Accelerated microevolution in an outer membrane protein (OMP) of the intracellular bacteria Wolbachia
© Baldo et al; licensee BioMed Central Ltd. 2010
Received: 20 November 2009
Accepted: 17 February 2010
Published: 17 February 2010
Outer membrane proteins (OMPs) of Gram-negative bacteria are key players in the biology of bacterial-host interactions. However, while considerable attention has been given to OMPs of vertebrate pathogens, relatively little is known about the role of these proteins in bacteria that primarily infect invertebrates. One such OMP is found in the intracellular bacteria Wolbachia, which are widespread symbionts of arthropods and filarial nematodes. Recent experimental studies have shown that the Wolbachia surface protein (WSP) can trigger host immune responses and control cell death programming in humans, suggesting a key role of WSP for establishment and persistence of the symbiosis in arthropods.
Here we performed an analysis of 515 unique alleles found in 831 Wolbachia isolates, to investigate WSP structure, microevolution and population genetics. WSP shows an eight-strand transmembrane β-barrel structure with four extracellular loops containing hypervariable regions (HVRs). A clustering approach based upon patterns of HVR haplotype diversity was used to group similar WSP sequences and to estimate the relative contribution of mutation and recombination during early stages of protein divergence. Results indicate that although point mutations generate most of the new protein haplotypes, recombination is a predominant force triggering diversity since the very first steps of protein evolution, causing at least 50% of the total amino acid variation observed in recently diverged proteins. Analysis of synonymous variants indicates that individual WSP protein types are subject to a very rapid turnover and that HVRs can accommodate a virtually unlimited repertoire of peptides. Overall distribution of WSP across hosts supports a non-random association of WSP with the host genus, although extensive horizontal transfer has occurred also in recent times.
In OMPs of vertebrate pathogens, large recombination impact, positive selection, reduced structural and compositional constraints, and extensive lateral gene transfer are considered hallmarks of evolution in response to the adaptive immune system. However, Wolbachia do not infect vertebrates. Here we predict that the rapid turnover of WSP loop motifs could aid in evading or inhibiting the invertebrate innate immune response. Overall, these features identify WSP as a strong candidate for future studies of host-Wolbachia interactions that affect establishment and persistence of this widespread endosymbiosis.
Outer membrane proteins (OMPs) of pathogenic bacteria are widely recognized as crucially involved in bacterial interactions with eukaryotic hosts . They have thus been the subject of extensive studies aimed to clarify how they evolve and whether their patterns of divergence are informative about the biology of host-bacteria dynamics. OMPs are involved in a large repertoire of functions, including bacterial invasion and defense, transportation of various molecules, adhesion and signaling pathways [1, 2]. OMPs of mammalian pathogenic Proteobacteria often function as antigens , and several of these proteins are currently targets for vaccine development against important human pathogens, such as Ehrlichia, Rickettsia, Haemophilus influenzae, and Neisseria meningitidis [4, 5].
OMPs are highly variable and among the fastest evolving microbial proteins [6, 7]. Despite a large diversity of composition and function, they do share genetic and structural features that allow their identification as surface proteins, primarily via bioinformatic prediction . OMPs show a characteristic transmembrane β-barrel structure, formed by an even number of antiparallel sheets, connected to loops of variable length at the extracellular side and to short turns containing both N and C termini at the periplasmic side. Given the key role of OMPs in the interactions with the host, a large number of studies have been devoted to uncover trends in the molecular evolution of these proteins [4, 8–10]. Typically, residues in the β-barrel show the highest conservation, while variability mainly affects the conformational domains located in the extracellular loops, which can function as receptors and can be highly antigenic - e.g. P28 OMPs of Ehrlichia, Opa proteins of Neisseria and MSP2 proteins of Anaplasma [11–14].
While considerable attention has been given to OMPs of vertebrate pathogens, the role and evolution of OMPs in the establishment and persistence of both pathogenic and non-pathogenic microbial associations found in invertebrates remain largely unknown. Recent studies have shown that both the vertebrate and invertebrate immune systems can confer specific protection against bacterial infections and share comparable defensive solutions , making OMP of invertebrate pathogens important candidates for investigating host/symbiont interaction dynamics. In addition, for those vertebrate pathogens that are vectored by arthropods, selection by the invertebrate innate immune system could shape the virulence of vertebrate pathogens vectored by invertebrates . Therefore, studies of OMPs in bacteria that infect invertebrates, but not vertebrates, could reveal to what extent selection in invertebrates shapes OMP diversity and evolution.
The Wolbachia surface protein (WSP) is an OMP found in the intracellular bacteria of the genus Wolbachia , a very widespread and important group of endosymbionts of arthropods and filarial nematodes. Current estimates indicate that around 60% of arthropod species worldwide are infected with this intracellular bacterium . Wolbachia belong to the Rickettsiales and relatives are important vertebrate pathogens within the genera Rickettsia, Ehrlichia and Anaplasma . However, in contrast to some notable members of these related genera, Wolbachia are not pathogens of vertebrates. Instead, theyare mostly known to be "reproductive parasites" of arthropods [20–22], while in filarial worms and some insects they are required for their hosts' survival and provide them with some benefits [23–25].
The function of WSP in Wolbachia remains unknown, although several lines of evidence suggest that it may be an important mediator of the host/symbiont interaction. First, WSP is a dominant protein constituent of infected Drosophila eggs . Experimental studies have shown that WSP can activate the innate immune response in humans via interaction with Toll-like receptors , and trigger a potent inflammatory response in both human and canine filariasis . Recently, WSP has been shown to delay apoptosis in human polymorphonuclear cells (PMNs), typically involved in the innate immune response against microbial pathogens . Finally, inoculation of WSP in BALB/mice induces the expression and production of nitric oxide, an important toxic component used by the immune response against bacteria . The above studies indicate that WSP can induce host immune responses and recent hypothesis predicts WSP as an important player in the establishment and persistence of the symbiosis via apoptosis inhibition .
WSP shows a heterogeneous pattern of amino acid diversity characteristic of other OMPs, marked by four distinct hypervariable regions (HVRs) interspaced by conserved strings of amino acids (CRs) . Variants at each HVR have been frequently exchanged across bacterial strains, generating highly chimeric proteins . While shuffling of HVR motifs is apparent, the primary source of such remarkable amino acid diversity at HVRs remains unknown. Furthermore, it is unclear whether this genetic diversity is adaptive; because arthropod Wolbachia do not infect vertebrate hosts, selection acting on the protein is not due to a response to the vertebrate adaptive immunity (e.g. antibodies). Other forces, therefore, are likely to be shaping the evolution of this protein.
