|Home | About | Journals | Submit | Contact Us | Français|
The obligate human pathogen Neisseria gonorrhoeae (Gc) has co-opted conserved recombination pathways to achieve immune evasion by way of antigenic variation. We show that both the RuvABC and RecG Holliday Junction (HJ) processing pathways are required for recombinational repair, each can act during genetic transfer, and both are required for pilin antigenic variation (Av). Analysis of double mutants shows that either the RecG or RuvAB HJ processing pathway must be functional for normal growth of Gc when RecA is expressed. HJ processing deficient survivors of RecA expression are enriched for nonpiliated bacteria that carry large deletions of the pilE gene. Mutations that prevent pilin variation such as recO, recQ, and a cis-acting pilE transposon insertion all rescue the RecA-dependent growth inhibition of a HJ processing deficient strain. These results show that pilin Av produces a recombination intermediate that must be processed by either one of the HJ pathways to retain viability, but requires both HJ processing pathways to yield pilin variants. The need for diversity generation through frequent recombination reactions creates a situation where the HJ processing machinery is essential for growth and presents a novel target for novel antimicrobials against gonorrhea.
Microbial pathogens encounter many different barriers that must be overcome to successfully establish infection. The innate and adaptive immune systems provide a multilayered array of defenses that successful pathogens must contend with. Opportunistic pathogens often require some level of immune dysfunction to establish infection. In contrast, professional pathogens possess one or more strategies to overcome immune surveillance. N. gonorrhoeae (Gc) is the sole etiologic agent of the sexually transmitted disease gonorrhea. Gonorrhea has been evident within humans for all recorded history (Morton, 1977; Rothenberg, 1993; Wain, 1947) and has evolved to specifically thrive within the human genital tract. However, even without antibiotic treatment, gonorrhea is usually self-limiting. Gc relies upon sexual networks to spread and persist within infected populations. Consistent with this continual movement between individuals is the observation that high-risk people can contract gonorrhea countless times and never demonstrate immunity to the disease. One reason for the lack of effective immunity to reinfection is the capacity of Gc to antigenically vary its surface at high frequency. The surface exposed Opa proteins and the lipoligosacharride both undergo high-frequency antigenic variation (Av) by altering nucleotide repeat tracts (Kline et al., 2003). In contrast, the pilus undergoes high frequency Av by homologous recombination reactions between one of many silent pilin loci and a single expressed pilin gene (Kline et al., 2003).
Pilin variation is mediated by the homologous recombination machinery and is dependent on RecA and RecX, the RecF-like recombination pathway, the chromosome organizing factor RdgC, and the Rep helicase (Kline and Seifert, 2005b; Mehr and Seifert, 1998; Seifert, 1997; Skaar et al., 2002; Stohl and Seifert, 2001). These conserved recombination proteins help mediate high frequency gene conversion reactions that transfer short stretches of variant sequences bordered by small segments of sequence identity between silent pilin loci and the expressed gene (Criss et al., 2005; Howell-Adams and Seifert, 2000). These homologous recombination processes also contribute to survival form DNA damaging agents and some to DNA transformation competence (Kline et al., 2003). A genetic screen for factors interfering with pilin variation revealed a number of genes that, when mutated, reduced the frequency of pilin variation. Two pilin Av deficient mutants isolated were inactivated for the HJ processing helicases RuvAB (ruvA inactivation) and RecG (recG inactivation) (Sechman et al., 2005). In addition, a transposon insertion in the noncoding region upstream of the pilE expression locus also disrupted pilin Av, apparently acting in cis (Sechman et al., 2005). Here we show that a ruvB and a ruvC mutant each is deficient for pilin Av. Surprisingly, when any of the three ruv mutations are combined with a recG mutation there was a severe RecA-dependent growth deficiency for each mutant strain. This growth deficiency could be overcome by a variety of cis- and trans-acting mutations that each disrupt pilin Av, proving that the process of pilin antigenic variation specifically makes the recG and ruv mutations synthetically lethal. We propose that pilin antigenic variation intermediates contain HJs and the inability to process HJs creates a structure that blocks chromosomal replication and results in a growth deficiency.
