Bacterial classifications derived from recA protein sequence comparisons

RecA protein sequences from 62 eubacterial sources were compared with one another and relative to one archaebacterial RecA-like and a number of eukaryotic RecA-like sequences. Pairwise similarity scores were determined by a novel method based on significant segment pair alignment. The sequences of different species were grouped on the basis of mutually high similarity scores within groups and consistency of score ranges in comparison to other groups. Following this protocol, the gamma-proteobacteria can be subclassified into two major groups, those of mostly vertebrate hosts and those of mostly soil habitat. The alpha-proteobacterial sequences also divide into two distinct groups, whereas classification of the beta-proteobacteria is more complex. The gram-positive bacterial sequences split into three groups of low and three groups of high G+C genome content. However, neither the combined low-G+C-content nor the combined high-G+C-content group nor the aggregate of all gram-positive bacteria form homogeneous groups. The mycoplasma sequences score best with the Bacillus subtilis sequence, consistent with their presumed origin from a gram-positive ancestor. The eukaryotic RAD proteins generally show a single high-scoring segment pair with the proteobacterial RecA sequences around the ATP-binding domain. The bacteriophage T4 UvsX protein aligns best with RecA sequences on two segments disjoint from the ATP-binding domain. The distribution of the most highly conserved regions shared between RecA and noneubacterial RecA-like sequences suggests a mosaic character and evolution of RecA. The discussion considers some questions on the validity and consistency of bacterial classifications derived from RecA sequence comparisons.