Here we investigated structure and molecular evolution of WSP using 515 distinct alleles found in 831 host isolates, representing the largest sequence dataset available to date for Wolbachia. The first an in silico prediction of the three-dimensional structure of WSP is presented. Using the predicted structure as framework, we investigated the microevolutionary forces that drive the early diversification of WSP proteins by means of clusters of closely related proteins based upon haplotype categories for individual HVRs. This approach eliminates the problems of alignment due to extensive divergence within the HVRs, and allows identification of closely related HVRs (in different protein variants), thus providing a ready classification of variation generated via point mutation versus recombination in recently divergent proteins. We found that WSP shows a rapid turnover of amino acid sequences via both high rates of recombination and positive selection typical of immune antigens under strong selection for diversification. These appear as hallmarks of an ongoing arms race between the host and Wolbachia and identify WSP as a strong candidate for future studies of host-Wolbachia interactions.
In silicoprediction of WSP structure
Discrimination of WSP among globular, inner and outer membrane proteins was assessed based on a position specific scoring matrix (PSSM) profiles approach, implemented in TMBETADISC-RBF . Prediction was confirmed by querying the complete WSP sequence from wMel to the HHomp database, available at http://toolkit.tuebingen.mpg.de/, which detects sequence homology to known outer membrane proteins (OMPs) based on sequence-profiles identified with Hidden Markov Models (HMMs). The WSP two and three-dimensional structures were then predicted using HHpred , which uses structurally related proteins as template. Results were inputted into MODELLER  using a multiple alignment for modeling of the tertiary structure. The three-dimensional model was visualized in cn3D version 4.1 . HHpred has been shown to perform better over simpler approaches (BLAST and PSIBLAST) when template-target similarity is lower than 40% , as is the case with WSP and homologous proteins. To correct for diversity among WSP proteins, we also generated three-dimensional models using a diverse set of divergent WSP genotypes, besides WSP from wMel. WSP topology with respect to the outer membrane lipid bilayer was predicted by the posterior decoding method using PRED-TMBB, software based on Hidden Markov model . Analysis of the hydrophobic and hydrophilic indexes was performed using the Kyte and Doolittle scale .
The nucleotide sequences were either generated during this study or retrieved from Genbank. New sequences were obtained using standard primers and protocols available at http://pubmlst.org/wolbachia/wsp/info/protocol.shtml. The procedure for Genbank retrieval was as follows: all nucleotide wsp sequences present in Genbank were downloaded and redundancy discarded. The set of unique wsp sequences that met the length requirement to be assigned to an allele (see below) and that showed no ambiguous sites were retained. Together with the sequences generated in this study a total of 515 distinct alleles were obtained. All sequences were then compared against the NCBI nucleotide database with BLASTN. For all exact matches (100% identity and coverage), all available host taxon and country of origin information were collected. Multiple entries having the same host species and same wsp allele were retained if they differed by locality information. Several alleles were sequenced during this study and their host information was combined with information collected from Genbank. Overall, 831 distinct records were collected (see additional file 1).
Analysis of nucleotide genetic diversity and GC content was performed using DNAsp version 4.50 [41, 42]. GC content at the third (synonymous only) position was calculated for each of the six largest WSP complexes (C1-C6, see below for complex definition) and for the five MLST genes. Amino acid divergence was calculated in PAUP version 4.0b10 .
To explore whether additional intragenomic sources might contribute to wsp variability (such as pseudogenes or other noncoding sequences), complete wsp and single HVR nucleotide sequences from Drosophila melanogaster and Culex pipiens were BLASTN against the Wolbachia genomes from these two host species.
Identification of protein complexes
We identified complexes of closely related WSP proteins using the clustering program eBURST V3  and employing the matrix of WSP profiles to depict evolutionary relationships among WSP genotypes. This approach is superior to conventional phylogenetic analyses, in this case, because it reveals localized recombination (i.e. involving short intragenic sequences) among otherwise similar WSP sequences, without incorrectly inferring such sequences as phylogenetically distant. The program eBURST is typically used for analysis of MLST allelic profiles (where each allele at each MLST locus is given a number and an allelic profile is the combination of allele numbers at the MLST loci) and assumes an epidemic model of population structure for building the clusters: that is, it assumes that a founder genotype (an allelic profile) initially rises in frequency and subsequently diversifies to produce minor variants (i.e. single locus variants, SLV), hence producing a "clonal complex" of closely related genotypes. Given a matrix of allelic profiles, eBURST predicts the ancestral genotype to be the profile with the greatest number of single locus variants. Often the founder is also the most frequent profile in a complex (in terms of isolates in which is found).
Here we applied the same clustering algorithm to a single locus, thus using WSP profiles. WSP profiles were clustered into "WSP complexes", where a WSP complex is a group of related WSP profiles that differ at a single HVR peptide (here named single HVR variant, SHV) with respect to the ancestral profile. The ancestral profile of each complex is predicted by eBURST to be the WSP profile with the highest number of SHVs. Profiles that could not be assigned to a group were named singletons.
Recombination versus mutation estimates
We used the complexes identified by eBURST to estimate the relative contribution of recombination versus mutation that give rise to new proteins. Within complexes, mutant profiles were discriminated from recombinant profiles (and associated alleles) by assessing whether SHVs diverged by mutation or recombination. The procedure was as follows. First, within each WSP complex, all WSP profiles were compared to the ancestral profile and the number of amino acid changes at their SHV was annotated. In case of complexes of two profiles we performed a simple pairwise comparison. This allowed a first screening for major recombinant proteins, defined as the WSP profiles that showed 4 or more amino acid polymorphic sites at their SHVs (including both amino acid substitutions and indels) with respect to the ancestral type. All other profiles within a complex, which then showed 1 to 3 polymorphic sites at their SHVs, were compared to the whole dataset and nucleotide substitution patterns inspected. If a SHV peptide was shared among profiles in different complexes, then mutation was assumed when the nucleotide substitution patterns at their SHV differed. In cases of matching substitution patterns, a) a single nucleotide substitution (thus nonsynonymous) was assumed to be arisen by convergence and the profiles retained as mutants, b) multiple shared nucleotide substitutions were considered sign of recombination and the recombinant profile was predicted to be the one with highest number of amino acid changes with respect to its ancestral profile. If equal, both profiles were labeled as recombinants. The relative contribution of recombination to mutation in generating protein diversity was measured as a ratio of recombination to mutation per protein and per HVR. We are aware that this approximation does not take into account mutation and recombination events that did not result in amino acid changes (i.e. synonymous substitutions); indeed our goal was to explore the contribution of the two forces in promoting protein diversity, and not synonymous allele diversity. In addition, we have also examined recombination occurrence among distinct alleles carrying only synonymous substitutions (see below). The recombination analyses were also expanded to profiles members of subgroups and subgroup founders (coded as double HVR variants, DHVs).
