Single-cell sequencing of the small-RNA transcriptome

Single cells were lysed directly in the preparation tube and 5.8S rRNA was masked for adapter ligation immediately by a complementary oligo. Then, the 3' adapter was ligated to the small RNAs and the free adapter was removed enzymatically prior to the 5' adapter ligation. The 5' adapter contains a unique molecular identifier (UMI) to remove PCR replicates. Sequences required for Illumina cluster generation and sample indexing were added through two rounds of PCR amplifications.

We designed seven synthetic RNA oligos that we subjected to library preparation and subsequently sequenced the resulting library. In the analyses, we treated their first eight nucleotides as UMIs and we calculated the percentage of mismatches that occurs during PCR amplifications and sequencing reaction. In the barplot we summarized the percentage of sequenced reads with mismatches before and after using the adjacency method to collapse reads to UMIs. The results show that the adjacency method correctly collapsed all single bp mismatches into molecules and partially corrected for the reads that harbored two or more mismatched within the UMIs. The error bars represent standard deviation. Adjacency procedure is available at https://github.com/CGATOxford/UMI-tools

(a) Numbers of genes per RNA species observed within the small RNAs libraries (18-40 nts). (b) Size distributions of mRNA-derived small RNAs profiled in single cells and per cell type. Error bars denote standard deviations. Peaks in length distribution were observed around 18 and 31 nts suggesting that these RNAs were processed products and were not the result of mRNA degradation.

(a) Fraction of differentially expressed miRNAs detected when comparing naive and primed hESCs in bulk RNA measured by q-PCR (Sperber et al. Nat Cell Biol 2015) and in single-cell small-RNA sequencing data (Fig 1h). (b) Scatter plot of miRNAs expression in single naive and primed hESCs (Fig 1h), colored according to their differential expression of bulk data in Sperber et al. Nat Cell Biol 2015. The miRNAs identified as differentially expressed in single-cell data got colored according to whether the same miRNA was identified as significantly higher expressed in Naive (green) or Primed (blue) hESCs. Red circles denote miRNAs that were found significantly differentially expressed in our study but not in Sperber et al.

Analyses of tsRNA, sdRNA and mRNA-derived small RNAs in naive and primed hESCs. Genes in red were statistically significant, (assessed using SCDE: Single Cell Differential Expression analysis) requiring adjusted p-values below 0.05 and a 2-fold or larger difference in molecule counts between cell types. As in Fig 1h, 97 naïve and 90 primed cells were used.

(a) Spearman coefficients are reported for the correlation of miRNA expression among pairs of single cells within each cell type, as boxplots. Line in the middle of boxes represents median, and notches represent confidence interval around the median. Primed ESCs showed less correlation that was statistically significant among single cells compared to naive hESCs (Welch Two Sample t-test). (b) miRNA expression distributions across naive and primed hESCs in absolute molecule counts, for the miRNAs that showed biological variation in primed hESCs (Fig 1i). These miRNAs were abundant in naive hESCs but were only detected at high levels in a few primed hESCs. (c) Similar to (b) but for a selection of miRNAs that were differentially expressed and also showed no significant variation within naive nor primed hESCs. (a-c) Naive cells were sub-sampled according to the molecule distribution of primed cells in order to adjust for their differences in cell size.

The absolute molecule counts data was normalized to adjust for the cell size difference, by sub-sampling molecules in primed cells to match the molecule distribution obtained for the smaller naive cells (see Materials and Methods for details). (a) Expression variation within primed and naïve hESCs (as Fig 1i) for tsRNAs and sdRNAs. (b) Distribution of molecule counts observed in single naive and primed hESCs, for the RNA genes with most variation within populations.

(a) Clustering of single cells (columns) based on expression of miRNAs (rows). A number of miRNAs (128) were expressed exclusively in naive hESCs but not in HEK293FT nor primed hESCs. (b) The miRNAs were found uniquely expressed in naive hESCs.

Clustering of primed and naïve hESCs, HEK293FT cells and glioblastoma cells using miRNA expression with t-distributed stochastic neighbor embedding (t-SNE). Analysis included total 398 cells and 585 miRNAs expressed (>=1 molecule) in at least 2 cells. The number of cells used for each cell type is indicated inside parenthesis in the legend. Independent experimental replicates were performed for primed and naïve hESCs in order to evaluate the batch effect.

Scatter plots of miRNA expression on diluted HEK293FT RNA (1ug, 100ng, 10ng, 1ng, 0.1 ng, 0.01ng). Three replicates of 1ug profiles on the y-axis were plotted against three replicates of profiles generated from lower amounts of starting total RNA (x-axis). In the last row, profiles from 1ng HEK293FT total RNA was compared to the profiles obtained on three individual HEK29FT cells. Spearman coefficients are reported in top right corners.

Small-RNA sequencing libraries were generated from 1000 ng of HEK293FT RNA and we excluded the size selection step in the NEBNext protocol to make it directly comparable to our data (as we do not use any size selection step). Scatter plots summarized the profiling of miRNAs (a), tsRNAs (b) and sdRNAs (c) using the NEBNext protocol (x-axis) against single-cell small-RNA sequencing data (y-axis). Comparisons against different replicates generated with the same protocol are shown for NEBNext (d) and small-RNA single-cell sequencing protocol (e). Spearman correlation was computed by adding pseudocount of 1 to all samples. Dropouts are seen as small RNAs only detected with one method. Genes with a 40-fold or larger difference in expression between methods were colored red (higher in single-cell small-RNA protocol) and blue (higher in NEBNext protocol).