A genetic screen previously revealed 30 genes that showed reduced levels of pilin variation when mutated, including the Gc recG and ruvA genes (Sechman et al., 2005). In bacteria there are two major helicases involved in the processing of HJs to promote branch migration and the formation of heteroduplexes, RuvAB and RecG. DNA duplexes joined by HJs and processed by RuvAB are subsequently separated by the action of the dedicated endonuclease RuvC (Sharples et al., 1999). Analysis of the sequenced gonococcal FA1090 genome revealed that genes encoding proteins similar to both RuvB and RuvC are also present in Gc (Genebank accession: AE004969). Mutants carrying a loss-of-function mutation of either of these genes also reduced pilin antigenic variation to levels similar to the ruvA and recG mutants, confirming that both pathways are required for pilin variation (Fig. 1A). The pilin Av deficiency could be complemented by supplying an expressed copy of the ruvB or ruvC genes in the mutant strains from an ectopic locus (Fig 1A). The three ruv mutants and the recG mutant all showed reduced survival when exposed to UV light supporting the hypothesis that HJ processing is important for DNA repair (Fig 1B) (Mehr and Seifert, 1998). The UV survival phenotypes were complemented in the ruvB and ruvC backgrounds (Fig 1B). We had previously reported that the ruvA and recG mutants were not affected for DNA transformation competence (Sechman et al., 2005). Since that report, it was found that a Neisseria meningitidis recG mutant showed reduced transformation competence (Sun et al., 2005). We therefore retested the DNA transformation competence of the HJ processing mutants. In contrast to the previous assays, in which DNA exposure to the bacteria was continuous (Sechman et al., 2005), the time the bacteria were exposed to transforming DNA was limited to 15 minutes with the addition of DNaseI . By this method, all of the HJ processing mutants showed a significant reduction in transformation competence and this phenotype was complemented by the functional gene (Fig 1C). We conclude that both the RecG and RuvABC HJ processing pathways contribute to pilin variation, DNA repair, and DNA transformation.
Each individual HJ processing mutant retained measurable abilities to conduct pilin Av, DNA repair and DNA transformation relative to the recA mutant (Fig 1). We therefore constructed double mutant strains deficient in both HJ processing pathways (recG and ruvA, ruvB, or ruvC) to determine whether a HJ processing deficient strain would be similar to a recA mutant. Surprisingly, the HJ double mutants all showed a severe growth defect when RecA was expressed (Fig. 2A and C). In the absence of RecA expression, the growth of HJ double mutants was similar to the single mutants and to the parental strain with functioning HJ processing systems (Fig 2A and C). The growth defect of the recG ruvB double mutant could be complemented by an over expressed copy of ruvB in an ectopic locus (Fig. 2C). As expected, there was no growth defect found with ruvA ruvC or ruvB ruvC double mutants, where only one HJ processing pathway was disrupted (Fig 2C). These results strongly suggest that RecA-catalyzed recombination results in a structure that must be acted on by one of the HJ processing systems, or growth becomes severely restricted. It was not clear from these data whether this growth deficiency was dependent on general recombination processes or was specific for a particular recombination pathway.
The HJ processing single mutants were originally isolated as deficient in producing nonpiliated variants (Sechman et al., 2005), a phenotype shared by the additional HJ mutants created in this work (Fig 3A). However, HJ double mutants surviving RecA induction were enriched for nonpiliated variants (Fig 3A). These HJ double mutant survivors of RecA expression were analyzed to determine why the loss of HJ processing led to a RecA-dependent growth defect. The mechanism underlying this increased frequency of nonpiliated variants in the HJ double mutant was revealed by a large increase in pilE deletions in the double mutant background (Fig 3B). Interestingly, the nonpiliated variants found after RecA expression in the HJ single mutants also showed increased pilE deletions, even though the overall frequency of nonpiliated variants in the HJ processing single mutants was reduced (Fig 3A). A HJ double mutant carrying a pilE deletion no longer showed a RecA-dependent growth deficiency, demonstrating that an intact pilE gene was required for the HJ processing-related growth deficiency. Furthermore, strains with an intact pilE and a nonpiliated colony morphology retained the RecA-dependent growth defect (Fig. 4), demonstrating that loss of pilE and not the change in colony morphology alleviated the growth deficiency. The remainder of the surviving population was piliated and did not show high frequency sequence changes at pilE (data not shown). It is likely that these surviving progeny represent bacteria that never initiated the process of pilin Av. It is likely that the variability observed for the fraction of surviving HJ deficient bacteria to RecA expression from experiment-to-experiment reflects when pilin Av was initiated and when a pilE deletion event occurred to present an escape mutant (Figs (Figs11 and and22).