Prediction of the ancestral alleles
Ancestral proteins predicted by eBURST can be coded by multiple alleles within a complex. To identify the ancestral allele and confirm eBURST prediction we proceeded as follows. First, within a complex, all alleles associated to mutant profiles, including synonymous alleles, were further analyzed at the nucleotide level to exclude recombination occurrence using the method of Betran et al , implemented in the DNAsp. Second, the ancestral allele of each complex was predicted using a statistical parsimony method  carried out in TCS software version 1.21 .
Analysis of selective pressures
Selection on mutant alleles was investigated using the complexes previously detected. For complexes that included at least four sequences (n = 25, including also those coding for DHVs) all alleles were tested for neutrality, using the Tajima's D . For complexes of three or more alleles for which the ancestral allele was identified (n = 29), rates of non-synonymous substitutions (dN) and synonymous substitutions (dS) per codon were estimated using the codeml program in PAML package [49, 50]. Specifically, for each complex we generated a simple star tree with a bifurcation at the deepest node to assign the root (corresponding to the ancestral allele) and used it as input file for codeml. Four models were tested: the nearly neutral model (M1), which assumes a proportion p0 of conserved sites with ω0 < 1 and p1 = 1 - p0 of neutral sites with ω = 1; the positive-selection model (M2), which includes an additional class of sites with frequency p2 = 1 - p0 - p1 and with ω2 estimated from the data; the model M7, which assumes that ω is β-distributed and provides a flexible null hypothesis for testing positive selection; and the model M8, which includes one additional ω class of sites with respect to M7, estimated from the data. We performed two likelihood ratio tests (LRTs) comparing likelihood scores of M1 vs M2 and M7 vs M8. The presence of positively selected sites within each complex was determined by concordance between the two best-selected models using the Bayes Empirical Bayes analysis (BEB).
To explore selective pressures acting on partitions of wsp, the nucleotide alignment including the 515 alleles was divided into seven sections corresponding to hvr1, CR2, hvr2, CR3, hvr3, CR4 and hvr4. For this particular analysis, each hvr (here non capitalized to distinguish them from sections used for typing) comprises only the strings of hypervariable amino acids, thus excluding any portions of the flanking conserved regions. Section boundaries for hvrs were based on the alignment in Baldo et al. , where a similar analysis was performed using a much smaller dataset. Within each of the seven sections, redundancy of nucleotide sequences was discarded and sequences were grouped using BlastClust, available at http://toolkit.tuebingen.mpg.de/, based on 95% cutoff of nucleotide identity and 100% matching length. This clustering approach considers related regions of wsp even if inserted in a recombinant background. For each wsp section, dN/dS ratios were estimated at each cluster of three or more sequences using the method of Nei and Gojobori (1986) , implemented in DNAsp. Average values across groups within each section were then calculated. For few groups within hvrs, dN/dS values were not available, as sequences diverged only by nonsynonymous substitutions returning no ratios; although suggestive of positive selection, these groups were conventionally assigned dN/dS = 1 as a conservative estimate.
Analysis of synonymous allele variants
For each of the 435 proteins in our dataset, number of allele variants, average synonymous diversity and number and type of polymorphisms were estimated using DNAsp. As a control, we compared these estimates to those obtained for the five MLST housekeeping genes, using a dataset of published and partly unpublished data. Statistical significance of difference in values between wsp and each of the five MLST genes was inferred performing a Wilcoxon two-sample test.
Statistical association of WSP sequences with host taxa
The rarefaction curve was built using the online calculator available at http://www2.biology.ualberta.ca/jbrzusto/rarefact.php. Curve fitting to predict the asymptotic number of WSP proteins was performed using the on-line regression analysis at site http://www.xuru.org/rt/NLR.asp#Manually. The 37 points along the curve were used for curve fitting to formulae allowing for three parameters.
We tested whether genetic distances between WSP protein sequences that are found in hosts within the same genus were more similar than they would be by random chance. We choose to analyze proteins instead of nucleotide sequences to facilitate generation of alignments and because WSP amino acid diversity largely reflects the nucleotide diversity. The procedure was as follows: the 515 wsp sequences were translated into amino acids, aligned using ClustalX  and manually curated in Bioedit vs22.214.171.124 . The HVR4 was eliminated from the analyses as difficult to align and highly recombinant. A distance matrix was generated based on the amino acid alignment using PAUP version 4.0b10 .
To determine whether WSP sequences were significantly associated with host genera, we sub-sampled the initial dataset to comprise only one host species per wsp allele, that is, the dataset included multiple representatives of the same species only in case they carried distinct wsp alleles. This avoids overrepresentation of a single wsp sequence found in the same host species but from different geographical regions. The final dataset included 732 entries. In order to create a null distribution, we generated 1000 pseudoreplicates by randomly resampling host taxa without replacement across all WSP sequences. For each pseudoreplicate we calculated the mean and median pairwise distances between WSP sequences within the taxonomic group being tested. We then compared the original mean and median values with the resulting pseudoreplicate distribution in a one-tailed test to determine p-values. For this association study we first performed a global analysis resampling all WSP pairwise distances within and between host genera separately and averaging the two sets of values; we then tested the association within individual genera.
Three-dimensional structure prediction of WSP: an eight β-barrel OMP
The analysis of the hydrophilic and hydrophobic pattern (Kyte & Doolittle scale) and prediction of WSP position on the lipid bilayer (Fig. 2B) confirmed that WSP shows a typical eight B-barrel structure with four hydrophilic extracellular loops and periplasmic turns connected to a β-barrel core containing predominantly the neutral or polar residues valine, alanine, glycine and tyrosine, in accordance to the typical composition of β-barrel proteins .
This computationally based structural analysis provided a framework to compare the patterns of protein evolution in different predicted functional domains.
