Parasitism and mutualism in Wolbachia: what the phylogenomic trees can and cannot say

Abstract

Ecological and evolutionary theories predict that parasitism and mutualism are not fixed endpoints of the symbiotic spectrum. Rather, parasitism and mutualism may be host or environment dependent, induced by the same genetic machinery, and shifted due to selection. These models presume the existence of genetic or environmental variation that can spur incipient changes in symbiotic lifestyle. However, for obligate intracellular bacteria whose genomes are highly reduced, studies specify that discrete symbiotic associations can be evolutionarily stable for hundreds of millions of years. Wolbachia is an inherited obligate, intracellular infection of invertebrates containing taxa that act broadly as both parasites in arthropods and mutualists in certain roundworms. Here, we analyze the ancestry of mutualism and parasitism in Wolbachia and the evolutionary trajectory of this variation in symbiotic lifestyle with a comprehensive, phylogenomic analysis. Contrary to previous claims, we show unequivocally that the transition in lifestyle cannot be reconstructed with current methods due to long-branch attraction (LBA) artifacts of the distant Anaplasma and Ehrlichia outgroups. Despite the use of 1) site-heterogenous phylogenomic methods that can overcome systematic error, 2) a taxonomically rich set of taxa, and 3) statistical assessments of the genes, tree topologies, and models of evolution, we conclude that the LBA artifact is serious enough to afflict past and recent claims including the root lies in the middle of the Wolbachia mutualists and parasites. We show that different inference methods yield different results and high bootstrap support did not equal phylogenetic accuracy. Recombination was rare among this taxonomically diverse data set, indicating that elevated levels of recombination in Wolbachia are restricted to specific coinfecting groups. In conclusion, we attribute the inability to root the tree to rate heterogeneity between the ingroup and outgroup. Site-heterogenous models of evolution did improve the placement of aberrant taxa in the ingroup phylogeny. Finally, in the unrooted topology, the distribution of parasitism and mutualism across the tree suggests that at least two interphylum transfers shaped the origins of nematode mutualism and arthropod parasitism. We suggest that the ancestry of mutualism and parasitism is not resolvable without more suitable outgroups or complete genome sequences from all Wolbachia supergroups.

FIG. 1.—

Phylogenomic trees showing the relationships of the Wolbachia supergroups and the placements of the root (arrow head symbol) from the Anaplasma and Ehrlichia outgroups to the Wolbachia ingroup. The taxon names of the Wolbachia ingroup denote those of the host species. Only support values below 100% are shown. (A) MP tree based upon concatenated nucleotide sequences of 21 protein-coding genes. (B) Common topology reconstructed from a Bayesian analysis of the amino acid translation of the concatenated proteins and a ML analysis of the concatenated genes. Support values above the nodes indicate Bayesian posterior probabilities/ML bootstrap values. The scale bar indicates the distances in substitutions per nucleotide.

FIG. 2.—

The unrooted ML tree showing the evolutionary relationships of the Wolbachia ingroup based on the concatenated alignment of 21 protein-coding genes and a tabular summary of an analysis of 12 different rooting positions. The numbers on the internal branches correspond to the rooting positions analyzed with the SH test. The lowest (best) −ln L scores for each data set are shown in descending order followed by the −ln L difference between the best score and the other data sets. Significance levels are based on full optimization. Taxon labels at the end of each terminal node refer to host species or to letters that correspond to the same supergroup letters in figure 1.

FIG. 3.—

Phylogenomic trees inferred from the site-heterogenous CAT mixture model. Only support values below 100% are shown. (A) PhyloBayes CAT tree based upon the concatenated 21 proteins. (B) PhyloBayes CAT tree following removal of six genes with compositionally deviant taxa. Support values above the nodes indicate Bayesian posterior probabilities. The scale bars indicate the distances in substitutions per nucleotide.

FIG. 4.—

Summary plot of Z scores for each gene's amino acid alignment in ascending order for (A) the current data set and (B) a previously published data set (Fenn et al. 2006). Protein alignments with Z scores greater than 2 have taxa that show significant compositional deviation (P value < 0.05) and are excluded in downstream analyses. Labels in each plot correspond to the respective gene labels used in each data set.

FIG. 5.—

Unrooted phylogenomic tree inferred using the site-heterogenous CAT mixture model after excluding 6 compositionally deviant protein sequences of the original 21 sequences. Support values at the nodes indicate Bayesian posterior probabilities. The scale bar indicates the distance in substitutions per nucleotide.

FIG. 6.—

Phylogenomic trees inferred from the site-heterogenous CAT mixture model following removal of genes with compositionally deviant taxa. (A) PhyloBayes CAT tree based upon 10 concatenated proteins from a previous data set (Fenn et al. 2006) that did not have compositionally deviant taxa. (B) PhyloBayes CAT tree based upon 25 combined proteins from the previous and current protein sequence data in this paper. Only Bayesian posterior probabilities values below 100% are shown.