While the requirement for an intact pilE suggested that the growth deficiency was related to pilin variation, we tested whether mutations known to interfere with pilin Av would alter the growth defect. Introduction of a recO, recJ, or recQ loss-of-function mutation into the HJ double mutant blocked the RecA-dependent growth defect (data not shown). These results show that a functional RecF-like recombination pathway is required for the growth defect and support the hypothesis that RecA acts in conjunction with the RecF-like recombination pathway to create HJs during pilin Av. However, since the RecF-like pathway participates in both pilin Av and DNA repair processes (Mehr and Seifert, 1998; Sechman et al., 2005; Skaar et al., 2002) we wanted to specifically ask whether the growth phenotype of the HJ processing mutant was directly linked to pilin Av. To this end, the effect of two transposons inserted immediately upstream of pilE that differentially affect pilin Av were tested (Fig 5A). One transposon insertion was previously shown to block pilin Av (pilE::Tn5#1) (Sechman et al., 2005) to produce a nonvarying colonial morphology (P+nv), and a second transposon insertion (pilE::Tn5#9) was found 83 bp upstream of the #1 transposon insertion that had no effect on pilin Av (Fig 5A) (K.A Kline, Criss, A., Wallace A. and H.S. Seifert, in preparation). Upon RecA induction in the HJ processing double mutant, the #1 transposon insertion prevented agrowth defect, while the #9 transposon insertion retained the RecA-dependent growth defect (Fig 5B). Neither transposon insertion altered the growth of the parental strains, are within or near an open reading frame, or alter pilE expression (Sechman et al., 2005). These results conclusively show that pilin antigenic variation is directly responsible for the growth defect in the HJ processing mutants.
HJs are central to all homologous recombination and many site specific recombination systems (Lilley and White, 2001). Usually one HJ processing system is required for recombination, or different HJ processing systems can substitute for one another to facilitate recombination. The bacterial HJ processing pathways are also proposed to act on collapsed replication forks to allow RecF or RecBCD pathway recombinases to create a structure that can be acted on by the replication restart factors (Michel, 2000). Since we have previously demonstrated that mutations in priA, the primary mediator of replication restart, do not affect pilin antigenic variation (Kline and Seifert, 2005a), it is unlikely that the link between HJ processing and growth is related to replication restart. The requirement of both the Ruv and RecG systems for pilin Av frequencies similar to wild type suggests that these two separate HJ processing systems act at different steps during recombination or upon different substrates formed during the recombination process.
This is not the first instance where HJ processing has been demonstrated to be required for an organism’s viability. While many bacterial species can tolerate having both the Ruv and RecG systems interrupted in the presence of an active recA gene with no reported growth defects (Asai and Kogoma, 1994; Beam et al., 2002), there are two previous examples where a HJ processing deficiency caused problems. Attempts to isolate transformants of Bacillus subtilis that were deficient in both ruvAB and recG, ruvAB and ruvU (a ruvC homologue), or ruvU recG were unsuccessful for unknown reasons (Sanchez et al., 2005). These mutations were not analyzed with respect to RecA expression and it may have been the process of DNA transformation that made constructing the double mutants problematic. Moreover, the inability to isolate a ruvAB, ruvU double mutant in B. subtilis suggests there are both commonalities and differences between HJ processing within B. subtilis and N. gonorrhoeae. In addition, E. coli engineered to produce high frequency double stranded breaks through the action of the SceI endonuclease, recG and ruvA mutations were synthetically lethal (Meddows et al., 2004). It is possible that the synthetic lethality of the SceI expressing E. coli and Gc undergoing pilin Av are are the result of similar events (see below). Taken together, these data suggest that loss of HJ processing may only be a problem for bacterial cells when hetroduplex is formed with the bacterial chromosome at a high enough frequency to cause lethality due to the structures that not resolved.
