Phylogenomic Evidence for a Myxococcal Contribution to the Mitochondrial Fatty Acid Beta-Oxidation

Background The origin of eukaryotes remains a fundamental question in evolutionary biology. Although it is clear that eukaryotic genomes are a chimeric combination of genes of eubacterial and archaebacterial ancestry, the specific ancestry of most eubacterial genes is still unknown. The growing availability of microbial genomes offers the possibility of analyzing the ancestry of eukaryotic genomes and testing previous hypotheses on their origins. Methodology/Principal Findings Here, we have applied a phylogenomic analysis to investigate a possible contribution of the Myxococcales to the first eukaryotes. We conducted a conservative pipeline with homologous sequence searches against a genomic sampling of 40 eukaryotic and 357 prokaryotic genomes. The phylogenetic reconstruction showed that several eukaryotic proteins traced to Myxococcales. Most of these proteins were associated with mitochondrial lipid intermediate pathways, particularly enzymes generating reducing equivalents with pivotal roles in fatty acid β-oxidation metabolism. Our data suggest that myxococcal species with the ability to oxidize fatty acids transferred several genes to eubacteria that eventually gave rise to the mitochondrial ancestor. Later, the eukaryotic nucleocytoplasmic lineage acquired those metabolic genes through endosymbiotic gene transfer. Conclusions/Significance Our results support a prokaryotic origin, different from α-proteobacteria, for several mitochondrial genes. Our data reinforce a fluid prokaryotic chromosome model in which the mitochondrion appears to be an important entry point for myxococcal genes to enter eukaryotes.


Introduction
The growing number of fully sequenced genomes from both prokaryotes and deep-branching eukaryotes offers the possibility of identifying genetic transfers that may have occurred when the first eukaryotes appeared at least 1.5 billion years ago [1].Eukaryotic cells are chimeric entities, with both mitochondria and chloroplasts derived from endosymbiotic precursors that were distinct from the nucleocytoplasmic lineage [2].Additionally, many of the original mitochondrially encoded genes were gradually transferred to the eukaryotic nuclear genome via endosymbiotic gene transfer [3].It has been estimated that the a-proteobacterial ancestor of mitochondria contributed at least 630 genes to the eukaryotic nuclear genome [4].However, thousands of eukaryotic nuclear genes of eubacterial ancestry are not derived from an aproteobacterial ancestor [5,6].The hydrogen hypothesis proposes that the first eukaryote was a consortium of a hydrogen-dependent archaeon and a hydrogen-producing a-proteobacterium, which was the ancestor of the mitochondrion [7].In this single eubacterial ancestry hypothesis, the mitochondrial ancestor would have assimilated other eubacterial genes via lateral gene transfer prior to the endosymbiotic event [5].The fluid chromosome model would help to explain such mixed prokaryotic sources of the ancestral mitochondrial genome.This model assumes fluid prokaryotic genomes shaped by gene losses and lateral gene transfers instead of static genomes.Thus, the expected phylogeny for a gene acquired from the mitochondrion would be common ancestry for all eukaryotes but not necessarily trace to aproteobacteria because the ancestor of mitochondria possessed an as yet unknown collection of genes.Alternatively, it has been recently demonstrated a high frequency gene transfer system in the a-proteobacteria Rhodobacter capsulatus based on a virus-like gene transfer agent (GTA), which would facilitate random transfers between species [8].It has been suggested that the GTA system was present in the last a-proteobacteria common ancestor that could explain not only in as much so many a-proteobacterial genes came to reside in the nucleus, but also why mitochondrial and nuclear genes show such mixed phylogenetic affinities [9].Several hypotheses have been suggested to account for more than a single (eubacterial) endosymbiont entity [5,10].One major hypothesis is metabolic symbiosis, also known as syntrophy, with a symbiont distinct from the ancestral mitochondrion symbiont.One syntrophy case is based upon the exchange of sulfur compounds between a spirochete and an archaeal thermoplasma species [11].Another case is the hydrogen-driven syntrophy hypothesis, which suggests that the eukaryotic ancestor was derived from symbiosis between an ancestral d-proteobacteria, specifically a sulfatereducing myxococcal species, and a methanogenic archaeon followed by the eventual incorporation of an a-proteobacterium.The