Single-cell full-length total RNA sequencing uncovers dynamics of recursive splicing and enhancer RNAs

We report a novel single-cell total RNA-seq method, RamDA-seq, by combining a novel reverse transcription (RT) technology, RT with random displacement amplification (RT-RamDA), and not-so-random (NSR) primers. RT-RamDA provides global cDNA amplification directly from RNA during RT without any universal adopters, which benefits RT efficiency, simplification of the procedure, avoiding the step of PCR amplification, and decontamination of genomic DNA. NSR enables random priming while preventing cDNA synthesis from rRNAs. RamDA-seq showed high sensitivity to non-poly(A) RNAs and full-length coverage even for extremely long transcripts (>10 kb). Moreover, RamDA-seq revealed recursive splicing, a multi-step, splicing, within > 300 kbp-long introns. Finally, RamDA-seq enabled the first genome-wide analysis of enhancer RNAs (eRNAs) in single cells. Overall design: Total 1,172 single cells (mouse ES, PrE and differentiating cells from ES cells) were sequenced by using RamDA-seq protocol: (1) Cell cycle samples (277 RamDA-seq samples with mouse ES cells from G1, S, and G2M phases) were sequenced. (2) Cell differentiation time series RamDA-seq samples (456 RamDA-seq samples with mouse ES cells collected 0, 12, 24, 48, and 72 h after the induction of cell differentiation into PrE cells ) were sequenced. (3) 55 C1-RamDA-seq samples with mouse ES and PrE cells were sequenced. (4) To evaluate plate-to-plate variability (batch effect) of RamDA-seq, two experimenters performed RamDA-seq with two plates (384 cells) of single-cell samples (consisting of mouse ES and PrE cells) on two separate days. For assesment of RamDA-seq, 9 averaged single-cell lysate samples and 16 purified total RNA samples diluted to 10 pg were sequenced by using RamDA-seq protocol. We also sequenced 4 SMART-Seq v4 samples using 10 pg of total RNA. Moreover, 6 poly(A) RNA-selected and 6 rRNA-depleted samples derived from 1 ug of total RNA were sequenced by using a commercially available library prepararion kit. Furthermore, to evaluate reproducibility, we sequenced ~96 diluted RNA samples (corresponding to the entire plate) with RamDA-seq, C1-RamDA-seq, and C1-SMART-seq v4. To compare RamDA-seq and rRNA-depleted RNA-seq prepared with equal amounts of input RNA, we sequenced RamDA-seq samples using 1 ng of diluted total RNA and rRNA-depleted RNA-seq samples using using 1 and 10 ng of diluted total RNA.