Neocortical layer 4 in adult mouse differs in major cell types and circuit organization between primary sensory areas

Abstract

22Layer 4 (L4) of mammalian neocortex plays a crucial role in cortical information processing, yet 23 a complete census of its cell types and connectivity remains elusive. Using whole-cell 24 recordings with morphological recovery, we identified one major excitatory and seven inhibitory 25 types of neurons in L4 of adult mouse visual cortex (V1). Nearly all excitatory neurons were 26 pyramidal and almost all Somatostatin-positive (SOM + ) neurons were Martinotti cells. In 27 contrast, in somatosensory cortex (S1), excitatory cells were mostly stellate and SOM + cells 28 were non-Martinotti. These morphologically distinct SOM + interneurons correspond to different 29 transcriptomic cell types and are differentially integrated into the local circuit with only S1 cells 30 receiving local excitatory input. Our results challenge the classical view of a canonical 31 microcircuit repeated through the neocortex. Instead we propose that cell-type specific circuit 32 motifs, such as the Martinotti/pyramidal pair, are optionally used across the cortex as building 33 blocks to assemble cortical circuits. 34 3 Main 35The mammalian sensory neocortex is organized in a six-layered structure. In this stereotypical 36 architecture, layer 4 (L4) is the main target of sensory inputs coming from the thalamus, thus 37 acting as the first level of cortical processing for sensory signals. Understanding how L4 38 implements its computations requires to characterize the diversity of its constituent neuronal 39 components and the connectivity among them. 40 41Most previous studies of L4 have focused on primary somatosensory cortex (S1) of young rats 42 and mice. Spiny stellate cells have been reported to be the dominant excitatory neural type, 43 both in rat 1,2 and in mouse after postnatal day five (as a result of sculpting of initially pyramidal 44 neurons during development) 3 . In contrast, inhibitory interneurons are highly diverse in terms of 45 their genetic markers, morphologies and electrophysiological properties. Previous studies have 46 reported three types of fast-spiking (FS), parvalbumin-positive (PV + ) interneurons 4 and five 47 types of non-FS interneurons 5 , all of which have distinct morphologies. Several recent studies 48 revealed that the somatostatin-positive (SOM + ) interneurons form a single morphological 49 population that has been called non-Martinotti cells 6 since their axons mainly target L4 7,8 , in 50 contrast to typical Martinotti cells, which target L1. Interneuron types exhibit type-specific 51 connectivity patterns. For example, PV + FS interneurons receive thalamic inputs 9-13 but SOM + 52 non-FS interneurons do not 14 . Both groups are reciprocally connected to local excitatory 53 neurons and between each other 4,8,10,12,15,16 , but PV + inhibit each other while SOM + do not 17 . 54 55Since most of these detailed studies were performed in S1 of young animals, it is unclear 56 whether the cellular components of L4 and their connectivity profile are the same in other 57 cortical areas and in adult anima...

Sparse reduced-rank regression for exploratory visualization of paired multivariate datasets

“…To illustrate the structure of such data sets and motivate the development of sparse RRR for exploratory visualization, we use principal component analysis (PCA) on the largest available Patch-seq data set (Scala et al, 2018). It contains n = 102 somatostatin-positive interneurons from layer 4 and layer 5 of primary visual and somatosensory cortex in mice ( Figure 2).…”

“…Here we developed sparse reduced-rank regression based on the elastic net penalty to obtain an interpretable and intuitive visualization of the relationship between high-dimensional single-cell transcriptomes and electrophysiological information obtained using techniques like Patch-seq. We used three existing Patch-seq data sets (Fuzik et al, 2016;Cadwell et al, 2016;Scala et al, 2018) to demonstrate and validate our method. Our sparse RRR method extends sparse RRR of Chen and Huang (2012) and, as we show, outperforms it on our data.…”

A systematic evaluation of interneuron morphology representations for cell type discrimination

“…In this study we only used the 11 cell types that included more than 5 neurons (remaining sample size n = 212). The cortical data consisted of inhibitory interneurons from primary visual cortex manually reconstructed based on biocytin stainings (Jiang et al, 2015;Scala et al, 2019). We analyzed the neurons separated by layer (V1 L2/3: n = 108 neurons in 7 classes, Figure 1B; V1 L4: n = 92 neurons in 7 classes, Figure 1C; V1 L5: n = 93 neurons in 6 classes, Figure 1D).…”

“…We used data from Helmstaedter et al (2013), Scala et al (2019), andJiang et al (2015), splitting the latter data set into two parts by cortical layer. All neurons were labelled by human experts in the original studies.…”

“…V1 Layer 4. Manually traced biocytin stainings of n = 92 inhibitory interneurons of 7 types in layer 4 (L4) of adult mouse primary visual cortex (Scala et al, 2019). We obtained the reconstructions (as .ASC file) and their cell type labels from the authors.…”