RecQ DNA helicase HRDC domains are critical determinants in Neisseria gonorrhoeae pilin antigenic variation and DNA repair

Summary

Neisseria gonorrhoeae (Gc), an obligate human bacterial pathogen, utilizes pilin antigenic variation to evade host immune defenses. Antigenic variation is driven by recombination between expressed (pilE) and silent (pilS) copies of the pilin gene, which encodes the major structural component of the type IV pilus. We have investigated the role of the Gc RecQ DNA helicase (GcRecQ) in this process. Whereas the vast majority of bacterial RecQ proteins encode a single “Helicase and RNaseD C-terminal” (HRDC) domain, GcRecQ encodes three tandem HRDC domains at its C-terminus. Gc mutants encoding versions of GcRecQ with either two or all three C-terminal HRDC domains removed are deficient in pilin variation and sensitized to ultraviolet-light induced DNA damage. Biochemical analysis of a GcRecQ protein variant lacking two HRDC domains, GcRecQΔHRDC2,3, shows it has decreased affinity for single-stranded and partial duplex DNA and reduced unwinding activity on a synthetic Holliday junction substrate relative to full-length GcRecQ in the presence of N. gonorrhoeae single-stranded DNA binding protein (GcSSB). Our results demonstrate that the multiple HRDC domain architecture in GcRecQ is critical for structure-specific DNA binding and unwinding, and suggest that these features are central to GcRecQ's roles in N. gonorrhoeae antigenic variation and DNA repair.

Introduction

Neisseria gonorrhoeae, the causative agent of the sexually transmitted disease gonorrhea, is able to evade the host immune system by undergoing phase and antigenic variation of surface molecules. In part, this ability allows individuals to suffer multiple N. gonorrhoeae infections since gonococcal exposure provides limited protective immunological memory to prevent future infections (Boslego et al., 1991, Schmidt et al., 2001). Several surface molecules are varied by N. gonorrhoeae to avoid the immune response including lipooligosaccharides, opacity proteins, and type IV pili (Gotschlich, 1994, Stern et al., 1986, Swanson, 1973). Type IV pili are extracellular structures that promote infection and participate in cell adherence, twitching motility, and DNA transformation (Cohen & Cannon, 1999, Kellogg et al., 1963, Swanson, 1973, Henrichsen, 1975, Wolfgang et al., 1998, Seifert et al., 1990, Sparling, 1966). Antigenic variation of the pilin subunits of N. gonorrhoeae pili occurs at a rate of greater than 4 × 10⁻³ events per cell per generation, which is the highest reported level of antigenic variation for a gene in a pathogenic organism (Criss et al., 2005).

Pilin antigenic variation occurs by recombination-driven changes to the pilin-encoding pilE gene. The Gc genome contains a single gene that expresses pilin (pilE) and multiple partial copies of the pilin gene found at several chromosomal loci (Haas et al., 1992, Hamrick et al., 2001). These partial copies are referred to as silent pilin genes, or pilS, because they lack the promoter and 5′ coding sequences of pilE that are needed for expression (Meyer et al., 1984). High-frequency recombination at pilE produces hybrid genes in which sequences from one or more pilS copies are recombined into the pilE locus, which allows the antigenic sequence to be varied (Haas & Meyer, 1986, Segal et al., 1986). Recombination between pilE and pilS loci requires recA (Koomey et al., 1987) and a number of genes homologous to the Escherichia coli RecF recombination and Holliday junction resolution pathway genes (Mehr & Seifert, 1998, Sechman et al., 2006, Sechman et al., 2005). In the E. coli RecF pathway, recombination substrates are generated by the joint activities of the RecQ DNA helicase and RecJ exonuclease, which produce 3′ single-stranded DNA (ssDNA) (Umezu et al., 1990, Lovett & Kolodner, 1989). RecF, RecO, and RecR proteins then mediate the loading of RecA onto the resulting ssDNA, which drives homologous recombination (Hegde et al., 1996, Ivancic-Bace et al., 2005, Morimatsu & Kowalczykowski, 2003, Kowalczykowski et al., 1994). Disruptions in any of the RecF pathway genes present in N. gonorrhoeae (recQ, recJ, recO, and recR (no recF homolog has been identified)) leads to deficiency in pilin variation (Mehr & Seifert, 1998). In addition, mutations of the N. gonorrhoeae ruvA, ruvB, ruvC, or recG genes that function in Holliday junction resolution show a similar defect in pilin variation (Sechman et al., 2006). Mutations in these genes also result in a recA-dependent growth defect suggesting they are necessary for processing DNA intermediates generated during recombination (Sechman et al., 2006). Therefore, the mechanisms behind pilin variation likely share some features with DNA recombination and repair pathways in E. coli, but clearly have evolved unique features that allow high-frequency pilE/pilS recombination.

GcRecQ provides an intriguing target for examining specialization of recombination proteins in N. gonorrhoeae. First, mutants of Gc recQ are not only deficient for pilin variation but also sensitized to a variety of DNA damaging agents (Mehr & Seifert, 1998, Stohl & Seifert, 2006). The latter phenotype is also observed upon mutation of recQ in recBC sbcBC E. coli in which the RecF pathway is the predominant means of recombination (Horii & Clark, 1973, Nakayama et al., 1984, Mendonca et al., 1995). Despite the presence of a functional RecBCD pathway in N. gonorrhoeae, the deleterious effects of a recQ mutation suggest that both RecBCD and RecF-like pathways are critical for DNA repair in Gc cells (Stohl & Seifert, 2006). Second, GcRecQ is one of only three identified RecQ homologs to encode three HRDC domains at its C-terminus (also in Neisseria meningitidis and Deinococcus radiodurans RecQs) (Figure 1). This is exceptional, since almost all RecQ family members are comprised of a Helicase, RecQ-conserved (RecQ-Ct), and single HRDC domain (Morozov et al., 1997) (Figure 1). In fact, no known eukaryotic RecQ homolog encodes multiple HRDC domains. Included in this family are the five human RecQ homologs in which mutations of the conserved domains of BLM, WRN, and RECQ4 result in Bloom's, Werner's, and Rothmund-Thomson syndromes, respectively (Bachrati & Hickson, 2003, Ellis et al., 1995, Kitao et al., 1999, Yu et al., 1996).

