Less Is More: Neisseria gonorrhoeae RecX Protein Stimulates Recombination by Inhibiting RecA*

Escherichia coli RecX (RecXEc) is a negative regulator of RecA activities both in the bacterial cell and in vitro. In contrast, the Neisseria gonorrhoeae RecX protein (RecXNg) enhances all RecA-related processes in N. gonorrhoeae. Surprisingly, the RecXNg protein is not a RecA protein activator in vitro. Instead, RecXNg is a much more potent inhibitor of all RecANg and RecAEc activities than is the E. coli RecX ortholog. A series of RecXNg mutant proteins representing a gradient of functional deficiencies provide a direct correlation between RecANg inhibition in vitro and the enhancement of RecANg function in N. gonorrhoeae. Unlike RecXEc, RecXNg does not simply cap the growing ends of RecA filaments, but it directly facilitates a more rapid RecA filament disassembly. Thus, in N. gonorrhoeae, recombinational processes are facilitated by RecXNg protein-mediated limitations on RecANg filament presence and/or length to achieve maximal function.

Cellular recombination systems have a range of functions in all cells. In bacteria, the central function is the repair of stalled or collapsed replication forks (1)(2)(3)(4)(5). When a replication fork encounters breaks or lesions in DNA, the fork stalls, and cell death can occur if replication is not restarted. Bacterial replication fork stalling may occur as often as every replication cycle even under normal growth conditions (5,6). In addition, recombination is used to promote horizontal genetic transfer of alleles between organisms and can be used to promote specialized recombination programs.
The 37.8-kDa, 352-amino acid Escherichia coli RecA (RecA Ec ) 2 protein is essential for the recombinational DNA repair needed to regenerate a functional replication fork. RecA Ec protein functions primarily as a dynamic, extended helical filament formed on DNA in the presence of an adenine nucleotide triphosphate co-factor (7,8). RecA Ec filament formation on single-stranded DNA (ssDNA) begins with a slow nucleation step, followed by a faster 5Ј to 3Ј extension phase (6, 9 -11). Once formed, the RecA Ec filaments predominantly assemble at the 3Ј end and disassemble at the 5Ј end with ATP hydrolysis occurring throughout the filament (12)(13)(14).
In the human-specific pathogen Neisseria gonorrhoeae (Gc), RecA is not only necessary for DNA repair but is also necessary for DNA transformation and pilus antigenic variation (15)(16)(17). Gc pili mediate attachment of the pathogen to the host epithelium, and antigenic variation of these pili is postulated to be necessary for immune avoidance (18). Antigenic variation occurs when silent copies of the pilS gene are transferred unidirectionally into the expressed pilE gene via RecA-mediated recombination (19,20). Change of the amino acid sequence of PilE, the main component of Gc pili (21) can result in changes in pilus-dependent colony morphology (phase variation) (22,23), providing a convenient way to estimate the frequency of pilin antigenic variation (23,24). RecA Ng and RecA Ec share 65% identity and 81% similarity. When expressed in N. gonorrhoeae, RecA Ec complements for pilin antigenic variation, partially complements for DNA transformation, but does not complement for DNA damage survival (25). Similarly, RecA Ng can complement a recA null mutant in E. coli for homologous recombination but not for survival to DNA damage. In vitro characterization of RecA Ng reveals a protein functionally similar to RecA Ec , although ATP hydrolysis by RecA Ng is somewhat faster. 3 Regulation of recombination is critical for DNA repair and genome stability, and RecA represents the primary target for such regulation. RecA Ec is regulated on three levels: transcriptional regulation of the recA gene within the SOS regulon (27), autoregulation by the C terminus of the RecA protein (28), and regulation by other proteins. Most allosteric RecA regulators characterized to date are involved in controlling the formation or the disassembly of RecA filaments as reviewed in Ref. 29). Not as much is known about RecA regulation in Gc. The two species both possess orthologs of most genes involved in recombinational DNA repair (19). Until recently, no classical SOS response had been identified in Gc (30). However, a small regulon controlled by a LexA homolog was recently discovered in N. gonorrhoeae that does not regulate RecA Ng . 4 Thus, the regulation of RecA in N. gonorrhoeae and E. coli might differ because RecA Ng is involved in a process absent in E. coli (pilus antigenic variation), and RecA Ec is involved in a process not well conserved in N. gonorrhoeae (SOS response).
