Comparative RNA-Seq analysis reveals pervasive tissue-specific alternative polyadenylation in Caenorhabditis elegans intestine and muscles

Abstract

Background: Tissue-specific RNA plasticity broadly impacts the development, tissue identity and adaptability of all organisms, but changes in composition, expression levels and its impact on gene regulation in different somatic tissues are largely unknown. Here we developed a new method, polyA-tagging and sequencing (PAT-Seq) to isolate high-quality tissue-specific mRNA from Caenorhabditis elegans intestine, pharynx and body muscle tissues and study changes in their tissue-specific transcriptomes and 3'UTRomes.

Results: We have identified thousands of novel genes and isoforms differentially expressed between these three tissues. The intestine transcriptome is expansive, expressing over 30% of C. elegans mRNAs, while muscle transcriptomes are smaller but contain characteristic unique gene signatures. Active promoter regions in all three tissues reveal both known and novel enriched tissue-specific elements, along with putative transcription factors, suggesting novel tissue-specific modes of transcription initiation. We have precisely mapped approximately 20,000 tissue-specific polyadenylation sites and discovered that about 30% of transcripts in somatic cells use alternative polyadenylation in a tissue-specific manner, with their 3'UTR isoforms significantly enriched with microRNA targets.

Conclusions: For the first time, PAT-Seq allowed us to directly study tissue specific gene expression changes in an in vivo setting and compare these changes between three somatic tissues from the same organism at single-base resolution within the same experiment. We pinpoint precise tissue-specific transcriptome rearrangements and for the first time link tissue-specific alternative polyadenylation to miRNA regulation, suggesting novel and unexplored tissue-specific post-transcriptional regulatory networks in somatic cells.

Figure 1

PAT-Seq of intestine and muscles tissues. (A) PAT-Seq uses Gateway-compatible (GW) entry vectors expressing the PolyA-Pull cassette in each tissue using tissue-specific (TS) promoters. (1) PolyA-Pull expressed in the intestine (ges-1 promoter), pharynx (myo-2), and body muscle (myo-3). (2) Expression of PolyA-Pull produces a 3 × FLAG-tag (light blue) fused to PAB-1 (blue), which specifically binds to the poly(A) tails of mRNAs (TS mRNAs). The complex is immunoprecipitated using α-FLAG beads. (3) Tissue-specific cDNA libraries are sequenced and mapped onto the WS190 gene model. (B) Detection of stable integration of the PolyA-Pull cassette. Left panel: Using PCR we detected genomic integration of the common portion of the PolyA-Pull cassette (2.6 kb band, red asterisk) in each tissue. The negative control, myo-2Δpab-1, was also integrated. Right panel: Western blotting using α-FLAG antibodies detected the in-frame PolyA-Pull fusion protein in lysates from transgenic worms expressing it in the pharynx (myo-2::pab-1) but not in lysate from wild type N2 worms. (C) Quantification of the specificity and sensitivity of the pull-down using RT-PCR: Left panel: myo-2 (lane 1) (*), ges-1 (lane 3) (**) and dpy-7 (lane 4) transcripts were detected in total RNA extracted from wild type N2 worms. Middle panel: Using immunoprecipitation, we successfully detected the presence of the muscle-specific gene myo-2 (lane 5) (*) and the exogenous unc-54 3’UTR (lane 6), but not the intestine-specific ges-1 (lane 7) (**) and the hypodermis-specific dpy-7 (lane 8). These transgenic worms expressed PolyA-Pull cassette in the pharynx, but not in our negative control in wild type N2 worms (lanes 9-12). The primers used to detect unc-54 3’UTR also detected 18S rRNAs (lane 2). This band was replaced with two unc-54 3’UTR isoforms (lane 6), suggesting that PolyA-Pull enriched for polyA+ RNAs. Right panel: We are unable to isolate tissue-specific RNA from worms lacking pab-1 (Δpab-1).