latter hypothesis argues for the full incorporation of a methanogenic archaebacterium within a myxococcal cytoplasm.The acquisition of mitochondria was independent and early in the existing myxococcal-archaeon consortia [12,13] The d-proteobacteria represent one of the most diverse groups of bacteria, exhibiting a wide array of metabolic strategies, including free-living, syntrophic and pathogenic forms.This group is characterized by large variations in genome size.While species of the genus Syntrophus have genomes of approximately 3 Mb, the genomes of myxococcal species are among the largest in prokaryotes, with sizes of approximately 9-13 Mb, which might account for their higher biological complexity [14].The Myxococcales are characterized by social behavior directed toward predation and the construction of a unique multicellular structure, the asexual fruiting body [15].The formation of these fruiting bodies requires coordinated cellular motility and cell signaling [15].Thus, the Myxococcales are of special interest because they represent one of the few prokaryotic lineages that have independently acquired some degree of cellular differentiation and multicellularity.In the Myxococcales, lipids play a role in developmental aggregation, signaling and morphogenesis of the fruiting body [15].In addition, lipids are major contributors to myxococcal physiology as an energy reservoir, which is also the case in fungi, plants and animals, but not in most other prokaryotes [16].The Myxococcales have also been shown to share other similarities with eukaryotes, such as the presence of eukaryotic-like protein kinases [17], Ras-like G-proteins and GTPase-activating proteins that function in regulating cell polarity [18].Here, we employed a phylogenomic analysis to investigate a possible contribution of Myxococcales to the origin of eukaryotes.Our data revealed that several genes encoding mitochondrial proteins, mostly from the fatty acid b-oxidation pathway, were acquired from Myxococcales.

Phylogenetic analysis
To identify putative genetic transfers between eukaryotes and Myxococcales, we conducted a large-scale comparative genomic analysis of 40 eukaryotic genomes, 27 archaeal genomes and 330 eubacterial genomes, including 6 myxococcal species: Anaeromyxobacter dehalogenans, Haliangium ochraceum, Myxococcus xanthus, Plesiocystis pacifica, Sorangium cellulosum and Stigmatella aurantiaca.A preliminary and conservative step consisting of BLAST (threshold E-value of e-40) homologous sequence searches and maximum likelihood (ML) phylogenetic tree reconstruction identified 93 eukaryotic proteins with a predicted myxococcal origin (Figure S1) followed by 40 trees with a-proteobacterial, 15 with archaeal, 8 with firmicutes, 7 with cyanobacterial and 7 with chlamydial origin.The remaining trees traced to other bacterial groups but had less than four trees or were unresolved, with several prokaryotic groups preceding eukaryotes.These 93 positives were further evaluated using HMMER searches, a tool based on hidden Markov models, against additional eukaryotic genomes, and the alignments were performed again with these new taxa.From the inferred ML and Bayesian trees, we identified 15 eukaryotic proteins of obvious myxococcal origin, all with strong or moderate statistical support (Figure 1; see Dataset S1 for all ML trees and Dataset S2 for all Bayesian trees).
Of note, 13 of these 15 genes of myxococcal ancestry were localized to the mitochondria in eukaryotes (see Table 1).We determined the organellar localization of these genes based on bibliographic references and by detecting predicted mitochondrial targeting sequences for several eukaryotes (Table 1).Eight of these proteins play a role in the formation of the acyl-CoA pool; which comprises pivotal intermediates in lipid metabolism (Figure 2).In particular, we detected a myxococcal ancestry for two eukaryotic acyl-CoA synthetases (ACSs) and, remarkably, six fatty acid boxidation enzymes that degrade the acyl-CoA pool: four acyl-CoA dehydrogenases (ACDs), one electron transport flavoprotein (ETF) and one acetyl-CoA acyltransferase with thiolase activity (see Table 1).The ACSs identified corresponded to ACS bubblegum family members 1 and 2 (ACSBG1-ACSBG2) and ACS family member 3 (ACSF3), which activates fatty acids to form acyl-CoA, thus allowing their transport and metabolism [19].The ACD protein family catalyzes the oxidation of diverse acyl-CoA compounds produced during the degradation of fat and protein to enoyl-CoA [20].ACD subfamilies are distinguished by the metabolic pathways in which they participate and by their substrate specificity.The four eukaryotic ACDs with myxococcal