The RuvA Homologues from Mycoplasma genitalium and Mycoplasma pneumoniae Exhibit Unique Functional Characteristics

The DNA recombination and repair machineries of Mycoplasma genitalium and Mycoplasma pneumoniae differ considerably from those of gram-positive and gram-negative bacteria. Most notably, M. pneumoniae is unable to express a functional RecU Holliday junction (HJ) resolvase. In addition, the RuvB homologues from both M. pneumoniae and M. genitalium only exhibit DNA helicase activity but not HJ branch migration activity in vitro. To identify a putative role of the RuvA homologues of these mycoplasmas in DNA recombination, both proteins (RuvAMpn and RuvAMge, respectively) were studied for their ability to bind DNA and to interact with RuvB and RecU. In spite of a high level of sequence conservation between RuvAMpn and RuvAMge (68.8% identity), substantial differences were found between these proteins in their activities. First, RuvAMge was found to preferentially bind to HJs, whereas RuvAMpn displayed similar affinities for both HJs and single-stranded DNA. Second, while RuvAMpn is able to form two distinct complexes with HJs, RuvAMge only produced a single HJ complex. Third, RuvAMge stimulated the DNA helicase and ATPase activities of RuvBMge, whereas RuvAMpn did not augment RuvB activity. Finally, while both RuvAMge and RecUMge efficiently bind to HJs, they did not compete with each other for HJ binding, but formed stable complexes with HJs over a wide protein concentration range. This interaction, however, resulted in inhibition of the HJ resolution activity of RecUMge.


Introduction
A significant proportion of the genomes of Mycoplasma pneumoniae and Mycoplasma genitalium (approximately 8% and 4%, respectively) is composed of repeated DNA elements. These elements are referred to as RepMP elements in M. pneumoniae [1,2,3] and MgPa repeats (MgPars) in M. genitalium [4,5,6]. Although the different variants of these elements show a high level of sequence homology, they are not identical. Moreover, one or more of these variants are contained within open reading frames (ORFs) that encode antigenic surface proteins. Among these proteins are P1, P40 and P90 of M. pneumoniae and MgPa and P110 of M. genitalium. As these proteins can display amino acid sequence variation within the regions encoded by the RepMP and MgPar sequences, it has been proposed that this variation originates from recombination between different variants of RepMP or MgPar [7,8,9,10,11,12,13]. Consequently, homologous recombination between the repeated DNA elements in both Mollicutes species may play a crucial role in immune evasion [14].
It has previously been suggested that the mechanism of recombination between repeated DNA elements in M. pneumoniae and M. genitalium is similar to that of general homologous DNA recombination in these species [15,16]. As a consequence, these processes may utilize the same enzymatic machinery. Recent studies that were aimed at elucidation of the mechanism of recombination between repeated DNA elements therefore focused on the characterization of Mycoplasma proteins predicted to be involved in homologous DNA recombination, such as RecA [15], single-stranded DNA-binding protein (SSB) [16], RuvA [17], RuvB [18] and RecU [19,20]. The RecA proteins from M. pneumoniae and M. genitalium (RecA Mpn and RecA Mge , respectively) and the SSB protein from M. pneumoniae (SSB Mpn ) were reported to possess similar activities as their counterparts from Escherichia coli [15,16]. Both RecA Mpn and RecA Mge were found to catalyze the exchange of homologous DNA strands in an ATP-and Mg 2+dependent fashion [15]. This activity was stimulated strongly by SSB Mpn , which is a tetrameric protein that selectively binds to single-stranded DNA (ssDNA) [16].
In contrast to the SSB and RecA proteins, the RecU, RuvA and RuvB proteins from M. pneumoniae and M. genitalium displayed in vitro activities that differed considerably from those of their counterparts from other bacterial classes. Specifically, the RecU protein from M. genitalium (RecU Mge ) was found to diverge from other Holliday junction (HJ) resolving enzymes in four major aspects [19]. First and foremost, RecU Mge only displayed HJ resolvase activity in the presence of Mn 2+ and not in the presence of Mg 2+ . In contrast, the RecU homologue from Bacillus subtilis (RecU Bsu ) and the RuvC Eco and RusA Eco resolvases from E. coli possess Mg 2+ -dependent resolvase activity. Second, RecU Mge has a unique target DNA sequence, cleaving HJ substrates at the sequence 59-G / T CQPyTPuG-39. This cleavage site differs from the cleavage sites of RecU Bsu , RuvC Eco and RusA Eco (59-G / T GQC A / C -39, 59-A / T TTQ G / C -39 and 59-QCC-3, respectively) [21,22,23,24,25]. Third, unlike the RecU Bsu protein [21], RecU Mge is unable to anneal circular ssDNA to homologous, linear double-stranded DNA (dsDNA). Fourth, RecU Mge does not stably bind to long ssDNA substrates, in contrast to the RecU Bsu protein [21].
