Getting the Entire Message: Progress in Isoform Sequencing

Abstract

The advent of second-generation sequencing and its application to RNA sequencing have revolutionized the field of genomics by allowing quantification of gene expression, as well as the definition of transcription start/end sites, exons, splice sites and RNA editing sites. However, due to the sequencing of fragments of cDNAs, these methods have not given a reliable picture of complete RNA isoforms. Third-generation sequencing has filled this gap and allows end-to-end sequencing of entire RNA/cDNA molecules. This approach to transcriptomics has been a "niche" technology for a couple of years but now is becoming mainstream with many different applications. Here, we review the background and progress made to date in this rapidly growing field. We start by reviewing the progressive realization that alternative splicing is omnipresent. We then focus on long-noncoding RNA isoforms and the distinct combination patterns of exons in noncoding and coding genes. We consider the implications of the recent technologies of direct RNA sequencing and single-cell isoform RNA sequencing. Finally, we discuss the parameters that define the success of long-read RNA sequencing experiments and strategies commonly used to make the most of such data.

Keywords: RNA; epitranscriptome; isoforms; long-read; splicing.

Figure 1

Progress in isoform sequencing. Timeline highlights some of the key milestones in the history of isoform sequencing, dating back to the advent of short-read RNA-seq back in 2008. Note that this is presented as a summary only and is not intended to be exhaustive of all work done in the field. RNA-seq: RNA sequencing; PacBio: Pacific Biosciences; SLR: synthetic long-read; lncRNA: long noncoding RNA; ONT: Oxford Nanopore Technologies.

Figure 2

Resolution of alternative splicing events with long-read sequencing. (A) Schematic illustration of the structure of a hypothetical gene undergoing alternative splicing. The gene contains two alternatively spliced exons (red and blue) separated by constitutive exons (gray). In theory, if we let ‘n’ equal the number of alternative exons, then there are 2 ⁿ different combinations of these exons. (B) Under random pairing, we would expect to see all of these 2 ⁿ combinations, each at a relative abundance of 1/2 ⁿ. In this case, short-read RNA-seq would be sufficient, as it can accurately quantify percent spliced-in (PSI) scores for individual exons. (C, D) However, coordinated exon pairing can result in a situation whereby the alternative exons are mutually exclusive (C) or mutually associated (D). (E) With short-read RNA-seq, these three scenarios are indistinguishable, as the information regarding the connectivity of the alternative exons is lost. Conversely, with long-read sequencing, it is trivial to determine which scenario is present.