Compartmentalized learning with coupled electrochemical adaptation

Overview

Project Summary

Chemical synapses are the most complex inter-cellular juntion in the mammalian body. Involving over 2000 different proteins, interacting across multiple spatial and temporal domains, activity-dependent alterations in intracellular molecular dynamics underlie the phenomenon of synaptic plasticity which, in turn, provides the physiological, mechanistic basis for information acquisition and long-term storage (learning and memory).
While the types and nature of these processes have been systematically unravelled in the past decades, these insights have largely failed to permeate into theoretical modelling and the development of functional learning algorithms. Ascribing a mechanistic role to the many different molecular players and pathways involved and understanding their interplay from a functional standpoint is a necessary step to interpret the vast amounts of empirical data and derive deeper insights into the neurobiological bases of learning and memory.

In this exploratory project, we take the first steps towards a fundamental re-conceptualization of how we model learning via synaptic plasticity and how we think about adaptation in cortical circuits. By emphasizing intracellular molecular dynamics, we propose that:

1. plasticity is compartmentalized and heterosynaptic, i.e., the induction of plasticity mechanisms within one synapse is coupled to the expression of synaptic change in neighboring synapses within restricted spatial domains (dendritic branches).

2. intracellular heterosynaptic interactions bridge plasticity in different synapse types, i.e., the induction of plasticity at a Glutamatergic synapse will yield the expression of plasticity at neighboring GABAergic synapses, resulting in a localized, detailed balance between excitation and inhibition.

3. electrochemical compartmentalization yields distinct, but coupled, forms of distributed, competitive plasticity, i.e., the mechanisms of homeostatic synaptic scaling, neuron-wide intrinsic plasticity and structural remodelling, among others, are all coupled and share the same triggers and effectors.

Establishing adequate mathematical descriptions of these processes, validating their biophysical compatibility and investigating their functional consequences will yield new theoretical and computational tools to consolidate existing knowledge on the interplay between synaptic, dendritic and neuronal adaptive and homeostatic processes as well as a deeper understanding of the biophysics of learning and memory.

Main Goals

This project aims to investigate different levels of electrochemical compartmentalization and how to use intracellular variables to couple adaptive processes bridging synaptic, dendritic and neuronal adaptation. The project embarks on a bottom-up approach and emphasizes postsynaptic molecular dynamics in search for explicit biophysical links among processes that are typically considered disjoint. However, the primary aim is to investigate functional consequences at the circuit-/population-level. To do so, we focus on minimal effective models that provide as compact a description as possible while retaining physiological and biophysical compatibility. This is necessary, on the one hand, to ensure scalability and, on the other, to yield generalizable and transferable insights into the principles underlying learning and memory via synaptic and neuronal adaptation.

Beyond validating individual models to ensure biophysical compatibility, we aim to systematically profile functional consequences: characteristic input-output transfer functions, statistics of network activity, stability of population dynamics, operating timescales and memory capacity (Duarte & Morrison, 2019), task learning (in a constrained set of tasks) and information processing capacity (Schulte to Brinke et al., 2023). Note that the infrastructure to perform the required numerical simulations and analyses is already in place and results from the PI’s work in the past years. The team has developed a comprehensive set of tools (Duarte et al., 2017; Duarte et al., 2021) that range from input generation, system simulation, data acquisition, fitting and post-processing. These tools have been developed with a modular design and are agnostic to the precise neuronal/synaptic models employed and, as such, can be used to perform all the high-level numerical experiments required for this project.

Objective 1: Coupling excitation and inhibition with compartmentalized heterosynaptic plasticity.

Local cross-talk between neighbouring synapses onto the same dendritic compartment modulates the expression of plasticity (Lee et al., 2016). Mechanistically, this is a consequence of shared intracellular signaling whereby the induction of synaptic plasticity at a given synapse triggers a variety of different effectors, which, in turn, result in long-term changes both in the synapse that caused their activation and in neighboring processes that are directly influenced by the same effectors. While the identification of individual effector molecules and signaling pathways is a fundamental effort, we aim to leverage the tremendous empirical and conceptual progress achieved in the last decades (for reviews see, e.g., Magee & Grienberger, 2020; Südhof, 2021) to derive useful mathematical models that encompass the main physiological effects at different levels of granularity. We start by considering electrotonically isolated dendritic compartments and investigate how to formulate synaptic plasticity rules whose shared postsynaptic expression explicitly couples all synapses onto the same compartment (heterosynaptic plasticity). To achieve that in a physiologically realistic manner, the dendritic compartment needs to be equipped with key electrogenic phenomena which are known to underlie some of the core events in the expression of synaptic plasticity and/or to amplify its effects locally.

Objective 2: Couple synaptic, dendritic and neuronal adaptation as a continuum.

From individual synapses to whole neurons, there are multiple physical barriers constraining electrochemical diffusivity and introducing selectivity and competition in the adaptive processes. Their existence and prevalence entails that activity-dependent adaptation unfolds in a segregated, compartmentalized manner (d’Aquin et al., 2022; Makino & Malinow, 2011; Murphy-Baum & Awatramani, 2022; Palmer, 2022). On the other hand, the proteins and complexes that do diffuse across these barriers trigger the induction and expression of different forms of plasticity across different spatial and temporal domains. With these observations in mind, this sub-project investigates the coupling between synaptic and intrinsic plasticity by considering the processes across these different levels of compartmentalization and their interactions as a set of coupled adaptive systems. Our aims are two-fold: (1) investigate competition and cooperation among different forms of plasticity across different spatiotemporal domains; (2) investigate the mechanisms linking synaptic plasticity with homeostatic scaling and other forms of neuron-wide adaptation.

Project Details