Figure 1. Domain schematic of RecQ family members.

Alignment of conserved domains present in selected RecQ family members including Human BLM, WRN, RECQ4, S. cerevisiae Sgs1, E. coli RecQ, D. radiodurans RecQ, and N. gonorrhoeae RecQ. The number of residues present in each protein is indicated at the right margin.

The Helicase and RecQ-Ct domains form a single structural unit in E. coli RecQ that is responsible for hydrolysis of ATP, binding to numerous DNA structures, and interactions with heterologous proteins (Bernstein & Keck, 2003, Bernstein et al., 2003, Hu et al., 2005, Liu et al., 2004, Shereda et al., 2007). In E. coli RecQ, the HRDC domain provides additional DNA binding specificity and can sensitively modulate the affinity of the enzyme for different structures (Bernstein & Keck, 2003, Bernstein & Keck, 2005). Removal of the HRDC domains from RecQ in E. coli and D. radiodurans has dramatic effects on the affinity of these proteins for particular DNA structures in vitro (Bernstein & Keck, 2003, Bernstein & Keck, 2005, Killoran & Keck, 2006, Bennett & Keck, 2004). The Saccharomyces cerevisiae RecQ (Sgs1) also exhibits this deficiency in addition to loss of the ability to complement topoisomerase gene mutations (Lu et al., 1996, Mullen et al., 2000). In addition, several point mutations in HRDC domains have been identified that alter binding or processing of particular nucleic acid structures in vitro and in vivo, underscoring the subtlety of HRDC domain specialization (Bernstein & Keck, 2005, Killoran & Keck, 2008, Midtgaard et al., 2006, Zuo et al., 2005). Therefore, the reliance of N. gonorrhoeae on GcRecQ for pilin variation and efficient DNA repair coupled with the presence of three HRDC domains in the protein raises the question of whether this unusual domain arrangement of GcRecQ is important for its function in N. gonorrhoeae.

In this study, we assess the physiological and biochemical roles of the multiple GcRecQ HRDC domains by examining the in vivo and in vitro activities of protein variants in which one or more HRDC domains are absent. Removal of HRDC #2 and #3 impairs GcRecQ function in pilin variation and resistance to UV radiation in vivo. These phenotypes can be complemented by expression of a second, wild-type Gc recQ gene. In contrast, expression of E. coli recQ, D. radiodurans recQ, or a mutant Gc recQ gene that encodes a catalytically-inactive enzyme fails to rescue these deficiencies. GcRecQ lacking HRDC #2 and #3 is defective in DNA binding in vitro relative to wild-type GcRecQ but displays normal levels of DNA-dependent ATP hydrolysis. Interestingly, this protein variant also has a weakened ability to unwind Holliday junction DNA in the presence of GcSSB, but not partial-duplex DNA. Together, these results indicate the multiple HRDC domains target GcRecQ activity to particular DNA structures that may arise in both pilin variation and DNA repair of UV-induced lesions in N. gonorrhoeae.

Results

Multiple HRDC domains of GcRecQ are required for pilin antigenic variation

The ability to identify N. gonorrhoeae colonies that have undergone pilin gene recombination through observation of pilus-dependent colony morphology has greatly facilitated the study of pilE recombination and the identification of recombination-deficient Gc mutants (Gibbs et al., 1989, Jonsson et al., 1991, Swanson et al., 1986, Sechman et al., 2005). In the process of pilin antigenic variation, non-piliated (P−) colonies can arise from piliated (P+) progenitors when recombination introduces a stop codon or alternate sequence into the pilE gene that prevents normal assembly of pili (Haas et al., 1987, Gibbs et al., 1989, Bergstrom et al., 1986). Therefore, measuring the frequency of P− gonococcal colonies provides quantitative insights into recombination events at pilE.

Earlier work has demonstrated that Gc recQ is involved in antigenic variation of type IV pili (Mehr & Seifert, 1998). It was unclear, however, whether the unusual triple-HRDC structure of GcRecQ (HRDC #1, #2, and #3 from N-terminus to C-terminus, respectively) was important for GcRecQ-mediated pilin antigenic variation. To test the roles of HRDC domains in GcRecQ, we constructed gonococcal strains that encode truncated versions of GcRecQ lacking one, two, or all three HRDC domains (Figure 2A). These strains were tested for pilin antigenic variation by measuring the percentage of P− colonies that arose from P+ colonies following 24 hours of growth (see Experimental procedures). Wild-type N. gonorrhoeae formed P− colonies with a frequency of 3.5 +/− 2.0 % (Figure 2B). This is in agreement with previously published measurements that range from ∼1-5% using the same methodology (Mehr & Seifert, 1998, Sechman et al., 2005). A strain expressing GcRecQ lacking HRDC #3 (Gc recQ-Opal2062) underwent pilin antigenic variation at a similar rate to the wild-type strain, exhibiting an average of 2.3 +/− 0.9% P− colonies (Figure 2B). In contrast, gonococcal strains expressing GcRecQ lacking HRDC #2 and #3 (Gc recQ-Opal1801) or all three HRDC domains (Gc recQ-Opal1564) were both reduced in their ability to undergo pilin antigenic variation, forming 0.5 +/− 0.4% and 0.5 +/- 0.3% P− colonies, respectively (Figure 2B). The decreased frequency of P− colonies in both of these strains showed a statistically significant difference from wild type cells (P-value < 0.05). The decreased variation in the HRDC #2 and HRDC #3 truncation strains was similar to that of a recQ interruption mutant in which an erythromycin resistance marker replaces 58% of the recQ coding sequence (strain recQ::erm) (Figure 2B). To ensure that the decreased appearance of P− variants in the strains expressing truncated RecQ proteins was not due to a growth defect in these mutants, we measured growth of these strains over time and found that these strains grew at the same rate as the wild type parent strain (Figure S1 (Supporting information)). As a control to determine whether the truncated RecQ proteins were simply non-functional, we introduced Gc recQ-Opal1801 at a distinct locus in otherwise wild-type (Gc recQ⁺) Gc cells. Interestingly, Gc recQ-Opal1801 was dominant negative over wild-type recQ, with the formation of P− colonies being reduced to 0.7 +/−0.6% (Figure 2B). These results demonstrate that the multiple-HRDC structure of GcRecQ is critical to RecQ's role in recombination-based variation of pili.

