Plasticity manifolds and ion-channel degeneracy govern circadian oscillations of neuronal intrinsic properties in the suprachiasmatic nucleus

Abstract

Motivation and methods: The suprachiasmatic nucleus (SCN) is the master circadian clock of the mammalian brain that sustains a neural code for circadian time through oscillations in the firing rate of constituent neurons. These cell-autonomous oscillations in intrinsic properties are mediated by plasticity in a subset of ion-channels expressed in SCN neurons and are maintained despite widespread neuron-to-neuron variability in ion channel expression profiles. How do SCN neurons undergo stable transitions and maintain precision in intrinsic properties spanning the day-night cycle if several ion channels change concomitantly in a heterogeneous neuronal population? Here, we address this important question using unbiased stochastic searches on the parametric and the plasticity spaces using populations of SCN models, each explored for multiple valid transitions spanning one complete circadian cycle (day-to-night followed by night-to-day transitions). Results: Our analyses provided three fundamental insights about the impact of heterogeneities on the circadian oscillations of SCN intrinsic properties. First, SCN neurons could achieve signature electrophysiological characteristics (day-like or night-like) despite pronounced heterogeneity in ion channel conductances, with weak pairwise correlations between their conductance values. This ion-channel degeneracy precluded the need to maintain precise ion-channel expression profiles for achieving characteristic electrophysiological signatures of SCN neurons, thus allowing for parametric heterogeneities despite functional precision. Second, it was not essential that specific conductances had to change by precise values for obtaining valid day-to-night or night-to-day transitions. This plasticity degeneracy, the ability of disparate combinations of ion-channel plasticity to yield the same functional transition, confers flexibility on individual neurons to take one of several routes to achieve valid transitions. Finally, we performed nonlinear dimensionality reduction analyses on the valid plasticity spaces and found the manifestation of a low-dimensional plasticity manifold in day-to-night transitions, but not in night-to-day transitions. These observations demonstrated that the concomitant changes in multiple ion channels are not arbitrary, but follow a structured plasticity manifold that provides a substrate for stability in achieving stable circadian oscillations. Implications: Our analyses unveil an elegant substrate, involving a synthesis of the degeneracy and the plasticity manifold frameworks, to effectuate stable circadian oscillations in a heterogeneous population of SCN neurons. Within this framework, the ability of multiple ion channels to change concomitantly provides robustness and flexibility to effectively achieve precise transitions despite widespread heterogeneities in ion-channel expression and plasticity.

Balancing complexity, performance and plausibility to meta learn plasticity rules in recurrent spiking networks

“…Running two (intrinsically stochastic) evolutionary searches from the same starting point on the familiarity task with an I-to-E small polynomial rule converged to two plasticity rules with dissimilar pre-post protocols (Fig.8A). This shows, in agreement with previous work in rate networks [44] [45, 46], that at least two and probably many plasticity rules from the same search space can solve this task. This conclusion is not unique to I-to-E plasticity (Fig.8B).…”

Ion-channel degeneracy and heterogeneities in the emergence of signature physiological characteristics of dentate gyrus granule cells

“…Finally, the use of such network models with morphologically realistic neurons that manifest degeneracy at different scales could be used to assess the multifarious and heterogeneous impact of different neuromodulators across different cells at different locations. In addition to changes mediated by neuromodulation, such morphologically realistic model populations could also be used to assess plasticity profile degeneracy (Anirudhan and Narayanan, 2015; Shridhar et al, 2022), plasticity heterogeneities (Shridhar et al, 2022), and plasticity degeneracy (Nagaraj and Narayanan, 2023) in implementing the encoding functions of the DG network.…”

“…The rise and decay time constants of AMPAR were τ 𝑟 (= 2 ms) and 𝜏 𝑑 (= 10 ms) (Ye et al, 2005). The AMPAR density (permeability value 𝑃 ̅ 𝐴𝑀𝑃𝐴𝑅 ) of individual synapses in the base model were adjusted (Narayanan and Chattarji, 2010;Basak and Narayanan, 2018;Roy and Narayanan, 2021) such that the propagated somatic EPSP amplitude, irrespective of dendritic location, was in the 0.2-0.3 mV range to match with unitary somatic EPSP amplitudes in DG granule cells (Krueppel et al, 2011). The same location-dependent density values were used across all valid GC models to assess heterogeneities in synaptic information transfer within granule cells.…”

“…Morphologically realistic models with a broad set of physiological constraints are essential because morphology sets strong structural constraints on neuronal physiology (Mainen and Sejnowski, 1996;Krichmar et al, 2002;Narayanan and Chattarji, 2010;Dhupia et al, 2015;Basak and Narayanan, 2020) and the number of physiological constraints contribute to model complexity (Yang et al, 2022;Schneider et al, 2023).…”

“…In addition, reliance on a single hand-tuned model would yield biased conclusions as the hand-tuned model is simply one instance of the several possible realizations of characteristic neuronal properties. The population-of-models approach is now an established route to systematically study heterogeneities and differential dependencies of functional outcomes in several biological neuronal subtypes (Foster et al, 1993; Goldman et al, 2001; Prinz et al, 2003; Marder and Taylor, 2011; Rathour and Narayanan, 2012; 2014; Basak and Narayanan, 2018; Migliore et al, 2018; Mittal and Narayanan, 2018; Rathour and Narayanan, 2019; Basak and Narayanan, 2020; Goaillard and Marder, 2021; Roy and Narayanan, 2021; Mittal and Narayanan, 2022; Arnaudon et al, 2023; Nagaraj and Narayanan, 2023; Reva et al, 2023; Roy and Narayanan, 2023), including DG granule cells (Beining et al, 2017b; Mishra and Narayanan, 2019; 2021c; Schneider et al, 2023).…”