Another crucial finding regarding the RecU orthologues from M. pneumoniae and M. genitalium was the inability of M. pneumoniae to produce a functional RecU protein [19,20]. While a subset of M. pneumoniae strains (so-called subtype 2 strains) is able to express a RecU homologue (RecU Mpn ), this protein was found to be inactive in HJ-binding and -cleavage in vitro. Moreover, the other major subset of M. pneumoniae strains (subtype 1 strains) was reported to be incapable of producing a full-length RecU homologue, due to the presence of a nonsense codon in the RecU gene [19]. The inability of M. pneumoniae to produce a functional RecU protein was suggested to be (one of) the causative factor(s) of the relatively low level of homologous DNA recombination in this bacterium [19].
Unique properties were recently also attributed to the RuvB homologues from M. genitalium and M. pneumoniae (RuvB Mge and RuvB FH , respectively). In contrast to the E. coli DNA branch migration motor protein RuvB Eco , both RuvB Mge and RuvB FH were found to have RuvA-independent DNA helicase activity [18]. The activity of RuvB Mge , however, was significantly lower than that of RuvB FH . Interestingly, RuvB FH is exclusively expressed by subtype 2 strains of M. pneumoniae. The RuvB protein expressed by subtype 1 strains (RuvB M129 ) displays only marginal levels of DNA helicase activity, due to a single amino acid substitution with respect to RuvB FH [18]. Although RuvB FH did not appear to be stimulated at all by M. pneumoniae RuvA (RuvA Mpn ), the helicase activity of the RuvB Mge protein was found to be promoted by M. genitalium RuvA (RuvA Mge ) under specific reaction conditions [18].
The apparent inability of RuvA Mpn to stimulate RuvB FH activity can be caused by specific, aberrant features of the RuvB FH protein in comparison with RuvB Eco . Alternatively, RuvA Mpn itself may be unable to interact with, and/or activate, RuvB FH . In this regard, it is interesting to note that RuvA Mpn did not stimulate the branch migration activity of RuvB Eco in vitro, and could not functionally substitute for RuvA Eco in vivo (in E. coli) [17]. Thus, while the E. coli RuvA protein has a vital role in the interaction with both RuvB and the HJ resolving enzyme RuvC (within the RuvABC resolvasome), the function of RuvA Mpn within a putative branch migration and resolution complex remains enigmatic.
In this study, the activities of RuvA Mpn and RuvA Mge are characterized and compared. We show that these proteins differ considerably in (i) their affinities for branched and non-branched DNA substrates, (ii) complex formation with HJs, and (ii) their interaction with other proteins from the DNA recombination machinery.

Results
M. pneumoniae ORF MPN535 and M. genitalium ORF MG358 encode RuvA homologues The MPN535 ORF of M. pneumoniae was previously shown to encode a RuvA homologue (RuvA Mpn ) [17]. A multiple amino acid sequence alignment indeed shows significant similarities between RuvA Mpn and other (putative) RuvA proteins from gram-negative and gram-positive bacteria (Fig. 1A). While the similarity between the sequences of RuvA Mpn and RuvA Eco is relatively low (23.6% identity), a high similarity is observed between the sequences of RuvA Mpn and RuvA Mge (68.8% identity). In contrast to other members of the putative DNA recombination apparatus of M. pneumoniae, i.e. RecU and RuvB [18,19], RuvA Mpn does not differ in sequence among subtype 1 and subtype 2 strains.
Within the RuvA sequences, a relatively high level of amino acid sequence conservation is found in two so-called helix-hairpinhelix (HhH) motifs (Fig. 1A) [26]. These motifs were previously identified within domain II of RuvA Eco and were shown to be crucial for sequence-independent DNA binding by interacting with the DNA phosphate backbone of Holliday junctions (HJs) [27,28,29]. The lowest level of sequence conservation was seen in the region defined as the 'flexible linker', which separates domain II from domain III in RuvA Eco [30].
RuvA Mge and RuvA Mpn can bind to synthetic oligonucleotide substrates Both RuvA Mge and RuvA Mpn were expressed in E. coli as poly histidine (H 10 )-tagged proteins and were purified to near homogeneity using similar protocols (as described in Materials and Methods). The H 10 -tagged proteins were found to have activities that were indistinguishable from that of their non-tagged counterparts (data not shown). Because the H 10 -tagged proteins were obtained at higher concentrations and at a higher purity than their 'native' versions (.95% versus ,90% homogeneity), they were used throughout this study. The estimated molecular masses of the purified proteins matched the theoretical molecular masses of 23.7 kDa for both RuvA Mge (Fig. 1B, lane 2) and RuvA Mpn (lane 3).
