Direct full-length RNA sequencing reveals unexpected transcriptome complexity during Caenorhabditis elegans development

Abstract

Massively parallel sequencing of the polyadenylated RNAs has played a key role in delineating transcriptome complexity, including alternative use of an exon, promoter, 5 ′ or 3 ′ splice site or polyadenylation site, and RNA modification. However, reads derived from the current RNA-seq technologies are usually short and deprived of information on modification, compromising their potential in defining transcriptome complexity. Here, we applied a direct RNA sequencing method with ultralong reads using Oxford Nanopore Technologies to study the transcriptome complexity in Caenorhabditis elegans. We generated approximately six million reads using native poly(A)-tailed mRNAs from three developmental stages, with average read lengths ranging from 900 to 1100 nt. Around half of the reads represent full-length transcripts. To utilize the full-length transcripts in defining transcriptome complexity, we devised a method to classify the long reads as the same as existing transcripts or as a novel transcript using sequence mapping tracks rather than existing intron/exon structures, which allowed us to identify roughly 57,000 novel isoforms and recover at least 26,000 out of the 33,500 existing isoforms. The sets of genes with differential expression versus differential isoform usage over development are largely different, implying a fine-tuned regulation at isoform level. We also observed an unexpected increase in putative RNA modification in all bases in the coding region relative to the UTR, suggesting their possible roles in translation. The RNA reads and the method for read classification are expected to deliver new insights into RNA processing and modification and their underlying biology in the future.

Nanopore direct RNA sequencing detects differential expression between human cell populations

“…dRNA doesn't utilise any amplification or fragmentation steps and has the potential to quantify both genes and isoforms in an unbiased manner, while also characterising the RNA modifications and polyA tail on each RNA. Studies to date have used dRNA sequencing to catalogue known and novel transcripts in yeast (Jenjaroenpun et al 2018), C.elegans (Roach et al 2020;Li et al 2020), Arabidopsis (Zhang et al 2020) and human cell lines (Workman et al 2019;Soneson et al 2019); characterise polyA tail lengths of individual transcripts (Roach et al 2020;Workman et al 2019); identify allele specific gene and isoform expression (Workman et al 2019); identify RNA base modifications (Workman et al 2019;Garalde et al 2018;Liu et al 2019;Lorenz et al 2019) and infer RNA structure (Stephenson et al 2020). Maximum read lengths for dRNA were also much longer than for PCR-based long-read cDNA sequencing (Workman et al 2019), demonstrating its potential to sequence long and complex RNA splice isoforms.…”

“…The unbiased nature of dRNA sequencing should allow accurate quantification and has previously shown good performance on synthetic control RNAs of known abundance (Garalde et al 2018;Sessegolo et al 2019). However, the ability of dRNA sequencing to identify differential gene and isoform expression has largely been performed on model organisms with much simpler transcriptomes than found in mammals and/or based on expression fold changes without statistical analysis (Jenjaroenpun et al 2018;Li et al 2020;Zhang et al 2020). Therefore, it remains essential to establish the effectiveness of dRNA sequencing to identify differential expression in complex organisms.…”

Alternative splicing of COQ-2 determines the choice between ubiquinone and rhodoquinone biosynthesis in helminths

“…112 We identified two distinct splice forms of C. elegans coq-2: coq-2a and coq-2e. These are annotated in the 113 genome and confirmed by RNA-seq, by nanopore sequencing, and by targeted validation studies (Ramani et al, 114 2011, Kuroyanagi, Takei & Suzuki, 2014, Roach et al, 2020, Li et al, 2020. These two isoforms differ by the 115 mutually exclusive splicing of two internal exons (6a and 6e), both of 134 nucleotides (see Figure 2A).…”

The long and the short of it: unlocking nanopore long-read RNA sequencing data with short-read tools

“…Reference-free methods, such as FLAIR (6), maps reads to the reference genome, clusters alignments into groups and collapses them into isoforms. Differential transcript usage (DTU) is another transcript-level analysis that is of great interest (6)(7)(8). DTU analyses examine differences in the relative proportions of expressed isoforms between two conditions.…”

“…Soneson et al (13) concluded that read coverage in native RNA libraries (∼0.5 million aligned reads per flow cell) were too low for gene-level analyses, resulting in low power and high variability. Li et al (7) worked around this by simply using fold-changes to identify differentially expressed genes for three ONT MinION direct RNA Caenorhabditis elegans samples, however the lack of statistical testing could lead to unreliable results. Jenjaroenpun et al (14) used DE-Seq2 (15) to perform differential expression analysis on direct RNA transcript-level counts, but gene-level expression was not studied.…”