Analysis of intronic and exonic reads in RNA-seq data characterizes transcriptional and post-transcriptional regulation

Abstract

A n A ly s i s RNA-seq experiments generate reads derived not only from mature RNA transcripts but also from pre-mRNA. Here we present a computational approach called exon-intron split analysis (EISA) that measures changes in mature RNA and pre-mRNA reads across different experimental conditions to quantify transcriptional and post-transcriptional regulation of gene expression. We apply EISA to 17 diverse data sets to show that most intronic reads arise from nuclear RNA and changes in intronic read counts accurately predict changes in transcriptional activity. Furthermore, changes in posttranscriptional regulation can be predicted from differences between exonic and intronic changes. EISA reveals both transcriptional and post-transcriptional contributions to expression changes, increasing the amount of information that can be gained from RNA-seq data sets.Cellular RNAs are regulated at multiple stages, including transcription, RNA maturation and degradation. Several analytic methods have been developed to measure these processes on a transcriptomewide scale. For example, global run-on sequencing (GRO-seq) 1 uses incorporation of a nucleotide analog to enrich for nascent RNA. In Nascent-seq 2,3 , newly transcribed RNAs are isolated by purification of their complex with proteins and the DNA template. Cellular fractionation techniques 4 have also been adapted to measure nascent transcripts, which are enriched in the nucleus. mRNA half-lives have been determined, for example, by blockage of transcription followed by transcriptional profiling 5 . RNA sequencing, the most widely used method for transcriptome analysis, has been applied in numerous studies to determine steady-state mRNA levels 6,7 and alternative splicing events 8 and to identify previously unknown transcripts and noncoding RNAs 9-11 . In general, these protocols aim to enrich for mature mRNA by selection of polyadenylated RNA or by depletion of ribosomal RNA.Many computational methods (reviewed in refs. 12,13) have been developed for the analysis of RNA-seq data, to enable spliced alignment 14,15 , transcript assembly 16,17 , transcript quantification 14,18 and differential expression analysis [19][20][21] . Although RNA-seq mostly generates reads that map to exons, it also captures less abundant intronic sequences 6 . However, their interpretation has remained controversial. Some have suggested that they originate from DNA contamination and can thus be used as a quality metric for RNA-seq data 22 (see also RNA-seq guidelines of the Roadmap Epigenomics Consortium, http://www.roadmapepigenomics.org/), whereas others have hypothesized that they stem from unknown exons or intronic enhancers 6,7 . In a study based on exon arrays, probes mapping to introns were used to investigate pre-mRNA dynamics 23 . Three recent studies based on RNA-seq provided evidence that intronic reads might correlate with transcriptional activity. In two of these, the read coverage along introns was related to nascent transcription in combination with co-transcriptional splicing e...

DRUID: a pipeline for transcriptome-wide measurements of mRNA stability

“…Although we initially envisioned using exogenous spike-ins for a normalization approach, we were surprised that this framework was surpassed by intron normalization. Despite using oligo(dT) selection to generate our libraries, we found that introns were abundant in our datasets; these reads are possibly derived from processing intermediates and are consistent with previous observations (Gaidatzis et al 2015). We found that intron-based normalization was effective for all datasets we examined, irrespective of the organism examined, even for datasets that were otherwise recalcitrant.…”

“…Consistent with previous observations (Gaidatzis et al 2015), we noted that introns were abundant in our libraries, especially at the earliest time points, where they made up 19% of the total reads ( Figure 3A). As with the Drosophila spike-ins, the overall proportion of reads mapping to introns exhibited a time-dependent decrease ( Figure 3A), and the relative abundance of individual introns did not show a large time-dependent decrease in similarity ( Figure S2C; r s = 0.90 to 0.95), indicating that equilibrium levels were generally reached before the first time point.…”

Circadian and feeding rhythms differentially affect rhythmic mRNA transcription and translation in mouse liver

“…1A). For RNA-Seq, intronic reads provide a good proxy for transcription rate and exonic reads for mRNA abundance (17). Owing to the large number of samples, we developed a simplified and faster protocol for ribosome profiling library generation, similar to that in ref.…”

“…6G). We took advantage of the RNA-Seq technology to decipher the role of transcriptional and posttranscriptional regulations (17) in the establishment of the diurnally rhythmic transcriptome in mouse liver. We concluded that transcription is the main regulator of rhythmic mRNA accumulation (66% of rhythmic mRNA are rhythmically transcribed) (Fig.…”

Isolation of the protein and RNA content of active sites of transcription from mammalian cells

“…The Integrative Genomics Viewer (www.broadinstitute.org/igv) was used to visualize genomic coverage, transcripts were assembled using Cufflinks 31 and annotated via the Genomatix software suite (Genomatix). It should be noted here that there now exist dedicated in silico pipelines 32,33 for assessing changes in gene expression based on intronic RNA levels (instead of exonic levels that standard algorithms use), and we anticipate that their implementation will further enhance the sensitivity of factory RNA-seq.…”