Extensive compensatory cis-trans regulation in the evolution of mouse gene expression

Abstract

Gene expression levels are thought to diverge primarily via regulatory mutations in trans within species, and in cis between species. To test this hypothesis in mammals we used RNA-sequencing to measure gene expression divergence between C57BL/6J and CAST/EiJ mouse strains and allele-specific expression in their F1 progeny. We identified 535 genes with parent-of-origin specific expression patterns, although few of these showed full allelic silencing. This suggests that the number of imprinted genes in a typical mouse somatic tissue is relatively small. In the set of nonimprinted genes, 32% showed evidence of divergent expression between the two strains. Of these, 2% could be attributed purely to variants acting in trans, while 43% were attributable only to variants acting in cis. The genes with expression divergence driven by changes in trans showed significantly higher sequence constraint than genes where the divergence was explained by variants acting in cis. The remaining genes with divergent patterns of expression (55%) were regulated by a combination of variants acting in cis and variants acting in trans. Intriguingly, the changes in expression induced by the cis and trans variants were in opposite directions more frequently than expected by chance, implying that compensatory regulation to stabilize gene expression levels is widespread. We propose that expression levels of genes regulated by this mechanism are fine-tuned by cis variants that arise following regulatory changes in trans, suggesting that many cis variants are not the primary targets of natural selection.

Figure 1.

Study design. Liver samples were collected from six adult male mice from each of four groups: C57BL/6J, CAST/EiJ, F1 initial cross hybrid of a C57BL/6J male with a CAST/EiJ female, and F1 reciprocal cross hybrid of a C57BL/6J female with a CAST/EiJ male. For each sample, the polyadenylated fraction of total RNA was sequenced on an Illumina GAIIx with 72-bp paired-end reads.

Figure 2.

Imprinted genes. After removing genes on the sex chromosomes, similar numbers of genes were found to be expressed from the maternal allele (290 genes, colored pink) and from the paternal allele (245 genes, colored blue). The average log₂ expression fold change between the two alleles in the initial cross hybrids (F1i) and between the two alleles in the reciprocal cross hybrids (F1r) is plotted on the y- and x-axis, respectively.

Figure 3.

Classification of genes according to their pattern of gene expression divergence. The average log₂ expression fold change between the alleles in the hybrids (F1) and between the parental strains (F0s) is plotted on the x- and y-axis, respectively. (A) Genes for which the expression levels have not diverged between the two strains are classified as conserved (colored black), while genes in which expression has diverged are classified as cis, trans, or cis and trans according to whether the divergence is explained by at least one regulatory variant acting in cis (colored yellow) or in trans (colored red), or by at least two regulatory variants, one in cis and one in trans (colored purple). (B) Subdivision of the cis and trans category. The regulatory variants can cause gene expression changes in the same direction with the regulatory variant in cis having a stronger effect on expression change than the regulatory variant in trans (blue), or the variant in trans having a stronger effect than the variant in cis (green). Expression changes can also be in opposite directions with the variant in cis having a stronger effect than the variant in trans (brown), or the variant in trans having a stronger effect than the variant in cis (orange).

Figure 4.

Exonic sequence conservation scores for the different classes of regulatory divergence. GERP conservation scores relative to all mammalian species in the Ensembl compara database were calculated for every exonic base. The proportion of bases above a GERP score of 1.4 in each exon was calculated for exons in each category. The mean conservation score for all exons is represented as a horizontal dashed blue line. (A) The conservation proportions for exons in the trans category are significantly higher than for genes in the cis category (P-value 9.7 × 10⁻⁷; t-test). Imprinted and conserved genes are also significantly more conserved than the cis and the cis and trans categories (Supplemental Table 10). (B) The cis and trans category is subdivided into four subcategories: cis and trans in the same direction with cis stronger (CIS + trans), cis and trans in the same direction with trans stronger (cis + TRANS), cis and trans in opposite directions with cis stronger (CIS − trans), and cis and trans in opposite directions with trans stronger (cis − TRANS). As in A, for the two categories where the cis and trans regulatory variants act in concert, the set of exons from genes for which the trans effect is stronger also show higher conservation than the set for which the cis effect is stronger (P-value 6.8 × 10⁻⁹; t-test). Supplemental Figures 10 and 11 show that the results do not change when different GERP conservation thresholds are used, or when promoter regions are considered.