To test and compare the DNA-binding characteristics of RuvA Mge and RuvA Mpn , both proteins were incubated with HJs, double-stranded (ds) and single-stranded (ss) oligonucleotide substrates, and analyzed by electrophoretic mobility shift assay (EMSA). As described before [17], two distinct complexes (complex I and complex II) were formed between RuvA Mpn and HJs in a protein-concentration dependent fashion ( Fig. 2A). Similar complexes were reported to be generated between RuvA Eco and HJs, and were found to consist of a single protein tetramer (complex I) or a double tetramer (complex II) bound to a HJ [30,31,32,33,34,35]. The HJ binding activity of both RuvA Mpn and RuvA Mge was strongly reduced in the presence of Mg 2+ (compare Fig. 2A to Fig. 2B, and Fig. 2C to Fig. 2D). A similar inhibitory effect of Mg 2+ on DNA-binding activity has previously also been observed for RuvA Eco [31,36]. In contrast to RuvA Eco and RuvA Mpn , RuvA Mge produced only a single complex with HJs (Fig. 2C, lane 6), even at protein concentrations up to 4 mM (see below). This complex migrated through the gels with a mobility similar to that of RuvA Mpn -HJ complex I. These data indicated that: (i) the RuvA Mge -HJ complex is composed of a tetramer of RuvA Mge bound to a HJ, and (ii) RuvA Mge may not stably bind to HJs as an octamer. These notions were supported by gel filtration chromatography data, which indicated that RuvA Mge exists as a single, major protein species with a molecular mass of ,108 kDa (Fig. S1). This molecular mass corresponds to the theoretical molecular mass of a tetramer of RuvA Mge (95 kDa). Thus, RuvA Mge primarily exists as a homo-tetramer in solution.
In contrast to RuvA Eco , RuvA Mpn was previously reported to form stable complexes with linear duplex oligonucleotides [17]. As shown in Fig. 2E and 2F, both RuvA Mpn and RuvA Mge are able to form DNA-protein complexes in the presence of ds oligonucleotides (substrate HJ11/HJ11rv). Interestingly, at least part of these complexes consisted of RuvA molecules bound to non-annealed, ss oligonucleotide HJ11, which was present as a minor 'contaminant' of the ds substrate; this oligonucleotide (designated 'Free ss' in Fig. 2E and 2F) was completely complexed by the RuvA proteins at the highest protein concentrations tested (Fig. 2E, lane 4-6 and Fig. 2F, lane 6). In a separate EMSA, we could confirm the binding of RuvA Mpn to oligonucleotide HJ11; this binding appeared to occur with an efficiency similar to that observed with the four-stranded HJ substrate (compare Fig. 2G to Fig. 2A). Conversely, while the RuvA Mge protein also displayed binding to the ssDNA (Fig. 2H), this binding was considerably less efficient than that observed with the HJ substrate (Fig. 2C).
The preferences of RuvA Mpn and RuvA Mge for binding to either ssDNA or HJ DNA were further investigated in DNA-binding competition experiments, in which a labeled DNA substrate was kept at a constant concentration and another, unlabeled substrate was included at different concentrations. As shown in Fig. 3A, the binding of RuvA Mge to the labeled HJ substrate was not significantly influenced by inclusion of up to a 20-fold excess of unlabeled ssDNA in the reaction (lanes 3-6). In contrast, the binding of RuvA Mpn to the HJ substrate was already clearly reduced in the presence of a 2.5-fold excess of unlabeled ssDNA in the binding reactions (Fig. 3B, lane 3). Although the dsDNA substrate also competed with the HJ substrate for binding by RuvA Mpn , this competition was less efficient than that observed with ssDNA (Fig. 3C). The high affinity of RuvA Mpn for ssDNA was further demonstrated in an experiment in which the binding of RuvA Mpn to labeled ssDNA was assayed in the presence of different concentrations of unlabeled HJ substrate. As shown in Fig. 3D, the ssDNA-binding of RuvA Mpn was only marginally  [27,28,30,33,54]. The position of the 'acidic pin', between b sheets 6 and 7 of RuvA Eco , two helix-hairpin-helix (HhH) motifs, and the flexible linker (between domain II and III), are also indicated. The multiple alignment was performed using Clustal W (http://www.ebi.ac.uk/Tools/msa/ clustalw2/). The program BOXSHADE 3.21 (http://www.ch.embnet.org/software/BOX_form.html) was used to generate white letters on black boxes (for residues that are identical in at least three out of five sequences) and white letters on grey boxes (for similar residues).

