Single-Cell Morphologies of Parvalbumin-Positive Interneurons in the Mouse Primary Somatosensory and Visual Cortices

Abstract

Classic approaches to defining cell types have looked to describe the elaborate size and shape characteristics of single cells.Classic approaches to neuroanatomy have remained a largely descriptive gold standard. However, to keep pace with the advances made by single-cell technologies, suitable data structures and analysis methods are needed to align the types of insights afforded by these two major techniques. Using open-access in situ hybridization (ISH) data from the Allen Institute Mouse Neuroanatomy database, and the Allen Cell Types database , we developed a method to quantify and characterize the cell morphologies of PV+ interneurons in the mouse primary somatosensory (S1) and visual (V1) cortices. Using a custom image analysis pipeline in CellProfiler, we obtained high-dimensional, cell morphology data for thousands of single-labelled cells. We defined and characterized cell morphologies using Robust-Sparse K-Means Clustering (RSKC) and the denSNE dimensionality reduction algorithm to determine the salient morphology features that distinguish among groups of labelled cells. We show that our approach can integrate data from ISH and whole-cell fill, establishing the multimodal potential of our pipeline.Furthermore, we apply our pipeline to probe the morphological correlates of transcriptomic classes of PV+ interneurons establishing the potential for morphological differences highlighting different transcriptomic classes. Our pipeline identifies morphological complexity comparable to the complexity seen in single-cell transcriptomics studies. Insights from our work overcome the limitations in how cell morphologies are presented, using a quantitative, data-driven and unsupervised approach.