Genomic evidence unveils the genetic architecture and evolution of the S-locus controlling heterostyly in Rubiaceae

The S-locus supergene controlling heterostyly has multiple angiosperm origins and is characterized by convergent evolution in form and function. The genetic architecture of floral polymorphism has been studied in several unrelated families, but not Rubiaceae, which has the largest number of heterostylous species. We assembled genomes of the long- and short-styled morphs of distylous Mussaenda lancipetala and investigated the structure and evolution of the S-locus in three Mussaenda species to evaluate evidence for convergence in genetic architecture and gene function in unrelated families. Genome assemblies and population genomics indicated that the S-locus in M. lancipetala is comprised of three hemizygous genes - MuIAA, MuGA3ox, and MuAPs - present only in the S-morph. In each species, MuIAA, involved in the auxin response pathway, was highly expressed in pistils and floral tubes of the S-morph, and is a likely candidate gene controlling style length. Molecular evolution analysis indicated that the S-haplotype has accumulated repetitive sequences, S-linked genes showed no evidence of relaxed purifying selection, and the evolutionary assembly of the S-locus involved stepwise duplication. Our findings provide support for the prevalence of hemizygosity in S-locus genes and have identified molecular pathways underpinning convergent morphological evolution of clear adaptive significance.