Genetic diversity of WSP
WSP protein and single HVR peptide genetic diversity based on 515 alleles
Amino acid length (range)
No. of prot or pept
Because WSP sequences show very low homology across the diversity within their individual HVRs, global phylogenetic reconstructions or a distance-related approach for studying WSP evolution is not practical. To circumvent the problem of such extensive variation among proteins and HVRs, we focused on recently diverged proteins that still retain true homology. This approach allowed us to identify recent events of recombination and mutation in evolving proteins.
Clusters of closely related WSP
Among identified complexes, 29 contained at least three distinct but related proteins for which the putative founder protein could be predicted (Additional file 2). We note that assignment of the founder genotype by eBURST does not take into account the number of allele variants coding for it, nor its genotype frequency across isolates. Nevertheless, within each complex, typically the predicted founder was also the protein with the highest number of allele variants, and the one found in the largest host taxonomical range (see below). These lines of evidence further support the accuracy of the founder sequence assignment. We used these complexes and predicted founders to assign directionality to the evolutionary changes and to estimate relative rates of mutation and recombination.
Recent evolution of WSP sequences: interplay of mutation and recombination
In terms of the relative contribution of recombination versus mutation to WSP genetic diversity (Fig. 4B), it is notable that in half of the recombinant cases detected (9/18) recombination introduced more than 10 amino acid changes. As a result, although mutation contributes the most to generating new proteins/HVR peptides (i.e. overall haplotype diversity: 88.46% mutation vs 11.53% recombination, Fig. 4A), the contribution of the two forces to WSP amino acid diversity is almost equal (50.56% vs 49.43%, Fig. 4B). As a striking example, although the relative contribution of mutation and recombination to HVR4 haplotype diversity is approximately 3:1 (Fig 4A), the relative contribution to HVR4 amino acid diversity is 1: 2 (Fig. 4B): 75% of the observed HVR4 amino acid diversity is in fact due to recombination.
When analyses were expanded to include subgroups within complexes (Fig. 3, subgroup founders in yellow), the proportion of recombinants increased from 11.53% to 28.6%: 10 out of 35 subgroup profiles were recombinants at one HVR at least, all 10 showing more than 10 amino acids changes with respect to the ancestral HVR.
A large impact of recombination among more divergent WSP
Most common HVR peptides in the dataset and genetic diversity of proteins and nucleotide alleles in which they are found.
No. of Prot/alleles in which they are found
Average Prot diversity (%)1
Diversifying selection acting on HVRs
Predicted founder alleles of major WSP complexes and corresponding profiles, estimated using TCS.
Founder WSP profile
No. of prot/alleles
In the second approach, we considered HVRs and CRs individually and identified sets of closely related sequences within each of the seven sections of WSP, separately (section borders are shown in Fig. 1A, excluding CR1 that was missing in our alignment). We then estimated the average dN/dS for each group within single sections and averaged these values across groups. This global analysis of dN/dS showed that the average dN/dS values for each HVR is typically >1 (see additional file 4), implying diversifying selection. The exception was HVR3, for which the average falls below 1 (0.73) although several groups of related sequences within HVR3 showed dN/dS >>1. In contrast, each CR appears to be under strong purifying selection, with dN/dS<<1 (average range across the three CRs is 0.07 - 0.14). Therefore, when comparing sequences among closely related WSPs (which have only recently diverged) we find only a weak signature of diversifying selection. However, when comparing similar HVR variants that occur in different WSP proteins, we find a strong signature of selection for amino acid substitutions. This finding suggests that those HVR variants that spread via recombination into different WSP proteins can become targets of diversifying selection.
WSP synonymous variants are evolutionary unstable
We finally tested whether protein genotypes are typically stable over long evolutionary times by investigating the diversity of nucleotide variants coding for a single protein sequence. The rationale is that, if a particular protein genotype is evolutionarily stable over time, it will progressively accumulate synonymous substitutions while discarding amino acid changes. Therefore, the occurrence and genetic diversity of synonymous variants should reflect how long a protein type has been around.
Allele diversity per protein type estimated for wsp and the five MLST genes
No. of alleles
No. of prot with multiple allele variants
No. of allele variants per prot (average)
Analysis of the polymorphic pattern within groups containing three or more allele variants per WSP protein type showed, in all cases, that all polymorphisms are unique to a single allele, (i.e. without homoplasies) indicating independent evolution through radiation from the ancestral allele.
Host population structure of WSP
Pairwise analysis for significant association (in bold) of similar WSP sequences within the same host genus (1000 replicates).
No. of Isolates
No. of Host species
No. of wspalleles
Of the 12 most common alleles in Fig. 6, 10 are predicted founders of WSP complexes (Table 3). Because of the accelerated evolution of this gene, such a broad distribution of a single allele variant appears to be the result of either a recent wsp lateral transfer coding for a particularly adaptive WSP haplotype or the rapid host-range expansion of a Wolbachia strain associated to such wsp sequences, rather than the persistence of an ancestral genotype.
Surface proteins found in intracellular bacteria are extremely interesting candidates for the study of symbiotic interactions, given their location at the interface between the bacteria and the host cell environment. Nevertheless functions and molecular processes driving their evolution in invertebrate hosts have been poorly investigated. Only recently, studies are providing examples of how adaptive immune response in insects can be linked to specific OMP polymorphic variants found in their bacterial symbionts , thus unveiling a largely underestimated specificity of the invertebrate immunity .
Here we investigated the molecular processes driving diversification of the outer membrane protein WSP, one of the most abundant Wolbachia proteins found in Drosophila eggs. Based on protein folding prediction, we have shown that WSP presents a typical eight β-barrel structure spanned through the outer membrane and connected to four extracellular loops. The four loops face on the same side of the barrel, and presumably make contact with the host cytoplasm or the vacuole intermembrane space that envelops Wolbachia. The lack of strong compositional and conformational constraints at the loops is consistent with their impressive diversification and putative function as receptors, as shown in several OMPs [11, 13, 14]. On the other hand, the large conservation of the WSP transmembrane core (up to 90% and no indels) likely reflects structural constraints, providing the architectural anchor of the protein to the membrane.
Microevolution of WSP is largely driven by recombination
WSP shows the most remarkable pattern of recombination seen among the Wolbachia proteins studied so far . While recombination involving shuffling of HVR motifs across WSP sequences has been well documented and explains the existence of a remarkable repertoire of WSP protein variants, it remains unclear how this motif diversity is generated and whether it is functional. Here we attempted to reconstruct the microevolutionary steps in WSP early diversification with the aim of uncovering the major forces at the basis of this variation. The dataset used, which comprises 515 alleles found in more than 831 isolates spanning a great taxonomical host range, reflects a substantial sampling diversity of WSP and allowed grouping of recently diverged WSP proteins and HVR motifs.
