A brief history of somatostatin interneuron taxonomy or: how many somatostatin subtypes are there, really?

We provide a brief (and unabashedly biased) overview of the pre-transcriptomic history of somatostatin interneuron taxonomy, followed by a chronological summary of the large-scale, NIH-supported effort over the last ten years to generate a comprehensive, single-cell RNA-seq-based taxonomy of cortical neurons. Focusing on somatostatin interneurons, we present the perspective of experimental neuroscientists trying to incorporate the new classification schemes into their own research while struggling to keep up with the ever-increasing number of proposed cell types, which seems to double every two years. We suggest that for experimental analysis, the most useful taxonomic level is the subdivision of somatostatin interneurons into ten or so "supertypes," which closely agrees with their more traditional classification by morphological, electrophysiological and neurochemical features. We argue that finer subdivisions ("t-types" or "clusters"), based on slight variations in gene expression profiles but lacking clear phenotypic differences, are less useful to researchers and may actually defeat the purpose of classifying neurons to begin with. We end by stressing the need for generating novel tools (mouse lines, viral vectors) for genetically targeting distinct supertypes for expression of fluorescent reporters, calcium sensors and excitatory or inhibitory opsins, allowing neuroscientists to chart the input and output synaptic connections of each proposed subtype, reveal the position they occupy in the cortical network and examine experimentally their roles in sensorimotor behaviors and cognitive brain functions.