Multiple origins of endosymbiosis within the Enterobacteriaceae (γ-Proteobacteria): convergence of complex phylogenetic approaches

Abstract

Background: The bacterial family Enterobacteriaceae gave rise to a variety of symbiotic forms, from the loosely associated commensals, often designated as secondary (S) symbionts, to obligate mutualists, called primary (P) symbionts. Determination of the evolutionary processes behind this phenomenon has long been hampered by the unreliability of phylogenetic reconstructions within this group of bacteria. The main reasons have been the absence of sufficient data, the highly derived nature of the symbiont genomes and lack of appropriate phylogenetic methods. Due to the extremely aberrant nature of their DNA, the symbiotic lineages within Enterobacteriaceae form long branches and tend to cluster as a monophyletic group. This state of phylogenetic uncertainty is now improving with an increasing number of complete bacterial genomes and development of new methods. In this study, we address the monophyly versus polyphyly of enterobacterial symbionts by exploring a multigene matrix within a complex phylogenetic framework.

Results: We assembled the richest taxon sampling of Enterobacteriaceae to date (50 taxa, 69 orthologous genes with no missing data) and analyzed both nucleic and amino acid data sets using several probabilistic methods. We particularly focused on the long-branch attraction-reducing methods, such as a nucleotide and amino acid data recoding and exclusion (including our new approach and slow-fast analysis), taxa exclusion and usage of complex evolutionary models, such as nonhomogeneous model and models accounting for site-specific features of protein evolution (CAT and CAT+GTR). Our data strongly suggest independent origins of four symbiotic clusters; the first is formed by Hamiltonella and Regiella (S-symbionts) placed as a sister clade to Yersinia, the second comprises Arsenophonus and Riesia (S- and P-symbionts) as a sister clade to Proteus, the third Sodalis, Baumannia, Blochmannia and Wigglesworthia (S- and P-symbionts) as a sister or paraphyletic clade to the Pectobacterium and Dickeya clade and, finally, Buchnera species and Ishikawaella (P-symbionts) clustering with the Erwinia and Pantoea clade.

Conclusions: The results of this study confirm the efficiency of several artifact-reducing methods and strongly point towards the polyphyly of P-symbionts within Enterobacteriaceae. Interestingly, the model species of symbiotic bacteria research, Buchnera and Wigglesworthia, originated from closely related, but different, ancestors. The possible origins of intracellular symbiotic bacteria from gut-associated or pathogenic bacteria are suggested, as well as the role of facultative secondary symbionts as a source of bacteria that can gradually become obligate maternally transferred symbionts.

Figure 1

Study design. General design of the study summarizing all analyses and results. Individual topologies show the gradient of acquired results; method names are written above and below the arrows. Notice an increasing number of independent origins of symbionts and decreasing number of phylogenetic artifacts along the continuum towards the 'derived' methods. 1+2: third codon positions excluded; AT/GC(BI/4-11): AT/GC datasets 4-11 analyzed by BI; BI: Bayesian inference; Dayhoff6/Dayhoff4/HP: amino acid recoded matrices; ML: maximum likelihood; nhPhyML: ML under nonhomogeneous model; MP: maximum parsimony; RY: purine/pyrimidine recoded matrix; SF: slow-fasted datasets.

Figure 2

MrBayes phylogram - 69 genes, nucleotide matrix. Phylogenetic tree inferred from the concatenated nucleotide matrix using BI under the GTR+I+Γ model. Asterisks designate nodes with posterior probabilities equal to 1.0, values next to species names represent GC content calculated from the 69-gene dataset, genomic GC content can be found in Additional file 4. BI: Bayesian inference.

Figure 3

MrBayes phylogram - 69 genes, amino acid matrix. Phylogram inferred from the concatenated amino acid matrix using BI under the WAG+I+Γ model. Values at nodes represent posterior probabilities (WAG+I+Γ model, GTR+I+Γ protein model) and bootstrap supports from ML analysis (LG+I+Γ model). Asterisks designate nodes with posterior probabilities or bootstrap supports equal to 1.0, dashes designate values lower than 0.5 or 50, values next to species names represent GC content calculated from the 69-gene dataset, genomic GC content can be found in Additional file 4. BI: Bayesian inference. ML: maximum likelihood.

Figure 4

PhyloBayes phylogram - 14 genes, amino acid matrix. Phylogram derived from concatenation of 14 genes (AceE, ArgS, AspS, EngA, GidA, GlyS, InfB, PheT, Pgi, Pnp, RpoB, RpoC, TrmE and YidC) using PhyloBayes under the CAT+GTR model. Asterisks designate nodes with posterior probabilities equal to 1.0, values next to species names represent GC content calculated from the 69-gene dataset, genomic GC content can be found in Additional file 4.

Figure 5

PhyloBayes cladogram - 69 genes, Dayhoff6 amino acid recoded matrix. Cladogram inferred from amino acid matrix recoded with Dayhoff6 scheme using PhyloBayes with the CAT and CAT+GTR model. Because of the length of symbiotic branches, phylogram is presented only as a preview (original phylogram can be found in Additional trees on our website). Values at nodes represent posterior probabilities from CAT and CAT+GTR analyses, respectively (asterisks designate nodes with posterior probabilities equal to 1.0). Values next to species names represent GC content calculated from the 69-gene dataset, genomic GC content can be found in Additional file 4.

Figure 6

nhPhyML phylogram - 69 genes, nucleotide matrix, third positions excluded. Phylogram inferred from the concatenated nucleotide matrix without third codon positions using the nonhomogeneous model of evolution as implemented in nhPhyML. Values at nodes and branches represent GC content.