobilized tadpoles while performing in vivo whole-cell recordings and network-level Ca++ imaging. To do this we have developed a new behavioral paradigm in which visual stimuli are projected directly to the tadpole's eye via a fiber optic bundle. Tadpoles are embedded in low-melt agarose such that their tails move freely while the tadpole remains fixed in place. By tracking tail movements we can detect when the tadpole performs an escape response. We then present a stimulus set consisting of visual stimuli with varying amounts of spatiotemporal information (e.g. whole field flash, expanding disk, moving grid, etc.). The goal is to build a stimulus library of visual stimuli with different degrees of behavioral relevance (Fig. 4). Then we present the same stimuli while performing in vivo electrophysiological recordings or Ca++ imaging of large neuronal populations in the tectum. This permits the measurement of the responses of the tectal circuit, as well as excitatory and inhibitory synaptic currents of tectal neurons in response to the various stimuli. This approach has allowed us to observe how stimuli with different levels of behavioral relevance are encoded in the tectum, and what strategies the tectum utilizes to differentiate between them. By comparing developmental stages, we are able to understand how the various changing circuit elements work together to encode these responses, and by perturbing normal development we can determine the ability of the network to adapt to these perturbations and the effect of these perturbations on normal behavior.
(see https://wiki.brown.edu/confluence/x/AIQEAQ for video of behavior)
Development of multisensory integration
The optic tectum not only receives visual input, but like its mammalian homologue, the superior colliculus, it receives input from a variety of sensory modalities. This sensory convergence is important for integrating multiple sensory cues and generating orienting behavior. Moreover, it is known that proper development of multisensory responses requires neural activity. In Xenopus we have developed an an experimental preparation in which we can independently activate different pathways representing different sensory modalities while performing whole-cell recordings or Ca++ imaging in an isolated brain [13]. This allows us to study the neural mechanisms underlying multisensory integration at a cellular level during development, as well as test the role of sensory activity in this process. In the long-term, we aim to integrate these single cell-level studies into a behavioral paradigm, which will allow us to directly link cellular events to behavioral output. This is one of the few preparations where the detailed cellular and network-level mechanisms underlying multisensory integration can be studied. Furthermore this preparation will also be useful towards understanding the various types of cross-modal plasticity that occur during development and their underlying mechanisms [8].
Tectal development as a model for neurodevelopmental disorders
Another advantage of our model system is that it can be a useful model for studying disease. Because we are continuously developing a deep understanding of the various aspects that regulate normal development of neural circuits, we are well placed to understand how perturbations (eg. genetic, environmental) can affect the system at all levels, serving as a useful model for a series neurodevelopmental disorders. Experimentally, Xenopus tadpoles are very amenable to genetic and pharmacological manipulation, and visually guided behavior can be used to screen for abnormal development patterns, which can then be examined in detail using anatomy and electrophysiology. In one project we have developed a model to study epilepsy and have discovered a novel neuroprotective role for polyamines, a class of endogenous molecules that becomes elevated after a seizure and ultimately results in reduction of seizure susceptibility, protecting the brain from further seizures and lasting damage [24]. In another we have examined the role of pro-inflammatory cytokines in neural circuit development [9]. Elevated cytokine levels have been found in the cerebrospinal fluid of individuals with autism, schizophrenia and other neurodevelopmental disorders. Many of these cytokines also play a role in regulating neural function. Using a behavioral screen we identified cytokines which affected tectal circuit development, among those TNF-alpha. We found that elevated levels of TNF-alpha result in abnormally early maturation of glutamatergic synapses, prematurely stabilizing dendritic arbors and preventing normal pruning, resulting in hyperconnectivity within the tectum, abnormal visual behavior and increased seizure susceptibility. This approach can be extended to a variety of molecules and manipulations involved in diverse neurodevelopmental disorders.