Figure 2. Pilin variation assays demonstrate the importance of GcRecQ HRDC #2 and #3.

(A) Domain schematic of GcRecQ marking the position of Opal stop codon insertions (triangles). Residue numbers with corresponding nucleotide positions in parenthesis are indicated. The percent of N. gonorrhoeae P− colonies for (B) wild-type cells and strains encoding truncated forms of GcRecQ and (C) the Gc recQ-Opal1564 strain tested for complementation with the indicated constructs. Asterisks indicate a statistically significant difference from wild-type (P-value <0.05). Experiments were performed a minimum of three times for each strain. Error bars represent one standard deviation from the mean.

To test the specificity of these defects, we introduced a wild-type copy of Gc recQ onto the gonococcal chromosome of the Gc recQ-Opal1564 strain. Gc recQ partially complements the recQ-Opal1564 defect in pilin variation; resulting in 1.7 +/− 0.5% P− colonies (P-value = 0.11) (Figure 2C). The increased frequency relative to the strain in which Gc recQ-Opal1801 is induced in wild-type (Gc recQ⁺) cells (Figure 2B, fifth bar) may be due to differences in expression caused by complementation with IPTG-inducible constructs. Although we were not able to detect native or induced GcRecQ constructs in cell lysates using several western blotting approaches, the simplest explanation for difference in P− colony formation between the complementation strains (Figure 2B, fifth bar and Figure 2C, second bar) is that there is increased expression of WT GcRecQ from the IPTG-induced construct versus that under native expression. We also tested complementation with a Gc recQ mutant encoding an arginine substitution at lysine 47. Mutation of the analogous lysine to arginine is a well-characterized substitution in the Walker A motif that abolishes catalytic activity in E. coli RecQ (EcRecQ), S. cerevisiae Sgs1, and the Human RecQ homologs RECQ5β and WRN (Walker et al., 1982, Xu et al., 2003, Lu et al., 1996, Gray et al., 1997, Garcia et al., 2004). In the latter, this mutation is also sufficient to induce a Werner's syndrome phenotype in mouse tail-derived fibroblasts (Wang et al., 2000). Gc recQK47R did not complement the defect in pilin antigenic variation of the Gc recQ-Opal1564 strain forming 0.7 +/− 0.2% of P− colonies (Figure 2C). These results demonstrate that catalytic activity of GcRecQ is required for pilin variation.

We next tested whether GcRecQ performs an activity during pilin variation that might be common amongst other bacterial RecQs by examining pilin variation in the Gc recQ-Opal1564 strain when recQ genes from E. coli or D. radiodurans are expressed from the chromosome. We hypothesized that if these bacterial RecQ homologs could complement strains lacking GcRecQ HRDC domains in pilin variation, it would point to a shared feature or function among these homologs. Contrary to this notion, expression of neither E. coli recQ nor D. radiodurans recQ in the Gc recQ-Opal1564 strain increased formation of P− colony variants (Figure 2C). These results further support the specific requirement for GcRecQ in pilin variation and demonstrate that other bacterial RecQs, including one with multiple HRDC domains, are not sufficient for this process.

GcRecQ HRDC domains are required for resistance to UV radiation

The deficiency of strains encoding truncated versions of GcRecQ in pilin variation raised the question of whether HRDC #2 and #3 are necessary for other functions of GcRecQ in the cell. Since a mutation in Gc recQ had previously been demonstrated to sensitize gonococci to DNA damaging agents (Mehr & Seifert, 1998), we tested each of our strains for the ability to survive UV irradiation. As was the case for pilin variation, the Gc recQ-Opal2062 strain showed resistance at levels similar to wild-type gonococci. In contrast, the Gc recQ-Opal1801 and Gc recQ-Opal1564 strains were significantly reduced in their ability to survive UV irradiation (Figure 3A) showing the same reduction in survival as the recQ::erm interruption mutant (data not shown). This sensitivity to UV irradiation indicates that HRDC #2 is critical for GcRecQ function in DNA damage repair, which parallels its importance in pilin variation. We were able to fully complement the defect in DNA damage repair in Gc recQ-Opal1564 by expressing a wild-type copy of Gc recQ in this assay (Figure 3B). However, expression of Gc recQK47R in the Gc recQ-Opal1564 strain was not able to complement this defect, indicating that the catalytic activity of GcRecQ is required for its function in repair of UV light-induced DNA damage as well (Figure 3B). In addition, we observed UV sensitivity in the Gc recQ-Opal1564 strains expressing E. coli or D. radiodurans recQ (Figure 3B). Given the involvement of recQ to UV resistance in E. coli (Nakayama et al., 1984), the UV sensitivity of these strains emphasizes the species specificity of these bacterial recQ genes. These experiments corroborate the results of our pilin variation assays, in which there is a specific requirement for the presence of HRDC #2 and catalytic activity of GcRecQ for its in vivo function.

Figure 3. HRDC #2 and #3 are necessary for GcRecQ function in UV-resistance.

Percent of survival relative to unirradiated strains at increasing doses of UV radiation for (A) wild-type and strains encoding truncated GcRecQ protein variants lacking its C-terminal HRDC domains and (B) the Gc recQ-Opal1564 strain tested for complementation with the indicated constructs. Labels for each strain are inset within the graph. Experiments were performed a minimum of three times for each strain. Error bars represent one standard deviation from the mean.

GcRecQ HRDC domains 2 and 3 are involved in DNA binding

The parallel defects in pilin variation and UV resistance caused by deletion of the C-terminal-most two HRDC domains led us to examine the biochemical effects of their absence in GcRecQ. The ability to bind nucleic acids is a conserved function among HRDC domains and their mutation or removal can modulate DNA binding affinity and specificity (Bernstein & Keck, 2005, Killoran & Keck, 2008, Killoran & Keck, 2006, Kitano et al., 2007, Liu et al., 1999, Midtgaard et al., 2006, Zuo et al., 2005). To determine whether truncation of HRDC #2 and #3 alters DNA binding, we compared the equilibrium DNA binding behavior of purified recombinant GcRecQ and GcRecQΔHRDC2,3 (Figure 4A) in vitro using a fluorescence anisotropy (FA) assay. In these experiments, fluorescein-labeled ssDNA, partial duplex DNA (dup-3′), or Holliday junction (HJ) DNA structures were combined with GcRecQ and changes in FA due to DNA binding were measured.

