Ancestral chromatin state constrains the functional landscape of bivalent domains in mammalian spermatogenesis.

Mammalian germ cells are enriched for bivalent chromatin, an epigenetic state defined by the dual presence of the activating H3K4me3 and repressive H3K27me3 histone modifications. Bivalency is evolutionarily conserved at developmentally important genes in germ cells but diverges at hundreds of additional loci, and evolutionary gains in bivalency have been proposed to reflect divergent somatic functions of the associated genes. Here, we sought to discover if evolutionary gains in bivalency occur selectively at genes with specific functions, and to better elucidate the role of bivalent chromatin in germ cells. By comparing genome-wide profiles for four histone modifications in spermatogenic cells of six mammalian species, we define a comprehensive set of mammalian bivalent domains and classify them based on conservation or divergence of chromatin state. We find that evolutionarily conserved bivalent regions exhibit canonical features of bivalency and maintain bivalency in embryonic stem cells. In contrast, bivalent domains emerging from a purely active or repressed ancestral chromatin state have atypical sequence and regulatory features and are frequently germ cell specific. Genes associated with these recent bivalent domains exhibit distinct somatic expression patterns that reflect their ancestral chromatin state in germ cells. Specifically, bivalent genes emerging from ancestrally active chromatin are more highly expressed in somatic tissues and are enriched for immune-related functions, while those emerging from ancestrally H3K27me3-only domains are lowly expressed in the soma and enriched for neurogenesis functions. We propose that recent bivalent regions demarcate sites of regulatory sequence change that preferentially impacts specific somatic lineages.