A comprehensive analysis of 3' end sequencing data sets reveals novel polyadenylation signals and the repressive role of heterogeneous ribonucleoprotein C on cleavage and polyadenylation

Abstract

Alternative polyadenylation (APA) is a general mechanism of transcript diversification in mammals, which has been recently linked to proliferative states and cancer. Different 3' untranslated region (3' UTR) isoforms interact with different RNA-binding proteins (RBPs), which modify the stability, translation, and subcellular localization of the corresponding transcripts. Although the heterogeneity of pre-mRNA 3' end processing has been established with high-throughput approaches, the mechanisms that underlie systematic changes in 3' UTR lengths remain to be characterized. Through a uniform analysis of a large number of 3' end sequencing data sets, we have uncovered 18 signals, six of which are novel, whose positioning with respect to pre-mRNA cleavage sites indicates a role in pre-mRNA 3' end processing in both mouse and human. With 3' end sequencing we have demonstrated that the heterogeneous ribonucleoprotein C (HNRNPC), which binds the poly(U) motif whose frequency also peaks in the vicinity of polyadenylation (poly(A)) sites, has a genome-wide effect on poly(A) site usage. HNRNPC-regulated 3' UTRs are enriched in ELAV-like RBP 1 (ELAVL1) binding sites and include those of the CD47 gene, which participate in the recently discovered mechanism of 3' UTR-dependent protein localization (UDPL). Our study thus establishes an up-to-date, high-confidence catalog of 3' end processing sites and poly(A) signals, and it uncovers an important role of HNRNPC in regulating 3' end processing. It further suggests that U-rich elements mediate interactions with multiple RBPs that regulate different stages in a transcript's life cycle.

Figure 1.

Hexamers with highly specific positioning upstream of human and mouse pre-mRNA 3′ end cleavage sites. (A) The frequency profiles of the 18 hexamers that showed the positional preference expected for poly(A) signals in both human and mouse. The known poly(A) signal, AAUAAA, had the highest frequency of occurrence (left). Apart from the 12 signals previously identified (AAUAAA and motifs with the purple frame) (Beaudoing et al. 2000), we have identified six additional motifs (orange frame) whose positional preference with respect to poly(A) sites suggests that they function as poly(A) signals and are conserved between human and mouse. (B) Sequence logos based on all occurrences of the entire set of poly(A) signals from the human (left) and mouse (right) atlas. (C) The (U)₆ motif, which is also enriched upstream of pre-mRNA cleavage sites, has a broader frequency profile and peaks upstream of the poly(A) signals, which are precisely positioned 20–22 nt upstream of the pre-mRNA cleavage sites (indicated by the dashed, vertical line).

Figure 2.

siRNA-mediated knock-down of HNRNPC leads to increased use of distal poly(A) sites. (A) Relative location of sites whose usage decreased (brown), did not change (blue) or increased (red) in response to HNRNPC knock-down within 3′ UTRs. We identified the 1000 poly(A) sites whose usage increased most, the 1000 whose usage decreased most, and the 1000 whose usage changed least upon HNRNPC knock-down; divided the associated terminal exons into five bins, each covering 20% of the exon's length; and computed the fraction of poly(A) sites that corresponded to each of the three categories within each position bin independently. Values represent means and SDs from the two replicate HNRNPC knock-down experiments. (B) Smoothened (±5 nt) density of nonoverlapping (U)₅ tracts in the vicinity of sites with a consistent behavior (increased, unchanged, decreased use) in the two HNRNPC knock-down experiments. (C) Cumulative density function of the percentage change in usage of the 250 poly(A) sites with the highest number of (U)₅ motifs within ±50 nt around their cleavage site (red) and of poly(A) sites that do not contain any (U)₅ tract within ±200 nt (blue), upon HNRNPC knock-down.

Figure 3.

The length, number, and location of poly(U) tracts with respect to poly(A) sites influence the change in poly(A) site use upon HNRNPC knock-down. (A) Mean change in the use of sites containing the highest number of (U)₅ motifs within 100-nt-long regions located at specific distances from the cleavage site (indicated on the x-axis) upon HNRNPC knock-down (KD). Shown are mean ± SEM in the two knock-down experiments. Two hundred fifty poly(A) sites with the highest density of (U)₅ motifs at each particular distance were considered. (B) Mean changes in the relative use of poly(A) sites that have 0, 1, 2, or more (≥3) nonoverlapping poly(U) tracts within ±50 nt from their cleavage site. Distributions of relative changes in the usage of specific types of sites were compared, and the P-values of the corresponding one-sided Mann-Whitney U tests are shown at the top of the panel.

Figure 4.

HNRNPC-responsive 3′ UTRs are enriched in ELAVL1 binding sites. (A) Fraction of HNRNPC-responding and not-responding 3′ UTR regions that contain one or more ELAVL1 CLIP sites. The P-value of the one-sided t-test is shown. (B) Density of ELAVL1 CLIP sites per kilobase (kb) in the 3′ UTR regions described above. The P-value of the one-sided t-test is shown. (C) Model of the impact of A/U-rich elements (ARE) in 3′ UTR regions on various aspects of mRNA fate (Berkovits and Mayr 2015). (D) Density of A-seq2 reads along the CD47 3′ UTR in cells, showing the increased use of the distal poly(A) site in si-HNRNPC compared with si-Control transfected cells. The density of ELAVL1 CLIP reads in this region is also shown.

Figure 5.

The knock-down of HNRNPC affects CD47 protein localization. (A) Indirect immunophenotyping of membrane-associated CD47 in HEK 293 cells that were treated either with an si-HNRNPC (blue) or with si-Control (red) siRNA. Mean, median, and mode of the Alexa Fluor 488 intensities computed for cells in each transfection set (top), with histograms shown in the bottom panel. (B) Immunofluorescence staining of permeabilized HEK 293 cells with CD47 antibody (left) or nuclear staining with Hoechst (right). Top and bottom panels correspond to cells that were treated with control siRNA and si-HNRNPC, respectively.

Figure 6.

HNRNPC knock-down leads to increased usage of intronic poly(A) sites. (A) The change in the relative use of intronic poly(A) sites that did not contain any (U)₅ within ±200 nt and of the top 250 intronic poly(A) sites according to the number of (U)₅ motifs within ±50 nt around the cleavage site, upon HNRNPC knock-down. (B) Relative location within the gene of the top 250 most-derepressed intronic poly(A) sites that have HNRNPC binding motifs within −200 to +100 nt around their cleavage site and of the 250 intronic poly(A) sites that changed least upon HNRNPC knock-down. (C) Screenshot of the KLHL3 gene, in which intronic cleavage and polyadenylation was strongly increased upon HNRNPC knock-down.