Figure 4. GcRecQΔHRDC2,3 is defective for binding some DNA structures.

(A) Purified GcRecQ and GcRecQΔHRDC2,3 protein shown with a protein marker (M) with masses indicated on the right edge of the gel. Fluorescence anisotropy of GcRecQ (diamonds) and GcRecQΔHRDC2,3 (squares) plotted at increasing protein concentrations for (B) ssDNA, (C) dup-3′, and (D) HJ substrates. Labels used for detection are indicated by a star on their respective substrate schematic inset within each graph. Error bars represent one standard deviation from the mean.

The DNA binding studies showed that deletion of HRDC domains #2 and #3 of GcRecQ weakened DNA binding in a structure-specific manner. GcRecQ bound ssDNA with an apparent K_d (K_d,app) of 1.9 +/−0.2 nM whereas GcRecQΔHRDC2,3 had 30-fold weaker binding (K_d,app, 60.6 +/−5.0 nM) (Figure 4B). Likewise, the affinity for dup-3′ was significantly decreased for GcRecQΔHRDC2,3, with a K_d,app 58.2 +/−3.9 nM for GcRecQ versus ∼1200 nM for GcRecQΔHRDC2,3 (Figure 4C). The K_d,app for GcRecQΔHRDC2,3 is an estimate since binding was not saturated over the protein concentration range tested. Interestingly, these large defects were not observed for HJ DNA where GcRecQ and GcRecQΔHRDC2,3 exhibited a biphasic shaped curve that may reflect multiple binding events on this substrate (Figure 4D). To compare the relative affinity of these proteins, we calculated K_d,app values for their binding curves at low and high protein concentrations. For GcRecQ, these K_d,app values were 0.9 +/− 0.1 nM and 270 +/− 134 nM, respectively. GcRecQΔHRDC2,3 had a slightly weaker affinity with K_d,app values of 1.9 +/− 0.7 nM and 470 +/− 3.6 nM. The gradual increase in anisotropy across a broad range of protein concentrations for the HJ substrate suggests that GcRecQ may have multiple modes of binding, consistent with our observations of D. radiodurans RecQ binding to this substrate (Killoran & Keck, 2008). These results indicate HRDC #2 and #3 significantly enhance GcRecQ binding to some, but not all DNA structures, which is similar to previous observations of RecQ HRDC domains preferentially binding ssDNA-containing substrates (Bernstein & Keck, 2005, Liu et al., 1999, Killoran & Keck, 2008, Killoran & Keck, 2006). The presence of HRDC domains #2 and #3 in GcRecQ may play an important role in targeting this enzyme to specific DNA structures in the cell. Moreover, the modest HJ binding differences indicate that other defects observed with GcRecQΔHRDC2,3 are not due to gross misfolding of the protein.

Deletion of HRDC #2 and #3 does not alter GcRecQ ATP hydrolysis

To determine whether the absence of HRDC #2 and #3 also alters the enzymatic activity of GcRecQ, we measured the ssDNA-dependent ATP hydrolysis rate of GcRecQ and GcRecQΔHRDC2,3. DNA stimulates ATP hydrolysis by GcRecQ and allows comparison of its catalytic activity with that of the GcRecQΔHRDC2,3 variant. Since GcRecQΔHRDC2,3 has a lower affinity for ssDNA, we expected that ssDNA stimulation of ATP hydrolysis might accordingly require higher DNA concentrations. However, both GcRecQ and GcRecQΔHRDC2,3 were stimulated similarly by the addition of dT₂₈ ssDNA and reached the same maximal ATP hydrolysis rate of ∼800 min⁻¹ at the highest DNA concentration tested (Figure S2). Interestingly, in similar assays removal of the single HRDC in E. coli and HRDC #2 and #3 in D. radiodurans RecQ increases their ssDNA-dependent ATP hydrolysis rate (Bernstein & Keck, 2003, Killoran & Keck, 2006). The results indicate the defect in ssDNA binding by GcRecQΔHRDC2,3 does not translate into a deficiency in ATP hydrolysis on ssDNA. A similar observation has been made for an EcRecQ variant lacking its HRDC domain (Bernstein & Keck, 2003). This may be due to differences in GcRecQ affinity for ssDNA upon ATP binding or hydrolysis or the way in which the HRDC domains associate with DNA during catalysis. Therefore, the enhancement of ssDNA binding by HRDC #2 and #3 may be minimized when GcRecQ is actively hydrolyzing ATP on DNA substrates.

GcRecQΔHRDC2,3 is defective for unwinding Holliday junctions in the presence of GcSSB

We next tested the ability of GcRecQ and GcRecQΔHRDC2,3 to unwind dup-3′ and HJ DNA substrates. Helicase activity coordinates both DNA binding and ATP hydrolysis, providing a critical test of each enzyme's function. Since DNA binding was weakened significantly for GcRecQΔHRDC2,3 on dup-3′ but not HJ DNA, we anticipated that helicase activity might parallel these results. In contrast, GcRecQΔHRDC2,3 processed the dup-3′ with similar efficiency as GcRecQ across a concentration range of 3-30 nM (Figure S3A). These results are analogous to our ATP hydrolysis assays that demonstrate similar activity among these proteins on DNA substrates. Similarly, an EcRecQ protein variant that lacks its HRDC domain also has a reduction in dup-3′ binding while displaying wild-type unwinding efficiency (Bernstein & Keck, 2003).