The interaction between RuvA Mge and RecU Mge on HJs
The RecU protein from M. pneumoniae (RecU Mpn ) was previously found to be inactive in HJ-binding and -cleavage [19]. In contrast, the M. genitalium RecU protein (RecU Mge ) was reported to be a potent HJ-resolving enzyme [19,20]. Because it is possible that RecU Mge functionally interacts with RuvA Mge in the processing of HJs, both proteins were included in HJ binding and resolution assays. The binding of RecU Mge to HJ substrate HJ 1.1 was previously demonstrated to result in a single DNA-protein complex [19,20]. Interestingly, at relatively high RecU Mge concentrations and at different binding conditions than those used previously (i.e., binding on ice instead of at room temperature and in the absence of BSA), a range of discrete RecU Mge -HJ DNAprotein complexes were generated, with an inverse correlation between protein concentration and mobility of the complexes through EMSA gels (Fig. 4A, lanes 2-4). At 500 nM of RecU Mge , three major DNA-protein complexes and one minor complex can  be discerned (lane 4). A similar range of complexes was previously also observed after binding of E. coli resolvase RusA to HJ substrates [25]. Due to the distinct nature of the RecU Mge -HJ complexes and their relative migration in the gel, we hypothesize that they represent different multimeric forms of RecU Mge , bound to a single HJ substrate. Upon addition of RuvA Mge to these complexes (after preincubation of RecU Mge with the HJ substrate), novel complexes were formed with a considerably slower mobility than the RecU Mge -HJ complexes (Fig. 4B, lanes 3-7). At the highest concentration of RuvA Mge used (4 mM), all RecU Mge -HJ complexes appeared to have shifted to a higher position in the gel (lane 7). Because the novel complexes had a slower mobility than the RuvA Mge -HJ complex (Fig. 4B, lane 8), it is likely that they represent HJs bound by both RecU Mge and RuvA Mge . This notion was corroborated by a reciprocal experiment in which the HJ substrate was preincubated with RuvA Mge (at 4 mM), followed by the addition of RecU Mge at concentrations ranging from 0 nM to 500 nM (Fig. 4C, lanes 2-7). Already at a RecU Mge concentration of 31 nM (lane 3), a 'supershift' of the RuvA Mge -HJ complex was observed; this supershift was virtually complete at a RecU Mge concentration of 250 nM (lane 6). At the latter concentration, a single major supershifted complex was observed. At 500 nM of RecU Mge , however, four discrete supershifted complexes were formed, which corresponded in mobility with the complexes generated in the previous experiment (Fig. 4B, lane 7). Again, the supershifted complexes displayed a slower mobility than did the RecU Mge -HJ and RuvA Mge -HJ complexes (Fig. 4C, lanes 2 and 8), indicating that they indeed represent RecU Mge -RuvA Mge -HJ complexes. The interactions between RecU Mge and RuvA Mge on HJ substrates differ significantly from those reported between RuvA Eco and the RuvC resolvase from E. coli (RuvC Eco ). Specifically, RuvA Eco appears to have a significantly higher affinity than RuvC Eco for HJ substrates, and a fully saturated RuvA Eco -HJ complex (complex II) cannot be bound detectably by RuvC Eco [32]. As a consequence, RuvAC Eco -HJ complexes are only observed at relatively low RuvA Eco concentrations (1-20 nM); at higher RuvA Eco concentrations, RuvAC Eco -HJ and RuvC Eco -HJ complexes are either not formed or rapidly dissociated [32]. In contrast, RecU Mge and RuvA Mge do not appear to compete with each other in HJ binding, but rather associate readily and stably on a HJ substrate at a wide range of concentrations of both RuvA Mge (Fig. 4B) and RecU Mge (Fig. 4C). As yet, the multimeric protein composition of the different RecU Mge -RuvA Mge -HJ complexes is unknown. Nevertheless, while a single stable complex is generated between RuvA Mge and HJs, it is likely that each of the RecU Mge -RuvA Mge -HJ complexes only contains a single tetramer of RuvA Mge .

RuvA Mge inhibits HJ resolution by RecU Mge
Because RuvA Mge readily binds to RecU Mge -HJ complexes, we investigated the influence of RuvA Mge on the activity of RecU Mge in HJ resolution assays. In these assays, substrate HJ 1.1 was preincubated on ice with either RecU Mge (at 0.2 mM; Fig. 5A) or RuvA Mge (at 0 to 4 mM; Fig. 5B), followed by the addition of the other protein. After incubation for 30 min at 37uC, the resolution products were analyzed by polyacrylamide gel electrophoresis. As shown in Fig. 5A and 5B, RuvA Mge inhibited the resolution activity of RecU Mge in a RuvA Mge concentration-dependent fashion. The inhibition of HJ resolution was most effective when RuvA Mge was added to the HJ substrate before RecU Mge (Fig. 5B and 5C). In that case, HJ resolution by RecU Mge was already inhibited by ,20% at a RuvA Mge concentration of 60 nM (Fig. 5B, lane 3 and Fig. 5C). At RuvA Mge concentrations of 1 mM or higher, RecU Mge activity was reduced by $80% (Fig. 5B, lanes 7-9). When the HJ substrate was incubated with RecU Mge before the addition of RuvA Mge , a significant inhibition of HJ resolution activity ($20%) was only observed at RuvA Mge concentrations of $250 nM (Fig. 5A and  5C). Moreover, inhibition levels of .80% were not observed at RuvA Mge concentrations lower than 4 mM. When RecU Mge and RuvA Mge were added simultaneously to the HJ substrates, a similar pattern of HJ resolution was observed as that shown in Fig. 5A  The RuvB protein that is expressed by M. pneumoniae subtype 2 strains, RuvB FH , was recently reported to act as a DNA helicase on specific, partially double-stranded DNA substrates [18]. Interestingly, while this activity of RuvB FH was not influenced by RuvA Mpn , the RuvB protein from M. genitalium, RuvB Mge , did show RuvA Mge -dependent helicase activity. The latter activity, however, was only detected on a single helicase substrate, i.e. Substrate IV from Fig. 6A [18]. To further delineate the functional interactions between the RuvA and RuvB proteins from M. pneumoniae and M. genitalium, the proteins were combined at various concentrations (including considerably higher RuvA concentrations than used previously) in DNA helicase or branch migration assays, using the DNA helicase substrates shown in Fig. 6A. While the helicase activity of RuvB FH was not influenced by RuvA Mpn (data not shown), the helicase activity of RuvB Mge on Substrate II ( Fig. 6B and 6C) and Substrate I (Fig. 6D) was stimulated in the presence of high concentrations of RuvA Mge . As expected, RuvA Mge alone did not display any DNA helicase activity (lane 7 in Fig. 6C and 6D). This stimulatory effect of RuvA Mge was observed at various concentrations of RuvB Mge , from 0.9 mM (Fig. 6B) to 2.7 mM (Fig. 6C). These results indicated that the activation of RuvB Mge by RuvA Mge is a general phenomenon that is not restricted to a specific DNA substrate. Nevertheless, irrespective of the presence of high concentrations of the RuvA proteins, both RuvB Mge and RuvB Mpn were unable to unwind small, double-stranded oligonucleotide substrates (data not shown).

