The origin of replication of C. crescentus is located between orthologs of hemE (CC_3763) and a conserved hypothetical protein, COG1806, and contains DnaA, CtrA and IHF binding sites 37: In R. prowazekii, the ori region is also located between the hemE (RP_885) and COG1806 orthologs.

Based on a slightly modified approach and findings of Brassinga et al. 37, the Wolbachia ori was predicted in silico. Our approach was based on four criteria: (a) position near either hemE or COG1806 orthologs; (b) an intergenic region (that is, the putative ori region) containing an appropriate number of binding sites for the DnaA, CtrA and IHF factors; (c) genome-wide searches confirming that no other appropriately sized intergenic region (>300 bp) contains a significant number of these three characteristic binding sites and (d) the AT-content of the predicted origins that is higher than the average for the respective genome.

First, the Wolbachia orthologs of hemE and COG1806 were searched in the recently published Wolbachia wMel genome 15 and identified as the loci WD_1028 and WD_0341, respectively. The great distance between the two genes in Wolbachia (>600 kb) indicated no conservation of the hemE-COG1806 region in Wolbachia. The lack of conservation is most likely due to a single chromosomal rearrangement resulting in two possible locations for ori: WD_0340-WD_0341 or WD_1027-WD_1028. The first region is between the heme exporter WD_0340 (ccmC) and the COG1806 homolog (WD_0341). The second region is between the CBS domain protein WD_1027 (referred to in the present study after its NCBI COG number, COG1253) and uroporphyrinogen decarboxylase WD_1028 (hemE).

Both regions were examined for the presence of DnaA-, CtrA-, and IHF-binding sites, essential components of ori 37. Only the intergenic region (IGR) between the COG1253 and hemE gene (position 988364–988765, 402 bp) was found to have all three characteristic binding sites. In addition, out of all 110 IGRs of greater than 300 bp, this region has the highest number of CtrA, DnaA and IHF binding sites (eight total, p = 1/110 = 0.009), the highest density of total binding domains per IGR bp (0.020 per bp). The next closest is the IGR between WD_0248-WD_0249, which has a much lower total number of binding sites (six) and binding site density (0.013 per bp). However, it lacks putative DnaA boxes, a prerequisite for an ori region. The putative ori region is one of only two IGRs > 300 bp that have all three binding site types (p = 2/110 = 0.018). However, the IGR between WD_0100-WD_0102 has only 4 total binding domains (compared to 8) and, due to it's larger size (545 bp) a binding site density 1/3 that of the putative ori (0.007 vs 0.020). These data, combined with the association of the WD_1027-WD_1028 IGR with the same genes shown to be flanking the ori region in C. crescentus and R. prowazekii, provides compelling evidence that this is most likely, the origin of replication. More specifically, there are 3 DnaA, 4 CtrA and 1 IHF binding sites in the 404 bp region between COG1253 and hemE. In contrast, the region between ccmC and COG1806 has only a single IHF binding site and is only 130 bp long, which is rather short compared to the ori of related bacteria (e.g. R. prowazekii and C. crescentus, have replication origins >400 bp 37).

Another feature of origins is that they typically contain an AT-rich region to facilitate dissociation of the DNA during replication initiation 17. As a result, the overall AT-content of ori is higher when compared to the respective genome. Consistent with this feature, the putative wMel origin is 76% AT-rich compared to the 65% average AT-content for the genome. In addition, the ori region is significantly more AT-rich than the average found for other intergenic regions of >300 bp. Of 110 intergenic regions >300 bp, only 6 have AT-content greater or equal to the ori (p = 6/110 = 0.055). Indeed, the AT-contents of all origins detected in this study were higher than their respective genome averages (Table 1). It has been proposed in E. coli that this exceptionally AT-rich DNA is the place where the initiator factor of DNA replication, the DnaA protein, first unwinds the origin 3839. These AT-rich sequences are also conspicuous because they are tandem repeats. For example, the E. coli oriC region is composed of three imperfect 13-mer sequences with the consensus sequence GATCTNTTNTTTT 38. Similar tandem repeats were detected in all bacterial species and strains studied except in A. marginale, E. canis and N. sennetsu. In most cases the repeats were 10–12 bp long and appeared in two copies (data not shown).