ancestry participate in the b-oxidation of fatty acids, with optimal activity for acyl-CoA substrates of specific lengths: short (ACADS), medium (ACADM), long unsaturated (ACAD9) or very long (ACADVL) [20,21].The other ACD subfamily identified is implicated in amino acid degradation.After removal of the amino groups from isoleucine, the remaining branched acyl-CoA is dehydrogenated by the short/branched chain acyl-CoA dehydrogenase (ACADSB).Electron transport flavoprotein A (ETFA) is, in addition to electron transport flavoprotein B (ETFB), the primary acceptor for reducing equivalents from the b-oxidation of acyl-CoA dehydrogenases [22].The phylogenetic trees of both electron transport flavoproteins (ETFA and ETFB) showed patchy distributions, with eukaryotes nested within different bacterial groups; however, the bacterial group first identified as preceding eukaryotes corresponded to the Myxococcales (see ETFA tree in Dataset S1 and S2).Remarkably, an acyl-CoA dehydrogenase gene (ACADM) is located next to the ETFA and ETFB genes in the genome of M. xanthus, indicating the importance of electron flow during the b-oxidation of acyl-CoA intermediates in M. xanthus.Moreover, the thiolase identified, acetyl-CoA acyltransferase 2 (ACAA2), performs the last thiolytic cleavage of the fatty acid b-oxidation spiral in mitochondria and it should be differentiated from the peroxisomal thiolase ACAA1 and from the thiolases ACAT1 and ACAT2 that can also act in the biosynthesis of eukaryotic ketone bodies and sterols through the condensation of Figure 1.Myxococcal ancestry of eukaryotic proteins.Bayesian phylogenetic trees of (i) the isocitrate dehydrogenase 3 NAD(+) alpha, beta and gamma genes, (ii) the acyl-CoA dehydrogenase C-2 to C-3 short chain gene (ACADS), and (iii) acetyl-CoA acyltransferase 2 (ACAA2).The Bayesian posterior probability (PP) until convergence diagnostic and 1000-replicate bootstrap values (BV) for ML trees are indicated if they were above 50%.A black dot indicates PP.0.95.Eukaryotic, myxococcal/d-proteobacterial and a-proteobacterial taxa are highlighted in blue, red and green, respectively.doi:10.1371/journal.pone.0021989.g001two acetyl-CoA molecules to form acetoacetyl-CoA [23].The ACAA2 thiolase has previously been reported to have a myxococcal origin [24].
The remaining proteins of unequivocal myxococcal ancestry belonged to different functional groups, and five of them also localized to the mitochondria (Table 1).These remaining proteins are as follows: i) three proteases: the M3 family peptidases neurolysin (NLN) and the thimet oligopeptidase (THOP1) and the cytosolic aminopeptidase NPEPL1; ii) a translation elongation factor G that is predicted to localize to mitochondria in organisms ranging from mammals to trypanosomatids and that performs GTP-dependent translocation of the ribosome during translation; iii) the Krebs cycle enzyme NAD(+)-dependent isocitrate dehydrogenase; iv) an arginil tRNA synthetase; v) an aminotransferase, 4-aminobutyrate aminotransferase; and vi) a ceramidase, N-acylsphingosine amidohydrolase 2, that catalyzes the hydrolysis of the N-acyl linkage of ceramide, a second messenger in a variety of cellular events, to produce sphingosine.Despite the lack of a predicted mitochondrial targeting sequence, this protein has been experimentally localized to the mitochondria of mice and humans [25,26].

Compositional analysis
The position of the Myxococcales as a sister group to eukaryotes in phylogenetic trees produced in this study implies the occurrence of lateral gene transfer events from Myxococcales to eukaryotes or vice versa (Figure 1).We investigated this possibility further by performing a sequence compositional analysis of the 15 eukaryotic genes of myxococcal ancestry using homologs from both Myxococcales and unicellular eukaryotes [27].In cases of recent lateral gene transfer, it has been suggested that the nucleotide composition of the transferred gene might be more similar to that of the donor species than to that of the recipient [28].The average G+C content of myxococcal species ranges from 69 to 74%.To examine the possibility of a transfer from Myxococcales to eukaryotes, we calculated the average G+C content in the coding sequence (CDS) of our 15 candidate genes in the unicellular eukaryotes and myxococcal homologs and compared them to the average whole-genome G+C content for each species.We found that the 15 CDSs from unicellular eukaryotes were consistently representative of their respective genomes and were thus adapted to function in the organism in which they resided, an observation that was inconsistent with the occurrence of recent lateral gene transfer events.