One well conserved RecA regulating protein in bacteria is RecX. The 19.4-kDa, 166-amino acid E. coli RecX (RecX Ec ) protein is encoded by the recX gene (also called oraA) located downstream of the recA gene. Cellular levels of RecX Ec , produced by transcriptional read-through from the recA promoter, are less than 5% of RecA Ec levels with numbers of less than 50 RecX Ec molecules/cell (31) and 15,000 RecA Ec molecules/cell (31)(32)(33). The only phenotype of an E. coli ⌬recX strain is a 5-fold decrease in UV resistance (31). Overexpression of recX inhibits SOS induction and leads to a decrease in UV resistance and P1 transduction frequency (31). These data indicate that in E. coli, high levels of RecX inhibit the RecA-mediated response to DNA damage and suggest that a low level of RecX is beneficial for conferring UV resistance.
Three-dimensional EM reconstruction shows that RecX Ec binds within the major helical groove of a RecA Ec filament, spanning the monomer-monomer interface (34). RecX Ec inhibits RecA Ec -mediated homologous recombination, co-protease, and ATPase activities in vitro at substoichiometric concentrations (31,35). Kinetic and electron microscopic analysis of the effects of RecX Ec on RecA Ec provided evidence for a filament capping mechanism (35). Net RecA Ec filament disassembly occurs because of RecX Ec blocking assembly at the 3Ј-proximal filament end, whereas disassembly continues at the 5Ј-proximal end. The crystal structure of RecX Ec reveals three bundles of three helices making up a crescent-shaped protein (36).
The 17.7-kDa, 153-amino acid N. gonorrhoeae RecX (RecX Ng ) protein is encoded by a recX gene that is not located downstream of recA as it is in many other bacteria (37)(38)(39)(40). The RecX Ec and RecX Ng proteins share 25% identity and 54% similarity. A Gc recX loss of function mutant shows a 6-fold decrease in pilus phase variation, a 5-fold decrease in DNA transformation, and a 4-fold decrease in DNA repair ability upon nalidixic acid exposure, whereas the levels of RecA Ng protein are not affected (41). This demonstrates that the Gc recX gene aids all RecA Ng -related activities in N. gonorrhoeae and indicates a positive effect of RecX Ng on RecA Ng function (41). It is not clear what role RecX Ng plays in these RecA-mediated processes, and no biochemical studies have been carried out.
In the present work, we have examined the effects of RecX Ng protein in light of its positive effects on RecA Ng function in N. gonorrhoeae. Surprisingly, we find that RecX Ng is not only an inhibitor of both RecA Ng and RecA Ec in vitro, but it is a much more potent inhibitor than RecX Ec . Our results offer important insight into the regulation of RecA-mediated processes by negative regulators and show that inhibition of RecA activities in vitro can result in stimulation of RecA-dependent processes in the bacterial cell. Enzymes, Biochemicals, and Buffers-The native E. coli wildtype RecA and single-stranded DNA binding (SSB) proteins were purified as described previously (44). The concentration of the purified RecA and SSB proteins was determined from the absorbance at 280 nm using the extinction coefficients of 2.23 ϫ 10 4 M Ϫ1 cm Ϫ1 (45) and 2.83 ϫ 10 4 M Ϫ1 cm Ϫ1 , respectively (46).

EXPERIMENTAL PROCEDURES
The N. gonorrhoeae RecA and SSB proteins were purified as described (26). The E. coli RecX protein was purified as described (47) with modifications described in the supplemental materials. The concentration of E. coli RecX was determined from the absorbance at 280 nm using the native extinction coefficient 2.57 ϫ 10 4 M Ϫ1 cm Ϫ1 (47).