These lines of in silico evidence suggest that Wolbachia has an origin of replication on its chromosome that lies between the COG1253 and hemE genes. One of the boundary genes, hemE (WD_1028), encodes for uroporphyrinogen decarboxylase – a component in heme biosynthesis. The other, COG1253 (WD_1027), encodes for a CBS domain protein. CBS domains have been shown to bind ligands with an adenosyl group such as AMP, ATP and S-AdoMet 40. Using the publicly available HMMer program 41, it was found that this protein has a CBS domain pair and a CorC_HlyC transporter associated domain.

The putative ori regions were also identified in other complete and/or partial publicly available Wolbachia genomes by using the above mentioned criteria. The orthologs of the COG1253 (WD_1027) and hemE (WD_1028) genes were identified by BLAST searches 42 of the genomes of the wAna 43, wSim 43, wWil 44, wPip 45 and wBm Wolbachia strains 14 (see Table 1). Both flanking genes are highly conserved in position and in sequence. Based on criteria set forth by Zyskind et al. 46, the regions of the wMel and wBm strains for which their genome sequence is available 1415, were compared indicating that: (a) 266 of 347, or 77%, of the amino acids are conserved, while 790 of 1044 (76%) of the nucleotides are conserved in the flanking gene hemE and (b) 216 of 279, or 77%, of the amino acids are conserved, while 609 of 819 (74%) of the nucleotides are conserved in the flanking gene COG1253. The ori regions of the two genomes present 73% identity at the nucleotide level. The putative ori-region-related binding sites for the wAna, wSim and wWil strains (all members of the A-supergroup), wPip strain (B-supergroup), and the wBm strain (D-supergroup) have been depicted [see Additional Table 3]. A schematic representation is depicted in Figure 1 while the length and genome coordinates are shown in Table 1 (where applicable).

Based on the consensus sequence of the flanking genes of the putative ori region in Wolbachia [see Additional Text File 1], a PCR strategy [see Additional Figure 1 and Additional Table 2] was developed to amplify and sequence the putative ori regions of other Wolbachia strains infecting different Drosophila species: wMelPop, wRi, wAu, wHa, wYak, wTei, wSan, wNo, wMa and wMau [see Additional Table 1] 4748495051525354555657. Further sequencing of the putative ori region of an additional 37 Wolbachia strains from diverse hosts confirmed the overall characteristics of this region both in size and sequence [see Figure 1 and Additional Table 1]. Wolbachia strains belong to eight supergroups A-H 58596061. The present study includes representative strains from supergroups A, B, D, E and F. All strains present three DnaA and a single IHF boxes while they differ in the number of CtrA binding sites (two to seven). The sequenced putative ori regions of some Wolbachia strains present a peculiar pattern of binding sites [see Figure 1 and Additional Table 1]. It is worth noting that the pseudoscorpion Wolbachia ori does not have any CtrA binding motifs while it presents three DnaA boxes and a single IHF binding site suggesting either the existence of a very diverged CtrA binding motif (which could not be identified when compared to the consensus sequence used in the present study) or the CtrA factor may not play an important regulatory role in the DNA replication of this Wolbachia strain.

We tested for selective constraints on the binding sites for DnaA, CtrA and IHF by comparing the frequencies of polymorphic base positions in binding domain sites versus non-binding domain sites within the same IGR in different clades containing closely related Wolbachia strains (B1, B2 and A1, figure 2). The proportion of positions showing polymorphism was lower in the binding domain sites than non-binding sites for all three clades (B1 0.08 vs 0.19, p = 0.035; B2 0 vs 0.12, p = 0.001; A1 0.02 vs 0.08, p = 0.081; Fisher Exact Test), with two showing significant reductions in polymorphisms within domain positions. Combining the data give 25.3 percent lower proportion of positions with polymorphisms in the binding domains relative to flanking non-binding positions (p < 0.0001 Fisher Exact Test), indicating some selective constraint on these positions.

