Single-cell isoform RNA sequencing (ScISOr-Seq) across thousands of cells reveals isoforms of cerebellar cell types.
Full-length isoform sequencing has advanced our knowledge of isoform biology1-11. However, apart from applying full-length isoform sequencing to very few single cells12,13, isoform sequencing has been limited to bulk tissue, cell lines, or sorted cells. Single splicing events have been described for <=200 single cells with great statistical success14,15, but these methods do not describe full-length mRNAs. Single cell short-read 3 sequencing has allowed identification of many cell sub-types16-23, but full-length isoforms for these cell types have not been profiled. Using our new method of single-cell-isoform-RNA-sequencing (ScISOr-Seq) we determine isoform-expression in thousands of individual cells from a heterogeneous bulk tissue (cerebellum), without specific antibody-fluorescence activated cell sorting. We elucidate isoform usage in high-level cell types such as neurons, astrocytes and microglia and finer sub-types, such as Purkinje cells and Granule cells, including the combination patterns of distant splice sites6-9,24,25, which for individual molecules requires long reads. We produce an enhanced genome annotation revealing cell-type specific expression of known and 16,872 novel (with respect to mouse Gencode version 10) isoforms (see isoformatlas.com).\n\nScISOr-Seq describes isoforms from >1,000 single cells from bulk tissue without cell sorting by leveraging two technologies in three steps: In step one, we employ microfluidics to produce amplified full-length cDNAs barcoded for their cell of origin. This cDNA is split into two pools: one pool for 3 sequencing to measure gene expression (step 2) and another pool for long-read sequencing and isoform expression (step 3). In step two, short-read 3-sequencing provides molecular counts for each gene and cell, which allows clustering cells and assigning a cell type using cell-type specific markers. In step three, an aliquot of the same cDNAs (each barcoded for the individual cell of origin) is sequenced using Pacific Biosciences (\"PacBio\")1,2,4,5,26 or Oxford Nanopore3. Since these long reads carry the single-cell barcodes identified in step two, one can determine the individual cell from which each long read originates. Since most single cells are assigned to a named cluster, we can also assign the cells cluster name (e.g. \"Purkinje cell\" or \"astrocyte\") to the long read in question (Fig 1A) - without losing the cell of origin of each long read.\n\nO_FIG O_LINKSMALLFIG WIDTH=180 HEIGHT=200 SRC=\"FIGDIR/small/364950_fig1.gif\" ALT=\"Figure 1\">\nView larger version (66K):\norg.highwire.dtl.DTLVardef@4df0cdorg.highwire.dtl.DTLVardef@fc4beborg.highwire.dtl.DTLVardef@1dc485forg.highwire.dtl.DTLVardef@1138a3e_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFigure 1:C_FLOATNO (A) Outline of our ScISOr-Seq approach. (B) TSNE-plot depicting cell clusters, marker genes and names given to clusters, including: Bergman glia (BG), External granule cell layer neurons (EGL), Internal granule cell layer and other neurons in the interior of the cerebellum (IGL), two clusters of Purkinje cell layer neurons (PCL), oligodendrocyte progenitor cells (OPCs), Atoh1+ neuronal progenitors, Ptf1a+ neuronal progenitors and other neuronal progenitors (NPCs) (C) In-situ hybridization images from the Allen Brain Atlas depicting expression of marker genes in specific layers. (D) Expression patterns of selected marker genes across cell types.\n\nC_FIG