To determine if ssDNA binding differences among GcRecQ protein variants could manifest in the presence of GcSSB, we also examined the effect of its addition to our reactions. EcSSB has been observed to stimulate EcRecQ helicase activity by both physically associating with the enzyme and trapping the liberated ssDNA following unwinding (Shereda et al., 2007, Harmon & Kowalczykowski, 2001). Therefore, we expected that GcSSB addition to our reactions would result in a similar stimulation. For these assays, we prepared a long dup-3′ substrate with a 30 base-pair duplex region and 70 base 3′-ssDNA extension. In contrast to the short duplex region of the dup-3′ substrate that can be separated by GcSSB alone, the long dup-3′ provides a binding site for GcSSB and allows assessment of GcRecQ-dependent unwinding of this substrate (Shereda et al., 2007). Using the long dup-3′ substrate in the absence of GcSSB, GcRecQ and GcRecQΔHRDC2,3 showed similar efficiency of unwinding (Figure S3B). Addition of 30 nM GcSSB to these reactions resulted in ∼5-7 fold stimulation of GcRecQ and GcRecQΔHRDC2,3 unwinding activity (Figure S3C). A similar 5-fold stimulation of EcRecQ by EcSSB on this substrate has been observed (Shereda et al., 2007). These results indicate GcRecQ is stimulated by GcSSB to a similar extent on dup-3′ DNA regardless of whether HRDC #2 and #3 are present.

Finally, we tested the helicase activity of these proteins on the synthetic HJ substrate. Both GcRecQ and GcRecQΔHRDC2,3 bound this substrate with similar affinity. We observed a small ∼1.5-fold increase in the estimated concentration necessary to unwind 50% of the HJ substrate for GcRecQΔHRDC2,3 relative to GcRecQ (Figure 5A, Figure S4). However, the presence of GcSSB further reduced the unwinding efficiency of GcRecQΔHRDC2,3 without any apparent effects for GcRecQ (Figure 5B, Figure S4). This difference in GcSSB stimulation among these protein variants is intriguing and shows that HRDC #2 and #3 in GcRecQ affect unwinding of this substrate differently when GcSSB is present. Since there was no difference in stimulation by GcSSB of these proteins on long dup-3′, these results suggest this is specific to the DNA structure being processed by GcRecQ. The ability to efficiently unwind Holliday junction DNA when GcSSB is present highlights a possible role for HRDC domains #2 and #3 in how this intermediate is processed by GcRecQ during DNA recombination. Interestingly, both BLM and EcRecQ require their single HRDC domains for dissolution of double HJ structures in vitro (Wu et al., 2005). Since the accumulation of these DNA structures are a diagnosing feature of Bloom's syndrome in human cells, the participation of HRDC domains in their unwinding may be a conserved feature and suggests the deficiency we observe for GcRecQΔHRDC2,3 may be further increased on this substrate.

Figure 5. GcRecQΔHRDC2,3 is deficient for HJ unwinding and attenuated by GcSSB addition.

GcRecQ or GcRecQΔHRDC2,3 was incubated with (A) HJ or (B) HJ + GcSSB (30 nM, monomers) at concentrations of 1 nM, 2.9 nM, 4.4 nM, 6.6 nM, 10 nM, and 20 nM. The substrate (S) and boiled substrate (B) are included and their positions indicated on the left margin. For (B), GcSSB is present in the substrate and boiled controls. Labels used for detection are indicated by a star on their respective substrate schematic inset within each graph.

Discussion

By generating heterogeneity in its extra-cellular structures, antigenic variation in N. gonorrhoeae is critical to this species' success as a human pathogen. GcRecQ-dependent recombination is essential for high-frequency recombination of pilS sequences into the pilE locus, which encodes the major structural component of pili (Sechman et al., 2006). The unusual domain architecture of GcRecQ led us to investigate the roles of its multiple HRDC domains in its function in N. gonorrhoeae. We find the C-terminal HRDC domains of GcRecQ are required for its cellular functions, as strains encoding GcRecQ protein variants without HRDC #2 and #3 exhibit low levels of pilin variation and sensitivity to UV radiation. These processes require GcRecQ ATPase activity and are not complemented by the RecQ enzymes from E. coli or D. radiodurans, indicating an adapted function for GcRecQ's HRDC domains for its in vivo activities. Biochemical analysis demonstrates a critical role for HRDC #2 and #3 in structure-specific DNA binding and unwinding. These results suggest the “additional” HRDC domains are determinants that help direct GcRecQ to specific DNA structures in the cell that are critical for DNA recombination and repair.

Multiple HRDC domains modulate GcRecQ function

The diversity of enzymes encoding HRDC domains, including DNA helicases, ribonucleases, topoisomerases, and polymerases, testifies to their utility in nucleic acid metabolism. We have demonstrated that the HRDC #2 and #3 in GcRecQ are critical determinants for structure-specific DNA binding; a truncated form lacking these domains has a striking 30-fold defect in binding ssDNA and 20-fold defect in binding dup-3′ while exhibiting no change in affinity for HJ DNA. Sequence comparison reveals only one out of the five basic and aromatic residues important to E. coli RecQ HRDC domain ssDNA binding are conserved in the GcRecQ HRDC domains (Bernstein & Keck, 2005). Likewise, only two of the eight residues in the Sgs1 HRDC domain implicated in DNA binding are conserved in GcRecQ (Liu et al., 1999). These differences are surprising given the 30-fold ssDNA binding defect for GcRecQ lacking HRDC #2 and #3 is the largest out of all the changes we observed for this protein. Since the highest degree of similarity among the HRDC domains of E. coli RecQ and GcRecQ is with the latter's HRDC #1 (53%), DNA binding by HRDC #2 and #3 could be mediated through a different arrangement of surface residues. It is possible HRDC #2 and #3 form a continuous DNA binding surface with the remainder of GcRecQ that not only enhances affinity but provides specificity as well. A similar model of nucleic acid binding has been proposed for E. coli RNaseD in which two HRDC domains are positioned around a central cavity that is thought to accommodate RNA (Zuo et al., 2005). Alternatively, these domains in GcRecQ may not directly bind to DNA but could modulate the structure of the remainder of the enzyme in a way that confers the structure-specific binding attributes demonstrated in this work.