The reconstruction of WSP relationships presents several technical issues that are common to the analysis of other recombinant OMPs (e.g. the multigene family of P28 OMPs of Ehrlichia). In particular there are challenges to detecting true homology via straightforward alignments, due to the high level of sequence and length variation, and issues relating to correcting for the recombination bias while attempting to detect and measure selection acting on the molecule [4, 55]. Indeed, most WSP and other OMPs-based studies have excluded the HVRs to avoid alignment issues, although this resulted in exclusion of crucial functional domains of the protein. On the other hand, studies that have relied on global alignments have faced extreme alignment problems, making assumptions on true homology difficult or impossible. Here we approached these issues using a profile-based method for proteins grouping, which does not base on alignments. This clustering method allowed the identification of several sets of recently diverged WSP proteins and their ancestral genotypes, thus providing statistical power to investigate sequence evolution on relative short-time scales. Specifically, by grouping similar haplotypes at HVRs we were able to discriminate between amino acid changes introduced by mutational versus recombinational events and analyze the contribution of the two forces to the actual protein (thus functional) diversity, and not to allele diversity. Results indicated that while mutation in WSP occurred at a higher frequency than recombination, as expected, recombination has had a remarkable impact both on the emergence of novel protein types and on the very rapid increase of genetic diversity among proteins, being responsible alone for about 50% of the total amino acid variation observed among recently diverged proteins. Shuffling of WSP portions among more divergent sequences is also frequent, as indicated by the sharing of identical HVRs among otherwise very divergent alleles. Such a pattern strongly suggests that recombination is ongoing and largely contributed to both the short- and long-term diversification and evolution of WSP.
How did this WSP mosaicism generate? A similar pattern of diversification is observed in other OMPs of vertebrate pathogens, such as in Neisseria Opa proteins [12, 56] and MSP2 of Anaplasma . Variability in these proteins is typically generated via a process of gene conversion involving modular cassettes of HVR motifs located in pseudogenes within the same genome [56, 57]. Unlike these proteins, wsp occurs as single copy in the genome, based on all published Wolbachia genomes and evidence from PCR amplifications using universal wsp primers, which typically return clear single sequences. We searched for presence of additional single HVR motifs in the published Wolbachia genomes from Drosophila melanogaster and Culex pipiens, but failed to detect any significant homology to full or partial HVR motifs (data not shown). This suggests that WSP chimeric structure is primarily due to recombination events involving foreign DNA rather than a modular exchange of hypervariable regions within a single genome. Modes of DNA transfer across Wolbachia strains remain unknown, although the widespread occurrence of coinfections of a single host clearly provides a suitable arena for DNA exchange.
Loop diversity is adaptive
The predicted four extracellular loops, which accommodate the HVR motifs, show an extreme plasticity and a mutational pattern that appears largely unpredictable. Indels, which represent one of the major sources of diversity among WSP sequences and only occur at HVRs, were absent among recently diverged proteins (i.e. in SHVs) but numerous among recombinants, suggesting that they are normally introduced via recombination. Among the four loops, L4 presents the largest variation, due to both mutation and recombination. There is no apparent restriction in L4 size, which can vary by as much as 19 amino acids in length. This loop versatility does not disrupt the reading frame, which was always conserved, and thus strongly suggests that this diversification is adaptive. In contrast, L3 showed the lowest variation in length (two amino acids difference) and the lowest dN/dS ratio, which could be due to larger structural constraints or simply to a lower recombination impact. Regardless of length plasticity, it is, however, interesting to note that all four loops show similar haplotype diversity (Table 1), with a potential to accommodate a very large repertoire of distinct peptides. Nevertheless, some compositional constraints are expected: 1) AT-biased codons are strongly favored, as indicated by the high AT content at third (synonymous only) codon positions of wsp (83%) as well as of Wolbachia housekeeping genes (76%), and 2) amino acid composition should account for a high percentage of hydrophilic amino acids, given that these sites are extracellularly exposed.
What are the types of selective pressures acting during the early diversification of WSP? Alleles appear to be under neutrality during recent divergence, although we cannot exclude statistical limits in detecting positive selection in very closely related sequences. A signature of selection becomes visible with increasing sequence divergence. HVRs and CRs are clearly evolving under very different selective pressures, as shown by average dN/dS values typically >1 for HVRs, and <<1 at the CRs. Although average dN/dS values for some HVRs approximate 1, thus suggesting neutrality, we note that these dN/dS values are averaged across codon sites, as well as across groups of sequences within single HVRs. On the other hand, CRs are clear targets of strong purifying selection, likely due to structural constraints.
Concordant with a large impact of recombination and diversifying selection, allele diversity per protein type indicates that WSP is a highly unstable protein. WSP shows the lowest number of synonymous variants per protein type and the lowest synonymous diversity when compared to any of the five MLST housekeeping genes, despite the use of a much larger dataset for WSP. This suggests that any single protein haplotype does not persist for long periods of time and that selection for amino acid diversification is likely ongoing. Indeed, all nucleotide substitutions observed among synonymous allele variants are unique, with no homoplasic events, supporting their recent and independent divergence. Similarly, identical HVRs motifs found in distinct proteins typically show a very low synonymous divergence, which suggests that either they have recombined quite recently, or more likely that HVR synonymous variants are particularly unstable after settling into a new allele and are soon target of nonsynonymous substitutions. This implies that those HVR variants that spread via recombination into different WSP proteins can rapidly become targets of diversifying selection and are thus adaptive to some extent. Evidence that OMP genetic diversity can play a crucial role in host-symbiont interaction comes from the native symbiont Sodalis of the tsetse fly Glossina morsitans, where polymorphisms at the exposed loop of the outer membrane protein OmpA were shown to mediate host tolerance, determining the host/symbiont type of interaction (pathogenic and not) . We speculate that the extensive variation at WSP extracellular loops could also play a similar role in escaping or down-regulating the immune system by means of rapid turnover of exposed amino acids.
