A dual role for mycobacterial RecO in RecA-dependent homologous recombination and RecA-independent single-strand annealing.

Mycobacteria have two genetically distinct pathways for the homology-directed repair of DNA double-strand breaks: homologous recombination (HR) and single-strand annealing (SSA). HR is abolished by deletion of RecA and reduced in the absence of the AdnAB helicase/nuclease. By contrast, SSA is RecA-independent and requires RecBCD. Here we examine the function of RecO in mycobacterial DNA recombination and repair. Loss of RecO elicits hypersensitivity to DNA damaging agents similar to that caused by deletion of RecA. We show that RecO participates in RecA-dependent HR in a pathway parallel to the AdnAB pathway. We also find that RecO plays a role in the RecA-independent SSA pathway. The mycobacterial RecO protein displays a zinc-dependent DNA binding activity in vitro and accelerates the annealing of SSB-coated single-stranded DNA. These findings establish a role for RecO in two pathways of mycobacterial DNA double-strand break repair and suggest an in vivo function for the DNA annealing activity of RecO proteins, thereby underscoring their similarity to eukaryal Rad52.

INTRODUCTION

The maintenance of genome integrity is critical to all living organisms, and the systems that repair DNA damage are diverse. Double-strand breaks (DSBs) are a particularly lethal form of DNA damage that prevent chromosome replication if not repaired. In most bacteria, homologous recombination (HR) is the dominant pathway of DSB repair and is mediated by the RecA protein, which executes homology search and strand invasion of the broken ends into an intact homologous template, resulting in faithful repair. Mycobacteria elaborate two additional DSB repair pathways: non-homologous end joining (NHEJ) and single-strand annealing (SSA) (1–4). The core components of the NHEJ pathway are Ku and DNA ligase D, which cooperate to repair DNA damage in late stationary phase caused by ionizing radiation or desiccation (5,6), to seal linearized plasmids (1,7), and to rectify I-SceI-induced chromosomal DSBs (3,6). Mycobacteria lacking RecBCD cannot execute repair via the SSA pathway, whereas cells lacking the AdnAB helicase/nuclease are partially defective for HR (3). The proteins that mediate the AdnAB-independent HR pathway, which accounts for ∼50% of HR events in wild type cells, are undefined.

The bacterial RecFOR machinery supports single-strand gap repair and restart of stalled replication forks (8–10). The contribution of RecFOR to the HR pathway of DSB repair in wild-type bacteria varies among taxa. In Escherichia coli, where RecBCD drives the dominant pathway of recombination by performing the end-resection and RecA loading functions required to initiate HR, RecFOR is relegated to a backup role evident when RecBCD is inactive and the bacterium acquires sbc suppressor mutations (11). By contrast, two parallel RecA-dependent HR pathways of recombination operate in Bacillus subtilis, one driven by the AddAB helicase/nuclease (which is homologous to mycobacterial AdnAB) and another by RecFOR. The AddAB and RecFOR pathways contribute equally to clastogen resistance in Bacillus, such that both pathways must be inactivated to ablate HR in vivo (12). In Deinococcus radiodurans, which lacks both RecBCD and AddAB, the RecFOR system constitutes the main pathway of recombinational repair of all DNA breaks, whether single-strand gaps or DSBs (13–15).

The RecO protein of E. coli is an essential component of the RecOR and RecFOR recombination mediator complexes that load RecA onto SSB-coated ssDNA (16–19). RecO can also anneal complementary ssDNA protected by its cognate SSB (20–22). Both the RecA loading and single-strand annealing activities of RecO depend on its interaction with the SSB C-terminal peptide (SSB-Ct) (18,22). These RecO functions resemble the activities of yeast Rad52, which mediates loading of Rad51 (the eukaryal RecA homolog) and can anneal complementary ssDNA coated with RPA (the eukaryal SSB) (23,24). In yeast, Rad52 is required for the Rad51-independent single-strand annealing pathway (25), suggesting an in vivo role for the annealing activity. However, the in vivo function of the single-strand annealing activity of RecO is less clear, as this activity should not be required for RecO to serve as a RecA mediator. Although RecET stimulated illegitimate recombination and plasmid recombination have been demonstrated to be RecO dependent in E. coli and appear to involve an SSA type mechanism (26,27), a bonafide SSA pathway that repairs double strand breaks flanked by repeats has not been demonstrated. One suggestion is that E. coli RecO might (by analogy to Rad52) promote second-end capture during HR, i.e., the annealing of the processed second DSB end to the single-strand of the D-loop created by RecA-dependent strand invasion (28).