What could be the means of specialization among HRDC domains? Structural analysis of the HRDC domains from WRN, Sgs1, E. coli RecQ, and D. radiodurans RecQ has revealed large changes to the surface charges of exposed residues of these domains that have important functional consequences (Bernstein & Keck, 2005, Killoran & Keck, 2008, Kitano et al., 2007, Liu et al., 1999). For example, mutation of the basic and aromatic residues of the E. coli RecQ HRDC domain can produce 10-fold defects in ssDNA binding and protein variants with dramatically altered structure-specific DNA binding properties (Bernstein & Keck, 2005). Likewise, acidic and hydrophobic residues on the surface of D. radiodurans RecQ, S. cerevisiae Rrp6, and BLM HRDC domains modulate their affinity and specificity for DNA in vitro and in vivo (Bernstein & Keck, 2005, Killoran & Keck, 2008, Midtgaard et al., 2006, Wu et al., 2005). Therefore, it is likely HRDC #2 and #3 have evolved to direct GcRecQ DNA binding through changes to their surface residues as well.

In addition to the large decrease in DNA binding observed in the GcRecQΔHRDC2,3 protein variant, there was also a modest defect in HJ unwinding. This is not the case for all DNA substrates, since this variant unwound dup-3′ and long dup-3′ with similar efficiency. Interestingly, GcRecQΔHRDC2,3 exhibited reduced HJ unwinding in the presence of GcSSB that contrasted with the stimulation observed on long dup-3′. This is likely due to the ssDNA tail of the long dup-3′ substrate providing a GcSSB binding site not initially present in the HJ substrate prior to GcRecQ addition. The lack of GcRecQ stimulation by GcSSB on HJ unwinding suggests that interactions with ssDNA formed during strand separation may be an important function of HRDC #2 and #3. Since we have shown that HRDC #2 and #3 enhances ssDNA affinity, their interaction with ssDNA liberated during unwinding might precede GcSSB binding and dampen its stimulatory effect. Therefore, the effect of HRDC domains in GcRecQ may be to regulate the unwinding efficiency of some DNA structures. The ability to modulate DNA structure-specificity appears to be a highly conserved function among HRDC domains across several types of nucleic acid modifying proteins. Parallels among similar RecQ protein variants highlight a need for HRDC domains to modulate unwinding of a variety of DNA substrates. For example, the HRDC domain of BLM is required to unwind double HJ substrates whereas the removal of HRDC #2 and #3 from D. radiodurans RecQ enhances its activity on dup-3′ and HJ substrates (Wu et al., 2005, Killoran & Keck, 2006). The differences in HJ unwinding by GcRecQ in the absence of HRDC #2 and #3 are more subtle and it is unclear whether the phenotypic effects we observe in vivo can be correlated with reduced efficiency of processing this substrate in vitro.

A redundant role for the RecF-like recombination pathway in recombination and pilin variation

RecQ-dependent unwinding of specific DNA structures is important for recombination initiation by the RecF pathway in E. coli. In vivo studies have highlighted critical roles for this pathway in the rescue of stalled replication forks, repair of gapped DNA structures, and in preventing illegitimate recombination among heterologous sequences (Courcelle & Hanawalt, 1999, Hanada et al., 1997, Morimatsu & Kowalczykowski, 2003). Sensitivity of N. gonorrhoeae to DNA damaging agents such as UV irradiation, hydrogen peroxide, and γ-irradiation in the absence of GcRecQ suggests it functions to monitor the genome in this organism as well (Mehr & Seifert, 1998, Sechman et al., 2006, Sechman et al., 2005, Stohl & Seifert, 2006). Our findings have demonstrated GcRecQ acts to bind and unwind several DNA structures that are common substrates in DNA repair and recombination. However, the question remains whether GcRecQ performs these activities during pilin antigenic variation as well.

Several models have been formed to account for the involvement of a RecF-like pathway in high frequency recombination at pilE. For example, genetic evidence supports a model wherein a specific sequence directs recombination initiation upstream of the expressed pilE gene by creating a substrate for GcRecQ unwinding (Sechman et al., 2006). However, using the E. coli paradigm of RecF-pathway function it is difficult to predict how GcRecQ may recognize such a locus in a sequence-specific manner to initiate recombination via this pathway in N. gonorrhoeae. Our results provide possible evidence that pilin variation and UV-induced DNA damage repair may overlap through a shared DNA substrate(s) processed by GcRecQ. Since GcRecQ lacking HRDC #2 and #3 is deficient for both of these pathways and has defective activities on only a subset of DNA substrates in vitro, we propose that pilin variation could rely upon a similar intermediate used in DNA damage repair. In N. gonorrhoeae, stalling of the replication fork at sites undergoing pilin variation could link GcRecQ function to the processing of these DNA structures. Alternatively, movement of the fork through pilE could generate a particular DNA substrate that attracts GcRecQ to a site near pilE, which would subsequently drive recombination. In such a scenario, where recombination from a replication fork is initiated by GcRecQ, the need for the structural specificity provided by HRDC #2 and #3 is apparent. A locus-specific recombination mechanism that utilizes structure-dependent initiation via a RecF-like pathway may have allowed N. gonorrhoeae to evolve a specialized role for its recombination machinery that required minimal changes to its pathways of DNA metabolism.

Experimental Procedures

Bacterial growth conditions

Bacterial strains are described in Table 1. Piliated, transparent gonococci were used for all experiments. N. gonorrhoeae strains were grown in GCBL (1.5% Proteose Peptone no. 3, 0.4% K₂HPO₄, 0.1% KH₂PO₄, 0.1% NaCl) with Kellogg's supplements (Kellogg et al., 1963) and 0.042% NaHCO₃ at 37°C with aeration or on GCB agar (Difco) plates under 5% CO₂. E. coli strains were grown either on Luria-Bertani (LB) agar plates or in LB broth at 37°C (Sambrook, 2001). Erythromycin was used at 500 μg/ml, and kanamycin was used at 40 μg/ml for E. coli. Chloramphenicol was used at 25 μg/ml for E. coli and at 10 μg/ml for gonococci. Isopropyl-β-D-thiogalactopyranoside (IPTG, Fisher) was used at a final concentration of 1 mM in GCB agar when needed.

Table 1.