To further examine the possibility of a eukaryote-to-Myxococcales transfer, we used several compositional methods to measure lateral gene transfer in bacteria: (i) Karlin's method based on codon usage [29,30], (ii) the Horizontal Gene Transfer Database (HGT-DB), which relies on G+C content, codon usage, gene position and amino acid composition [31], and (iii) the Island Viewer server, which identifies genomic islands or clusters by integrating sequence composition and comparative genomic approaches.None of the 15 myxococcal ancestors of eukaryotic genes was predicted to have undergone lateral gene transfer from eukaryotes.This analysis only identified overlaps of 3 and 6 bp in ACADS and ACSBG, respectively, in the M. xanthus genome (Figure S2).Thus, our findings on the base composition of the 15 CDS candidates are not consistent with recent lateral gene transfers.Instead, our results suggest that ancestral genetic transfers occurred between Myxococcales and eukaryotes, and the incorporated genes might have since converged with the bulk of the genome through a process of amelioration [32].

Myxococcal origin of part of the mitochondrial b-oxidation pathway
Although we used a conservative pipeline (threshold E-value of e-40) and despite few myxococcal genomes being currently available, our results clearly indicate that several eukaryotic genes have a myxococcal ancestry.Half of the genes identified encoded mitochondrial enzymes involved in acyl-CoA intermediate metabolism, primarily in fatty acid b-oxidation (Table 1 and Figure 2).Among the enzymes involved in b-oxidation with a clear myxococcal ancestry were four acyl-CoA dehydrogenases (ACDs), which catalyze the oxidation of diverse acyl-CoA compounds produced during the degradation of fat and amino acid to enoyl-CoA in a substrate-specific manner [20].Fatty acids can be degraded in hydrogen-driven syntrophy with symbiotic partners under anaerobic conditions [33,34].There are natural examples of fatty acid syntrophies with d-proteobacteria.For instance, the fermenting Syntrophaceae family has the ability to grow on fatty acids in syntrophy with methanogens [34,35].Notably, the syntrophic oxidation of fatty acids involves hydrogen production from high potential electron donors, such as the acyl-CoA intermediates, that are oxidized by members of the ACD family [36].Indeed, a genomic analysis of the fatty acid degrading syntrophic bacterium Syntrophus aciditrophicus has suggested that the oxidation of acyl-CoA intermediates, which plays a crucial role in generating reducing equivalents, might be specific to syntrophic metabolism [33].The production of hydrogen from the ACD reaction is thermodynamically unfavorable and can occur only with energy input by a process known as reverse electron transfer.With regard to this mechanism, it has been suggested that ETF could transfer electrons from acyl-CoA intermediates oxidized by ACD to membrane redox complexes, such as the complex that includes the membrane-bound iron-sulfur oxidoreductase present in S. aciditrophicus [33,36].
Our list of putative fatty acid b-oxidation genes with a myxococcal ancestry did not include the genes encoding the central multifunctional proteins involved in b-oxidation (HADHA and HADHB).A possible explanation for this absence may lie in the fact that the human HADHA/HADHB multienzyme complex is formed by a gene fusion between 3-hydroxyacyl-CoA dehydrogenase and enoyl-CoA hydratase of HADHA and the noncovalent interaction of the thiolase activity of HADHB.This arrangement probably resulted from the combination of monospecific enzymatic functions [37].Thus, the gene fusion present in the human genome from which we retrieved myxococcal homologous sequences (step 1 in our pipeline) might have masked ancestral monofunctional subunits that are still functional in other eukaryote lineages (for example, in Euglena gracilis [37]).Indeed, we identified a monofunctional enzyme with 3-hydroxyacyl-CoA dehydrogenase activity with myxococcal ancestry, although the statistical support for this inference was low (38% ML and 0.91 BPP).This protein is absent in Metazoa, but Capsaspora, Chromalveolata, Dictyostelium, Naegleria and Fungi all encode homologs of this gene (Figure S3).Our results contradict those of studies suggesting that the a-proteobacterial ancestor of mitochondria was the donor of multiple genes involved in acyl-CoA metabolism, including the b-oxidation pathway, to the nucleocytoplasmic lineage [4].This is most likely because our analysis involved a higher number of myxococcal genomes and included a relatively large taxon sampling.However, we cannot exclude the possibility that the a-proteobacterial ancestor of mitochondria was also a fatty acid degrading bacteria [38], with the possibility that some eukaryotic b-oxidation enzymes could have been acquired from a-proteobacterial genomes.Indeed, our results indicate that some mitochondrial b-oxidation enzymes did not descend from a-proteobacterial homologs because they were absent in the reported trees (see the ACADSB tree in Figure 2 and Dataset S1 and S2) or their topology was unrelated to the eukaryotic clade (see ACAA2, ACADM, ACADS, ACADVL-ACAD9 and ACSBG trees in Figure 2 and Dataset S1 and S2).