Applying the same criteria, the origin of DNA replication was predicted in silico for the closely related A. marginale St. Maries, A. phagocytophilum HZ, E. ruminantium Welgevonden, E. ruminantium Gardel, E. canis Jake and E. chaffeensis Arkansas (Table 1). Annotation, BLAST analysis and/or domain searches (using HMMer) were used to identify the two flanking genes of the putative ori regions (Table 1).

The length of the predicted ori region differs markedly by 300 bp between the two Anaplasma spp 3436. In A. marginale, the ori is 620 bp; in A. phagocytophilum it is just 385 bp (Table 1). Both species have a single CtrA binding site and a single IHF binding site that differ in both sequence and position. A. marginale has two DnaA boxes while A. phagocytophilum has four [see Figure 1 and Additional Table 1].

Origin of replication features of the different Ehrlichia spp. examined in this study are summarized in Table 1. Multiple possible IHF binding sites are found for each member (depicted in Figure 1). Of special interest is the absence of CtrA binding sites in E. ruminantium. However, if two mismatches are allowed two putative CtrA sites are found.

The putative ori region of the four different Rickettsia spp. examined here was predicted by searching for the previously described hemE/COG1806 37 [see Figure 1, Table 1 and Additional Table 1]. Following the same type of analysis, two putative ori regions can be identified in N. sennetsu that are unlike other bacteria examined in this study [see Table 1 and Additional Table 1]. The first region is located between COG1806 and COG1253, is 553 bp long, and has no DnaA, no CtrA and no IHF binding site. This intergenic region contains a hypothetical gene (92 amino acids long) and a predicted tRNA-Arg. The second region is located between hemE and an uncharacterized phage protein, is 408 bp long, and has no DnaA, no CtrA and two IHF binding sites. Although less clearly defined by the binding sites, this latter region is close to the shift in GC-skew 36 and its AT-content is higher when compared to the respective genome. Additionally, when it is compared with the first one, it appears to have more IHF binding sites in its immediate neighborhood. The above data indicate that the second region may be the putative ori region of N. sennetsu (Figure 1); however, it is possible that the proposed search criteria may not be appropriate for the identification of the origin of replication of N. sennetsu.

Taken together, the in silico approach can predict the putative origin of replication of most bacterial species closely related to Wolbachia including Ehrlichia, Anaplasma, and Rickettsia. In addition, the Maximum Likelihood derived trees demonstrate a clear concordance of the ori region and 16S rRNA phylogenies for the Rickettsia genus, Anaplasma/Ehrlichia genera, and Wolbachia supergroups [see Additional Figure 2].

Significant recombination events at the ori region of Wolbachia were detected within both supergroup A and B and between the two supergroups (MaxChi, P < 0.001). Most of the recombination breakpoints fell at one of the two edges of the intergenic region (see graph in Figure 3A). The majority of binding sites (excluding DnaA_2 and CtrA_2) occur within a region that experienced a similar low number of recombination events, suggesting that usually these sites, when recombining, are exchanged as a unique sequence tract. A likely recombination breakpoint occurred in the nucleotide region between IHF_1 and DnaA_2 (see arrow, Fig. 3A) indicating a potential shuffling among binding sites at this region.

