Genetically Defined Subtypes of Somatostatin-Containing Cortical Interneurons

Abstract

Of the four main subclasses of inhibitory cortical interneurons, somatostatin-containing (SOM) interneurons are the most diverse. Earlier studies identified layer 1-projecting (Martinotti) cells in layer 5/6 of the X98 and the Chrna2-cre transgenic lines, and two groups of non-Martinotti cells - long-range projecting SOM cells in layers 2 and 6, and layer 4-projecting X94 cells in layers 4 and 5. Later in-vivo and ex-vivo studies described two morphological types of Martinotti cells which appear to have opposing roles in behaving animals. More recently, large-scale transcriptomic studies attempting to classify all cortical neurons by their gene expression profiles and by their morphological and electrophysiological phenotypes divided all SOM interneurons into 13 morpho-electro-transcriptomic (MET) types. It remains unclear, however, how the previously identified SOM subtypes relate to each other, and how they map onto the suggested MET classification scheme. Importantly, only a small number of Cre or Flp driver line are available to target SOM interneurons, and there are currently no genetic tools to target the majority of the proposed MET types for recording, imaging or optogenetic manipulations, severely hindering progress on understanding the roles SOM interneurons play in sensorimotor processing or in learning and memory. To begin to overcome these barriers, we undertook a systematic examination of SOM interneuron subtypes in layer 5 of mouse somatosensory cortex. We generated 4 intersectional triple-transgenic genotypes, by crossing the Sst-IRES-Flp line with 4 different Cre lines and with a dual-color reporter that labels all Cre expressing SOM cells with GFP and all other SOM cells in the same brain with tdTomato. Brains from adult mice of both sexes were retrogradely labeled by epipial dye deposits, processed histologically, and immunostained for 3 marker proteins known to be expressed in different SOM subsets. By correlating fluorescent protein expression, retrograde label and marker proteins in the same neurons, we found that Cre-expressing SOM cells in the Calb2-IRES-Cre and in the Chrna2-Cre lines, and GFP expressing neurons in the X94 line, comprise three non-overlapping SOM populations which together account for about half of all SOM cell in layer 5. Using whole-cell recordings ex-vivo, we show that they also exhibit electrophysiological properties which are distinctly different from each other. This multimodal convergence of axonal projection target, marker protein expression and electrophysiological properties strongly suggests that these three populations can be considered bona-fide SOM subtypes. Indeed, each of the three subtypes appears to map onto a unique MET type. Our findings call for a renewed effort to generate additional driver lines that can be used combinatorially to provide genetic access to the many remaining SOM subtypes and uncover their roles in cortical computations.

Integrating EM and Patch-seq data: Synaptic connectivity and target specificity of predicted Sst transcriptomic types

“…Observed complementary networks of inhibition (L4 vs L5) Sst+ cells in L4 and 5 with distinct axonal morphology have distinct intrinsic physiology 21,69 and even opposite behavior during whisking 37 . Some studies found that Sst cells specifically innervate excitatory targets in one layer but not the other (Sst+ cells innervating L4 will target L4 and not L5 pyramidal cells and vice versa) 32,38,39 .…”

PhysMAP - interpretablein vivoneuronal cell type identification using multi-modal analysis of electrophysiological data

“…3C, left, many trends immediately pop out: PV + cells seem to be uniformly narrow no matter the layer; SOM + cells occupy a gradient of widths and overlap with E-5 cells at their widest; and the widest spiking cells are E-4. This finding regarding SOM + cells occupying a range of spike widths is likely because it is composed of several subtypes of varying widths (Hostetler et al, 2023) which have distinct computational roles (Wu et al, 2023). Waveform peak-to-trough ratios also differ between cell types Fig.…”

“…Again, as in the juxtacellular dataset, the PhysMAP approach provided interpretable representations. To illustrate this point, we identified two sub-populations of SOM + cells not documented in the original publication of this dataset (Lakunina et al, 2020) but predicted by the literature (Hostetler et al, 2023; Wu et al, 2023; Urban-Ciecko and Barth, 2016). These wide-spiking SOM + (WS-SOM + , Fig.…”

“…In particular, many SOM + cells demonstrate a fastspiking phenotype in rodents and are known to express fast-spiking potassium channels (Urban-Ciecko and Barth, 2016). In contrast other SOM + cells, such as Calb2, exhibit broad-spiking phenotypes (Hostetler et al, 2023).…”

“…In particular, many SOM + cells demonstrate a fast-spiking phenotype in rodents and are known to express fast-spiking potassium channels (Urban-Ciecko and Barth, 2016). In contrast other SOM + cells, such as Calb2, exhibit broad-spiking phenotypes (Hostetler et al, 2023). Similarly, in the juxtacellular dataset, excitatory neurons formed a broad class and layer 4 excitatory neurons had different and distinct electrophysiological properties from layer 5 excitatory neurons.…”