The ATPase activity of RuvB Mge is stimulated by RuvA Mge
While RuvB FH and RuvB Mge were previously found to possess intrinsic ATPase activity, this activity was significantly higher for RuvB FH than for RuvB Mge [18]. To investigate whether the ATPase activities of the RuvB proteins can be modulated by their corresponding RuvA proteins, ATPase assays were carried out in which the RuvA and RuvB proteins were tested together. In accordance with previous findings [18], RuvB FH was found to possess a significantly higher ATPase activity than RuvB Mge (Fig. 7). However, while the activity of RuvB FH was not significantly influenced by RuvA Mpn , the activity of RuvB Mge was strongly stimulated by RuvA Mge . Thus, the ATPase activities of RuvB FH and RuvB Mge directly reflect the DNA helicase activities of these proteins in two important aspects. First, the intrinsic enzymatic activity of RuvB FH is higher than that of RuvB Mge . Second, RuvB Mge activity can be stimulated by RuvA Mge , whereas RuvB FH activity is not influenced by RuvA Mpn . As expected, both RuvA Mpn and RuvA Mge did not show any ATPase activity on their own (Fig. 7).

Discussion
The DNA recombination and repair machineries of mycoplasmas differ considerably from those of gram-positive and gramnegative bacteria. Most importantly, in contrast to the latter micro-organisms, mycoplasmas do not possess homologues of LexA, RecBCD, AddAB, RecQ, RecJ and RecF [37,38]. In addition, some components of the putative DNA recombination machineries of M. pneumoniae and M. genitalium were found to have characteristics that diverge from those of their homologues from other bacterial classes. These components include the RecU and RuvB proteins [18,19,20]. In Table 1, the characteristics of these as well as the other (putative) components of the DNA recombination machineries of M. pneumoniae and M. genitalium are listed and compared.
We here report that the RuvA proteins from both Mycoplasma spp., RuvA Mge and RuvA Mpn , also possess exceptional properties as opposed to their well-characterized counterpart from E. coli, RuvA Eco . While both RuvA Mge and RuvA Eco [31,39] preferentially bind to HJs, RuvA Mpn displayed a high affinity for both HJ and ssDNA. In addition, while RuvA Mpn and RuvA Eco are both able to form two distinct complexes with HJ substrates, RuvA Mge only formed a single complex with HJs. As this RuvA Mge -HJ complex had a similar mobility through polyacrylamide gels as RuvA Mpn -HJ complex I and RuvA Eco -HJ complex I [17,30,31,32,33,34,35], and because RuvA Mge is a tetramer in solution, it is highly likely that this complex is composed of a tetramer of RuvA Mge bound to a single HJ. This implies that RuvA Mge may only stably bind to HJs as a tetramer. This notion can have important consequences for the interaction of the RuvA Mge -HJ complex with other proteins that are potentially targeted to HJs, such as RuvB Mge and RecU Mge . It was previously reported that the ability of RuvA Eco to form stable octamers on HJs was vital for full activity of the protein. This notion was inferred from the activities of four different octamerization-deficient RuvA Eco mutants [34,35,40]. Three of these mutants carried amino acid substitutions in a protein region known to be involved in tetramer-tetramer interactions [34,35,40]. This region was identified within the crystal structure of HJ-bound octamers of the Mycobacterium leprae RuvA protein (RuvA Mle ) [33]. Within this structure, the two RuvA tetramers make direct protein-protein contacts through specific amino acid side chain interactions at four equivalent points, which are localized to the a6 helix of domain II (Fig. 1A). The interacting a6 helices from two RuvA monomers are in an antiparallel configuration, such that ion pair interactions are formed between three pairs of amino acid residues. On the basis of sequence alignments, we predict that only two of such pairs may be formed between two antiparallel a6 helices of both RuvA Mge and RuvA Mpn . In RuvA Mpn , these pairs would consist of Lys121-Asp133 and Arg124-Glu130, whereas in RuvA Mge , they would consist of Lys121-Glu133 and Arg124-Glu130. While this prediction emphasizes the sequence similarity between RuvA Mpn and RuvA Mge , it does not provide an explanation why RuvA Mpn is able to form stable octameric complexes with HJs, and RuvA Mge is not. It should be considered, however, that the octamerization signals of RuvA Mpn (which are absent from RuvA Mge ) may differ considerably from those of RuvA Mle , and are not (solely) determined by contacts between amino acid residues located in the a6 helix. In this regard, it is relevant to note that one of the reported RuvA Eco mutants that is unable to form stable octamers on HJs, RuvAz87, does not carry mutations in helix a6, but in two other regions of the protein, i.e. in the region between helices a2 and a3 and in helix a4 [40].