Most of the recombination breakpoints per sequence were single and thus involved recombination of sequences encompassing one of the two halves of the alignment. Clear examples of recombinant tracts can be visualized by simple examination of the shared nucleotide polymorphisms among triplets of sequences (see example in Figure 3B). For instance, the ori region of wAenc_B, a strain from the host A. encedon belonging to supergroup B, shows a clear recombinant pattern between A- and B-type sequences. Specifically, the first 210 bp of the sequence from the wAenc_B strain shares most of the polymorphisms (70) with the B strain wMa_B, from D. simulans, while the remaining sequence portion shares most of polymorphisms (42) with the A strain wMel-A, from D. melanogaster. As a result, wAenc_B places as a deep branch of supergroup B, separated by a great distance from other B-strains (bootstrap value, P = 82; data not shown); this apparent phylogenetic divergence is actually due to recombination within the sequence. A similar case of recombination is found for the wCalt_B strain and involves B-type sequences. The first 500 bp of the wCalt_B sequence shares most of the polymorphisms with the wVul sequence, while the following 173 bp share most polymorphisms with the wOscaB sequence. Other instances of recombination involve the Wolbachia A-strains from A. albopictus, A. sparsa, C. pennsylvanicus and S. invicta, denoted A2 in figure 2. These four ori sequences are recombinant between A and B-types sequences (MaxChi, P < 0.001).

As a non-coding region, the ori region is often overlooked and not annotated in genomes. However, this is a significant issue since it has an essential role in DNA replication. Therefore, we have established several criteria that should be used to identify the origin from genome sequencing data including: (a) boundary genes that are homologous to those of closely related bacteria, (b) an intergenic region 200–1000 bp in size, (c) presence of appropriate binding sites (primarily for DnaA and IHF), (d) an increased distribution of the appropriate binding sites when compared to other intergenic regions, (e) increased AT content relative to the genome, (f) increased homology to closely related sequences relative to the genome, and (g) a shift in GC-skew. Although all seven elements may not be present in all genomes, a significant combination of these elements should allow for a better prediction of this important region.

The presumed ori regions of the bacterial species and strains of the present study are characterized by the presence of DnaA, CtrA and IHF boxes (the exception being N. sennetsu). The DnaA and IHF boxes are present in all bacterial ori regions characterized. However, the CtrA boxes seem to be restricted, as yet, to the ori regions of the α-Proteobacteria. In C. crescentus, CtrA is a global cell cycle regulator 26 and more specifically, the response regulator protein of a two-component signal transduction signal. Upon phosphorylation by a sensor histidine kinase, CtrA binds to its corresponding binding sites and represses chromosomal replication 63. CtrA's binding sites are overlapping with a DnaA box and the IHF binding site. Thus, binding of CtrA prevents binding of DnaA and IHF 37. CtrA is degraded before the onset of the S phase by the protease ClpXP allowing DnaA and IHF binding 6465. Despite A. marginale CtrA having 59% amino acid identity with its wMel ortholog, A. marginale and A. phagocytophilum ori regions have only a single CtrA binding site, as opposed to four in wMel. E. ruminantium has no CtrA binding site in its putative ori region while its CtrA protein is 80% identical to the A. marginale ortholog. Similarly, the putative ori regions of the pseudoscorpion C. scorpioides Wolbachia strain, E. chaffeensis, E. canis, R. conorii and the filarial nematode B. malayi Wolbachia strain (wBm) have zero, one, three, two and two CtrA binding sites respectively. These observations suggest that CtrA may be dispensable in these bacteria. Interestingly, the bacteria with zero to three CtrA binding sites in their ori region are not associated with insects. In contrast, insect-associated Wolbachia and Rickettsia bacteria usually present high number of CtrA binding sites present in their ori region ranging from four to seven (the exception being R. felis which has three CtrA binding sites). Whether bacterial growth control is host-dependent and is regulated through the number of the CtrA binding sites in the origin of replication region awaits experimental confirmation.

It is worth noting that there are orthologs to dnaA, ctrA and ihfAB in the bacterial genomes studied except N. sennetsu and A. marginale. They are both missing IHF-β but have retained the IHF-α subunit [see Additional Table 3]. Another interesting observation is that the filarial nematode B. malayi Wolbachia strain (wBm) and A. phagocytophilum are missing both parA and parB, which are involved in partitioning the chromosomes 1436. This suggests that Wolbachia and Anaplasma replication may be quite interesting; experimental work is needed in order to clarify the roles of replication-associated proteins in binding the ori and initiating DNA replication initiation process.