In this study, we examine the function of RecO in mycobacteria. Our results indicate that RecO has dual functions in mycobacterial DSB repair: in a pathway of mycobacterial HR parallel to that mediated by AdnAB, and also in the mycobacterial SSA pathway. Consistent with these dual in vivo functions, the biochemical properties of mycobacterial RecO differ from those of the E. coli homolog. Mycobacterial RecO displays strong zinc-dependent DNA binding and anneals complementary SSB-coated DNA without interacting physically with the SSB-Ct.

MATERIALS AND METHODS

recO knockout and complementation

In-frame deletion of Mycobacterium smegmatis recO in the ΔadnAB ΔrecBCD strain was achieved by a two-step allelic exchange process that uses a suicide vector containing a hygromycin-resistance marker and the counterselectable marker sacB, as described previously (1). recO was deleted in WT, ΔadnAB and ΔrecBCD backgrounds by specialized transduction using a temperature-sensitive mycobacteriophage. The hygR marker (flanked by loxP sites) at the disrupted locus was subsequently excised by expressing Cre recombinase, to generate the unmarked strains ΔrecO, ΔrecO ΔadnAB and ΔrecO ΔrecBCD. Southern blotting was performed to confirm recO knockout using either 5′ or 3′ flanking DNA sequence as the probe, and recBCD and adnAB loci were verified using 5′ and 3′ flanking DNA sequences as the probes, respectively. For complementation, the M. smegmatis recO gene along with its 5′ untranslated region (202 bp) containing the presumed promoter was cloned in the mycobacterial integrative vector pMV306kan, to generate pRGM30. The ΔrecO and ΔrecO ΔadnAB strains were complemented with recO by integrating pRGM30 at the attB site of the chromosome by plasmid transformation. Rescue of the ΔrecO mutant phenotypes was tested by assaying clastogen sensitivity. WT, ΔrecO and ΔrecO ΔadnAB strains harbouring pMV306kan in the chromosomal attB locus served as controls in these experiments.

Chromosomal DSB repair assay

Competent M. smegmatis cells of the specified genotypes harbouring the chromosomally integrated lacZ reporter construct for I-SceI-mediated DSB repair assay (3) were transformed separately with the same molar amount of the I-SceI plasmid and the control vector plasmid to determine the frequency of the DSB repair outcomes and net % survival as described previously (3). For each strain, the experiment was performed at least thrice using different batches of competent cells, and results are expressed as mean values. Among the blue colonies, the gene conversion (GC) and SSA events were distinguished by scoring for kanamycin resistance; kanR (kanamycin-resistant) colonies represent GC and kanS (kanamycin-sensitive) colonies denote SSA. Because the % survival values remained constant between different genetic backgrounds, frequencies of different repair outcomes were compared by calculating the relative repair fractions, as follows: relative GC frequency = (GC events/blue events) × (fraction blue); relative SSA frequency = (SSA events/blue events) × (fraction blue).

Growth studies

M. smegmatis strains were revived from frozen glycerol stocks by repeated sub-culturing in LB medium (supplemented with 0.5% glycerol, 0.5% dextrose and 0.1% Tween 80) to ensure that all cells had entered log phase. To study growth kinetics, the strains were re-inoculated into fresh medium to A600 of 0.1, incubated at 37°C with constant shaking (150 rpm) and aliquots were removed at regular intervals to measure A600 and determine colony forming units (CFUs) by serial-dilution plating. Change in absorbance or CFUs was plotted against time. Doubling times were calculated using the formula: doubling time = time (t − t0)/number of generations (G), where G = (log[#bacteria or A600 at time(t)] − log[#bacteria or A600 at time(t0)])/0.301.

Clastogen sensitivity assays

Mycobacterium smegmatis cultures were grown at 37°C in LB medium supplemented with 0.5% glycerol, 0.5% dextrose and 0.1% Tween 80 to log-phase (A600 of 0.4–0.5) and subjected to treatment with different DNA-damaging agents, i.e. UV, Methyl methanesulfonate (MMS) and IR as described previously (3). Ten-fold serial dilutions of the treated and untreated cells (derived from the same culture) were then spotted on LB agar plates or LB agar plates containing 20 µg/ml kanamycin (in case of recO complementation experiments), and % survival was calculated compared with untreated control cells.

