Abstract
When a free-living bacterium transitions to a host-beneficial endosymbiotic lifestyle, it almost invariably loses a large fraction of its genome [1, 2]. The resulting small genomes often become stable in size, structure, and coding capacity [3-5], as exemplified by Sulcia muelleri, a nutritional endosymbiont of cicadas. Sulcia's partner endosymbiont, Hodgkinia cicadicola, similarly remains co-linear in some cicadas diverged by millions of years [6, 7]. But in the long-lived periodical cicada Magicicada tredecim, the Hodgkinia genome has split into dozens of tiny, gene-sparse circles that sometimes reside in distinct Hodgkinia cells [8]. Previous data suggested that all other Magicicada species harbor complex Hodgkinia populations, but the timing, number of origins, and outcomes of the splitting process were unknown. Here, by sequencing Hodgkinia metagenomes from the remaining six Magicicada and two sister species, we show that each Magicicada species harbors Hodgkinia populations of at least 20 genomic circles. We find little synteny among the 256 Hodgkinia circles analyzed except between the most closely related cicada species. Gene phylogenies show multiple Hodgkinia lineages in the common ancestor of Magicicada and its closest known relatives but that most splitting has occurred within Magicicada and has given rise to highly variable Hodgkinia gene dosages among species. These data show that Hodgkinia genome degradation has proceeded down different paths in different Magicicada species and support a model of genomic degradation that is stochastic in outcome and nonadaptive for the host. These patterns mirror the genomic instability seen in some mitochondria.