The Spatial Basis of Continuous Neuronal Identities in the Striatum
Geoffrey Stanley1, Özgun Gökce2,5, Norma Neff3, Thomas C. Südhof2,4**, Stephen R. Quake3**
1Program in Biophysics; 2Department of Molecular and Cellular Physiology; 3Department of Bioengineering; 4Howard Hughes Medical Institute, Stanford University, Stanford, CA.;5Institute for Stroke and Dementia Research, Klinikum der Universitat Munchen, Ludwig-Maximilians-Universitat LMU, 81377 Munich, Germany
The brain makes use of the largest variety of cell types and anatomical structures of any organ in the mammalian body, and this diversity is crucial to brain function. Single-cell mRNA sequencing has given us more information than ever about the identity of brain cells, but it is unresolved how to rigorously turn this information into a comprehensive atlas of brain cells. Previous approaches have assumed that neuronal identity in the adult brain is discrete and used clustering algorithms to define neuron types. However, it has been observed that not all neurons are well described by a single identity. Here, we present a rigorous framework for dissecting neuronal identity with a linear, iterative approach, and apply it to single D1 and D2 medium spiny neuron transcriptomes. Careful analysis of gene expression signatures revealed two broad classes of neuronal identity in the striatum: discrete identities, defined by distinct modes of gene expression that are not connected by intermediate cells, and continuous identities, where cells are connected along gradients of gene expression. We find that the discrete subtypes of striatal medium spiny neurons are defined by the canonical D1 and D2 markers. Within each of the discrete subtypes, we found multiple dimensions of continuous heterogeneity. Using quantitative RNA in situ staining, we show that these transcriptional continua encode anatomical location. Each neuron’s location in transcriptional space corresponds directly to its anatomical location, and anatomically adjacent regions are continuous in gene expression space. We validated ten such regions with unique expression patterns, ranging from the ventral islands of Calleja to the dorsal striosomes, and thus present a comprehensive cell atlas of striatal MSNs. We find that the genes expressed along these continua are largely conserved between the D1 and D2 discrete subtypes, indicating a degree of independence between discrete D1/D2 identity and spatial identity. Since spatial location is a continuous property, discrete identities should not be able to fully encode spatial information. Indeed, within anatomical regions, the discrete D1 and D2 MSN subtypes are spatially interspersed among each other. The remarkable concordance between continuous transcriptional identity and anatomical location in the striatum suggests a general principal for a brain cell atlas: regions consist of intermingled discrete subtypes, within which spatial location is encoded by conserved gene expression gradients. Future work will be needed to understand how these gradients arise, how they are maintained epigenetically, and what their functional role is in the adult brain.