Transcriptome-Wide Analysis of Polyadenylation Sites in Mantle Cell Lymphoma Using Polyadenylation Signal Sequencing,

Abstract 3478

In mantle cell lymphoma (MCL) cyclin D1 overexpression is caused by the translocation t(11;14) and mutations in the 3'UTR of cyclin D1 further enhance its expression. Notably, higher levels of the shorter cyclin D1 isoform correlate with inferior clinical outcome. Alternative polyadenylation (APA) is an important post-transcriptional mechanism that controls mRNA stability, subcellular localization, and protein translation. 3'UTR shortening has been associated with proliferation and oncogenic transformation.

We analyzed the polyadenylome in 10 MCL cell lines, in CD19+ selected tumor cells from 2 leukemic and 1 nodal MCL sample, in core biopsies from 2 nodal MCL, and in peripheral blood B cells from 2 healthy donors by polyadenylation (signal) sequencing (PA-seq), which results in short sequencing reads comprising the polyadenylation (PA) site. Reads per sample were scaled to the median of total reads in order to normalize samples for cross comparison. We called a PA site if the sum of reads in a window of −2 to +25nt located in the 3'UTR contributed ≥ 10% of total reads for the gene in one sample, and ≥ 2 samples had ≥ 5 reads at a given location. A gene was considered to have multiple PA sites if > 1 site met the minimum criteria in > 1 sample. First, we identified 13,057 PA sites representing 7,115 genes mapped to the plus strand. Of these genes 25% had 2, 14% 3 and 9% ≥ 4 PA sites, respectively. Owing to the generally repressive nature of the 3'UTR, shortening of this region may give rise to a gain of function. We therefore sought to identify genes in which tumor cells display shorter mRNA transcripts than normal B cells. We identified 217 transcripts where the most proximal PA site of a gene was used by tumor but not normal cells. These genes belong to a wide range of different functional groups. Analyses into the possible role of these genes in MCL pathogenesis and into the mechanism contributing to APA are ongoing.

3'UTR shortening of cyclin D1 transcripts in MCL as a mechanism for proto-oncogene activation has been previously reported. However, in a substantial number of cases in which short cyclin D1 transcripts were detected the mechanism of APA has remained elusive. We were therefore particularly interested in cyclin D1. Using the criteria described above we identified 3 PA sites, a proximal site previously identified in tumor cells, the consensus site, about 3.5kb distal to the first site and an additional, previously unreported site further 3' that was used in only one cell line. Five of 10 MCL cell lines and 2 of 5 primary tumor samples preferentially used the proximal PA site. Consistent with the role of the 3'UTR in destabilizing cyclin D1 mRNA, the 7 MCL samples with preferential use of the proximal site had on average 3-fold higher expression than those that preferentially used the distal site (p=0.02). The only exception is the Mino cell line that had the highest cyclin D1 expression across all samples despite preferential usage of the distal site. Moreover, we found that normal B cells also used the proximal PA site albeit with very low numbers of cyclin D1 transcripts overall. Next we compiled the sequences at the PA sites to test for mutations that may give rise to APA. The canonical AAUAAA hexamer represents the most frequent PA signal but hexamer variants have been described. Mutations in MCL and the MCL-derived cell line Z138 have been reported to create an aberrant proximal AAUAAA hexamer in cyclin D1 3'UTR. We identified the previously described A insertion in Z138 cells, but there was no mutation in any of the other samples. The absence of AAUAAA and variant sequences suggest that the pentamer sequence AAUAA may mediate APA under select circumstances. Why this site can be used by some but not other cells will be interesting to address and may reveal an additional layer of regulation.

In summary, we have begun to systematically address the role of APA as a gain of function mechanism in MCL. Hundreds of genes give rise to shorter transcripts in MCL cells than in normal B cells, and it will be interesting to investigate whether these genes contribute to MCL pathogenesis. The analysis of APA in cyclin D1 suggests that under certain circumstances, the exact conditions remain to be defined, pentameric PA sequences may be used efficiently and can give rise to a gain of function genotype.