Purification of RecO and SSB

His-tagged M. smegmatis RecO (MsRecO) protein was cloned in pMCSG7, and expressed in E. coli, and purified using methods described previously for the Deinococcus radiodurans and E. coli RecO proteins (22). E. coli SSB (EcSSB) and a C-terminal truncation SSBΔC EcSSBΔC were purified as described (29). M. smegmatis SSB (MsSSB) was expressed from plasmid Pet21b-smegSSB (30), which was modified to introduce a translational stop codon before the polyhistidine tag such that the expressed SSB protein has no affinity tag. M. smegmatis SSB was produced in E. coli and purified by sequential polyethylenimine and ammonium sulfate precipitation steps, followed by heparin affinity and size exclusion chromatography steps (29). The EcSSBΔC expression plasmid was a gift from Dr M. Cox. The purity of all proteins was verified by SDS-PAGE and is shown in Supplementary Figure S2.

DNA and SSB-Ct binding assays

Assay components were prepared as previously described (22). Proteins were dialysed or diluted into buffer A (50 mM NaCl, 25% (v/v) glycerol, 50 mM HEPES pH 7.5, 1 mM Tris(2-carboxyethyl)phosphine (TCEP)), buffer B (100 mM NaCl, 25% glycerol, 20 mM Bis-Tris-Propane, pH 7.5, 1 mM TCEP), buffer C (50 mM NaCl, 25% glycerol, 20 mM Bis-Tris-Propane, pH 7.5, 1 mM TCEP) or buffer D (200 mM NaCl, 25% glycerol, 20 mM Bis-Tris-Propane, pH 7.5, 1 mM TCEP). Protein concentrations were determined by absorbance at 280 nm and with extinction coefficients of 24 595 M−1 cm−1 (EcRecO) or 30 668 M−1 cm−1 (MsRecO). Lyophilized fluorescein 5′-labelled (FAM) dT15 and dT35 oligonucleotides were solubilized with Milli-Q H2O. Double-stranded DNA was generated by mixing FAM-labelled 15mer (5′-TATCCGCAGAGTTGG) with an equimolar amount of complementary unlabelled 15mer (5′-CCAACTCTGCGGATA) and incubating at 95°C for 1 min, followed by annealing at room temperature for 30 min. FAM-labelled EcSSB-Ct peptide [WMDFDDDIPF (Genscript)] was solubilized with dimethylformamide and diluted into assay buffer; peptide concentration was measured by tryptophan absorbance at 280 nm with an extinction coefficient of 5500 M−1 cm−1. The concentration of FAM-labelled MsSSB-Ct peptide [GADDEPPF (Lifetein Inc.)] was measured by FAM absorbance at 485 nm using EcSSB-Ct FAM absorption as a reference. Fluorescence anisotropy assays were performed by serially diluting RecO in the indicated buffer with or without Zn(OAc)2. FAM-labelled DNA (5 nM) or SSB-Ct peptide (20 nM) was added to RecO in a final volume of 100 µl and incubated for 30 min. Binding was measured with excitation/emission of 485/528 nm at room temperature using a BioTek Synergy 2 plate reader. Anisotropy values were normalized using equation: A = (Ai − A0)/A0, where A0 and Ai are anisotropy values of free and protein-bound FAM-labelled DNA or SSB-Ct, respectively. Binding constants were calculated using BioKin Dynafit using simple binding equations [RecO] + [DNA or SSB-Ct] <=> [RecO.DNA or SSB-Ct].

DNA annealing assay

RecO-mediated DNA annealing was assayed as described previously (22). SSB concentrations were measured by absorbance at 280 nm and with extinction coefficients of 18 975 M−1 cm−1 (EcSSB), 18 010 M−1 cm−1 (EcSSBΔC) or 17 401 M−1 cm−1 (MsSSB). 49mer 5′-FAM-labelled and complementary 3′-Dabcyl-labelled oligonucleotides at 20 nM were incubated separately with 200 nM of SSB for 30 min at room temperature in buffer D. Complementary oligo•SSB complexes were mixed, and RecO was added and incubated for 30 min at room temperature in the presence or absence of 50 µM Zn(OAc)2. Annealing was measured by a decrease of FAM emission (excitation/emission of 485/528 nm) as a function of time for a total of 21 min. Final annealing amplitude was normalized using equation: I = −1 × (Ii − I0)/I0, where I0 and Ii are intensity values of ssDNA and annealed dsDNA.