Population structure of WSP
Previous studies have shown that WSP-based relationships are typically incongruent with inferences based on other Wolbachia housekeeping genes, suggesting that WSP is often horizontally transferred as a single gene, uncoupled from the rest of the genome . While the same or similar wsp alleles can often occur in otherwise very divergent strains, and therefore reconstructions of strain relationships based solely on wsp should not be trusted, on average closely related host taxa tended to harbor strains with significantly closer WSP sequences than observed between strains from more phylogenetically distant hosts. This appears the result of a preferential transfer of the entire Wolbachia strain among closely related hosts (or codivergence of strains) rather than WSP alone, as supported by previous studies using MLST data [59, 60]. However, because wsp represents the only genetic information for the majority of isolates included in this study, the two scenarios cannot be discriminated at this time.
Despite an overall non-random association of WSP with the host genus, few wsp haplotypes (e.g. wsp-23 and wsp-10) were widespread across a large host taxonomical range. It is noteworthy that wsp-23 and 10 alleles are typically found in two of the most widespread Wolbachia strains identified by MLST (ST-13 and ST-19 respectively, [57, 61]), suggesting that wsp distribution in this case largely reflects the distribution of these two Wolbachia strains.
Large recombination impact, diversifying selection, lack of strong compositional and structural constraints in WSP extracellular loops, and frequent horizontal gene transfer are signature features of adaptive evolution. These features are typically found in proteins targeted by the adaptive immune system (such as OMPs of vertebrate pathogens) [3, 13]; however, Wolbachia infect invertebrates only. There is growing interest in understanding whether and how Wolbachia escape or down-regulate the host immune responses so that they can exist within host cells. By combining the structural analysis with the microevolutionary analysis of WSP, we can speculate that the extracellular loops contain peptide motifs that serve to evade or inhibit host detection, aiding in the early settlement and persistence of Wolbachia into a new host. Biochemical investigations exploring binding properties of WSP are currently ongoing and will help elucidating the role of WSP in the invertebrate hosts.
We would like to thank Thomas Girke (University of Riverside, CA) for the valuable help with bioinformatic tools. This project was funded by the U.S. National Science Foundation grant EF-0328363 to John H. Werren.
- Lin J, Huang S, Zhang Q: Outer membrane proteins: key players for bacterial adaptation in host niches. Microbes Infect. 2002, 4 (3): 325-331. 10.1016/S1286-4579(02)01545-9.View ArticlePubMedGoogle Scholar
- Koebnik R, Locher KP, Van Gelder P: Structure and function of bacterial outer membrane proteins: barrels in a nutshell. Mol Microbiol. 2000, 37 (2): 239-253. 10.1046/j.1365-2958.2000.01983.x.View ArticlePubMedGoogle Scholar
- Ohashi N, Zhi N, Zhang Y, Rikihisa Y: Immunodominant major outer membrane proteins of Ehrlichia chaffeensis are encoded by a polymorphic multigene family. Infect Immun. 1998, 66 (1): 132-139.PubMed CentralPubMedGoogle Scholar
- Mes TH, van Putten JP: Positively selected codons in immune-exposed loops of the vaccine candidate OMP-P1 of Haemophilus influenzae. J Mol Evol. 2007, 64 (4): 411-422. 10.1007/s00239-006-0021-2.PubMed CentralView ArticlePubMedGoogle Scholar
- Vandeputte-Rutten L, Bos MP, Tommassen J, Gros P: Crystal structure of Neisserial surface protein A (NspA), a conserved outer membrane protein with vaccine potential. J Biol Chem. 2003, 278 (27): 24825-24830. 10.1074/jbc.M302803200.View ArticlePubMedGoogle Scholar
- Zheng Y, Roberts RJ, Kasif S: Identification of genes with fast-evolving regions in microbial genomes. Nucleic Acids Res. 2004, 32 (21): 6347-6357. 10.1093/nar/gkh935.PubMed CentralView ArticlePubMedGoogle Scholar
- Wimley WC: The versatile beta-barrel membrane protein. Curr Opin Struct Biol. 2003, 13 (4): 404-411. 10.1016/S0959-440X(03)00099-X.View ArticlePubMedGoogle Scholar
- Haake DA, Suchard MA, Kelley MM, Dundoo M, Alt DP, Zuerner RL: Molecular evolution and mosaicism of leptospiral outer membrane proteins involves horizontal DNA transfer. J Bacteriol. 2004, 186 (9): 2818-2828. 10.1128/JB.186.9.2818-2828.2004.PubMed CentralView ArticlePubMedGoogle Scholar
- Jiggins FM: Adaptive evolution and recombination of Rickettsia antigens. J Mol Evol. 2006, 62 (1): 99-110. 10.1007/s00239-005-0080-9.PubMed CentralView ArticlePubMedGoogle Scholar
- Malorny B, Morelli G, Kusecek B, Kolberg J, Achtman M: Sequence diversity, predicted two-dimensional protein structure, and epitope mapping of neisserial Opa proteins. J Bacteriol. 1998, 180 (5): 1323-1330.PubMed CentralPubMedGoogle Scholar
- Zhang JZ, Guo H, Winslow GM, Yu XJ: Expression of members of the 28-kilodalton major outer membrane protein family of Ehrlichia chaffeensis during persistent infection. Infect Immun. 2004, 72 (8): 4336-4343. 10.1128/IAI.72.8.4336-4343.2004.PubMed CentralView ArticlePubMedGoogle Scholar
- Hobbs MM, Seiler A, Achtman M, Cannon JG: Microevolution within a clonal population of pathogenic bacteria: recombination, gene duplication and horizontal genetic exchange in the opa gene family of Neisseria meningitidis. Mol Microbiol. 1994, 12 (2): 171-180. 10.1111/j.1365-2958.1994.tb01006.x.View ArticlePubMedGoogle Scholar