Despite its inability to octamerize on HJs in a stable fashion, RuvA Mge was found to stimulate the DNA helicase and ATPase activities of RuvB Mge . The octamerization-competent RuvA Mpn protein, however, did not augment RuvB FH activity. It is possible that the relatively high intrinsic DNA helicase activity of RuvB FH obscured the observation of any additional stimulatory effect on this protein by RuvA Mpn . An alternative explanation for the inability of RuvA Mpn to boost RuvB FH activity is that these proteins are unable to physically interact. In agreement with this notion, we have not yet been able to detect direct or indirect interactions between these proteins in DNA-binding studies.
Another unique feature of RuvA Mge is the mode in which this protein forms tripartite complexes with HJ resolvase RecU Mge and HJs. This is the first report to demonstrate an interaction between a member of the RecU protein family and a RuvA protein.
RuvA Mge and RecU Mge were found to associate readily and stably on HJ substrates at a broad protein concentration range. In contrast, tripartite complexes of RuvA Eco , RuvC Eco and HJs were only observed at relatively low concentrations of RuvA Eco , because the latter protein has a higher affinity than RuvC Eco for HJ DNA [32]. At relatively high RuvA Eco concentrations, the HJ DNA will be saturated with protein, such that two RuvA Eco tetramers are bound to opposite faces of the junction. Thus, the binding of RuvC Eco to the junction is excluded [32,33,41]. At low RuvA Eco concentrations, however, the main protein-HJ complex that is formed is complex I, which consists of a single tetramer of RuvA Eco bound to a single face of the junction. This structure may allow the binding of a RuvC Eco dimer to the other face of the DNA substrate, thereby generating a tripartite RuvAC Eco -HJ complex [32]. In analogy with this model, a tetramer of RuvA Mge bound to one side of a HJ may permit the binding of (multimers of) RecU Mge at the opposite side of the junction. Because RuvA Mge is unable to form stable octameric-HJ complexes, as discussed above, the tetrameric RuvA Mge -HJ complex may always be accessible, at one face of the junction, for binding by RecU Mge . This may explain why RecU Mge and RuvA Mge do not compete with each other for binding to HJs, but rather interact readily by forming a stable tripartite complex. This interaction does, however, lead to inhibition of the HJ resolution activity of RecU Mge , a phenomenon that parallels the inhibition of RuvC Eco -catalyzed HJ resolution by RuvA Eco [32]. It remains to be determined whether the RecU Mge -RuvA Mge -HJ complexes are stabilized exclusively by protein-DNA interactions or also by RecU Mge -RuvA Mge interactions; experiments aimed at the detection of such protein-protein interactions have hitherto not produced conclusive results. In addition, it is clear that the physiological role will have to be established of the RecU Mge -RuvA Mge interaction and the RuvA Mge -mediated inhibition of the HJ resolution activity of RecU Mge . Nevertheless, it is likely that a functional coupling exists between these proteins and that the combined activities of a complex of RuvB Mge and RuvA Mge may be linked to the resolvase activity of RecU Mge . Such a situation could be similar to that in E. coli, in which the RuvAB DNA branch migration complex is coupled to the RuvC resolvase in a RuvABC Eco resolvasome complex. In this regard, it is also interesting to note that a close association between RecU Mge and RuvA Mge (plus RuvB Mge ) is also reflected in the genome of M. genitalium, in which the ORF encoding RecU Mge (MG352) is localized in the vicinity of the ORFs encoding RuvA Mge (MG358) and RuvB Mge (MG359).
Another issue that remains to be addressed is the nature of the four different RecU Mge -HJ complexes that were formed at relatively high concentrations of RecU Mge . In previous studies on this protein, only a single RecU Mge -HJ complex was observed due to the use of different DNA binding conditions [19,20]. It was shown by protein crystallography and structure determination that the RecU homologues from Bacillus subtilis [42] and Bacillus stearothermophilus [43] exist as dimers. Based on this information, we speculate that the four RecU Mge -HJ complexes that were observed in this study consist of HJs bound by dimers, tetramers, hexamers and octamers, respectively, of RecU Mge . How the larger multimers would be accommodated on a single HJ, and how these would also leave room for binding of a RuvA Mge tetramer, which was observed for each of the four RecU Mge -HJ complexes, are challenging questions. The formation of large assemblies of proteins bound to a junction, however, is not unprecedented, as RuvA Eco mutant RuvA3m was reported to generate HJ-protein complexes consisting of six protein tetramers [35].