RESULTS

recO is not required for M. smegmatis viability

Early bioinformatic surveys of the predicted mycobacterial proteome failed to identify RecO (31,32), but later analyses did (33). An alignment of the primary structure of the putative M. smegmatis RecO protein (a 280-aa polypeptide encoded by MSMEG_4491) with the RecO proteins from D. radiodurans, Bacillus subtilis, Helicobacter pylori, Neisseria gonorrhea and E. coli is shown in Supplementary Figure S1. M. smegmatis RecO is 24% identical/38% similar to E. coli RecO, 26% identical/46% similar to B. subtilis RecO, and 31% identical/47% similar to D. radiodurans RecO. Mycobacterial RecO contains a tetracysteine motif (see green boxes in Supplementary Figure S1) that is also found in the RecO proteins from D. radiodurans and B. subtilis, but not in E. coli RecO. M. smegmatis RecO is highly conserved among mycobacteria—with homologs in M. marinum (83% identical), M. tuberculosis (81% identical), M. intracellulare (82% identical), M. leprae (82% identical), M. gilvum (84% identical), M. avium (80% identical), M. abscessus (83% identical) and M. bovis (79% identical). RecO is also conserved in other Actinomycetales such as Rhodococcus equi (78% identical) and Nocardia abscessus (75% identical). Despite the limited amino acid sequence conservation between the E. coli and D. radiodurans RecO proteins (30% identity), the secondary structural elements of the two proteins, indicated above and below the alignment in Supplementary Figure S1, are well conserved.

To interrogate the role of RecO in mycobacterial DSB repair, we deleted the recO gene (MSMEG_4491) from the M. smegmatis chromosome by allelic exchange. To interrogate potential epistatic relationships between recO and the adnAB-dependent HR and recBCD-dependent SSA pathways, the ΔrecO allele was also introduced into the ΔadnAB, ΔrecBCD, and ΔadnAB ΔrecBCD strains. The strategy for genotyping by Southern blotting depicted in Figure 1A demonstrates the deletion of recO in the ΔadnAB ΔrecBCD strain background, leaving an unmarked ΔrecO allele. Specialized transduction of a ΔrecO::loxP-hyg-loxP allele was used to delete recO in the three other strain backgrounds (Figure 1B). The hygromycin-resistance cassette was then excised using the Cre recombinase to yield unmarked ΔrecO, ΔrecO ΔadnAB, and ΔrecO ΔrecBCD mutants (Figure 1B). The presence of the ΔrecBCD and ΔadnAB alleles was confirmed by diagnostic Southern blotting (data not shown).

Figure 1.

Deletion of recO in Mycobacterium smegmatis (A) Generation of the ΔrecBCD ΔadnAB ΔrecO strain by two-step allelic exchange. Map of the wild-type and ΔrecO loci with NcoI restriction sites, probe location and predicted fragment sizes for WT (11683 bp) and ΔrecO (2065 bp). Below the restriction maps is a Southern hybridization of chromosomal DNA isolated from ΔrecBCD ΔadnAB M. smegmatis (labelled WT with regard to the recO locus), seven sucrose-resistant, hygromycin-sensitive recombinants, and the parental recO duplication strain (recO dup) which contains the wild-type allele of recO and the ΔrecO deletion allele at the 3′-end of the recO locus, separated by the hyg and sacB markers. The lane marked with the asterisk denotes the strain used for further studies. (B) Deletion of recO by specialized transduction. Map of the wild-type and ΔrecO::loxP-hyg-loxP loci with SmaI restriction sites, probe location, and predicted fragment sizes for WT (2245 bp) and ΔrecO::loxP-hyg-loxP (7012 bp) after SmaI restriction of chromosomal DNA. Below the restriction maps is a Southern hybridization of chromosomal DNA from wild-type M. smegmatis (WT), ΔrecO::hyg (lanes 2, 5, 8 in the indicated strain backgrounds) or recO::loxP (after Cre recombinase-mediated excision of the hyg marker, lanes 3, 4, 6, 7).

ΔrecO strains have diminished growth rate

We examined the growth characteristics of ΔrecO strains by measuring A600 in logarithmic cultures and comparing these cell density measurements to the number of viable bacteria. The apparent doubling times of the ΔrecO mutants were determined after repeated serial dilution of logarithmic phase cultures. In comparison to wild type (doubling time 174 min), the ΔrecO (212 min) and ΔrecO ΔadnAB (231 min) strains grew slower and their growth rates were similar to that observed for the ΔrecA strain (226 min). This difference in apparent doubling time was not the result of altered cell shape, as the doubling times of ΔrecO (215 min), ΔrecO ΔadnAB (222 min) and ΔrecA (213 min) calculated from determination of viable bacteria were greater than that of the wild type strain (155 min). The slow growth phenotype of the mutant strains is also manifest as small colony size when cultured on agar media (Figure 2B).

Figure 2.

