Regulatory Divergence of Transcript Isoforms in a Mammalian Model System

Abstract

Phenotypic differences between species are driven by changes in gene expression and, by extension, by modifications in the regulation of the transcriptome. Investigation of mammalian transcriptome divergence has been restricted to analysis of bulk gene expression levels and gene-internal splicing. Using allele-specific expression analysis in inter-strain hybrids of Mus musculus, we determined the contribution of multiple cellular regulatory systems to transcriptome divergence, including: alternative promoter usage, transcription start site selection, cassette exon usage, alternative last exon usage, and alternative polyadenylation site choice. Between mouse strains, a fifth of genes have variations in isoform usage that contribute to transcriptomic changes, half of which alter encoded amino acid sequence. Virtually all divergence in isoform usage altered the post-transcriptional regulatory instructions in gene UTRs. Furthermore, most genes with isoform differences between strains contain changes originating from multiple regulatory systems. This result indicates widespread cross-talk and coordination exists among different regulatory systems. Overall, isoform usage diverges in parallel with and independently to gene expression evolution, and the cis and trans regulatory contribution to each differs significantly.

Fig 1. Divergent Isoform Usage (DIU) between closely related mouse subspecies.

(A) Experiments interrogated DIU by comparing the parental F0 strains, and both directional crosses of F1 mice. Illustrative examples are shown of purely cis and trans driven divergence of isoforms. (B) Divergence of transcript expression between liver transcriptomes of male BL6 and CAST mice. Each point is one gene expressing two transcripts: the x-axis is the proportion of total gene expression in F0 BL6 which is derived from one transcript; the y-axis is the proportion of total gene expression in F0 CAST which arises from the same transcript. (C) Histogram of the number of genes (y-axis) binned by the number of expressed transcripts observed in male mouse liver (x-axis). Genes expressing only two transcripts were studied (black bar) to detect divergent isoform usage (DIU). Venn diagram callout shows the overlap of genes expressing exactly two transcripts and levels of Divergent Gene Expression (DGE) in the same sample set [28].

Fig 2. Divergent Isoform Usage of a single gene most often involves mechanistic contributions from multiple regulatory systems.

Genes with differential isoform usage were categorized according to the differences in transcript structure between the two expressed isoforms: Alternative First Exon (AFE), Transcription Start Site (TSS), Internal Splicing (INT), Alternative Last Exon (ALE), and Alternative Poly-Adenylation (APA). All five categories of structural change are illustrated on the left, and the number of isoform pairs with each combination of structural differences is shown by columns (black indicates the presence of the structural change, white indicates the absence). For example, a gene expressing two isoforms which differ by both alternative first exon and alternative polyadenylation site usage has both AFE and APA and therefore is counted in the 5^th column from left, and in total there were 44 genes like this. The summary statistic at the bottom of each grey panels indicates the number of genes with any combination of 3 structural changes, 2 structural changes or only a single change.

Fig 3. Divergent Isoform Usage is caused equally by regulatory changes in cis and in trans.

Genes were classified according to the mechanism underlying their DIU: conserved, regulatory changes in cis, in trans, in cis & in trans, or genes where no model was significantly favored over the rest (A) Scatterplot shows each heterozygous gene expressing exactly two isoforms in liver, plotting the mean fold change in the ratio of CAST to BL6 transcript expression in the F0 (F0 BL6 v F0 CAST) against the F1 (BL6 allele in F1 v CAST allele in F1), weighted by the inverse of the estimate variances (B) The relative contribution of cis and trans mechanisms towards transcriptome changes differed significantly between divergent isoform usage (DIU) and divergent gene expression (DGE) in the same sample set [28]. (C) Divergent isoform usage is encoded in cis to the Commd5 gene. In F1 hybrid offspring, the BL6 allele expresses a single transcript (Commd5-001, black) and the CAST allele expresses two transcripts (Commd5-001, white, and Commd5-002, hatched). Commd5-001 and Commd5-002 utilise different transcription start sites (>), alternative internal splicing, and discrete polyadenylation sites (<). SNV between BL6 and CAST are indicated in red. * Indicates rs32416751, predicted to disrupt the 5’ splice site in Commd5.

Fig 4. Allele-specific isoform divergence was validated pyrosequencing in the F1.

The contribution of the BL6 allele to gene and transcript expression in the F1 hybrids was validated by site-specific pyrosequencing. For each of the eight genes tested, two independent SNVs were assayed: one SNV measured the contribution of the BL6 allele to total gene expression (S), the other assayed the BL6 contribution to one of the two expressed transcripts (U). The pyrosequencing results measuring BL6 contribution to total gene expression (S) and to transcript 2 only (U) should both be in agreement with the RNASeq/MMSeq Expression Estimates. Good agreement was observed for 7 of the 8 genes (Rcn1, Ptpna, Zfyve21, Ascc2, Zfp691, Rpa1, Fam149a).