Strains and plasmids

Strain or plasmid	Properties	Source or reference
Plasmids
pET28.b	Cloning vector (KanR)	Novagen
pMK202	recQ in pET28.b (KanR)	This work
pMK214	pMK202 with additional 0.8 kb of genomic DNA downstream of recQ (KanR)	This work
pMK215	pMK214 with a nonsense mutation at base 1564 of recQ and an EcoRI site at the mutation (KanR)	This work
pMK216	pMK214 with a nonsense mutation at base 1801 of recQ and an EcoRV site at the mutation (KanR)	This work
pMK217	pMK214 with a nonsense mutation at base 2062 of recQ and a PstI site at the mutation (KanR)	This work
pKH35	Complementation vector (CmR)	(Hamilton et al., 2005)
pMK143	E. coli recQ in pKH35 (CmR)	This work
pMK144	D. radiodurans recQ in pKH35 (CmR)	This work
pMK145	N. gonorrhoeae recQ in pKH35 (CmR)	This work
pMK146	N. gonorrhoeae recQ K47R mutant in pKH35 (CmR)	This work
pPK1014	pMK215 with an erythromycin resistance marker, ermC, replacing bases 229 to 1564 of recQ	This work
pIDN1	Cloning vector (ErmR)	(Hamilton et al., 2001)
Gonococcal strains
MS11	Laboratory strain of N. gonorrhoeae	(Swanson et al., 1971)
VD300	MS 11 ΔpilE2	(Koomey & Falkow, 1987)
PK133	VD300 encoding an allele of recQ with a stop codon inserted before HRDC #3 (Gc recQ-Opal2062)	This work
PK134	VD300 encoding an allele of recQ with a stop codon inserted before HRDC #2,3 (Gc recQ-Opal1801)	This work
PK131	VD300 encoding an allele of recQ with a stop codon inserted before HRDC #1,2,3 (Gc recQ-Opal1564)	This work
PK151	PK131 transformed with pMK145; N. gonorrhoeae recQ complement (CmR)	This work
PK155	PK131 transformed with pMK143; E. coli recQ Complement (CmR)	This work
PK152	PK131 transformed with pMK144; D. radiodurans recQ complement (CmR)	This work
PK158	PK131 transformed with pMK146; N. gonorrhoeae recQK47R complement (CmR)	This work
PK172	PK131 transformed with pMK148; N gonorrhoeae recQ-Opal1801 complement (CmR)	This work
recQ::erm	VD300 transformed with pPK1014 (EmR)	This work

Construction of plasmids and strains

The recQ open reading frame was amplified from N. gonorrhoeae strain FA1090 genomic DNA by the polymerase chain reaction (PCR) and cloned into the pET28.b bacterial expression vector (Novagen) using NdeI and XhoI restriction sites to create pMK202. An additional 0.8 kilobases (kb) of genomic DNA directly downstream of the recQ gene required for recombination was amplified by PCR and cloned into the XhoI restriction site of pMK202 to create pMK214. Nonsense mutations were then made in pMK214 according to the Quikchange mutagenesis protocol (Stratagene) at positions 1564 (using oMK197), 1801 (using oMK199), or 2062 (using oMK201) of N. gonorrhoeae recQ to make plasmids pMK215, pMK216, and pMK217 with the respective restriction sites EcoRI, EcoRV, and PstI at the sites of the mutations (Table S1 (Supporting information)). These plasmids encode versions of GcRecQ truncated at residues 522, 600, and 687, respectively. N. gonorrhoeae strain VD300 was used as the parent strain for making truncations in Gc recQ. VD300 is a derivative of gonococcal strain MS11 and has only one copy of the pilE locus. Strain VD300 was transformed with pMK215, pMK216, or pMK217 using the method of Gunn and Stein (Gunn & Stein, 1996) in order to introduce nonsense mutations into Gc recQ. To screen for the mutations, the Gc recQ locus was amplified by PCR and the resulting PCR product was digested with the appropriate restriction enzyme to identify incorporation of the desired nonsense mutation. The pilE gene of each strain was sequenced to confirm that all strains used for pilin variation assays carried the same allele of pilE.

To construct a strain containing an erythromycin resistance marker replacing most of the recQ coding sequence, the erythromycin resistance gene was cut out of pIDN1 with EcoRI and BsrBI and subcloned into pMK215, restriction digested with EcoRI and MscI. Strain VD300 was transformed with the resulting plasmid, pPK1014, and transformants were selected for resistance to erythromycin. The resulting strain was designated recQ::erm.

Plasmids used for complementation analysis were constructed by amplifying the appropriate recQ gene from N. gonorrhoeae strain FA1090, E. coli strain K12, or D. radiodurans strain R1 by PCR and cloning into the PstI and SacI sites of pKH35. Sequences of all plasmids were confirmed by DNA sequencing. The complementation plasmids were used to transform gonococcal strain PK131 and transformants were selected by growing gonococci on medium containing chloramphenicol. The pilE gene of each strain was sequenced to ensure that all strains carried the same pilE allele.

Phase variation assay

The percent of phase variation of pili was determined as described previously (Sechman et al., 2005). Briefly, strains were streaked on GCB plates and grown for 24 hours. One colony exhibiting the piliated morphology (P+) of each strain was restreaked and grown on GCB agar plates. After 24 hours of growth, sterile filter paper triangles were used to lift five P+ colonies of each strain off of the plate. The lifted colonies were resuspended in GCBL and dilutions were plated to GCB agar. After 18-20 hours of growth the colonies were counted, and the percentage of P− colonies was calculated. Measurements were determined to be significantly different if their calculated P-value was less than 0.05 using Student's t-test. For complementation strains, each stage of gonococcal growth was carried out on GCB plates containing 1 mM IPTG. Inclusion of IPTG in the growth medium did not significantly affect pilin antigenic variation in WT strains (data not shown).

UV radiation survival assay

Gonococci were grown in GCBL for 2 hours at a starting optical density of 0.2 at 540 nm. Cultures were serially diluted, and 100 μl of the 10⁻³, 10⁻⁴, and 10⁻⁵ dilutions was plated to GCB agar and plates were allowed to dry for 1 minute. The plates were irradiated with a UV Stratalinker 1800 (Stratagene) at energy doses of 0-16 J/m² and incubated at 37°C overnight. Colonies were counted after 18-20 hours of growth and the percent survival relative to unirradiated (0 J/ m²) controls were calculated for each UV dosage.