ΔrecO phenocopies ΔrecA in clastogen resistance. (A) Survival curves of WT M. smegmatis (open circle), ΔadnAB (open square), ΔrecA (filled circle), ΔrecO (filled triangle), ΔrecO ΔadnAB (inverted open triangle) exposed to escalating doses of UV light. Survival is plotted on a log scale and is calculated compared with an unexposed control for each strain. (B) Ionizing radiation survival curve using the same strain key as in panel A. The plate pictures below the graph are a representative unexposed (top panel) and 224 Gy exposed (bottom panel) 10-fold dilutions of the indicated strains. (C) MMS-induced killing. The x-axis represents the time of MMS exposure and survival is plotted on a log scale on the y-axis.

RecO is required for survival upon DNA damage

Our previous results indicated that the M. smegmatis ΔadnAB strain is sensitized to killing by diverse clastogens, whereas ΔrecBCD cells are not (3). However, the ΔadnAB strain is less sensitive to DNA damage than a ΔrecA strain, indicating that there are AdnAB-independent pathways of homology-directed repair. To investigate if RecO is involved in DNA repair in mycobacteria, we studied the survival of the ΔrecO strains after exposure to UV, IR and the alkylating agent MMS. The deletion of recO conferred severe sensitivity to UV radiation (Figure 2A), ionizing radiation (Figure 2B) and MMS (Figure 2C). The clastogen sensitivity of the ΔrecO strain was indistinguishable from ΔrecA (Figures 2A–C). Note that the ΔrecO ΔadnAB, ΔrecO ΔrecBCD and ΔrecO ΔadnAB ΔrecBCD strains were no more sensitive to UV, IR or MMS than the ΔrecO strain (Figure 2, Supplementary Figure S2). These results suggest that RecO and RecA function in the same pathway of DNA repair, conceivably by loading RecA onto single-stranded DNA at damage sites.

In the M. smegmatis chromosome, the recO gene appears to be in an operon with three other genes, namely MSMEG_4490 (uppS, encoding undecaprenyl diphosphate synthase), MSMEG_4489 (encoding a conserved hypothetical protein) and MSMEG_4488 (encoding a nudix family hydrolase) (Figure 1). The predicted translational start codon of uppS overlaps the 3′-end of recO by 53 nucleotides, raising the possibility that the recO deletion in our strains might disrupt uppS expression. Although we were careful to leave the coding sequence of uppS intact, we performed a complementation experiment to confirm that the clastogen sensitivity of our ΔrecO strain was due to loss of RecO function. Complementation of the deletion mutant with a single recO copy at the phage L5 integration site in the chromosome restored the clastogen resistance of the ΔrecO strain to wild-type levels (Figure 3A and B). Complementation of the ΔrecO ΔadnAB strain with recO restored the clastogen resistance to the level of the ΔadnAB strain (Figure 3). Together, these results verify that the phenotype of the recO deletion is due to loss of recO and not due to a polar effect in the recO operon or a spontaneous mutation elsewhere in the chromosome.

Figure 3.

The clastogen susceptibility of the ΔrecO strain is due to loss of recO. (A) Survival curves of WT M. smegmatis (open circle), ΔadnAB (open square), ΔrecO (filled triangle), ΔrecO ΔadnAB (inverted open triangle), ΔrecO ΔadnAB complemented with a single copy of recO (inverted triangle), or ΔrecO complemented with a single copy of recO (open triangle) exposed to escalating doses of UV light. Survival is plotted on a log scale and is calculated compared to an unexposed control for each strain. The plate pictures below the graph show unexposed and 20 mJ/cm2 exposed 10-fold serial dilutions of the indicated strains. (B) Survival curves of the same strains as in (A) exposed to escalating doses of ionizing radiation. Survival is plotted on a log scale and is calculated compared to an unexposed control for each strain. The plate pictures below the graph show unexposed and 224 Gy-exposed 10-fold serial dilutions of the indicated strains.

