Modeling reveals human-rodent differences in h-current kinetics influencing resonance in cortical layer 5 neurons

Abstract

15Most existing multi-compartment, mammalian neuron models are built from rodent data. However, 16 our increasing knowledge of differences between human and rodent neurons suggests that, to 17 understand the cellular basis of human brain function, we should build models from human data. 18 Here, we present the first full spiking, multi-compartment model of a human layer 5 cortical 19 pyramidal neuron. Model development balanced prioritizing current clamp data from the neuron 20 providing the model's morphology, while also ensuring the model's generalizability via preservation 21 of spiking properties observed in a secondary population of neurons, by "cycling" between these 22 data sets. The model was successfully validated against electrophysiological data not used in 23 model development, including experimentally observed subthreshold resonance characteristics. 24 Our model highlights kinetic differences in the h-current across species, with the unique 25 relationship between the model and experimental data allowing for a detailed investigation of the 26 relationship between the h-current and subthreshold resonance. 27 28 30 (Womelsdorf et al., 2014) within the six-layered neocortex stems from invasive and in vitro studies 31 in rodents and non-human primates. Whether or not such principles can be extended to human 32 neocortex remains speculative at best. Despite the significant transcriptomic convergence of 33 human and mouse neurons (Hodge et al., 2019), significant differences between human and rodent 34 cell-type properties exist. In vitro studies have identified differences between mouse and human 35 neurons in morphology (Mohan et al., 2015), dendritic integration (Beaulieu-Laroche et al., 2018; 36 Eyal et al., 2016), synaptic properties (Verhoog et al., 2013), and collective dynamics (McGinn and 37 Valiante, 2014; Molnár et al., 2008; Florez et al., 2013). However, less explored are the active 38 1 of 36 Manuscript submitted to eLife membrane properties of human cortical neurons, which together with their passive and synaptic 39 properties underlie oscillations which are of likely physiological relevance (Akam and Kullmann, 40 2014; Womelsdorf et al., 2014; Fries, 2005; Anastassiou et al., 2011; Hanslmayr et al., 2019; Vaz 41 et al., 2019). 42 Recently it has been shown that increased expression of hyperpolarization activated cation chan-43 nels (h-channels) contribute to the observed subthreshold resonance in supragranular layer human 44 pyramidal cells not seen in their rodent counterparts (Kalmbach et al., 2018). Such differential 45 expression of h-channels also appears to be present between superficial and deep layer neurons 46 of human cortex, with layer 5 (L5) pyramidal cells demonstrating a larger sag voltage mediated 47 65 et al., 2013; Beaulieu-Laroche et al., 2018) leads to two important questions for computational 66neuroscientists: in what settings is it appropriate to utilize rodent neuron models to glean insights 67 into the human brain, and when such approximations are u...

Age-dependent increased sag amplitude in human pyramidal neurons dampens baseline cortical activity

“…3D, E , paired-sample t-test, p < 0.001, Cohen’s d : -8.7). We also tested this effect when changing the h-channel mechanism kinetics to values that have been used in previous modeling work (Hay & Segev, 2015; Rich et al, 2021), and consistently obtained decreased baseline Pyr neuron spike rates. This decrease in baseline spike rate could be recovered and even increased by changing dendritic h-channel density and passive membrane parameter values in the older Pyr neuron model dendrites to the values used in the younger Pyr neuron model dendrites ( Fig.…”

“…There is increasing data of h-current properties in human neurons as measured by transcriptomics and sag voltage recordings. Recent studies showed that HCN channel subunits are more ubiquitously expressed in pyramidal neurons across cortical layers of humans relative to rodents, with larger sag voltage in deeper layers compared to superficial layers (Kalmbach et al, 2018; Chameh et al, 2021; Rich et al, 2021). However, it is unknown if sag changes with age in human neurons as suggested by the above studies in monkeys.…”

The structural aspects of neural dynamics and information flow

“…While our understanding of human neurons is often inferred from models based on rodent data, the morphological and biophysical differences between human and rodent neurons may influence neural dynamics and information processing in various ways. The recent development of computational models for human layer 5 cortical neurons by Rich and colleagues lay a foundation for quantitative analysis on the inter-species differences [56].…”

Resilience through diversity: Loss of neuronal heterogeneity in epileptogenic human tissue impairs network resilience to sudden changes in synchrony

“…Potential future work involves the use of more biophysically detailed human inspired neuron and network models, allowing for the implementation and study of additional types of heterogeneity (including multiple, diverse inhibitory populations) and/or the study of model seizures. Such studies will be facilitated by our recent development of a biophysically-detailed computational model of a human L5 cortical pyramidal neuron (Rich et al, 2021), allowing them to be more directly applicable to potential clinical applications for the treatment of human epilepsy. In this vein, while we do not model seizures per se in this work, the two most common types of seizure onsets observed in intracranial recordings are the low-voltage fast (Lee et al, 2000) and hyper-synchronous onsets (Velascol et al, 1999).…”

“…These limitations warrant the development of biophysically detailed, human inspired neuron and network models, allowing for the study of additional types of heterogeneity. Such studies will benefit from our recent development of a biophysically-detailed computational model of a human L5 cortical pyramidal neuron (Rich et al, 2021). In this vein, while we do not model seizures per se in this work, the two most common types of seizure onsets observed in intracranial recordings are the low-voltage fast (Lee et al, 2000) and hyper-synchronous onsets (Velascol et al, 1999).…”