– AstroComp

Focus in theoretical neuroscience has traditionally been directed at neurons and synaptic connectivity as the primary mediators of information processing in the brain.
This view is beginning to be reshaped by a recognition of multicellular contributions to neural computation, including by astrocytes. These cells actively influence synaptic transmission, neuronal excitability, and blood flow, suggesting they play not just a supportive role, but may be pivot points in the propagation or gating of functionally salient signals throughout the brain. Indeed, the molecular diversity of astrocytes and their spatiotemporally hierarchical interactions with neurons makes them potentially uniquely suited to significantly impact plasticity and other key aspects of learning and information processing. However, their multifaceted input-output relationships also make the study of astrocytes difficult through a purely experimental approach. Computational modeling may thus be an ideal platform to explore how astrocytes enhance neural circuit dynamics and function. In this spirit, our proposal will pair modeling at circuit and network scales with advanced methods for recording and manipulating astrocytes, toward exploring and validating new theories regarding neural-astrocyte interaction and brain function. Our study is based on a hypothesis we have termed ‘contextual guidance,’ which views astrocytes as multiplexers or switchboards that transmit signals about the environment and physiological state to neurons and networks thereof. Indeed, it is well documented that astrocytes respond to contextual inputs by adjusting extracellular conditions, energetic supply to neurons, and efficacy of synaptic transmission. As such, we posit that astrocytes are positioned to serve as a force multiplier by expanding the repertoire of neuronal dynamics and hence computation. We aim to explore two hypotheses within the contextual guidance framework, (a) that astrocytes actively control network dynamics, conveying environmental inputs to neurons, and (b) this modulation enables information processing and facilitates learning across contexts. To test these hypotheses, we propose two aims. In Aim 1, we will systematically elucidate astrocytes' roles and mechanisms in neuromodulatory systems, investigating how their distinct spatial and temporal scales may contribute to adaptation in neural and synaptic dynamics. In Aim 2, we will examine neural-astrocyte interplay at a network level, exploring features like 'tiling,' wherein astrocytes overlay specific neuron clusters, potentially influencing signal routing. Our goal is to connect these models to functions such as resolving contextual ambiguity.
The specific aims will be pursued through a comprehensive modeling approach alongside experimental validation using a battery of modern tools for recording neural- astrocyte interactions and causally manipulating astrocytes. The experimental approach will allow for tight theory-experiment feedback and direct validation of key premises and predictions arising from our neural-astrocyte interaction models. The proposed research may have broader impacts in brain-inspired computing and specifically on the construction of new artificial neural network designs that incorporate astrocytes. For instance, our proposed neural astrocyte network dynamics may surpass existing AI constructs in expressiveness, potentially enhancing features like gating for enacting context- and history-dependent learning. Furthermore, the proposed effort includes initiatives to engage trainees in interdisciplinary neuroscience research through workshops in the astrocyte biology and computational neuroscience communities, thus helping create a new generation of scholars able to bridge between theory and experiments.