RecO participates in a parallel pathway to AdnAB in RecA-dependent HR

Because UV, IR and MMS produce diverse types of DNA damage, the clastogen sensitivity of the recO mutant does not suffice for conclusions about the function of recO in specific pathways of DNA repair. To interrogate the participation of RecO in DSB repair, we used an I-SceI recombination system developed previously (3). In this system, a recombination substrate is integrated into the chromosome of M. smegmatis at the L5 phage attachment site. This recombination substrate contains two defective lacZ alleles, lacZ-I-SceI and lacZ-ΔN, which are separated by a kanamycin-resistance cassette. lacZ-I-SceI is interrupted by two recognition sites for the homing endonuclease I-SceI such that cleavage by I-SceI produces a chromosomal break with incompatible 3′ overhangs. Repair of this break can be achieved by three mechanisms: NHEJ, gene conversion (GC) or SSA (Figure 4A). NHEJ (which requires ku and ligD) ligates the broken ends after varying degrees of end-resection and end-healing, and therefore does not reconstitute an active lacZ gene. Thus, repair via NHEJ yields a white colony that is kanamycin-resistant in most cases, unless an especially long end-resection invades the kanamycin-resistance gene. GC (which is recA-dependent) uses the downstream lacZ-ΔN as a homologous template to repair the DSB. This mechanism of repair reconstitutes a functional lacZ gene while leaving the kanamycin-resistance cassette intact, yielding a blue, kanamycin-resistant colony. Finally, repair by SSA (which is recBCD-dependent) results from bidirectional single-strand resection from the I-SceI sites, revealing complementary single strands that can anneal and reconstitute an intact lacZ allele with a deletion that eliminates the kanamycin-resistance marker. Thus, all three repair outcomes can be scored on agar plates containing X-gal and counterscreening for kanamycin resistance, as depicted in Figure 4A. Repair by SSA and GC is then confirmed by diagnostic PCR amplification (3).

Figure 4.

RecO is required for adnAB-independent GC and SSA. (A) Schematic of DSB repair outcomes in the I-SceI system, adapted from (3). An I-SceI-induced DSB is initiated by transformation of a plasmid encoding I-SceI into a strain bearing a previously described I-SceI recombination substrate. Repair of this incompatible DSB yields either blue or white colonies. White colonies can result either from NHEJ-mediated repair, or inactivation of the I-SceI enzyme through mutation. Blue colonies indicate that the defective lacZ coding sequence has been restored either through SSA or GC. DNA resection that occurs during SSA results in deletion of the kanamycin marker, whereas GC does not. (B) Each panel shows M. smegmatis of the indicated genotype transformed with the I-SceI encoding plasmid and cultured on agar plates containing hygromycin and X-gal. (C) Graph of GC frequency according to strain genotype. For each strain, the relative GC frequency is calculated from the genotyping algorithm in (A) and the raw data presented in the table in (E). The graphed ΔadnAB data is from (3). (D) SSA frequency for the strains indicated. The graphed ΔadnAB data is from (3). (E) Survival, % blue and pathway outcome for each strain. #blue is the number of blue colonies genotyped to give the GC and SSA numbers. Relative GC frequency is given as (%blue) (#GC/#blue). SSA frequency is given as (%blue) (#SSA/#blue).

We applied this system to understand the function of RecO in DSB repair. Induction of a DSB by transformation of an I-SceI encoding plasmid in wild-type cells yields an approximate 50:50 mixture of white and blue colonies, as reported previously (3). Of these blue colonies, 80% are GC (40% of the total outcomes) and 20% are SSA (10% of the total) (Figure 4B–E). In the ΔrecO strain, we observed a 5-fold decrement in the yield of blue colonies after I-SceI transformation (Figure 4B and E). Testing kanamycin resistance and genotyping the blue colonies indicated that gene conversion outcomes were 4-fold less frequent than in wild-type cells (relative GC frequency in WT versus ΔrecO, 37% versus 9.5%), compared with the 2-fold decrement observed in the ΔadnAB strain in our prior study (3). We then examined the GC frequency in the ΔadnAB ΔrecO double mutant and observed a near complete abolition of GC outcomes (relative GC frequency in ΔadnAB ΔrecO = 0.077%). These data indicate that adnAB and recO define parallel pathways of recA-dependent gene conversion.

Requirement for RecO in the RecBCD-dependent SSA pathway

We next examined the role of RecO in the mycobacterial SSA pathway. In wild-type cells, SSA outcomes accounted for 26% of the blue colonies and 12.9% of the total outcomes (Figure 4D and E). In the ΔrecO strain, we observed a 15-fold decrement in SSA outcomes (relative SSA frequency in ΔrecO = 0.86%), compared with the complete loss (0/120) of SSA previously observed in the ΔrecBCD strain. As predicted, there were no blue colonies in the ΔrecO ΔrecBCD ΔadnAB strain, consistent with the loss of both GC and SSA pathways (Figure 4E). These assays reveal a genetic requirement for the RecO protein in the mycobacterial SSA pathway.