- Callaghan MJ, Buckee CO, Jolley KA, Kriz P, Maiden MC, Gupta S: The effect of immune selection on the structure of the meningococcal opa protein repertoire. PLoS Pathog. 2008, 4 (3): e1000020-10.1371/journal.ppat.1000020.PubMed CentralView ArticlePubMedGoogle Scholar
- Barbet AF, Lundgren A, Yi J, Rurangirwa FR, Palmer GH: Antigenic variation of Anaplasma marginale by expression of MSP2 mosaics. Infect Immun. 2000, 68 (11): 6133-6138. 10.1128/IAI.68.11.6133-6138.2000.PubMed CentralView ArticlePubMedGoogle Scholar
- Sadd BM, Schmid-Hempel P: Insect immunity shows specificity in protection upon secondary pathogen exposure. Curr Biol. 2006, 16 (12): 1206-1210. 10.1016/j.cub.2006.04.047.View ArticlePubMedGoogle Scholar
- Waterfield NR, Wren BW, Ffrench-Constant RH: Invertebrates as a source of emerging human pathogens. Nat Rev Microbiol. 2004, 2 (10): 833-841. 10.1038/nrmicro1008.View ArticlePubMedGoogle Scholar
- Braig HR, Zhou W, Dobson SL, O'Neill SL: Cloning and characterization of a gene encoding the major surface protein of the bacterial endosymbiont Wolbachia pipientis. J Bacteriol. 1998, 180 (9): 2373-2378.PubMed CentralPubMedGoogle Scholar
- Hilgenboecker K, Hammerstein P, Schlattmann P, Telschow A, Werren JH: How many species are infected with Wolbachia?--A statistical analysis of current data. FEMS Microbiol Lett. 2008, 281 (2): 215-220. 10.1111/j.1574-6968.2008.01110.x.PubMed CentralView ArticlePubMedGoogle Scholar
- Hotopp JC, Lin M, Madupu R, Crabtree J, Angiuoli SV, Eisen JA, Seshadri R, Ren Q, Wu M, Utterback TR, et al: Comparative genomics of emerging human ehrlichiosis agents. PLoS Genet. 2006, 2 (2): e21-10.1371/journal.pgen.0020021.View ArticleGoogle Scholar
- Werren JH: Biology of Wolbachia. Annu Rev Entomol. 1997, 42: 587-609. 10.1146/annurev.ento.42.1.587.View ArticlePubMedGoogle Scholar
- Werren JH, Baldo L, Clark ME: Wolbachia: Master Manipulators of Invertebrate Biology. Nature Reviews Microbiology. 2008, 6 (10): 741-51. 10.1038/nrmicro1969.View ArticlePubMedGoogle Scholar
- Serbus LR, Casper-Lindley C, Landmann F, Sullivan W: The Genetic and Cell Biology of Wolbachia-host Interactions. Annu Rev Genet. 2008, 42: 683-707. 10.1146/annurev.genet.41.110306.130354.View ArticlePubMedGoogle Scholar
- Bandi C, Trees AJ, Brattig NW: Wolbachia in filarial nematodes: evolutionary aspects and implications for the pathogenesis and treatment of filarial diseases. Vet Parasitol. 2001, 98 (1-3): 215-238. 10.1016/S0304-4017(01)00432-0.View ArticlePubMedGoogle Scholar
- Hoerauf A, Mand S, Adjei O, Fleischer B, Buttner DW: Depletion of wolbachia endobacteria in Onchocerca volvulus by doxycycline and microfilaridermia after ivermectin treatment. Lancet. 2001, 357 (9266): 1415-1416. 10.1016/S0140-6736(00)04581-5.View ArticlePubMedGoogle Scholar
- Hosokawa T, Koga R, Kikuchi Y, Meng XY, Fukatsu T: Wolbachia as a bacteriocyte-associated nutritional mutualist. Proc Natl Acad Sci USA. 107 (2): 769-774. 10.1073/pnas.0911476107.
- Brattig NW, Bazzocchi C, Kirschning CJ, Reiling N, Buttner DW, Ceciliani F, Geisinger F, Hochrein H, Ernst M, Wagner H, et al: The major surface protein of Wolbachia endosymbionts in filarial nematodes elicits immune responses through TLR2 and TLR4. J Immunol. 2004, 173 (1): 437-445.View ArticlePubMedGoogle Scholar
- Porksakorn C, Nuchprayoon S, Park K, Scott AL: Proinflammatory cytokine gene expression by murine macrophages in response to Brugia malayi Wolbachia surface protein. Mediators Inflamm. 2007, 2007: 84318-10.1155/2007/84318.PubMed CentralView ArticlePubMedGoogle Scholar
- Bazzocchi C, Comazzi S, Santoni R, Bandi C, Genchi C, Mortarino M: Wolbachia surface protein (WSP) inhibits apoptosis in human neutrophils. Parasite Immunol. 2007, 29 (2): 73-79. 10.1111/j.1365-3024.2006.00915.x.View ArticlePubMedGoogle Scholar
- Morchon R, Bazzocchi C, Lopez-Belmonte J, Martin-Pacho JR, Kramer LH, Grandi G, Simon F: iNOs expression is stimulated by the major surface protein (rWSP) from Wolbachia bacterial endosymbiont of Dirofilaria immitis following subcutaneous injection in mice. Parasitol Int. 2007, 56 (1): 71-75. 10.1016/j.parint.2006.10.003.View ArticlePubMedGoogle Scholar
- Siozios S, Sapountzis P, Ioannidis P, Bourtzis K: Wolbachia symbiosis and insect immune response. Insect Science. 2008, 15 (1): 89-100.View ArticleGoogle Scholar
- Baldo L, Lo N, Werren JH: Mosaic nature of the wolbachia surface protein. J Bacteriol. 2005, 187 (15): 5406-5418. 10.1128/JB.187.15.5406-5418.2005.PubMed CentralView ArticlePubMedGoogle Scholar
- Ou YY, Gromiha MM, Chen SA, Suwa M: TMBETADISC-RBF: Discrimination of beta-barrel membrane proteins using RBF networks and PSSM profiles. Comput Biol Chem. 2008, 32 (3): 227-231. 10.1016/j.compbiolchem.2008.03.002.View ArticlePubMedGoogle Scholar
- Soding J, Biegert A, Lupas AN: The HHpred interactive server for protein homology detection and structure prediction. Nucleic Acids Res. 2005, W244-248. 10.1093/nar/gki408. 33 Web Server
- Eswar N, Webb B, Marti-Renom MA, Madhusudhan MS, Eramian D, Shen MY, Pieper U, Sali A: Comparative protein structure modeling using MODELLER. Curr Protoc Protein Sci. 2007, Chapter 2 (Unit 2): 9-PubMedGoogle Scholar
- Wang Y, Geer LY, Chappey C, Kans JA, Bryant SH: Cn3D: sequence and structure views for Entrez. Trends Biochem Sci. 2000, 25 (6): 300-302. 10.1016/S0968-0004(00)01561-9.View ArticlePubMedGoogle Scholar
- Sadowski MI, Jones DT: Benchmarking template selection and model quality assessment for high-resolution comparative modeling. Proteins. 2007, 69 (3): 476-485. 10.1002/prot.21531.View ArticlePubMedGoogle Scholar