These results, along with the previous studies into the mechanisms of pilin Av, allow us to make predictions about the DNA recombination processes mediating pilin Av. We propose that an unknown initiator acting on the DNA upstream of pilE begins the process of pilin Av, and that the action of the initiator is blocked by the pilE::Tn5#1 transposon insertion (Fig. 6A). This proposition is based on the fact that the transposon insertion does not map within any open reading frame and does not affect pilE expression (Sechman et al., 2005) and that it exerts its effect prior to the action of the recombination factors. The most likely activity for an initiator of recombination is an endonuclease which produces a double stranded break or a single stranded nick. Since pilin Av depends on the RecF-like pathway (Mehr and Seifert, 1998), which acts on single-stranded substrates (REF), the simplest model of initiation proposes a single stranded nick, although an initiating double stranded break cannot be ruled out. If a nick initiates, components of the RecF-like recombination pathway (ie, RecJ and/or RecQ), or other enzymes such as Rep could process the pilE DNA into a gapped substrate, which would be the product of a single stranded nick (Fig 6B). RecA can be loaded onto this gapped substrate with the help of the RecOR complex (Morimatsu and Kowalczykowski, 2003). The RecX protein can also aid RecA to promote recombination with a homologous double stranded molecule (Stohl and Seifert, 2001). The identity of this double stranded substrate is presently unknown, but could either be the previously postulated hybrid intermediate which shares sequence identity with the pilE upstream region (Howell-Adams and Seifert, 2000), or a silent locus sequence from a donor chromosome (not shown). This RecAXOR mediated recombination event would form HJs, which would be acted upon by both the RuvAB and RecG helicases to promote branch migration and produce extended regions of heteroduplex (Fig 6D). Alternatively, one of these HJ processing systems could be dedicated to producing the hybrid intermediate, while the other is responsible for processing the final HJs (Fig 6C). We do not favor this hypothesis since inactivation of a HJ processing system that specifically alters hybrid intermediate formation would not be predicted to interfere with growth. Regardless of which of these possibilities is correct, both HJ processing systems are required for pilin Av, and in their absence, a HJ structure is formed that severely limits Gc growth. We have not determined the precise structure of this HJ-containing intermediate, but since there is more than one complete copy of the gonococcal chromosome within each monococcus (Tobiason and Seifert, 2006), it is intriguing to speculate that the lethal structure is a fusion between two chromosomes. We can only conclude from the data presented in this work that at least one HJ, linking two pil-containing DNA duplexes, forms a structure that inhibits growth.
In all cells, recombination processes can be stimulated to enable DNA damage repair and specialized programs like meiosis, antibody diversification, mating type inter-conversion, or antigenic variation. The frequency of Gc pilin variation has been measured to be at least 4 × 10-3 observable recombination events per cell per generation (Criss et al. 2005) and may be higher (Rohrer et al., 2005). Assuming that each pilin Av initiation event results in a structure that must be acted on by one of the HJ processing systems to maintain viability, we can use the RecA-dependent change in cfu of the HJ deficient strain to estimate the frequency of pilin Av initiation. The parental, HJ processing proficient strain produced about 1.05 × 106 cfu/colony after 24 hrs growth, which represents 20 generations, assuming each colony arose from a single cfu. recG, ruvB and recG, ruvA strains produced between 2 and 5% of the cfu/colony relative to the parental strain (Fig 2B). To produce this observed reduction in cfu/colony after 20 generations one cfu would have to be lost every two or three generations. This calculation strongly suggests that pilin variation is initiated every second or third generation, but that only 1% of the time does initiation result in a variant pilE . We assume that this frequent initiation of pilin Av is necessary to accumulate the requisite level of diversity to continually propagate Gc within human populations. Any energy cost from frequent recombination initiation would be balanced by the selection for survival from immune surveillance.
The increasing appearance of antibiotic resistant gonococci and the loss of effective treatment strategies have created a critical need for new ways to target antimicrobial agents. The requirement of HJ processing for antigenic variation, coupled with the significant inhibition of growth by mutation of both HJ processing pathways suggests that blocking HJ processing in vivo would inhibit growth of the organism. Furthermore, by blocking pilin Av, bacteria would be more antigenically stable and acquired immune responses would be more effective. Since peptides inhibiting multiple HJ processing enzymes have been identified (Kepple et al., 2005), a small molecule that could interfere with HJ processing may allow a unique type of antimicrobial to be developed. If efficacious in people or animals, these HJ processing inhibitors may also be useful against diseases caused by other bacteria with high frequency antigenic variation systems such as, Meningococcal meningitis or Lyme disease (van der Woude and Baumler, 2004).