DNA binding by mycobacterial RecO is stimulated by zinc

The genetic data indicate a role for mycobacterial RecO in RecA-dependent GC and in SSA, dual functions not reported previously for RecO proteins. To correlate these in vivo roles with the biochemical properties of the mycobacterial RecO protein, we produced His-tagged M. smegmatis RecO in E. coli and purified it by nickel affinity chromatography (Supplementary Figure S3). We first examined the interaction of RecO with single-stranded DNA under equilibrium conditions by fluorescent anisotropy. Initial experiments revealed a strong dependence of the MsRecO interaction with ssDNA on the presence of zinc. In buffer A with 50 mM NaCl, MsRecO interacted weakly with ssDNA with an apparent dissociation constant (Kd) of >5 µM (Figure 5A). E. coli RecO (EcRecO) had a noticeably stronger affinity to ssDNA with an apparent Kd of 0.45 (±0.11) µM (Figure 5A). However, addition of 10 µM Zn(OAc)2 significantly stimulated MsRecO binding to ssDNA with a Kd of MsRecO was 0.036 (±0.004) µM. In contrast, zinc had no or a slightly inhibitory effect on binding of EcRecO to ssDNA (Figure 5A). Previously, we demonstrated a strong inhibitory effect of increased salt concentration on DNA binding by EcRecO (22). Consistent with these prior results, EcRecO did not appreciably bind ssDNA at 100 mM NaCl, whereas MsRecO retained strong interaction with ssDNA in the presence of 100 mM NaCl plus Zn(OAc)2 with Kd of 1 (±0.03) µM (Figure 5B). Magnesium did not stimulate DNA binding (data not shown). We also found that MsRecO bound dsDNA with equivalent affinity to ssDNA of the same length (Supplementary Figure S4). These results indicate that mycobacterial RecO binds ssDNA more avidly than E. coli RecO at moderate ionic strength and that this DNA binding is zinc-dependent.

Figure 5.

MsRecO binding to ssDNA is stimulated by zinc. (A) Equilibrium binding isotherms as measured by fluorescence anisotropy of FAM-dT15 (5 nM) upon titration by MsRecO in buffer A (25% glycerol, 50 mM NaCl, 50 mM HEPES pH 7.5, 1 mM TCEP) without Zn(OAc)2 (black dashed lines) and in the presence of 10 µM Zn(OAc)2 (black solid lines). Titration by EcRecO in buffer A without Zn(OAc)2 (gray dashed lines) and in the presence of 10 µM Zn(OAc)2 (gray solid lines). (B) Anisotropy of FAM-dT15 (5 nM) upon titration by MsRecO (black) and EcRecO (grey) in buffer B (Buffer A with 100 mM NaCl) in the presence of 50 µM Zn(OAc)2. (C) Anisotropy of FAM-dT15 (5 nM) in the presence of MsRecO (500 nM) upon titration by Zn(OAc)2 in buffer A.

The zinc requirement for DNA binding can be attributed to the presence of the 4xCys zinc-binding motif in the MsRecO primary structure (Supplementary Figure S1), similar to that of D. radiodurans RecO (DrRecO). DrRecO binds Zn2+ with sufficient affinity to carry the bound ion through multiple purification and crystallization steps involving the use of zinc-free solutions (34). The requirement for exogenous Zn2+ for MsRecO to bind DNA possibly indicates weaker interaction with the metal ion. We used an ssDNA-binding assay to measure Zn2+ binding by MsRecO. The Kd of MsRecO for Zn2+ was estimated to be 0.15 (±0.001) µM (Figure 5C). Therefore, interaction of MsRecO with DNA is modulated by the presence of micromolar concentrations of zinc ion, which might serve as a regulator of RecO DNA binding in vivo.

Mycobacterial RecO does not interact with C-terminal tail of SSB

SSB exhibits one of the highest affinities for ssDNA of any protein and is critical for protecting it from nucleases and for mediating interactions with other DNA-binding proteins (35). The majority of such interactions are mediated by a conserved C-terminal tail of SSB (SSB-Ct), which is essential for the viability of E. coli (36). Interaction of EcRecO with the conserved hydrophobic residues of SSB-Ct is important for binding of RecO to SSB-coated ssDNA and DNA annealing at moderate salt concentrations (22) as well as for loading of RecA by RecOR (18). However, this SSB-Ct binding is not conserved among all RecO proteins. For example, the SSB-Ct binding site is not conserved in DrRecO, which also contains a 4xCys zinc-binding motif, similar to MsRecO. We therefore sought to determine whether MsRecO interacts with either E. coli SSB-Ct or mycobacterial SSB-Ct and to interrogate the requirement for the RecO•SSB-Ct interaction for RecO function. An alignment of the SSB C-termini from E. coli and M. smegmatis is shown in Figure 6A. In comparison to the E. coli peptide, the M. smegmatis peptide is two amino acids shorter and differs by substitution of a proline in place of isoleucine in third position from the C-terminus and glutamate substituting aspartate in the adjacent position.