In conclusion, the studies of the RuvA, RuvB and RecU homologues from mycoplasmas have revealed that these proteins each have distinctive properties as opposed to their counterparts from other bacterial classes. It is possible that these unique features have emerged as a consequence of the evolutionary reduction that the genomes of the mycoplasmas are believed to have undergone. Specifically, the loss of a significant portion of an ancestral set of DNA recombination and repair enzymes may have required an accompanying modification of the function of the RuvA, RuvB and RecU proteins in order to preserve certain functionalities of the recombination and repair system. Nevertheless, the complete set of functions of this system in mycoplasmas is yet to be determined. In this regard, it is particularly interesting to learn how DNA recombination processes are achieved in M. pneumoniae in the absence of a functional RecU resolvase [19]. Although HJ resolvase activities may be exerted by other proteins, such proteins have not yet been identified in M. pneumoniae. Moreover, the lack of a functional RecU was proposed as a possible cause of the relatively low frequency of homologous DNA recombination events in M. pneumoniae [19]. Also, the HJ resolvase deficiency of M. pneumoniae may be associated with the difference between M. pneumoniae and M. genitalium in the specific mechanism by which homologous DNA recombination events occur in these species. In M. genitalium, the repeated DNA elements appear to recombine predominantly in a reciprocal fashion [7,8,12], whereas in M. pneumoniae such elements seem to recombine via a gene conversionlike mechanism, in which donor sequences are copied to the acceptor site and the original acceptor sequence is lost [9,10,11,13,14]. To address these and other issues related to the mechanism of homologous recombination in M. pneumoniae and M. genitalium, it is crucial that the entire set of putative DNA recombination and repair enzymes of these species be delineated. This will therefore be the goal of future studies.

Materials and Methods
Cloning of the M. pneumoniae MPN535 gene and M. genitalium MG358 gene Bacterial DNA was purified from cultures of M. pneumoniae strain M129 (ATCCH no. 29342 TM ) and M. genitalium strain G37 (ATCCH no. 33530 TM ), as described previously [16,44]. The MPN535 ORF of M. pneumoniae strain M129, which encodes a RuvA homologue, was amplified by PCR. The PCR reaction was performed using the following primers: RuvAmpn_fw (59-GGTCGTCATATGATTGCTTCAATTTTTGGAA-39, which overlaps with the translation initiation codon [underlined] of MPN535) and primer RuvAmpn_rev (59-GCAGCCGGATCCT-TAGGCGGTTTTATTTGTAAC-39, which overlaps with the antisense sequence of the translation termination codon [underlined] of the gene). The resulting 0.6-kilobase pairs (kb) PCR fragment was digested with NdeI and BamHI (the recognition sites for these enzymes are indicated in italics in the sequences of primers RuvAmpn_fw and RuvAmpn_rev, respectively), and cloned into NdeIand BamHI-digested E. coli protein expression vectors, i.e. pET-11c and pET-16b (Novagen), generating plasmids pET-11c-RuvA Mpn and pET-16b-RuvA Mpn , respectively. Plasmid pET-11c-RuvA Mpn was used for expression of native RuvA Mpn , while plasmid pET-16b-RuvA Mpn was employed for expression of RuvA Mpn as an N-terminally poly histidine (H 10 )tagged protein in E. coli.
Before cloning of the MG358 ORF of M. genitalium into E. coli protein expression vectors, a TGA codon within the ORF (encoding the Trp residue at position 27 of RuvA Mge ) was changed into a TGG codon using a PCR-based mutagenesis procedure [19]. Following mutagenesis, MG358 was amplified by PCR using the primers RuvAmg_pETfw (59-CGTCACATATGATTACATC-TATCTTTGG -39, which includes an NdeI restriction site [in italics] and the translation initiation codon of MG358 [underlined]) and RuvAmg_pETrv 59-CGTCAGGATCCGGTAT-TAGGCGGTTTTATTTG-39, which includes a BamHI site [in italics] and the antisense sequence of the translation termination codon [underlined] of the gene). The 0.6-kb PCR product was digested with NdeI and BamHI, and ligated into NdeIand BamHIdigested vectors pET-11c and pET-16b, resulting in plasmids pET-11c-RuvA Mge and pET-16b-RuvA Mge , respectively. These plasmids were used for expression of native and H 10 -tagged RuvA Mge , respectively, in E. coli. The integrity of all DNA constructs used in this study was checked by dideoxy sequencing, as described before [15].  ). The ATPase activity was determined using an NADH-coupled assay. In this assay, the activity is calculated from the stationary velocities of ATP hydrolysis, as determined by monitoring the absorption of NADH at 340 nm [15,46]. The 'no protein' reaction (6)   Protein expression and purification The various pET-11c-and pET-16b-derived vectors were introduced into E. coli BL21(DE3) and the resulting strains were grown overnight at 37uC in LB medium containing 100 mg/ml ampicillin. The cultures were diluted 1:100 in 300 ml LB medium with ampicillin and grown at 37uC to an optical density at 600 nm of 0.6. Protein expression was then induced by the addition of isopropyl-b-D-thiogalactopyranoside (IPTG) to a final concentration of 0.5 mM. After incubation for 2 hr at 30uC, the bacteria were harvested by centrifugation and stored at 220uC.
