IRFinder: assessing the impact of intron retention on mammalian gene expression.

Middleton R, Gao D, Thomas A, Singh B, Au A, Wong JJ, Bomane A, Cosson B, Eyras E, Rasko JE, Ritchie W. · 2017

Read IRFinder: assessing the impact of intron retention on mammalian gene expression. by Middleton R, Gao D, Thomas A, Singh B, Au A, Wong JJ, Bomane A, Cosson B, Eyras E, Rasko JE, Ritchie W. (2017) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5353968.txt.

Introduction

Robert Middleton and co-authors built a tool to study kept introns in gene messages, in Genome Biology in 2017. When a gene is read, spacer chunks called introns are cut from the draft message. At times a spacer stays in the final message, a quirk called intron retention. It is the least studied form of message editing, since kept spacers are hard to gauge with trust.

Their program, IRFinder, spots kept introns in reading data with care. Run on more than two thousand and five hundred samples, it found the quirk in all tissues tested, touching most coding genes. Kept spacers cluster in the same genes and carry set binding sites for RNA guide proteins. The quirk tracks cell maturing and the cell cycle. Messages that keep a spacer tend to make less protein, and those that dodge the cell's waste watch still sit idle, untranslated.

Catalog record

CREATORMiddleton R, Gao D, Thomas A, Singh B, Au A, Wong JJ, Bomane A, Cosson B, Eyras E, Rasko JE, Ritchie W.
YEAR2017
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC5353968