IHF binding sites were found in all bacteria using the consensus WATCAN₅WTR 37. In some Wolbachia strains, two to three such candidate sites were found within their putative ori regions. In these cases, only the common one was retained 37. In contrast, A. marginale and A. phagocytophilum had different IHF binding sites in both in sequence and in position. Two conserved putative IHF binding sites were detected in the putative ori region of E. canis and the two E. ruminantium strains, with all of them being similar both in sequence and position while a single putative IHF binding site was present in E. chaffeensis which was the only common IHF binding site found in all four Ehrlichia strains. IHF positioning varied between the four Rickettsia species. The only exception appears to be the conservation in position as well as in sequence between R. prowazekii and R. typhi. The IHF-β subunit is missing from the genome of A. marginale. The only IHF identified, AM_006, is 43.2% identical (amino acid level) to the wMel IHF-α subunit (WD_0057). A. phagocytophilum, all four Ehrlichia, and all four Rickettsia species have both IHF subunits. N. sennetsu also lacks IHF-β [see Additional Table 3]. Whether the absence of the IHF-β subunit is somehow correlated with the divergence/absence of IHF binding sites in Anaplasma and Neorickettsia is not known.

The difference in the boundary genes observed between the closely related Anaplasma, Ehrlichia, Wolbachia and the other α-Proteobacteria is likely due to chromosomal rearrangements that have taken place in the ori region. All of the origins examined have as boundaries either the COG1253 – hemE pair (Wolbachia, Ehrlichia and Anaplasma), or the hemE – COG1806 pair (Caulobacter and Rickettsia), or the hemE – uncharacterized phage protein pair (N. sennetsu). Overlaying these observations on the phylogeny of the α-Proteobacteria 36, the most likely ancestral configuration is that found in Caulobacter and Rickettsia. Sometime after the branching of the Anaplasmataceae, a rearrangement took place that repositioned the COG1253 gene to the position of COG1806 orthologs in Wolbachia, Ehrlichia, and Anaplasma with a second rearrangement occurring in N. sennetsu. The exact nature of these rearrangements cannot be determined since the synteny of the chromosomes has been lost. As discussed earlier, the distribution of ori-specific binding sites (DnaA, CtrA and IHF) is consistent with the currently accepted phylogeny of Wolbachia.

It has been widely accepted that bacterial infection levels are positively correlated with the virulent and/or pathogenic properties of bacterial pathogens 66. Similarly, several studies have shown that Wolbachia's ability to induce reproductive and/or virulent phenotypes is positively correlated with their intra-host infection levels 1112135467. Furthermore, intracellular bacteria that are transmitted through the eggs of their hosts must be "prudent replicators" in order to ensure their transmission to future generations by preventing overgrowth that could lead to host death prior to reproduction. This raises the question: is the DNA replication and growth of an obligate intracellular bacterium under bacterial control, host control, phage control or some combination? Overreplication by the "popcorn" strain of Wolbachia in D. melanogaster suggests that the bacterial strain has a strong influence on replication rates within the host 54, although effects of the host also occur 12 and effects of phage have not been tested 10.

