Abstract
Understanding the evolution of the free-living, cyanobacterial, diazotroph Trichodesmium is of great importance due to its important role in oceanic biogeochemistry and primary production. Unlike the other >150 available genomes of free-living cyanobacteria, only 63.8% of the Trichodesmium erythraeum IMS 101 (IMS101) genome is predicted to encode protein, which is 20-25% less than the average for other cyanobacteria and non-pathogenic, free-living bacteria. We use both distinctive isolates and metagenomic data to show that low coding density observed in IMS101 is a common feature of the Trichodesmium genus both in culture and in situ. Transcriptome analysis indicates that 86% of the non-coding space is expressed, although the function of these transcripts is unclear. When normalized per intergenic kilobase, a similar density of noncoding, possibly regulatory elements were predicted in Trichodesmium as in the more gene dense genomes of the sympatric cyanobacteria Synechococcus and about twice as dense as in Prochlorococcus. Conserved Trichodesmium ncRNA secondary structures were predicted in both culture and metagenomic sequences lending support to the predicted structural conservation. Conservation of these intergenic regions in spatiotemporally separated Trichodesmium populations suggests possible genus-wide selection for large intergenic spacers. These large intergenic spacers may have developed during intervals of strong genetic drift caused by periodic blooms of a subset of genotypes, which may have reduced effective population size. These data suggest that transposition of selfish DNA, low effective population size, and high fidelity replication allowed the unusual 'inflation' of noncoding sequence observed in Trichodesmium despite its oligotrophic lifestyle.