Figure 1.
Hypothesis of microRNA regulation on alternative polyadenylation.
Here, we illustrate an example of our hypothesis. Left and right panels represent the situations under normal and tumor cells, respectively. Initially, the APA gene produces long and short isoforms in both normal and tumor cells. However, due to modulation of tumor-up-regulated microRNA on the aUTR, the long isoforms in tumor cells are greatly degraded (transcripts with dashed outlines). The expression ratio between long isoforms and short isoforms (LSR) is thereby lower in tumor than normal cells.
Figure 2.
Correlation among expression level, shortened fraction and LSR.
(A) Box plot for log2 ratios of mRNA expression levels between MCF-7 and MCF-10A cells of the shortened genes reported by Fu et al., 2011. Expression level is the total read count of all isoforms for each gene. (B) Box plots for log2 ratios of mRNA expression levels between MCF-7 and MCF-10A cells against shortened fraction in MCF-7 cells. For each gene, shortened fraction is calculated by dividing average shifted length in MCF-7 by weighted 3′ UTR length in MCF-10A. All data originate from Fu et al., 2011. (C) Scatter plot for mRNA expression level versus LSR between MCF-7 and MCF-10A cells of all APA genes (y-axis: log2 ratio of mRNA expression level between MCF-7 and MCF-10A cells; x-axis: log2 ratio of LSR, where the lower the value, the higher the proportion of short isoform in MCF-7 versus MCF-10A). A positive correlation can be observed. The red dashed line denotes the linear regression line.
Figure 3.
MicroRNAs are associated with LSR of their target gene sets.
(A) Cumulative density plot of LSR for target genes (black) and non-target genes (gray) of the miR-17-5p/20/93.mr/106/519.d family as an example. The KS-test D value reflects the LSR difference between target and non-target genes. A negative D value means that the overall LSR of target genes is lower. The left figure consists of the target genes with one or more target sites in aUTR, while the right figure consists of those with two or more sites. (B) Box plot for D values of the microRNAs whose target gene LSRs decrease significantly (P<0.05, KS-test, 101 out of 130 total microRNAs) (left). The D values of these microRNAs decrease further if two or more target sites are required (right).
Figure 4.
MicroRNA differential expression versus LSR change of targets and preference of target sites in aUTRs.
(A) Cumulative density plot of LSR change for target genes of each microRNA. Z-score represents an estimator of target gene LSR change compared to background (see Materials and Methods). A negative z-score means that the overall LSR is lower in MCF-7 cells. Target genes of up-regulated microRNAs (red) show a significantly decreased LSR in MCF-7 versus the down-regulated ones (green). (B) An illustration for estimating the preference of microRNA target sites to appear in aUTRs. A higher value means that the target sites are more likely to appear in the aUTR of the APA gene. (C) Cumulative density plot of the preference of microRNA target sites to appear in aUTRs. Each point represents an aUTR target gene with the preference value exemplified in (B). MCF-7-up-regulated microRNA target sites (red) exhibit a significantly higher preference than MCF-7-down-regulated microRNA target sites (green).
Figure 5.
Correlation between target site abundance of up-regulated microRNAs and LSR change in MCF-7 cells.
Box plots for log2 ratios of (A) LSR, (B) long-isoform expression, and (C) short-isoform expression change between MCF-7 and MCF-10A cells against target-site abundance of MCF-7-up-regulated microRNAs in aUTRs. The genes with more target sites encounter a greater LSR decrease and long-isoform expression decline in MCF-7 cells. However, differential expression of the short isoform shows little correlation with target site abundance.