The results of the present study clearly indicate that no major differences could be detected between the ori regions of Wolbachia strains differing in ability to induce CI, virulence, parthenogenesis or feminization. All Wolbachia strains presented the same organization in their putative origin of replication: COG1253 and hemE as boundary genes, three DnaA boxes, two to seven CtrA boxes and a single IHF binding site. A notable exception is the pseudoscorpion Wolbachia strain which presents zero CtrA binding sites. The possibility that DNA binding specificity of CtrA has changed in these bacteria is an intriguing hypothesis. CtrA contains two domains, a receiver domain at the N-terminus and a transcriptional regulatory (DNA binding) effector domain at the C-terminus. These domains are nearly identical between wMel and wBm with only one positive amino acid substitution in each domain. All other changes are located at the far C-terminus where no domains are found. Since a single amino acid change can greatly affect DNA binding, it is conceivable that these two positive substitutions can alter CtrA's binding specificity. A second hypothesis is that CtrA's DNA-binding specificity did not change in wBm, and only the number of CtrA binding sites reduced. This CtrA binding site reduction may be associated with the mutualistic nature of the Wolbachia – B. malayi association and may have resulted in a novel control of bacterial replication. These hypotheses need to be tested experimentally.

Evidence presented here for the prediction of ori location, is based on the boundary genes and on the in silico finding of characteristic DnaA, CtrA, and IHF binding sites that have been experimentally confirmed only in C. cresentus and partially in R. prowazekii 37. Lack of a robust genetic transformation system for Wolbachia, Ehrlichia, Neorickettsia and most Anaplasma and Rickettsia precludes experimental verification in the bacterial species of the present study. However, significant progress has been made toward the development of a robust genetic transformation for A. phagocytophilum and R. prowazekii using homologous recombination-based and transposon-based approaches 686970. The fact that these bacteria (both Wolbachia and relatives) can be maintained in different cell lines 37172, the availability of complete genomic information 141534353673747576 and, the presence in some of them, such as Wolbachia, of endogenous phages and insertion sequences 7778 will certainly facilitate current efforts for the genetic transformation of these intracellular bacteria.

Various insect hosts and Wolbachia strains were used for DNA extraction in the present study [see Additional Table 1]. Fly stocks were reared on standard corn flour – sugar – yeast medium at 25°C. Bacterial DNA was extracted using the DNeasy Tissue Kit (Qiagen) according to the manufacturer's instructions.

A PCR strategy [see Results, Additional Figure 1 and Additional Table 2] was developed to amplify the ori region from the following Wolbachia strains belonging to supergroup A and B: wMel, wMelPop, wAu, wRi, wHa, wYak, wTei, wSan, wNo, wMa and wMau. Various PCR primers were used for the amplification reactions [see Additional Table 1]. PCR reaction mixtures contained 1× Taq buffer (750 mM Tris-Cl pH 8.8, 200 mM (NH₄)₂SO₄, 0.1% Tween 20), 2 mM MgCl₂, 125 μM dNTPs, 12.5 pmol of each primer, 1 unit Taq polymerase (Promega) and 25 ng/μl template DNA and were cycled with an initial denaturing step at 94°C for 10 minutes, 35 cycles of 94°C for 30 seconds, 54°C for 30 seconds and 72°C for 3 minutes followed by a final extension at 72°C for 10 minutes. PCR reactions were purified with Qiagen nucleotide removal kit or Qiagen gel extraction kit depending on the existence or not, of byproducts. Sequencing was performed by Macrogen (Korea). Sequence trace files from sequencing reactions were processed using the DNAStar 5.0 suite of programs.

A smaller region of ori was amplified and sequenced from a greater variety of Wolbachia strains according to the methods mentioned in 79. Reactions to generate these smaller amplicons were attempted for 52 Wolbachia strains from almost all described supergroups using standard PCR conditions with HotStarTaq (Qiagen), according to the manufacturer's suggestions, with 0.5 μM of each WD_1027_R and WD_1028_R. Reactions were initiated with a 15 minute incubation at 95°C followed by 50 cycles of 95°C for 30 seconds, 55°C for 30 seconds, 72°C for 1 min. The primers WD_1027_R and WD_1028_R had 5'-tags of M13 forward and reverse primer sequences, respectively, to serve as anchors for the degenerate primers in later stages of amplification and were later used for sequencing. Amplification reactions (8 μL) were treated with 0.5 U shrimp alkaline phosphatase and 1.0 U exonuclease I (Amersham) in the supplied buffer. Sequencing reactions were performed at the J. Craig Venter Joint Technology Center (Rockville, MD) with M13 forward and reverse primers. Assembly was done with the TIGR assembler and manually curated with Cloe (ClosureEditor, a TIGR program for editing assemblies). The alignment was initially generated using CLUSTALW and curated in Bioedit v. 7.0.4.1.