Cloning, Overexpression, and Purification of the Native N. gonorrhoeae Wild-type RecX (RecX Ng ) Protein-The Gc recX gene was amplified from the FA1090 genome by PCR using primers GCRXPETFOR and GCRXPETREV (supplemental Table S3). The resulting 470-bp DNA fragment was cloned into vector pBLUNT (Invitrogen), sequenced, and subcloned into pET21a yielding construct pET21a-RecX Ng .
Competent cells of E. coli strain BL21(DE3) were transformed with pET21a-RecX Ng . Ten liters of culture were grown at 37°C in LB medium to an A 600 of 0.52. RecX Ng protein expression was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to 0.4 mM. Following a 3.75-h incubation at 37°C, ϳ20 g of cells were harvested by centrifugation, flash frozen in liquid N 2 , and stored at Ϫ80°C. The RecX Ng protein was purified as described in the supplemental materials. The concentration of RecX Ng was determined from the absorbance at 280 nm using the native extinction coefficient 1.857 ϫ 10 4 M Ϫ1 cm Ϫ1 . The RecX Ng extinction coefficient was determined during the course of the present work using procedures published elsewhere (48,49).
Two liters (4 liters for R130A/F132A and D54A) of culture were grown, induced, harvested, and stored as described for wild-type RecX Ng . The HisRecX Ng HA point mutants were purified as described in the supplemental materials. The concentration of each point mutant was calculated from the absorbance at 280 nm using the calculated extinction coefficient 2.54 ϫ 10 4 M Ϫ1 cm Ϫ1 .
DNA Substrates-The cssDNA from bacteriophage M13mp18 (7249 nucleotides) was prepared as described (50,51) with the following modifications. Chemically competent E. coli JM101 cells were transfected with gel-purified RFI M13mp18 DNA from New England Biolabs. The CsCl banding was done twice in a Beckman Ti60 rotor at 37,000 rpm for 15-20 h. The concentration of cssDNA was determined by absorbance at 260 nm, using 36 g/ml A 260 Ϫ1 as the conversion factor. All of the DNA concentrations are given in M nucleotides with the exception of the fluorescence anisotropy experiments, where the DNA concentrations are given in M molecules. The oligonucleotides used in the fluorescence anisotropy experiments were purchased from Integrated DNA Technologies and resuspended and diluted in 10 mM Tris-acetate (80% cation) and 1 mM EDTA.
ATPase Assay-A coupled spectrophotometric enzyme assay (52, 53) as described in Ref. 54  Electron Microscopy-A modified Alcian method was used to visualize RecA filaments. Activated grids were prepared as described previously (28).
Samples for electron microscopy analysis were prepared as follows. All of the incubations were carried out at 37°C. RecA Ec or RecA Ng (3 M) was preincubated with 6 M M13mp18 cssDNA in RecA buffer for 10 min. An ATP regeneration system of 10 units/ml creatine phosphokinase and 12 mM phosphocreatine was included in the incubation. ATP and SSB were added to 3 mM and 0.6 M, respectively, and the reaction was incubated for another 15 min. RecX Ng or RecX Ec was added to concentrations of 0.11 and 0.1 M, respectively, and incubated for various times as indicated. ATP␥S was then added to 3 mM to stabilize the filaments, followed by another 3-min incubation. The sample was prepared for analysis as described previously (54).
To determine the proportion of molecules that were either fully or partially coated by RecA or bound only by SSB at different times of incubation with RecX, a minimum of 500 molecules from two independent experiments and two separate regions of each grid were counted at an identical magnification. A RecA filament was considered gapped if the ssDNA molecule to which it was bound had a detectable region of SSB-coated DNA of any size. The relative length of the remaining filament and the length of SSB-coated DNA were used to classify the gaps into three categories: small gaps, medium gaps, and big gaps (see supplemental Fig. S3 for examples and results section for descriptions). Linearized DNA molecules originating likely from shearing force during pipetting were also counted. With the total number of filaments counted as 100%, the percentage of each type of nucleoprotein filament was calculated.
Imaging and photography were carried out as described previously (54). Images were obtained of the two types of nucleoprotein filaments showing the highest percentage of all of the filaments counted. Where possible, both of the filament types were captured in the same image. Where this was impossible, the type of filament occurring at the highest percentage was chosen for photography.
