Duplicate gene retention is partly repeatable across independent whole-genome duplications in plants, fungi and animals

Abstract

Whole-genome duplication (WGD) has recurred throughout eukaryotic evolution. After each event most duplicated gene pairs (ohnologs) have returned to single copy, but a non-random subset has been retained. Whether the same ancestral genes are retained again after independent WGDs, and why, has not been quantified gene by gene. Using hierarchical orthologous groups, we follow ancestral genes through seven independent WGDs in plants, fungi and animals (42 descendant genomes). A gene retained after one reference angiosperm WGD is 1.67–1.73 times as likely to be retained after another. A liability-threshold model attributes 29% (95% CI 25–36%) of gene-level variation in retention propensity to a component shared across WGDs, and 76 gene features account for 33–40% of the association between WGDs. In prespecified tests with predictions hash-recorded before outcomes were examined, a score from three angiosperm WGDs predicts retention after the apple–pear and budding-yeast WGDs (Spearman ρ = 0.169 and 0.280; P < 0.001), whereas the teleost test does not meet its criterion. Comparing each gene across genomes descended from one WGD, holding the gene fixed, shows that yeast genomes lose physical complex partners together (z = 8.1), as dosage balance predicts, whereas plant and fish genomes lose neighbouring genes together, a lineage-specific process that does not recur between events. Sequence constraint measured before duplication accounts for 30% of the plant-to-yeast transfer. Thus a gene-intrinsic propensity linked to dosage balance and constraint recurs across kingdoms, whereas deletions and lineage history decide where losses fall.