We thank the members of the Seifert laboratory for comments on this work and manuscript. This work was supported by Public Health Service grants R37 AI033493 and R01 AI044239 to H.S.S.
E. coli and Gc strains were grown as previously described (Sechman et al., 2005). Gonococcal strains were grown on Gc Medium Base (Difco) plus Kellogg supplements (GCB) [22.2 mM glucose, 0.68 mM glutamine, 0.45 mM co-carboxylase, 1.23 mM Fe(NO3)3: all from Sigma] (Kellogg et al., 1963) at 37°C in 5% CO2 or in GCB liquid (GCBL) medium [1.5% protease peptone no. 3 (Difco), 0.4% K2HPO4, 0.1% KH2PO4, 0.1% NaCl] with Kellogg supplements and 0.042% sodium bicarbonate. When appropriate 1mM isopropyl-β-D-thiogalactopyranoside (IPTG, Diagnostic Chemicals, Ltd.) was added to the media to induce expression from the recA6 locus or the nics6 complementing locus.
Molecular biology techniques were performed as previously described (Sechman et al., 2005). recG and ruvA mutants were isolated as previously described (Sechman et al., 2005). The Gc ruvB and ruvC loci were cloned into the pBLUNT vector (Invitrogen) using the primers RUVBFOR (CCATTCCGCCCCCGACATA), RUVBREV (GCTGATGTGGGTCAACCCC), RUVCFOR2 (GGCGAATGTCGAAAACAATAAAT), and RUVCREV2 (CAAATAATGCTTATTGCGGTAG) and mutated using a deletion/insertion strategy. Briefly, 925 bp of the ruvB gene was deleted using the BbsI (NEB) restriction endonuclease (RE) and an ermC cassette was inserted between the BbsI sites. The ruvC gene was mutated in a similar fashion, however the 270 bp ClaI and NdeI (NEB) fragment was deleted and replaced with the ermC gene. Complementation was performed by amplifying fragments carrying the ruvB and ruvC genes from the Gc chromosome using primers RUVBFOR2 (TGCCGTCTGAAACGCGCCG), RUVBREV2 (CAAACGTCTGATAACAATGCCG), RUVCFOR3 (CTGGGACTGAACCGCAATAC), and RUVCREV3 (ATTTCATCTCGGTACACATTTTC) and expressing the wild type genes from lacIO-regulated complementation locus. Double mutants were generated by transforming isolated genomic DNA from ruvB and ruvC mutants into recG and ruvA mutants. Southern blot and pilE sequence analysis was performed on all mutants.
Pilin antigenic variation was measured using the surrogate pilus-dependent colony morphology change assay (PDCMC). Gc were grown on agar medium and colony variation was scored after 22, 24, 26, 28, and 30 hours of growth by observing the number of P- outgrowths on the same 3 to 7 colonies per time-point using a stereomicroscope (Sechman et al., 2005). No P- outgrowths produced a score of 0, one P- outgrowth a score of 1, etc. Colonies with four or more P- sectors was given a score of 4 and the scores from individual colonies for each time point were averaged.
Gc were collected using a Dacron swab to a concentration of ~108 cfu/ml, immediately serially diluted from 10-1 – 10-6, plated on agar medium, and exposed to 0 to 80 mJ/m2 UV radiation in a Stratalinker 1800. After 20 hrs of growth, colonies were counted and percent survival was calculated.
Transformation efficiency was assayed as previously described using pSY6 (Sechman et al., 2005), with the modification of 50 μgm/ml of DNaseI added to the cells 15 min after addition of transforming DNA and incubated for 10 mins to degrade extracellular DNA. The transformation efficiency is the mean number of NalR transformants/cfu.
Gc colonies were grown on agar medium for 24 hours. Five representative colonies were collected using sterile filter paper (Whatman) and placed into 1 ml of GCBL. After mixing vigorously, serial dilutions were spotted onto agar plates to determine cfu/colony.
Individual P- variants that arose from different P+ progenitors were lysed as described (Sechman et al., 2005) and PCR reactions performed with the pilE specific PILRBS and SP3A oligonucleotides (Wright et al., 1994). The presence of a ~650 bp PCR product indicated that the pilE gene was intact, while the absence of that PCR product represented a pilE deletion.