A cytochrome c method was used to visualize M13mp18 cssDNA to check the purity of the DNA. The samples were prepared as described previously (54,55).
Fluorescence Anisotropy-The experiments were carried out at 25°C with a Beacon fluorescence polarization system (Invitrogen). Serial dilutions of either RecX Ng or RecX Ec were incubated with 1 nM fluorescein-labeled oligonucleotide den7 (supplemental Table S3). The final reaction volume was 100 l, containing RecA buffer, 3 mM ATP, and compensating amounts of RecX Ec or RecX Ng storage buffer. The reactions were incubated for 30 min shielded from light before measurements were done. The K d values were calculated using Prism software fitting a nonlinear regression and one site total binding. The data are reported in triplicate.
Kinetic Pilus-dependent Colony Morphology Change Assay-This assay measures the number of visible pilus-dependent colony morphology changes that occur over time and is a very reliable method to report pilin antigenic variation frequencies (56). Briefly, gonococcal strains were revived onto Gc medium base from frozen freezer stocks. After ϳ20 h of growth, five individual colonies were passaged onto a plate. Colony variation of 20 colonies of each strain was scored after 18, 20, 22, 24, and 26 h by observing the number of P regions arising from each colony. The scores for the individual colonies at each time point were averaged.

RESULTS
Experimental Design-An array of biochemical approaches, including kinetic assays utilizing RecA-mediated ATP hydrolysis, electron microscopy, and fluorescence anisotropy, were used to analyze the effects of RecX Ng and its mutant variants on RecA Ng and RecA Ec function in vitro. This work was complemented by an analysis of RecX Ng and its mutant variants on pilin variation in N. gonorrhoeae. Most in vitro experiments described below were carried out using RecX Ng and its cognate RecA and SSB proteins and also RecX Ec with its cognate RecA and SSB. The SSB protein used is always cognate to the RecA protein used. We also tested the effect of RecX Ng on RecA Ec (with SSB Ec ) for three reasons: RecA Ec is very well characterized, many mutants are available to use as tools, and we could directly compare RecX Ng with RecX Ec .  (Fig. 1A), consistent with previous results (35). RecX Ec has almost no effect on the ATPase activity of RecA Ng at this same 1:24 ratio (Fig. 1C). In contrast, RecX Ng completely abolishes the ATPase activity of either RecA Ng or RecA Ec on cssDNA within 3-5 min at a 1:24 ratio of RecX (110 nM) to RecA (2.4 M) (Fig. 1, B and D). The percentage of ATPase activity remaining 9 min after RecX addition, averaged over three independent experiments, is detailed in supplemental Table S1. The potent inhibition of RecA Ng by RecX Ng is not due to an intrinsically faster disassembly of RecA Ng because RecX Ng has a similar if not a stronger effect on RecA Ec . The filament capping mechanism is insufficient to explain this rapid inhibitory effect.
RecX Point Mutants Affect RecX Function in Gc as Well as in Vitro Functions-Given the positive effect of RecX Ng on RecA Ng function in the bacterial cell (41), we further explored the unexpectedly inhibitory effects of RecX Ng in vitro. An amino acid alignment of RecX homologs from a variety of Gram-negative and Gram-positive bacteria and from the plant model organism Arabidopsis thaliana was carried out to guide the design of mutant proteins (supplemental Fig. S1). We noted residues that appeared to be conserved across species (Asp-54, Tyr-57, Arg-130, and Phe-132), and residues that appeared to differ specifically between Gc and the other species (Gly-24 and Gln-81). To determine the importance of these residues to RecX Ng function, the conserved residues were mutated to alanines (D54A, Y57A, and R130A/F132A), and the residues that differed in Gc were mutated to the prevailing amino acid in the alignment (G24E and Q81K). RecX Ng point mutant constructs were subsequently recombined into the FA1090⌬recX chromosome (see "Experimental Procedures").