- Bagos PG, Liakopoulos TD, Spyropoulos IC, Hamodrakas SJ: PRED-TMBB: a web server for predicting the topology of beta-barrel outer membrane proteins. Nucleic Acids Res. 2004, W400-404. 10.1093/nar/gkh417. 32 Web Server
- Kyte J, Doolittle RF: A simple method for displaying the hydropathic character of a protein. J Mol Biol. 1982, 157 (1): 105-132. 10.1016/0022-2836(82)90515-0.View ArticlePubMedGoogle Scholar
- Baldo L, Dunning Hotopp JC, Jolley KA, Bordenstein SR, Biber SA, Choudhury RR, Hayashi C, Maiden MC, Tettelin H, Werren JH: Multilocus sequence typing system for the endosymbiont Wolbachia pipientis. Appl Environ Microbiol. 2006, 72 (11): 7098-7110. 10.1128/AEM.00731-06.PubMed CentralView ArticlePubMedGoogle Scholar
- The Wolbachia wsp database. [http://pubmlst.org/wolbachia/wsp/]
- Rozas J, Rozas R: DnaSP, DNA sequence polymorphism: an interactive program for estimating population genetics parameters from DNA sequence data. Comput Appl Biosci. 1995, 11 (6): 621-625.PubMedGoogle Scholar
- Rozas J, Sanchez-DelBarrio JC, Messeguer X, Rozas R: DnaSP, DNA polymorphism analyses by the coalescent and other methods. Bioinformatics. 2003, 19 (18): 2496-2497. 10.1093/bioinformatics/btg359.View ArticlePubMedGoogle Scholar
- Swofford DL: PAUP*: Phylogenetic Analysis Using Parsimony (*and other methods). 2000, Sunderland, Mass: Sinauer AssociatesGoogle Scholar
- Feil EJ, Li BC, Aanensen DM, Hanage WP, Spratt BG: eBURST: inferring patterns of evolutionary descent among clusters of related bacterial genotypes from multilocus sequence typing data. J Bacteriol. 2004, 186 (5): 1518-1530. 10.1128/JB.186.5.1518-1530.2004.PubMed CentralView ArticlePubMedGoogle Scholar
- Betran E, Rozas J, Navarro A, Barbadilla A: The estimation of the number and the length distribution of gene conversion tracts from population DNA sequence data. Genetics. 1997, 146 (1): 89-99.PubMed CentralPubMedGoogle Scholar
- Templeton AR, Crandall KA, Sing CF: A cladistic analysis of phenotypic associations with haplotypes inferred from restriction endonuclease mapping and DNA sequence data. III. Cladogram estimation. Genetics. 1992, 132 (2): 619-633.PubMed CentralPubMedGoogle Scholar
- Clement M, Posada D, Crandall KA: TCS: a computer program to estimate gene genealogies. Mol Ecol. 2000, 9 (10): 1657-1659. 10.1046/j.1365-294x.2000.01020.x.View ArticlePubMedGoogle Scholar
- Tajima F: Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics. 1989, 123 (3): 585-595.PubMed CentralPubMedGoogle Scholar
- Yang Z: PAML: a program package for phylogenetic analysis by maximum likelihood. Comput Appl Biosci. 1997, 13 (5): 555-556.PubMedGoogle Scholar
- Yang Z, Nielsen R: Codon-substitution models for detecting molecular adaptation at individual sites along specific lineages. Mol Biol Evol. 2002, 19 (6): 908-917.View ArticlePubMedGoogle Scholar
- Nei M, Gojobori T: Simple methods for estimating the numbers of synonymous and nonsynonymous nucleotide substitutions. Mol Biol Evol. 1986, 3 (5): 418-426.PubMedGoogle Scholar
- Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Higgins DG: The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 1997, 25 (24): 4876-4882. 10.1093/nar/25.24.4876.PubMed CentralView ArticlePubMedGoogle Scholar
- Hall TA: BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucl Acids Symp Ser. 1999, 41: 95-98.Google Scholar
- Weiss BL, Wu Y, Schwank JJ, Tolwinski NS, Aksoy S: An insect symbiosis is influenced by bacterium-specific polymorphisms in outer-membrane protein A. Proc Natl Acad Sci USA. 2008, 105 (39): 15088-15093. 10.1073/pnas.0805666105.PubMed CentralView ArticlePubMedGoogle Scholar
- Anisimova M, Nielsen R, Yang Z: Effect of recombination on the accuracy of the likelihood method for detecting positive selection at amino acid sites. Genetics. 2003, 164 (3): 1229-1236.PubMed CentralPubMedGoogle Scholar
- Hobbs MM, Malorny B, Prasad P, Morelli G, Kusecek B, Heckels JE, Cannon JG, Achtman M: Recombinational reassortment among opa genes from ET-37 complex Neisseria meningitidis isolates of diverse geographical origins. Microbiology. 1998, 144 (Pt 1): 157-166. 10.1099/00221287-144-1-157.View ArticlePubMedGoogle Scholar
- Palmer GH, Futse JE, Knowles DP, Brayton KA: Insights into mechanisms of bacterial antigenic variation derived from the complete genome sequence of Anaplasma marginale. Ann N Y Acad Sci. 2006, 1078: 15-25. 10.1196/annals.1374.002.View ArticlePubMedGoogle Scholar
- Baldo L, Werren JH: Revisiting Wolbachia supergroup typing based on WSP: spurious lineages and discordance with MLST. Curr Microbiol. 2007, 55 (1): 81-87. 10.1007/s00284-007-0055-8.View ArticlePubMedGoogle Scholar
- Russell JA, Goldman-Huertas B, Moreau CS, Baldo L, Stahlhut JK, Werren JH, Pierce NE: Specialization and Geographic Isolation among Wolbachia Symbionts from Ants and Lycaenid Butterflies. Evolution. 2008, 63 (3): 624-40. 10.1111/j.1558-5646.2008.00579.x.View ArticlePubMedGoogle Scholar
- Baldo L, Ayoub NA, Hayashi CY, Russell JA, Stahlhut JK, Werren JH: Insight into the routes of Wolbachia invasion: high levels of horizontal transfer in the spider genus Agelenopsis revealed by Wolbachia strain and mitochondrial DNA diversity. Mol Ecol. 2008, 17 (2): 557-569.View ArticlePubMedGoogle Scholar
- Stahlhut JK, Desjardins CA, Clark ME, Baldo L, Russell JA, Werren JH, Jaenike J: The mushroom habitat as an ecological arena for global exchange of Wolbachia. Mol Ecol.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.