Tuned inhibitory firing rate and connection weights as emergent network properties

Abstract

Excitatory cortical neurons show clear tuning to stimulus features, but the tuning properties of inhibitory neurons are ambiguous and have been the subject of a long debate. While inhibitory neurons have been considered to be largely untuned [1–4], recent studies show that some parvalbumin expressing (PV) neurons do show feature selectivity and participate in co-tuned subnetworks with pyramidal cells in which PV cells show high response similarity to the excitatory (E) neurons [5, 6]. Given shared input from layer 4 that drives feature tuning in excitatory subnetworks, we demonstrate that homeostatic regulation of postsynaptic firing rate governing the synaptic dynamics of the connections from PV to E cells, in combination with heterogeneity in the excitatory postsynaptic potentials (EPSP) that impinge on PV cells, results in the self-organization of PV subnetworks. We reconcile different experimental findings by showing that feature tuning of PV cells is an emerging network property that may be driven by synaptic heterogeneity, and can be inferred using population-level measures, while pairwise individual-level measures may fail to reveal inhibitory tuning. We show that such co-tuning can enhance network stability at the cost of response salience.

Geometry and dynamics of representations in a precisely balanced memory network related to olfactory cortex

“…E/I co-tuning has been observed experimentally in cortical brain areas (Bhatia et al, 2019;Froemke et al, 2007;Okun and Lampl, 2008;Rupprecht and Friedrich, 2018;Wehr and Zador, 2003) and emerged in simulations that included spike-timing dependent plasticity at I synapses (Lagzi and Fairhall, 2022;Litwin-Kumar and Doiron, 2014;Vogels et al, 2011;Zenke et al, 2015). In simulations, E/I co-tuning can be established by including I neurons in assemblies, resulting in "E/I assemblies" where I neurons track activity of E neurons (Barron et al, 2017;Lagzi and Fairhall, 2022;Mackwood et al, 2021). Exploring the structural basis of E/I co-tuning in biological networks is challenging because it requires the dense reconstruction of large neuronal circuits at synaptic resolution (Friedrich and Wanner, 2021).…”

“…The resulting state of "precise" synaptic balance stabilizes firing rates because inhomogeneities or fluctuations in excitation are tracked by correlated inhibition. E/I co-tuning has been observed experimentally in cortical brain areas (Bhatia et al, 2019;Froemke et al, 2007;Okun and Lampl, 2008;Rupprecht and Friedrich, 2018;Wehr and Zador, 2003) and emerged in simulations that included spike-timing dependent plasticity at I synapses (Lagzi and Fairhall, 2022;Litwin-Kumar and Doiron, 2014;Vogels et al, 2011;Zenke et al, 2015). In simulations, E/I co-tuning can be established by including I neurons in assemblies that track activity of E neurons in "E/I assemblies" (Barron et al, 2017;Lagzi and Fairhall, 2022;Mackwood et al, 2021).…”

Unraveling Functional Diversity of Cortical Synaptic Architecture Through the Lens of Population Coding

“…One limitation of our approach is that our model did not account for many aspects of cortical networks such as stimulus-dependent correlations, recurrent connections, biophysics, inhibitory subnetworks, or nonlinear dendritic integration (Polsky et al, 2004 ; Lavzin et al, 2012 ; Goetz et al, 2021 ; Lagzi and Fairhall, 2022 ). The utility of our approach is that it demonstrates that even a simple linear model can produce heterogenous weights, without incorporating elements that introduce nonlinearities.…”

Computational functions of precisely balanced neuronal assemblies in an olfactory memory network

“…Coordinated E-I connectivity underlying precise synaptic balance may be established by different motifs including EI assemblies, which can be generated by biologically plausible learning rules 19,76 . However, experimental evidence for EI assemblies remains indirect.…”

“…Because small variations in the E/I current ratio cause large firing rate changes, network stability requires co-tuning of E and I synaptic inputs in individual neurons across external stimuli and time, which is referred to as "precise synaptic balance" 15,16 . E/I co-tuning requires specific higher-order connectivity that can emerge from activity-dependent synaptic plasticity in computational models [17][18][19] . Experimentally, E/I co-tuning has been observed in multiple brain areas including sensory cortices [20][21][22][23][24][25] but the underlying network organization remains unclear.…”