We tested the effect of each RecX point mutant using the well established pilus-dependent colony morphology change assay, which measures observable changes in colony morphology resulting from the appearance of nonpiliated blebs over time (56) and is a widely used method to measure frequencies of Gc pilin antigenic variation (23, 24) ( Fig. 2A). We found that strains R130A/F132A and D54A showed statistically the same amount of variation as the recX null strain. Strains Y57A and G24E showed an intermediate level of variation, statistically distinct from both the recX null mutant and the ⌬recX/ recXHA complement strain. Strain Q81K showed a statistically higher amount of antigenic variation than strain ⌬recX/ recXHA ( Fig. 2A). The RecX protein expression levels in the parent (⌬recX/recXHA) and mutant strains were measured by Western blot (data not shown) and varied from levels equal to that of the parent RecXHA protein to as low as 3% of the parent protein. However, the differences in protein expression levels of the mutant proteins do not appear to be responsible for the observed differences in phase variation. Constructs created to drive the expression of the parent RecXHA protein from two different promoters resulted in a 252-fold difference in RecXHA protein levels, as measured by Western blot, and spanned the various point mutant expression levels, yet both promoter constructs mediated identical levels of phase variation in Gc (data not shown). Therefore, we conclude that these amino acid residues all influence RecX activity in Gc, with the relative importance of the residues being: R130A/F132A Ͼ D54A Ͼ Y57A Ͼ G24E Ͼ Q81K.
We purified His-tagged versions of the same point mutants and characterized their effect on RecA using ATPase assays. Circular dichroism spectra indicate that all of the mutants are folded correctly except for R130A/F132A, which was slightly misfolded (supplemental Fig. S2). The His-tagged wild-type RecX Ng has the same inhibitory effects on RecA Ec and RecA Ng filaments as the native protein, as seen in the ATPase assay (Fig.  2B). ATPase assays showed that the RecX Ng point mutants that lead to decreased pilus phase variation similar to ⌬recX strains (R130A/F132A and D54A) are also deficient in inhibiting RecA-catalyzed ATP hydrolysis (Fig. 2B). The Q81K mutant was proficient in RecA ATPase inhibition, whereas the Y57A and G24E mutants showed intermediate activities in both assays (Fig. 2B). A quantitative measure of the percentage of ATPase activity 12 min after RecX addition compared with a reaction where no RecX was added is shown in Fig. 2C. The RecX Ng mutants exhibited deficiencies in RecA ATPase activity inhibition in the order: R130A/F132A Ͼ G24E Ͼ D54A Ͼ Y57A Ͼ Q81K. This is the same trend seen for deficiencies in pilus phase variation, with only the G24E mutation exhibiting somewhat greater effects in vitro than in N. gonorrhoeae. This indicates that the inhibitory effect of RecX Ng on RecA activities in vitro corresponds to a positive effect on pilus phase and antigenic variation in the bacterial cell.
The Ratio of RecX to RecA Proteins in N. gonorrhoeae Is Higher than in E. coli-RecX Ng is able to inhibit the RecA Ng ATPase activity at far substoichiometric levels in vitro (Fig. 1).
To determine the relative levels of RecA and RecX in Gc, we performed a semi-quantitative Western blot analysis of the two proteins. Serial dilutions of purified RecX and RecA were used as standards for comparison against cell extracts made from Gc strain FA1090. We found that there averaged one RecX molecule for every 5-10 RecA molecules (data not shown), a higher RecX Ng to RecA ratio than used in all of the assays shown above and a much higher ratio than the ratio of the E. coli proteins in the bacterial cell (approximately one RecX to 300 RecA) (31)(32)(33). These data suggest that the strong inhibition of RecA by RecX is important for optimal RecA function in Gc. Fig. 1 and the correlation between RecX activity in N. gonorrhoeae and the in vitro activities of RecX Ng shown above prompted us to investigate the mechanism by which RecX Ng inhibits RecA activities and compare it with the established filament capping mechanism for RecX Ec . We used EM to determine whether the addition of RecX Ng leads to RecA filament disassembly or whether RecX Ng inhibits RecA ATPase activity, whereas RecA remains bound to DNA. The addition of ATP␥S at different time points after RecX addition allows us to fix RecA filaments for visualization of the filament state at that time. In our experiments we observed many different states of RecA filaments and categorized them into five main groups for analysis. Representative images of each category are shown in supplemental Fig. S3. The categories are: full filaments where the cssDNA is coated with RecA (supplemental Fig. S3A), small gaps where a short region of SSB-coated DNA interrupts RecA filaments (supplemental Fig. S3B; generally more than 80% RecA coverage), medium gaps where SSB coats a larger region of the DNA and the RecA filaments are considerably shorter than in a full filament (supplemental Fig. S3C; corresponding roughly to 20 -80% coverage), big gaps where the RecA filaments are very short (Ͻ20% coverage), the rest of the DNA molecule, which is coated by SSB (supplemental Fig. S3D), and SSB-coated DNA where the whole DNA molecule is coated with SSB (supplemental Fig. S3E).

