Alternative TSSs are co‐regulated in single cells in the mouse brain

Abstract Alternative transcription start sites (TSSs) have been extensively studied genome‐wide for many cell types and have been shown to be important during development and to regulate transcript abundance between cell types. Likewise, single‐cell gene expression has been extensively studied for many cell types. However, how single cells use TSSs has not yet been examined. In particular, it is unknown whether alternative TSSs are independently expressed, or whether they are co‐activated or even mutually exclusive in single cells. Here, we use a previously published single‐cell RNA‐seq dataset, comprising thousands of cells, to study alternative TSS usage. We find that alternative TSS usage is a regulated process, and the correlation between two TSSs expressed in single cells of the same cell type is surprisingly high. Our findings indicate that TSSs generally are regulated by common factors rather than being independently regulated or stochastically expressed.

Expression of example genes across the full annotated transcript from UCSC browser for CA1 neurons. Expression is shown as bars where the Y-axis for single cells has a limit of 5 molecules and for all cells combined has a limit of 500 molecules. Major and minor TSS are marked in red, while other CAGE peaks associated with a gene are marked in black. CAGE peaks not associated with a gene are marked in green and TSS from other genes than the gene in focus are marked in blue. Annotations of genes other than the gene in focus are also marked blue. Cells were chosen based on high total TSS expression. Appendix Figure S2 -Distribution of molecules around coding start and end sites A and B| X-axis shows distribution in base pairs around (A) coding start positions and (B) coding end position. Y-axis shows total CA1-neuron molecules in black and the number of (A) TSS starting or (B) TTS ending in red. The coding start and end positions were taken from RefSeq genes downloaded from UCSC table browser.
Appendix Figure S3 -Correlation between genes or major TSS in single cells A | Difference in Pearson correlation between total gene and major TSS expression for all genes of the same cell for CA1 neurons. X-axis shows total TSS expression for the two cells being compared, gene expression is not shown. Each dot represents the correlation between two cells and a total of 872 such correlations were made. Only genes with two valid TSS and where annotation existed in both FANTOM5 and RefSeq were included (n=873) when doing the correlations. The average number of molecules mapping to the included RefSeq genes were 8000 and for the two TSS combined 2000.
B | Pearson correlation for single cells using major TSS expression in relation to the combined major TSS expression for the two cells being compared. Each dot is a correlation between two cells as in 3A.
C | Pearson correlation for single cells using genes expression in relation to combined gene expression for the two cells being compared. Each dot is a correlation between two cells as in 3A.
D | Same as supplemental figure 4A, but instead of comparing correlations between cells across genes, correlations are calculated comparing two genes across all cells.
Genes are chosen pairwise from the list of genes with valid TSS pairs (in supplemental  table 2) E | Same as 3B but comparing between pairs of genes across all cells F | Same as 3C but comparing between pairs of genes across all cells Appendix Figure S4: Example of higher correlation using TSS instead of gene count Shows two genes, the genes have a higher correlation if the expression value used is based on read counts from the major TSS (r=0.49), than if the expression value is based read counts from the full gene (r = 0.12). Expression is shown as bars where the Y-axis for single cells has a limit of 5 molecules and for all cells combined has a limit of 500 molecules.  Appendix Figure S7: TSS expression for example genes Same figure as supplemental figure 1, but zoomed in on the TSS. The major TSS is colored red, the minor blue and other TSS are colored black. Expression is shown as bars where the Y-axis for single cells has a limit of 5 molecules and for all cells combined has a limit of 1000 molecules. Pink bars shows expression higher than the limit. Note that some cells don't express Dcn.

Appendix Figure S8: Major TSS consistently higher expressed
Examples of TSS expression in single CA1 neuron cells. Plots show the number of mRNA molecules detected from the major and minor TSSs in single cells presented to highlight instances of minor TSS preference. Each line connects the read count of the major to the log-read count of the minor TSS of a single cell. Black, major TSS was higher. Red, minor TSS was higher. The number next to the gene symbol shows number of cells where the total expression for the TSSs was greater than zero. The number next to "Major" shows number of cells where cells the major TSS had higher expression than the minor, and the number next to "Minor" shows number of cells where the minor TSS had higher expression. Note that each line can contain more than one cell if two cells have the same expression of the major and minor TSS and that red lines are three times thicker than black lines. Son (869) Snap25 (724) Minor (1) Minor (15) Major (868) Major (709) Minor (328) Major (373) Minor (20) Major (

C. Gelsolin (Gsn)
Full length gene expression confirms expression of multiple TSS. The results of PacBio sequencing is shown for 3 genes, Arpc1a, Cnp and Gsn for six individual Oligodendrocyte cells, as well as annotated Fantom 5 TSS peaks and genomic positions with more than 100 molecules from all combined Oligodendrocyte STRT reads. Cells with no expression isn't shown.