Variable selection pressures across lineages in Trichodesmium and related cyanobacteria based on the heterocyst...

Due to the irreversible inhibition of nitrogenase by O2, N2 fixation is incompatible with the oxygenic photosynthesis of cyanobacteria. These organisms have therefore evolved various strategies for growing diazotrophically. One group of N2-fixing cyanobacteria has specialized cells, heterocysts, which contain the nitrogenase, lack the oxygenic photosystem II, and are virtually anoxic inside as the result of respiratory activity and a thick glycolipid cell wall. The hetR gene encodes a serine protease which is thought to be involved in the regulation of heterocyst development and in DNA binding. Although hetR is also present in many non-heterocystous N2-fixing cyanobacteria, its function in these organisms is unknown. In this study, hetR sequences of the N2-fixing, non-heterocystous cyanobacterium Trichodesmium spp. and related genera were examined for signatures of selection. In parsimony- or distance-based hetR phylogenies, the filamentous non-heterocystous cyanobacteria Symploca sp. and Leptolyngbya sp. were closest to Trichodesmium sp. However, accommodating molecular attributes of hetR such as nucleotide frequencies and rate heterogeneity in phylogenetic analyses suggested that many other genera could not be excluded as sister taxa of Trichodesmium. Maximum likelihood analysis of the dN/dS ratio (omega) showed that-irrespective of the use of Symploca, Leptolyngbya, or more distant taxa as an outgroup-the lineage between an outgroup and Trichodesmium (omega1=0.02-0.05) and a lineage leading to Trichodesmium erythraeum (omega1=0.02) were under much stronger purifying selection than the other lineages in Trichodesmium (omega0=0.13-0.32). Although the results from the maximum likelihood analyses are most trustworthy because of codon usage bias in Trichodesmium, the results from a simpler tree-based McDonald-Kreitman test were in general agreement. Due to their quite different assumptions, the combination of these two methods of analysis circumvents multiple testing which, in general, is problematic when using branch models. Although the causal selective forces underlying the substitution patterns in hetR have not yet been identified, these findings parallel the variety of physiological, molecular, and behavioral differences in cyanobacteria related to N2 fixation. The heterogeneity of selection pressures in Trichodesmium is more surprising, because multiple adaptation mechanisms have not been described in this genus.