Evolutionary conservation and divergence of human brain co-expression networks

Abstract

Mouse models have allowed for the direct interrogation of genetic effects on molecular, physiological and behavioral brain phenotypes. However, it is unknown to what extent neurological or psychiatric traits may be human or primate-specific and therefore, which components can be faithfully recapitulated in mouse models. We identify robust co-expression modules reflecting whole brain and regional patterns of gene expression and compare conservation of co-expression in 116 independent data sets derived from human, mouse and non-human primate representing more than 15,000 total samples. We observe greater co-expression changes occurring on the human lineage than mouse, and substantial regional variation that highlights cerebral cortex as the most diverged region. Cell type specific modules are the most divergent across the brain, compared with those that represent basic metabolic processes. Among these, glia are the most divergent, three times that of neurons. We show that regulatory sequence divergence explains a significant fraction of co-expression divergence. Similarly, protein coding sequence constraint parallels co-expression conservation, such that genes with loss of function intolerance are enriched in neuronal, rather than glial modules. We also identify dozens of human disease risk genes, such as COMT, PSEN-1, LRRK2, and SNCA, with highly divergent co-expression between mouse and primates or human. We show that 3D human brain organoids recapitulate in vivo co-expression modules representing several human cell types, which along with our analysis of human-mouse disease gene divergence, serve as a foundational resource to guide disease modeling and its interpretation.