The nucleotide sequences reported in this study have been deposited in GenBank under accession numbers DQ498834 – DQ498882.

Sequence information was used from publicly available genome data (GenBank) of Wolbachia and other closely related bacterial species and strains (Table 1).

For ori-boundary identification, genes homologous to COG1253 (see results for the new proposed gene name of the CBS domain protein) and hemE were identified using BLAST. Moreover, the CBS domain protein gene was searched for conserved domains using the HMMer program 41 and the Pfam_fs hidden markov model (HMM) database.

Binding sites were identified using local perl scripts based on the previously determined consensus sequences 37. For the CtrA binding site (TTAA-N₇-TTAA), a single mismatch was allowed only in the A's and only if these mismatches followed the looser consensus TTWW-N₇-TTWW described previously 37. For the DnaA binding site, the consensus TTATNCACA was used 80. For the IHF-binding site, the consensus WATCAN₅WTR was used 8182.

In order to examine the distribution of the CtrA-, IHF-, and DnaA-binding sites in the intergenic regions of each genome were identified using fuzznuc from the EMBOSS package 83 and locally developed scripts.

The alignment for the figure 2 tree was generated using ClustalX 1.83 for Windows 84 with the default parameters and trimmed using BioEdit for Windows 85. Maximum likelihood (ML) methods were used to infer phylogenetic relationships. Prior to ML analysis, a DNA substitution model was selected using Modeltest v3.06 and the Akaike information criterion (AIC). The selected model of evolution was based on a 382 bp ori region with indels excluded (TVM+G). ML heuristic searches were performed using 100 random taxon addition replicates with tree bisection and reconnection (TBR) branch swapping. ML bootstrap support was determined using 100 bootstrap replicates, each using 10 random taxon addition replicates with TBR branch swapping. Searches were performed in parallel on a Beowulf cluster using a clusterpaup program and PAUP version 4.0b10 86.

Conservation of binding domains in the Wolbachia ori region was investigated within three clades of closely related Wolbachia (B1, B2 and A1). It should be noted here that the A2 clade (see Figure 2) was not used because it is recombinant between the A and B supergroups in the ori region (see text for details). Binding site positions were identified based on a reference strain within each clade; this was straightforward due to relatively low variation within each clade. Positions were then classified as binding or non-binding. Positions within binding domains with free nucleotide designations (N) were classified as non-binding sites. In cases where binding domains were overlapping, if both positions were N sites, then the position was classified as non-binding, otherwise it was classified as a binding position. All positions within the IGR were then scored as polymorphic (containing at least one polymorphism) or not. Fisher Exat Tests were used to compare the proportions of polymorphic and non-polymorphic positions in binding sites versus non-binding sites.

The ori region of Wolbachia was searched for recombination signature by using the Maximum Chi Square (MaxChi2) program, implemented in the RDP2 program 87. MaxChi is a local method that uses a sliding window approach to search for putative recombination breakpoints in a set of aligned DNA sequences. Significant discrepancies between the two partitions of the window are calculated based on the difference in the number of variable sites (VI) on either sides of the central partition. A Chi-square statistics is applied. The step size was set to 20 nucleotides and the window size set to 20 VI, gaps were included and a Bonferroni correction was applied. The highest acceptable P-value cut-off was set to 0.001 and 1000 permutation were generated. We analyzed only strains belonging to supergroup A and B and having sequences of equal length encompassing a portion of both flanking genes. The final alignment was 673 base pairs long and included 38 strains (19 A- and 19 B-strains). The alignment was partitioned in dataset A and B (from the two supergroups) and recombination analyses were run on both single and combined datasets.