Efficient phase coding in hippocampal place cells

Abstract

Hippocampal place cells encode space through phase precession, whereby neuronal spike phase progressively advances during place-field traversals. What neural constraints are essential for achieving efficient transfer of information through such phase codes, while concomitantly maintaining signature neuronal excitability? Here, we developed a conductance-based model for phase precession within the temporal sequence compression framework, and defined phase-coding efficiency using information theory. We recruited an unbiased stochastic search strategy to generate thousands of models, each with distinct intrinsic properties but receiving inputs with identical temporal structure. We found phase precession and associated efficiency to be critically reliant on neuronal intrinsic properties.Despite this, disparate parametric combinations with weak pair-wise correlations resulted in models with similar high-efficiency phase codes and similar excitability characteristics.Mechanistically, the emergence of such parametric degeneracy was dependent on two factors.First, the dependence of phase-coding efficiency on individual ion channels was differential and variable across models. Second, phase-coding efficiency manifested weak dependence independently on either intrinsic excitability or synaptic strength, instead emerging through synergistic interactions between synaptic and intrinsic properties. Despite these variable dependencies, our analyses predicted a dominant role for calcium-activated potassium channels in regulating phase precession and coding efficiency. Finally, we demonstrated that a change in afferent statistics, manifesting as input asymmetry, introduces an adaptive shift in the phase code that preserved its efficiency. Our study unveils a critical role for neuronal intrinsic properties in achieving phase-coding efficiency, while postulating degeneracy as a framework to attain the twin goals of efficient encoding and robust homeostasis.

Spatial information transfer in hippocampal place cells depends on trial-to-trial variability, symmetry of place-field firing and biophysical heterogeneities

“…As the virtual animal traversed through the place field the presynaptic neurons fired action potentials. Their firing rates were modeled in a stochastic manner, driven by a Gaussian modulated cosinusoidal function, mimicking place cell inputs to the neuron (Basak & Narayanan, 2018Seenivasan & Narayanan, 2020). The presynaptic firing drove the opening of the colocalized synaptic NMDAR and AMPARs, resulting in synaptic currents (equations 8-16) flowing into the model neuron.…”

“…The presynaptic firing drove the opening of the colocalized synaptic NMDAR and AMPARs, resulting in synaptic currents (equations 8-16) flowing into the model neuron. The Gaussian modulated cosinusoidal function that defined the probability of occurrence of a presynaptic spike to each synapse in the neuron was computed as (Basak & Narayanan, 2018Seenivasan & Narayanan, 2020):…”

“…In introducing experience-dependent asymmetry in place-field firing (Harvey, Collman, Dombeck, & Tank, 2009;Mehta, Barnes, & McNaughton, 1997;Mehta, Lee, & Wilson, 2002;Mehta, Quirk, & Wilson, 2000), we replaced the symmetric Gaussian profile in equation 17by a horizontally reflected Erlang distribution to construct an asymmetric place-field envelope (Seenivasan & Narayanan, 2020). In this scenario, the Erlang-modulated cosinusoidal function that defined the probability of occurence of a presynaptic spike to each synapse in the neuron was computed as:…”

“…In equation 18, the parameters (=5) and k (=25) governed the extent of asymmetry (Seenivasan & Narayanan, 2020).…”

“…The voltage trace at the soma was filtered using a 50 ms wide median filter, to eliminate spikes but retain theta-frequency temporal modulation, and the Fourier spectrum of the filtered signal was computed. The power at 8 Hz of this power spectrum was employed as theta power (Basak & Narayanan, 2018Seenivasan & Narayanan, 2020).…”