RecX Ng Addition Leads to Disassembly of RecA Filaments within 3-5 min-The data presented in
When no RecX is added, most RecA filaments show full filaments and no regions of SSB-coated DNA (Fig. 3, A-C). When 100 nM RecX Ec was added, RecA Ec filaments were considerably shorter after 10 min, and after 20 min only big gaps (short RecAcoated DNA segments) remained. The addition of 110 nM RecX Ng led to a complete disassembly of RecA Ec or RecA Ng filaments after 3-5 min. A panel with representative images of the two most prominent categories of filaments for each time point is shown in Fig. 3. The rapid disassembly of RecA filaments observed via EM correlates well with the decrease in ATPase activity (Fig.  1) Table S2). Using a rate of disassembly of 60 -70 RecA Ec monomers/ minute (57), 7-12 gaps would be needed to account for complete disassembly within 3-5 min. The average number of visible gaps per filament for each time point is plotted in Fig. 4. The differences between the number of gaps produced by the two RecX proteins is small at best. At none of the time points do the gaps/filament average more than 2.5. This number is far below the number of gaps that would be expected (7)(8)(9)(10)(11)(12) if RecX Ng was simply creating more ends from which RecA would disassemble at 60 -70 monomers/minute. The small differences evident in Fig. 4 argue that creating more disassembling RecA filament ends accounts at best for only a small part of the increase in inhibition by RecX Ng . We thus continued to explore other potential effects of RecX, focusing on whether a direct facilitation of filament disassembly could play a role in the mechanism of RecX Ng action. RecX Ng and RecX Ec Bind to ssDNA with Similar Affinity-We tested whether RecX Ng binds ssDNA with higher affinity than RecX Ec to find an explanation for the strong RecA-mediated ATPase inhibition by RecX Ng and for the possibility that RecX Ng creates slightly more disassembling ends. Fluorescence anisotropy experiments with fluorescein-labeled 50-nucleotide ssDNA show an increase in anisotropy as the RecX concentration is increased (Fig. 5). For RecX Ng , an apparent dissociation constant (K d ) of 88.6 Ϯ 7.1 nM was extracted from analysis of the binding curve. The apparent K d for RecX Ec using the same DNA substrate is 81.0 Ϯ 6.4 nM. These two values are within standard error from each other, and we can conclude an almost identical affinity for ssDNA for both RecX proteins. Therefore, the stronger inhibitory effect of RecX Ng on RecA ATP hydrolysis is not due to RecX Ng binding more tightly to ssDNA than RecX Ec . We note that this assay reveals a stronger binding of RecX Ec to ssDNA than reported previously (35) (see "Discussion").
RecX Ng and RecX Ec Do Not Facilitate Disassembly of the ATPase-deficient RecA Mutant E38K K72R-To further investigate how RecX Ng causes such an abrupt RecA inhibition, we used the RecA Ec E38K K72R double mutant, and the fact that RecA-mediated ATP hydrolysis is coupled to RecA disassembly, as a tool to study the effect of RecX Ng on RecA in the absence of RecA ATP hydrolysis (13,57). RecA E38K K72R is proficient in forming filaments on cssDNA (54), but it hydrolyzes little or no ATP and is thus deficient in disassembly. Elec-tron micrographs show that preformed RecA E38K K72R filaments on cssDNA are not substantially affected after a 15-min incubation with either RecX Ng or RecX Ec (Fig. 6). This indicates that neither of the RecX proteins can cause disassembly of a RecA in the absence of ATP hydrolysis. Therefore, ATP hydrolysis by RecA is necessary for RecX to mediate RecA disassembly.