Synthetic DNA substrates

Oligonucleotide sequences are summarized in Table S1. 3′ fluorescein-labeled o18, 3′ fluorescein-labeled o30, and 5′ fluorescein-labeled oHolliday1 were synthesized and purified by The University of Wisconsin-Madison Biotechnology Center. Oligonucleotides o30-2 and o100 were synthesized and purified by Integrated DNA Technologies. All substrates were created by combining the appropriate oligonucleotides in equimolar amounts, boiling for 5 minutes, and slow cooling to room temperature. The dup-3′ substrate was created by combining o18 and o30 to form an 18 base pair (bp) duplex region with a 12 base 3′ single-stranded extension (Bernstein & Keck, 2003). For the long dup-3′ substrate, oligonucleotides o30-2 and o100 were annealed to form a 30 bp duplex and 70 base single-stranded extension (Shereda et al., 2007). The HJ substrate combined oHolliday1, oHolliday2, oHolliday3, and oHolliday4 to form a central four-way junction from which four 20 bp duplex arms extended (Killoran & Keck, 2006). For substrates used in fluorescence anisotropy experiments, fluorescein-labeled o30 was used for ssDNA, and 3′ fluorescein-labeled o18 or oHolliday1 were incorporated into the dup-3′ and HJ substrates, respectively.

Protein purification

Expression of His-tagged GcRecQ and GcRecQΔHRDC2,3 was carried out by transforming E. coli BL21(DE3) cells with plasmids expressing these constructs. Strains were grown in LB medium supplemented with 50 μg/ml kanamycin at 37 °C with orbital shaking. Cells were induced to express protein upon reaching OD₆₀₀=0.6 with the addition of 1 mM IPTG (final concentration). After 4 hours of induction, cells were harvested by centrifugation (10 minutes at 13,000g) and frozen overnight at −80 °C. Pellets were resuspended in lysis buffer (20 mM Tris, pH 8.0, 500 mM NaCl, 10% (v/v) glycerol, 1 mM 2-mercaptoethanol, 1 mM phenylmethylsulphonyl fluoride (PMSF), 100 mM glucose, and 15 mM imidazole and lysed by two passes through a French Press. All subsequent steps were carried out at 4°C. Cell lysate was clarified by centrifugation (30 minutes at 28,000g), loaded onto a Ni²⁺-NTA column (Qiagen), and washed with lysis buffer. His-tagged protein was eluted in a single-step by the addition of lysis buffer containing 250 mM imidazole, diluted 10-fold with Dilution buffer (5 mM Tris, pH 8.0, 20% (v/v) glycerol, 1 mM 2-mercaptoethanol, 1 mM EDTA), and loaded onto a Q Sepharose Fast Flow ion exchange column (GE Healthcare) equilibrated in QFF Buffer (10 mM Tris, pH 8.0, 50 mM NaCl, 10% (v/v) glycerol, 1 mM 2-mercaptoethanol, 1 mM EDTA). Protein was eluted by gradually increasing the concentration of NaCl to 1M. Fractions containing pure protein as assessed by SDS-PAGE were pooled, concentrated to 10 mg/mL, dialyzed against storage buffer (10 mM Tris, pH 8.0, 400 mM NaCl, 40% (v/v) glycerol, 1 mM 2-mercaptoethanol, 1 mM EDTA), and stored at −80 °C. Purified GcSSB was a gift from the laboratory of Michael Cox, University of Wisconsin-Madison.

DNA binding assays

Concentrated stocks of either GcRecQ or GcRecQΔHRDC2,3 were serially diluted in 20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM 2-mercaptoethanol, 1 mM MgCl₂, 0.1 g/L bovine serum albumin, 4% (v/v) glycerol and incubated with 1 nM fluorescein-labeled o30, dup-3′, or HJ DNA substrates at room temperature for 10 minutes. The fluorescence anisotropy of each sample was measured at 25 °C using a Panvera Beacon 2000 fluorescence polarization system set at 490 nm excitation and 535 nm emission wavelengths. For the range of protein concentrations tested the fluorescence intensity for substrates did not vary by more than 15 %, indicating protein binding did not influence the quantum yield of measurements. Each reaction was repeated a minimum of three times and plotted as an average value with error bars representing one standard deviation. Apparent K_d values were determined for each individual data set using the Curve Expert software program after which the average, standard-deviation, and P-values were calculated using the MedCalc software program.

DNA-dependent ATP hydrolysis assays

ATPase assays were performed as previously described (Bernstein & Keck, 2003). The ATP hydrolysis activity of GcRecQ and GcRecQΔHRDC2,3 was stimulated by the addition of dT₂₈ across a range of concentrations from 0.028-2800 nM (nucleotides). Each reaction was performed a minimum of three times and the rate of ATP hydrolysis at each DNA concentration was plotted as an average value with error bars representing one standard deviation.

DNA unwinding assays

Substrates were created as described above with the exception that the 5′ end of either o18, o30-2, or oHolliday1 was phosphorylated by T4 polynucleotide kinase (New England Biolabs) with [γ-³²P]-ATP prior to gel-purification by 10% native PAGE and electroelution of annealed substrates. GcRecQ and GcRecQΔHRDC2,3 were incubated with ∼1 nM substrate (molecules) for 30 minutes at room temperature in 20 mM Tris, pH 8.0, 50 mM NaCl, 1 mM 2-mercaptoethanol, 1 mM MgCl₂, 1 mM ATP, and 0.1 g/L BSA. Where indicated, 30 nM GcSSB (monomers) was included in the reaction mixture. Reactions containing dup-3′ were quenched by the addition of 11% (v/v) glycerol, 0.28% SDS, and 5 ng of unlabeled o18. For the long dup-3′ and HJ substrates with and without GcSSB, reactions were quenched by the addition of 9.6 μg proteinase K (Sigma-Aldrich), 0.25% SDS, 30 mM EDTA, and 0.75 ng unlabeled o30-2 or oHolliday1 and incubated at 37 °C for 30 minutes prior to separation by PAGE. Products for all assays were resolved through a 10-12% native polyacrylamide gel, dried onto Whatman paper, imaged using a Storm 820 Phosphorimager (Amersham Biosciences), and quantified using ImageQuant 5.1 software.