P294 Inhibitory-Targeted Plasticity in Developing Thalamocortical Networks
Matthew P. Szuromi*1,2, Gabriel K. Ocker2,3
1Graduate Program for Neuroscience, Boston University, Boston, USA 2Department of Mathematics and Statistics, Boston University, Boston, USA 3Center for Systems Neuroscience, Boston University, Boston, USA
*Email: mszuromi@bu.edu
Introduction
The maturation of thalamocortical (TC) afferents is a key feature of critical periods (CPs) for primary sensory cortices [1]. Bienenstock-Cooper-Munro (BCM) theory for synaptic plasticity has been effective in describing thalamic projections onto pyramidal (Pyr) neurons in layer 4 (L4) of primary visual cortex (V1) [2]. However, these models often consider only a homogeneous population of cortical neurons, neglecting the recurrent connectivity within cortex and the various cell types innervated by TC axons, such as parvalbumin+ (PV+) interneurons. To address this, we develop an excitatory-inhibitory thalamocortical network model equipped with triplet BCM spike-timing-dependent plasticity (STDP) and rigorously describe its dynamics. Methods Our model comprises three neuronal populations: cortical excitatory (E), cortical inhibitory (I), and thalamic (X), the latter of which can have correlated spike trains. Neurons are modeled as a mutually exciting Hawkes process [3]. We examine systems whereXtoE,XtoI, andEtoIsynapses can be plastic and update according to a triplet BCM STDP rule [4, 5]. Using standard separation of timescales, we derive dynamics for the mean interpopulation synaptic weights in terms of moments of the neural activity, calculated by the path integral formalism [6, 7, 8, 9]. We then apply numerical methods to assess how parameters (static weights, correlations, and STDP parameters) affect the stability and strength of the interpopulation weights. Results When only TC synapses are plastic, TC weights strengthen in response to increased thalamic correlations. Further, we find that corticocortical inhibition must be sufficiently strong (i.e., the ratio of the meanItoEweight to the meanEtoIweight must be sufficiently large) for bothXtoEandXtoIweights to stabilize at nonzero values. Additionally, we analyze the network whenEtoIsynapses are also plastic. We determine how parameters of the STDP rule and the network influence the trajectories and equilibria of the synaptic weight dynamics in response to varied thalamic correlations. Particularly, we describe regimes where the trajectory of the meanEtoIweight either mimics or opposes the trajectories of the TC weights. Discussion In this work, we extend models using triplet BCM STDP to excitatory-inhibitory networks. In L4 of V1, inhibitory synapses from PV+ interneurons onto Pyr neurons strengthen prior to the CP [10]. Our results suggest a possible explanation: strong inhibitory synapses are necessary for TC synapses to potentiate and stabilize. Experiments have also indicated that during the CP for V1, visual deprivation induces simultaneous TC depression and potentiation of Pyr to PV+ synapses [11]. Our results describe parameter regimes where this phenomenon can occur, suggesting potential plasticity rules for synapses onto PV+ cells during the CP.
Acknowledgements M.P.S. acknowledges the Neurophotonics Center at Boston University for their support. References 1. https://doi.org/10.1016/j.neuron.2020.01.031 2. https://doi.org/10.1038/381526a0 3. https://doi.org/10.1093/biomet/58.1.83 4. https://doi.org/10.1523/JNEUROSCI.1425-06.2006 5. https://doi.org/10.1073/pnas.1105933108 6. https://doi.org/10.1103/PhysRevE.59.4498 7. https://doi.org/10.1093/cercor/bhy001 8. https://doi.org/10.1371/journal.pcbi.1005583 9. https://doi.org/10.1103/PhysRevX.13.041047 10. https://doi.org/10.1523/JNEUROSCI.2979-10.2010 11. https://doi.org/10.7554/eLife.38846
