Alternative polyadenylation (APA), which is regulated by both cis-elements and trans-factors, is widespread across all eukaryotic species and is recognized as a major mechanism of gene regulation. It could change the 3'UTR of an mRNA transcript affecting its stability, translation efficiency, nuclear export and mRNA or translated protein localization, or, if an exonic/intronic polyadenylation site (PAS) upstream of the stop codon is used, it could affect a gene's coding region to produce different protein isoforms with distinct properties. Accumulating evidence suggests that global APA-mediated 3'UTR length change might play an important role in oncogenic transformation, pluripotency, lymphocyte activation, neuronal stimulation and in embryonic development and differentiation. However, recent studies found limited effects of 3'UTRs in most genes compared to other regulatory elements located in 5'UTRs or coding sequence. APA as a molecular trait is a low-level phenotype in the hierarchy of biological organization, and might only exert very limited effects on organismal fitness. Therefore, some researchers proposed the “error hypothesis”, stating that most observed APA is noise and that APA diversity within and between tissues is generally neutral or deleterious, and not functional. Similarly, it has been suggested that APA divergence between species is largely non-adaptive. This scenario would be consistent with the (nearly) neutral theory of molecular evolution, which predicts that genes under relaxed selective constraints accumulate neutral (or slightly deleterious) changes at a faster rate than those under stronger purifying selection. In order to clarify the general and tissue-dependent function and regulation of APA and its evolution in mammals, we applied 3'mRNA sequencing for multiple tissues of an F1 hybrid between the C57BL/6J (Mus musculus) and SPRET/EiJ (Mus spretus) mouse strains. We analyzed the factors regulating APA diversity and addressed the question whether APA is generally non-adaptive as proposed by the error hypothesis. In this study, we quantified all annotated PASs in nine tissues of the F1 hybrid mouse and comprehensively characterized different features of single-PAS genes and multi-PAS genes. Next, we checked the positional effects on PAS strength and discussed the functional difference between rank 1 and rank 2 PASs among distinct gene groups. By quantifying PAS usage in each allele, we studied the genes with divergent major PAS expression level and dN/dS ratio difference, and unveiled different evolutionary patterns between APA patterns and gene expression (mRNA levels). We found that in general APA of multi-PAS genes is consistent with the error hypothesis, and that most APA diversity within and between tissues appears to reflect noise, resulting from molecular error due to weak cis-regulation. However, we did not find different selective constraint in dN/dS between genes with high and with low APA diversity, but found strong correlation between mRNA abundance and APA accuracy. The minor and major relative PAS usage is also affected by PAS position. In addition to most major PAS, many minor PASs appear to have functional importance. They are highly conserved and can compete with the major PASs. Last, we found a small fraction of genes exhibits strongly tissue-regulated APA patterns. In these genes, PAS usage is under intensive trans-regulation between the C57BL/6J and SPRET/EiJ alleles in the F1 hybrid mouse. Whereas many divergent PASs exist between the two alleles in genes with low expression level and under relax selective constraints, comparing these with genes showing allelic mRNA transcript level differences, we unveiled different evolutionary patterns between APA and gene expression.