Lateral gene transfer from Myxococcales to eukaryotes or fluid chromosome?
Our results are compatible with the occurrence of multiple ancient lateral gene transfer events from Myxococcales to eukaryotes.We propose three possible explanations for the observation that most of the proteins with myxococcal ancestry found in this study are targeted to the mitochondria.First, myxococcal bacteria may have transferred genes to the nucleocytoplasmic lineage using direct and independent lateral gene transfers.Second, lateral transfer may have occurred via endosymbiotic gene transfer from a myxococcal symbiont to the nucleocytoplasmic lineage.This second proposal is compatible with the d-proteobacteria syntrophy hypothesis, which suggests that the eukaryotic ancestor was derived from a symbiosis between an ancestral sulfate-reducing myxococcal species and a methanogenic archaeon [12,13].However, this hypothesis assumes an unrelated incorporation of the myxococcal and mitochondrial ancestors, which does not support the majority of the myxococcal ancestry proteins found in this study being targeted to the mitochondria.The d-proteobacteria syntrophy hypothesis argues that a myxococcal endosymbiosis occurred prior and independently of the mitochondrion acquisition by an amitochondriate eukaryote.We believe that our findings are more compatible with a simultaneous origin for myxococcal and mitochondria proteins that might underlie the reason why proteins with myxococcal origin are preferably targeted to the mitochondrion.A third possibility is that ancestral lateral gene transfer events occurred between Myxococcales and the mitochondrial ancestor endosymbiont prior to the endosymbiotic event that gave rise to the mitochondrion.In this latter scenario, the myxococcal genes would have already been present in the ancestral mitochondrial genome.This option is compatible with the fluid chromosome model [39,40].Thus, myxococcal transfers to eukaryotic genomes would have originated vertically from the mitochondrial progenitor, which was not necessarily an a-proteobacterium because the first mitochondrial genome possessed an as yet unknown collection of genes.
Taken together, we believe that the most parsimonious event might have been that a hydrogen-producer, a fatty acid degrading d-proteobacteria, transferred acyl-CoA related enzymes to the hydrogen producing mitochondrial ancestor (Figure 3.1).According to the hydrogen hypothesis on the origin of eukaryotes, eukaryotes arose from a hydrogen-driven syntrophy between the hydrogen producing a-proteobacteria symbiont and a hydrogenconsuming methanogenic archaeon [7].In this scenario, harboring several ACD reducing equivalents producers acquired from Myxococcales could be of advantage for the hydrogen producing mitochondrial ancestor, fostering hydrogen, acetate and CO 2 production.Once the host methane production became irreversibly dependent on the symbiont hydrogen source, the methanogenic archaebacterium would have maximized its contact with the surface area of the symbiont by surrounding it, eventually engulfing the hydrogen producing symbiont.Over time, genes would have been transferred through endosymbiotic gene transfer to the eukaryotic nucleocytoplasmic lineage (Figure 3.2).If this scenario is true, we would expect that at least some mitochondrially encoded genes would trace to Myxococcales.To test this possibility, we analyzed the myxococcal ancestry of mitochondrially encoded proteins from 2209 mitochondrial genomes.Homologous sequence searches and subsequent phylogenetic reconstruction indicated that none of the mitochondrially encoded proteins had a myxococcal origin (data not shown).Taking into account this result, we cannot discard the possibility of a myxococcal endosymbiosis.The lack of myxococcal ancestry in mitochondrially encoded proteins might apparently support an independent myxococcal endosymbiotic event, but then the genes transferred to the host should be targeted not only to the mitochondrion but colonizing other subcellular compartments.As far as the myxococcal proteins identified are preferably targeted to the mitochondria, we believe that the most compatible scenario is the simultaneous origin for mitochondrial and myxococcal proteins.According to that, there are two possibilities: i) that the endosymbiotic event that gave rise to the mitochondrion also involved a myxoccocal bacterial partner that subsequently transferred a number of metabolic pathways to the host via endosymbiotic gene transfer; or ii) ancestral lateral gene transfer events took place between Myxococcales and the mitochondrial ancestor endosymbiont prior to the endosymbiotic event that gave rise to the mitochondrion.Supporting the last option, the mitochondrial genome seems to retain specific functional genes that do not descend from the myxococcal genomes.In this sense, the presence of a mitochondrial genome appears to correlate with increased respiratory capacity and ATP availability within organisms [41,42].Therefore, it has been suggested that this relationship could be the driving force behind the selective pressure to preferentially retain the genes encoding the electron transport chain in the mitochondrial genome [41,42].These genes are mostly derived from a-proteobacterial genomes; in contrast, the myxobacterially derived metabolic genes would have been transferred to the eukaryotic nucleocytoplasmic lineage.This transfer might account for the absence of mitochondrially encoded genes with myxococcal ancestry.
