ParD-ParE coevolution and phylogeny

ParD-ParE specificity residues, those residues showing strongest coevolution between proteins, were identified using GREMLIN (http://gremlin.bakerlab.org) using Mesorhizobium opportunistum ParD3 and ParE3 as input. We performed eight iterations with an E-value cutoff of 1 × 10⁻²⁰ and isolated residue pairings with a probability score >0.95. To construct a ParE protein tree, we first used HMMER (http://hmmer.org) to identify and align all α-proteobacteria homologs of M. opportunistum ParE2 in the ProGenomes Database. We then built a protein tree from all homologs using a trimmed alignment (positions represented in <50% of sequences were removed) in FastTree 2 (Price et al., 2010). The protein tree allowed us to categorize the resulting homologs as orthologs of ParE2, ParE3, or neither based on membership in either the ParE2 or ParE3 sister taxa. A species tree was generated for a concatenated alignment of 15 highly conserved single-copy bacterial genes (atpD, dnaA, rpsK, rpsD, ruvA, leuS, yeaZ, rpsF, rplI, radA, tsf, pyrH, yhhF, coaD, frr). HMMER was used to identify and align orthologs of these genes in α-proteobacteria species in the ProGenomes Database. The concatenated alignment was manually trimmed to remove positions represented in <50% of sequences and positions with <25% conservation, and a tree was generated using FastTree two for relevant species.