The H 10 -tagged RuvA Mpn and RuvA Mge proteins were both purified using the following protocol. Bacterial pellets were resuspended in 10 ml of buffer A (20 mM Tris-HCl pH 8.0, 1 M NaCl) containing 0.5 mg/ml of lysozyme. The suspension was sonicated on ice and clarified by centrifugation for 20 min at 12,0006 g (at 4uC). To the supernatant, imidazole was added to a final concentration of 5 mM. Then, the supernatant was loaded onto a column containing 1 ml of Ni 2+ -nitroloacetic acid (Ni-NTA)-agarose (Qiagen), which was equilibrated previously in buffer A containing 5 mM imidazole. The column was washed with 5 ml of buffer A plus 5 mM imidazole and with 5 ml of buffer A plus 20 mM imidazole. The specifically bound proteins were eluted from the column with 8 ml of buffer A containing 250 mM imidazole. Fractions of 0.5 ml were collected, analyzed by SDSpolyacrylamide gel electrophoresis (SDS-PAGE), pooled, and dialyzed against a solution of 20 mM Tris-HCl (pH 7.4), 0.2 M NaCl, 0.1 mM EDTA, 1 mM DTT and 50% glycerol (buffer B). Aliquots of purified protein, which had an estimated homogeneity of .95%, were stored at 220uC.
The native RuvA proteins were purified by solubilization of the bacterial pellets in a buffer containing 20 mM Tris-HCl pH 7.5, 1 mM EDTA, 1 mM DTT and 0.5 mg/ml of lysozyme. After sonication and centrifugation (using similar procedures as described above), the RuvA proteins were precipitated with ammonium sulphate and resuspended in 20 mM Tris-HCl pH 7.4, 0.1 M NaCl, 0.1 mM EDTA, 1 mM DTT. The proteins were then subjected to affinity chromatography using Heparin Sepharose 6 Fast Flow (GE Healthcare). Proteins were eluted from the column material with a linear gradient from 0 M to 1 M NaCl in 20 mM Tris-HCl pH 7.4, 0.1 mM EDTA and 1 mM DTT. The RuvA-containing fractions were pooled, dialyzed against buffer B, and stored at 220uC.

DNA-binding assays
Binding of the RuvA proteins to various DNA substrates was carried out in 10-ml volumes and included 20 mM Tris-HCl pH 7.5, 1 mM DTT, 1 mM EDTA, 12.3 nM oligonucleotide substrate and various concentrations of RuvA proteins. After incubation on ice for 10 min, 1 ml was added of a solution containing 40% glycerol and 0.25% bromophenol blue. Then, the reaction mixtures were electrophoresed through 8% polyacrylamide gels in 0.56 TBE buffer (45 mM Tris, 45 mM boric acid, 1 mM EDTA). Following electrophoresis, the polyacrylamide gels were analyzed by fluorometry, using a Typhoon Trio TM 9200 Variable Mode Imager (GE Healthcare) in combination with the Typhoon Scanner Control v4.0 software (Amersham Bioscience). Images were processed using Quantity OneH 1-D Analysis Software.

Holliday junction (HJ) resolution assays
HJ resolution assays were carried out as described by Sluijter et al. [19]. Reactions were analyzed by electrophoresis through 12% polyacrylamide/16 TBE mini-gels. The relative RecU Mge (resolution) activity (Fig. 5C) was expressed as percentage of the protein's activity in the absence of RuvA Mge .
DNA helicase and ATPase assays DNA helicase assays were performed similarly as described before [18]. After deproteinization, the reactions mixtures were analyzed by electrophoresis through 12% polyacrylamide/16 TBE mini-gel and fluorometry. The ATPase activities of RuvB FH and RuvB Mge were determined by using a b-nicotinamide adenine dinucleotide reduced form (NADH)-coupled assay on a VersaMax Tunable Microplate Reader (Molecular Devices) [15,46]. Figure S1 RuvA Mge is a tetramer in solution. (A) Gel filtration analysis of RuvA Mge . Gel filtration chromatography was performed in a similar fashion as described previously [16], using a Sephadex G-150 column (length, 1.0 m; inner diameter, 1.0 cm). The column was run at 4 ml/h in 50 mM Tris-HCl (pH 7.5)/ 135 mM NaCl, and calibrated with blue dextran (2,000 kDa), bovine serum albumin (BSA, 66.4 kDa), ovalbumin (42.9 kDa), and cytochrome C (12.3 kDa). Fractions of 1.0 ml were collected and monitored by measuring the optical density at 280 nm (OD280, Y-axis at the left-hand side of the graph). The fractions eluted from a subsequent run, containing 15 mg of RuvA Mge , were precipitated with trichloroacetic acid, and separated on 12% SDS-PAGE gels. Gels were silver-stained and recorded using the GelDoc XR system. RuvA Mge was quantified by densitometry using Quantity OneH 1-D Analysis Software (Bio-Rad). The relative concentration of RuvA Mge (Y-axis on the righthand side, in arbitrary units) is shown for column fractions 23 to 39. In all other fractions, RuvA Mge was not detected. (B) Calibration curve obtained from the gel filtration experiment shown in (A). The molecular weight of protein size standards (¤) is plotted against the elution volume (V e ) divided by the void volume (V 0 ) of the column (V e /V 0 ). V 0 was determined with blue dextran. The V e /V 0 of RuvA Mge is marked on the calibration curve (6). (TIF)