It is worth mentioning that because Myxococcales harbors some of the largest eubacterial genomes, lineage sorting could influence our topologies.However, most of our trees have representation from a wide sampling of other eubacterial genomes.Our results indicate that the mitochondrion might be an important entry point of eubacterial genes that are different from a-proteobacterial genes.Thus, the genome of the mitochondrial ancestor should be considered a mix of different eubacterial genes, some of them still conserved in the extant a-proteobacterial and myxococcal genomes.Further, it is tempting to speculate that the large genomes of the Myxococcales could be a repository of ancestral genes from many prokaryotic lineages, including that of the mitochondrial ancestor.
In conclusion, our data indicate the myxococcal origin of 15 nuclear eukaryotic genes that are not a-proteobacterial, some of which have key roles in acyl-CoA intermediate metabolism.Many other genes may also have a myxococcal origin, but the lack of a phylogenetic signal and/or our extremely conservative pipeline did not allow us to recover them.We propose that a fatty acid degrading d-proteobacterium donated some genes to the mitochondrial ancestor prior to the endosymbiotic event, and these genes were subsequently transferred to the nucleocytoplasmic lineage via endosymbiotic gene transfer.Thus, our results support a version of the fluid chromosome model as the most plausible scenario, in which Myxococcales contributed key metabolic genes to the first eukaryotes.

Data retrieval
Proteomes encoded by 355 publicly available complete genomes of eubacteria and archaea were obtained from the National Center for Biotechnology Information (NCBI) FTP server (ftp://ftp.ncbi.nih.gov/genomes/Bacteria), including four complete myxococcal genomes.In addition, we included the draft assembly genomes of the Myxococcales species P. pacifica and S. aurantiaca.We sampled 40 and 49 eukaryotic genomes for the first and second analyses, respectively, with the goal of encompassing the widest possible eukaryotic diversity.Sequence analyses were performed for a range of members of the Amoebozoa (Dictyostelium discoideum), the Archaeplastida (Arabidopsis thaliana, Chlamydomonas reinhardtii, Cyanidioschyzon merolae, Oryza sativa and Ostreococcus tauri),

Figure 3 .
Figure 3. Model of transfer of myxococcal genes to eukaryotes.(1) A hydrogen-consuming methanogenic archaeon established symbiosis with the hydrogen-producing mitochondrial ancestor.Previously, a fatty acid degrading myxococcal species transferred genes to the mitochondrial ancestor before the endosymbiotic event.(2)The methanogenic archaeon engulfed the symbiont that gave rise to the mitochondrion.Over time, genes were transferred to the eukaryotic nuclear genome via endosymbiotic gene transfer.Proteins with myxococcal, mitochondrial and archaeal ancestries are depicted in green, blue and yellow boxes, respectively.doi:10.1371/journal.pone.0021989.g003

Table 1 .
[52] of eukaryotic proteins with a predicted myxococcal origin.Symbol corresponds to the human annotation as found in NCBI.Support indicates the statistical support for a monophyletic myxococcal+eukaryotic group as obtained from the 1000-replicate ML bootstrap values (BV, in %) and the Bayesian posterior probability (PP).Localization indicates the localization determined for most of the studied eukaryotes.Taxa shown in the last column are the eukaryotes in which the TargetP program (http://www.cbs.dtu.dk/services/TargetP/)[52]detected a putative mitochondria-targeting peptide.When generic lineage names, such as Archaeplastida, Excavata, Fungi, Metazoa or Trypanosomatids, are shown, at least two different species of this group were found to have a predicted signal peptide.doi:10.1371/journal.pone.0021989.t001