Promoter architecture links gene duplication with transcriptional divergence

Summary Gene duplication is thought to be a central mechanism in evolution to gain new functions, but gene families vary greatly in their rates of gene duplication and long-term retention. Here, we discover a link between the promoter architecture of vertebrate genes and their rate of duplication: Genes that harbor CpG Islands in their promoters (CGI genes) – nearly 60% of our genes – have rarely duplicated in recent evolutionary times, and most CGI gene duplication events predate the emergence of CGI as a major regulatory element of vertebrate genes. In contrast, CGI-less genes predominate duplications that have occurred since the divergence of vertebrates. Furthermore, CGI-less paralogs are transcriptionally more divergent than CGI paralogs, even when comparing CGI and CGI-less paralogs that have duplicated at similar evolutionary times – suggesting greater capacity of CGI-less promoters to enable divergence in expression. This higher divergence between CGI-less paralogs is also reflected in lower similarity of transcription factors that bind to the promoters of CGI-less paralog pairs in comparison with CGI paralogs. Finally, CGI-less paralogs have a greater tendency to sub- and neo-functionalize, and they transcriptionally diversify faster following duplication. Our results highlight the links between promoter architecture, gene expression plasticity and their impact on gene expansion, and unravel an unappreciated role of CGI elements in shaping genome evolution. Significance statement Gene duplication generates extra gene copies, providing material for evolution of new functions. However, many duplicated genes are eliminated due to functional and regulatory constraints. The evolutionary processes that govern the elimination and persistence of duplicated genes are not well understood. Here, we focus on CpG Islands (CGIs) – important elements that occur in the majority of gene promoters. We show that genes with CGIs in their promoters have duplicated almost exclusively in ancient times, and nearly all recent duplications involve CGI-less genes. Furthermore, duplicated CGI-less genes diverge more in expression and display more distinctive transcription and cis-regulation compared to duplicated CGI-genes. Our results demonstrate how promoter structure influences transcriptional evolvability and, in turn, the retention of new genes.