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

IRFinder: assessing the impact of intron retention on mammalian gene expression By Robert Middleton, Dadi Gao, Aubin Thomas, Babita Singh, Amy Au, Justin J-L Wong, Alexandra Bomane, Bertrand Cosson, Eduardo Eyras, John E. J. Rasko, William Ritchie Published in: Genome Biology DOI: 10.1186/s13059-017-1184-4 ABSTRACT Intron retention (IR) occurs when an intron is transcribed into pre-mRNA and remains in the final mRNA. We have developed a program and database called IRFinder to accurately detect IR from mRNA sequencing data. Analysis of 2573 samples showed that IR occurs in all tissues analyzed, affects over 80% of all coding genes and is associated with cell differentiation and the cell cycle. Frequently retained introns are enriched for specific RNA binding protein sites and are often retained in clusters in the same gene. IR is associated with lower protein levels and intron-retaining transcripts that escape nonsense-mediated decay are not actively translated. Electronic supplementary material The online version of this article (doi:10.1186/s13059-017-1184-4) contains supplementary material, which is available to authorized users. BACKGROUND Alternative splicing (AS) affects up to 95% of multi-exonic genes in humans [1]. The three main types of AS are exon skipping, alternative 5′ or 3′ usage and intron retention (IR). IR occurs when an intron is transcribed into pre-mRNA and remains in the final mRNA. It constitutes a class of AS that is often neglected because these events are difficult to measure reliably. IR can introduce functional elements within mRNAs [2] or alternatively may lead to the introduction of premature termination codons, resulting in degradation of the mRNA by a surveillance mechanism called nonsense-mediated decay (NMD) [3]. This process of IR followed by NMD can downregulate up to 35% of alternatively spliced transcripts in mammals [4]. The NMD pathway is absolutely essential for post-implantation embryonic development as shown by Upf1 nullizygosity [5]. However, an obvious consequence of NMD is that mRNAs containing introns with premature termination codons are degraded and therefore difficult to quantify. Consequently, the role of IR in specific eukaryotic biological pathways has been poorly defined prior to the availability of ultra-deep sequencing technologies. We recently discovered that IR combined with the NMD pathway is not a by-product of faulty splicing but rather a major driver of the cellular differentiation of granulocytes [6]. In this pioneering study we developed an approach to correctly identify introns that are differentially retained during granulocytic differentiation from sequencing data. Our approach was subsequently used to uncover the role of IR in stem cell reprogramming where it regulated demethylation genes at specific stages of reprogramming [7]. Since then, other publications have highlighted the importance of IR in gene regulation [8], differentiation [9], and cancer [10]. Despite increasing evidence that IR can regulate hundreds of genes in numerous systems, current studies still fail to identify IR events in their transcriptomic data. Here we have developed a significantly enhanced program in terms of sensitivity and speed for detecting retained introns and filtering samples that are inappropriate for IR analysis in terms of library preparation and quality. This novel program was validated using quantitative RT-PCR (RT-qPCR) against retained introns and a NMD knockdown experiment. IRFinder correctly identified IR events and measured the ratio of retained introns to correctly spliced introns with great accuracy. Using IRFinder, we analyzed 3435 human samples, of which 2573 were suitable for analysis. We found that IR occurs in hundreds of genes in all tissues analyzed, affects over 80% of all coding genes, and is associated with cell differentiation and cell cycle. Retained introns in the same genes are frequently adjacent to each other and intron-retaining genes cluster closely on the genome, suggesting a global mechanism that regulates multiple introns simultaneously. By analyzing mass spectrometry and ribosome sequencing data we discovered that intron-retaining genes have lower protein output and that IR transcripts that escape NMD are not actively translated. Finally, by comparing introns that are frequently retained amongst the 2573 samples with introns that are rarely retained, we discovered a distinct primary sequence signature amongst frequently retained introns that is enriched in RNA binding protein sites. These proteins modulate the level of IR and thus the level of repression of frequently retained introns. Our program to analyze IR from sequencing data is available at GitHub (https://github.com/williamritchie/IRFinder) and a database of IR in over 2000 human samples is freely available at IRBase (http://mimirna.centenary.org.au/irfinder/database/). Implementation Fair measurement of intronic expression is challenged by numerous factors. Within introns, highly expressed features such as snoRNAs, microRNAs, or unannotated exons may erroneously inflate count-based measures of intronic expression. Conversely, low complexity regions, common in introns, prevent unique mapping of reads. Because retained introns are generally expressed at a fraction of their flanking exons, uncorrected biases can massively disrupt IR estimation. IRFinder implements an end-to-end analysis of retained introns from mRNA sequencing data in multiple species. It includes alignment via the STAR algorithm, quality controls on the sample analyzed, IR detection, and quantification and statistics for comparing IR levels between multiple samples. We provide standalone scripts for each of these steps so they can be used independently and provide command line tools to chain them together for complete analysis. On pre-aligned sequencing data, our program can run on a desktop computer with less than 2 GB of memory and takes approximately 10 minutes to detect IR events. Because we use STAR to align reads, our end-to-end analysis with raw reads requires at least 48 GB of memory and depends mainly on STAR runtime. Our tool facilitates the analysis of large amounts of online data by automatically testing samples for their suitability for IR detection. Unsuitable samples either have high levels of DNA contamination or have been mislabeled as mRNA sequencing when in fact they are other types of experiments such as genome sequencing or ChIP-seq. Although this information should be available from online repositories, we found that only 68% (2573/3774) were suitable for analysis in this study. This was mainly due to the incorrect use of the term “mRNA-Seq”, which was frequently used in whole RNA experiments and CHIP-Seq experiments. Our approach also uses a series of programmatically fast steps written in C++ that automatically detect and trim adapters from sequencing reads. Of the 3435 initial samples we analyzed, 3096 (91%) still had adapters in the sequencing reads despite having been already processed by an adaptor-trimming algorithm. IRFinder was capable of estimating IR events with low coverage or low mappability as confirmed by RT-qPCR (Additional file 1: Figure S1a–e). When compared with the currently available tools MISO and DEXseq, our IRFinder had higher accuracy and precision (Additional file 2: Text, Figure S2, Tables S2 and S3). The IRFinder algorithm and instruction manual are available at GitHub (https://github.com/williamritchie/IRFinder). RESULTS IRFinder detects IR in over 80% of coding genes and in all samples tested To demonstrate the functionality of our method, we downloaded 3774 human samples from the European Nucleotide Archive (http://www.ebi.ac.uk/ena). These were annotated as mRNA-Seq experiments in the archive with over 50 million reads each. The IRFinder quality control filter (see “Methods”) labeled 1201 of these as unfit for further analysis because they were either not RNA-seq experiments or not enriched for poly(A) tailed mRNAs. We searched for retained introns in 2573 remaining samples. We focused on introns that were retained

Read the complete text