Figure 2

Distribution of tissue -specific gene expression and alternative polyadenylation in intestine, pharynx and body muscle. Top panel: Tissue-specific genes identified by PAT-Seq and the distribution of their expression levels between each tissue. A large pool of 4,091 genes is uniquely expressed in the intestine, while a smaller portion of 312 and 329 genes is expressed uniquely in the pharynx and body muscle, respectively. We have detected a common set of 1,556 genes expressed in all three tissues. Edges represent the presence of transcripts in each tissue, and color-coding indicates expression levels of genes in tissues (legend). Bottom panel: The 1,556 genes shared in all three tissues were further sorted based on the 3’UTR isoform and their expression levels. Approximately 25% to 30% of these genes use common 3’ends in these three tissues, while the remaining 70% use tissue-specific 3’UTR isoforms.

Figure 3

3 ’UTR poly- A site mapping in tissue datasets. We used the raw sequencing reads to map high-quality polyA sites onto the WS190 worm annotation and compared our results with two published C. elegans 3’UTRome datasets. (A) The number of polyA clusters mapped from polyA-containing sequencing reads (total), the portion of those that mapped to the WS190 worm genome annotation (mapped), the number of genes with polyA sites mapped (closest gene to the polyA cluster) and the number of isoforms resulting from distinct mapping of polyA clusters (isoforms). (B) The majority of mapped 3’UTR isoforms are supported by two published 3’UTRomes and almost 90% of them are supported by at least one dataset. Left panel: the percentage of isoforms mapped to either of two published 3’UTRomes (green and red), to both (blue), and those not present in either 3’UTRome dataset (purple). Right panel: the distribution of 3’UTR length for all 3’UTR isoforms found in each tissue dataset, along with the median (vertical dashed red line) and the average length.

Figure 4

Abundance of APA in C . elegans tissues. A finalized list of genes with mapped 3’UTR isoforms was generated for each tissue and used to compare the abundance of 3’UTR isoforms between tissues. (A) Proportion of genes subject to alternative polyadenylation in each tissue. The intestine expressed significantly more genes containing more than one 3’UTR isoform, while the muscle tissues expressed similar proportions of genes with more than one 3’UTR isoform. (B) The average number of 3’UTR isoforms detected for each gene/tissue. The number of genes and isoforms (frequency) are displayed in each column (left x-axis). We calculated and displayed the change in 3’UTR isoform to gene ratio (right x-axis) between each tissue (green trend line). We detected slightly more APA in the intestine and pharynx, when compared with the body muscle tissue.

Figure 5

APA is pervasive between C . elegans tissues. We have followed 3’UTR length changes in genes with only one 3’UTR isoform between intestine, pharynx and body muscle tissues. Length comparison between the same genes expressed between (A) intestine and pharynx, (B) pharynx and body muscle and (C) intestine and body muscle tissues. Shaded circles represent those genes expressed with proximal 3’UTR isoforms in the intestine (black), pharynx (red) or body muscle (blue), where the distal isoform was detected in the other corresponding tissue in each graph. Genes with 3’UTR isoforms that were the same length between each tissue are represented in grey as noted in the legend. (D) Distribution of unique 3’UTR isoforms for genes detected in all three tissues. The majority of these 3’UTRs are common in all three tissues (blue). Genes with a 3’UTR isoform in the intestine distinct from muscle tissues are also abundant (muscle shared). Only 2% of these genes express different 3’UTR isoforms between all three tissues (distinct).

Figure 6

Analysis of tissue-specific 3’UTR isoforms. We calculated the proportions of genes in each tissue that have tissue-specific 3’UTR isoforms and how many of these 3’UTRs have predicted microRNA targets. (A) Charts displaying the proportion of genes containing tissue-specific 3’UTR isoforms (blue). The intestine expresses approximately two times as many tissue-specific 3’UTR isoforms as muscle tissues. (B) We compared the proportion of microRNA targeted genes with tissue-specific 3’UTRs (blue) to the same number of randomly selected genes (grey) in each tissue. Significantly more genes with tissue-specific 3’UTR isoforms have microRNA targets. microRNA targets were predicted using PicTar Software, using three species and five species conservation criteria, and from ALG-1 pull-down experiments (Zisoulis et al. 2010 [32]). *P-value <0.05, **P < 0.01, based on two-tailed Student’s t-test.