RecX Ng Inhibits the ATPase Activity of RecA Ec E38K and the dATPase Activity of Wild-type RecA Ec , whereas RecX Ec Does Not-Certain conditions are known in which RecA Ec is ATPase-proficient, but disassembly is uncoupled from ATP hydrolysis. For example, in the presence of dATP, the wild-type RecA Ec does   not disassemble significantly from cssDNA (13). A similar observation is made when reactions are carried out with RecA Ec E38K, using ATP as a nucleotide co-factor. 5 When RecX Ec is added to RecA Ec in the presence of dATP or RecA Ec E38K, it causes no effect on dATP or ATP hydrolysis, respectively (Fig.  7). These results are consistent with the proposed filament capping mechanism for RecX Ec , in which RecX Ec acts to cap the growing ends of RecA filaments. If no end-dependent filament disassembly occurs, no change in ATPase activity can be observed. Contrary to RecX Ec , RecX Ng rapidly inhibits both the dATPase activity of wild-type RecA Ec (Fig. 7A) and the ATPase activity of RecA Ec E38K (Fig. 7B). EM experiments under these conditions show that RecA filaments are disassembled when RecX Ng is present (data not shown), strongly suggesting that the inhibition of ATPase activity by RecX Ng results from RecA disassembly. This again indicates that the RecX Ng protein has an effect on RecA function beyond that of filament capping. We note that the slight increase in the number of gaps per filaments might be related to and may be symptomatic of a facilitated disassembly process.

DISCUSSION
There are two main conclusions to this work. First, RecX Ng , a protein with a strong inhibitory effect on RecA protein filaments, has a decidedly positive effect on the function of RecA Ng protein in the cell. This indicates that optimal RecA function requires the substantial constraint of RecA filament length and/or the length of time the filament exists. For recombinational DNA repair, DNA transformation, and pilus antigenic variation, less can be more. Second, the RecX Ng protein is a more potent inhibitor than the previously characterized RecX Ec protein and may reveal a broader range of potential RecX functionality. It is likely that RecX Ng accelerates end-dependent RecA filament disassembly using mechanisms that are absent or cryptic in the RecX Ec protein.
The general result that RecX Ng inhibits in vitro while enhancing recombination in N. gonorrhoeae is strengthened by the effects of a series of RecX Ng mutant proteins exhibiting a range of functional deficiencies. There is a strong correlation between the capacity of a RecX Ng variant to inhibit the RecA Ng ATPase and the capacity of that same mutant to support cellular functions such as RecA Ng -mediated pilus phase variation. Two mutations that affect residues conserved among all RecX proteins (R130A/F132A and D54A; see supplemental Fig. S2 for an alignment) are deficient in RecA ATPase inhibition and show the lowest frequency of pilus phase variation. These residues therefore are likely important for RecX Ng activity. Mutations that convert residues that are highly conserved in most RecX proteins (but are altered in RecX Ng ) to the consensus residue generally have intermediate effects. Overall, the trends allow us to conclude that strong inhibition of RecA ATPase activity corresponds to high pilus phase variation frequency.
Mechanistically, the RecX Ng protein imposes substantial inhibition on RecA filaments (both RecA Ng and RecA Ec ) at rates too rapid to be explained by the filament capping mechanism proposed for RecX Ec (35). The decline in ATP hydrolysis in all cases corresponds closely to filament disassembly as seen by EM. Thus, RecX Ng is somehow facilitating this more rapid disassembly. This could occur by 1) a displacement mechanism in which RecX actively removes RecA from the DNA (a RecXbased activity that does not rely on any action of RecA); 2) a disruption of subunit-subunit interactions to create and cap more filament ends, the wedge model, leading to more rapid disassembly at the now more numerous disassembling ends; or 3) a facilitation of ATPase-and end-dependent RecA filament disassembly (an active RecX mechanism, but one that relies on RecA-mediated ATP hydrolysis). In mechanism 3, RecX Ng could simply enhance the normal RecA disassembly process, perhaps by means of conformation changes brought about by RecX Ng binding in the filament groove.
We argue against active removal of RecA (mechanism 1) in a manner that does not rely on any function of RecA, because RecX Ng is unable to displace the RecA K72R E38K mutant protein from the DNA. This mutant does not hydrolyze ATP, indicating that RecA-mediated ATP hydrolysis is required for RecA filament disassembly whether RecX Ng is present or not. Mech-5 R. Britt, unpublished results. anism 2, the wedge model, is an elaboration of the capping idea, in which filament ends are both created and capped by RecX Ng . The small apparent increase in filament gaps observed by EM when RecX Ng is present relative to those present with RecX Ec supports the idea that RecX Ng may have a limited capacity to disrupt subunit-subunit interactions in a RecA filament. However, the observed increase in filament gaps ( Fig. 4 and supplemental Fig. S4) is small at best and insufficient to explain the enhancement of activity with RecX Ng . Although some RecX Nggenerated gaps may have been missed, we do not think the potential underestimation is sufficient to allow for a complete explanation of the effects of RecX Ng . If mechanism 2 is insufficient to explain RecX Ng activity, then there may be a direct facilitation of RecA dissociation by RecX (mechanism 3), a possibility we currently favor based on these data (Fig. 8). In all cases, the mechanism by which RecX Ng promotes RecA filament disassembly requires ATP hydrolysis by RecA.
When disassembly is uncoupled from ATP hydrolysis, using dATP (13) or RecA E38K, 5 RecX Ng addition leads to RecA disassembly, whereas RecX Ec addition does not. This is one of the key differences in the mechanisms between the two RecX proteins and allows us to argue that complex formation of RecX Ng with RecA Ng may result in more rapid but still ATPase-dependent RecA disassembly. In contrast, RecX Ec relies on the intrinsic disassembly rate of RecA. A more active facilitation of RecA disassembly by the RecX Ec protein, similar to mechanism 3, was previously proposed by Ragone et al. (36) based on surface plasmon resonance measurements of very short RecA filaments. We have found no evidence for an active RecX Ec mechanism with our much longer RecA filaments. However, the new results with RecX Ng suggest that such an active mechanism may be well within the functional potential of RecX proteins, and the assays we are using for RecX Ec studies may simply obscure it. Mechanisms 2 and 3 both have the virtue of providing a role for the RecX protein previously shown to bind along the major groove of a RecA filament (34).
When assayed by fluorescence anisotropy, RecX Ec has a more than 10-fold lower K d for ssDNA than had previously been reported (35). These assays are likely more reflective of the actual dissociation constants than the ones reported previously using electrophoretic mobility shift assays. Other cellular factors may greatly influence the interaction of RecX with ssDNA. For example the presence of a RecA filament on ssDNA could increase the affinity of RecX to the DNA. Alternatively, the presence of SSB could decrease the affinity, which would be consistent with the competition of RecX Ec with SSB Ec (26) and RecX Ng with SSB Ng (data not shown). In sum, ssDNA binding by RecX may play a more important role in the mechanisms of either RecX protein than assumed so far. However, the fact that RecX Ng has slightly lower ssDNA binding affinity than RecX Ec implies that ssDNA binding is not the reason for the much stronger inhibitory effect on RecA by RecX Ng .
We acknowledge that the strong inhibition of RecA may not directly affect pilus phase variation. There might be another enzyme or protein not yet discovered that could shed light on this apparently counterintuitive correlation. For example, another protein or proteins could partially counter the effects of RecX and perhaps cooperate with it to create the ideal RecA filament species. However, based on the evidence to date, we suggest that the key function for RecX is to keep RecA filaments at an optimal length. The process of pilus phase variation involves the recombination of very short DNA sequences, which may require shorter RecA filaments. Longer RecA filaments might inhibit the process of pilus phase variation by producing nonproductive products. Regardless, the very different mechanism-of-action of the RecX Ng and RecX Ec proteins shows how these two Gram-negative bacteria have optimized their recombination processes to respond to different selective pressures.