Figure 6.

MsRecO does not bind SSB-Ct. (A) Sequence of the E. coli SSB and the M. smegmatis SSB C-terminal peptide FAM conjugates. (B) Anisotropy of FAM-MsSSB-Ct (20 nM) and FAM-EcSSB-Ct (20 nM) upon titration by EcRecO (gray and black circles, respectively) or by MsRecO (cyan and blue circles, respectively) in buffer A. Anisotropy of FAM-EcSSB-Ct (20 nM) titrated by MsRecO in the absence (magenta squares) or presence of 10 µM Zn(OAc)2 (red squares) and of MsSSB-Ct in the presence of 10 µM Zn(OAc)2 (green squares) in buffer C (200 mM NaCl). Labels correspond to RecO/SSB-Ct.

We measured the strength of the RecO/SSB-Ct interaction by anisotropy of a fluorescein-labelled peptide composed of the C-terminal amino acids of E. coli SSB (FAM-WMDFDDDIPF) or M. smegmatis SSB (FAM-GADDEPPF). MsRecO did not bind SSB-Ct either from E. coli or M. smegmatis (Kd >10 µM, Figure 6B) regardless of whether zinc was present. EcRecO interacted with both peptides, albeit with lower affinity for MsSSB-Ct (Kd of 3.0 µM) compared with EcSSB-Ct (Kd of 0.12 µM) (Figure 6B). The lack of interaction between MsRecO and MsSSB-Ct raises questions about whether and how mycobacterial RecO can bind ssDNA when SSB is present.

Mycobacterial RecO can anneal SSBΔC-coated ssDNA

The ΔrecO M. smegmatis strain is defective for DSB repair via the SSA pathway (Figure 4), which suggests that MsRecO may aid in annealing SSB-coated DNA at resected DSB ends. In E. coli, annealing of ssDNA by RecO requires the hydrophobic interaction between RecO and the SSB-Ct at moderate salt concentration (22). We investigated the ability of EcRecO and MsRecO to anneal SSB-coated complementary single-strand DNAs using a fluorescence quenching annealing MsSSB-coated ssDNA (Figure 7). This annealing activity was abolished when truncated EcSSBΔC was included in the reaction in lieu of full length SSB (Figure 7A and B), indicating that the annealing activity of EcRecO depends on interaction with the SSB-Ct. MsRecO also promoted annealing of ssDNA coated with either EcSSB or MsSSB when Zn2+ was present in the annealing reaction. However, in contrast to EcRecO, the annealing activity of MsRecO was unaffected by deletion of the SSB-Ct (Figure 7). Therefore, mycobacterial RecO is able to anneal complementary SSB-coated DNA strands independent of an interaction with the SSB-Ct.

Figure 7.

MsRecO anneals SSB-coated DNA at moderate salt without interacting with the SSB C-terminus. (A) Relative degree of FAM quenching corresponding to annealing of FAM-labelled 49mer ssDNA and Dabcyl-labelled 49mer complementary ssDNA (20 nM) after 21 min. Each strand was incubated separately with either full length or truncated E. coli SSB or full length M. smegmatis SSB (200 nM). After 30 min, strands were mixed and RecO was added (2µM). All reactions were performed in buffer D. Proteins and presence of 50 µM Zn(OAc)2 in each reaction are shown below bar graph. (B) Time course of annealing using the indicated combinations of E. coli or M. smegmatis RecO/SSB, with or without zinc, as indicated to the right of each annealing curve.

DISCUSSION

We have examined the genetic and biochemical functions of mycobacterial RecO and found dual in vivo functions in the HR and SSA pathways. The critical role of RecO in the RecA-dependent recombination pathway is clearly distinct from the function of RecO in E. coli, but similar to B. subtilis, in which the RecFOR system has an important role in recombination in wild-type cells. Our studies also clearly indicate a RecO-independent, AdnAB-dependent HR pathway. The genetic requirement for RecO in the mycobacterial HR pathway may reflect a RecA loading function of RecO in association with RecF and/or RecR. This model is based on the function of the RecFOR complex in E. coli as a RecA mediator (16). Alternatively, the RecO protein may be required not for RecA loading, but rather for second-end capture via its annealing activity, as documented for yeast Rad52 (28).

Our studies indicate that mycobacterial RecO differs from E. coli RecO in its interaction with SSB. Mycobacterial RecO does not interact directly with the SSB C-terminal tail. It has been reported that the mycobacterial SSB C-terminal tail interacts directly with RecA both in vitro and in vivo (37), an interaction not described for other bacterial RecA proteins (35,38). This direct interaction between SSB-Ct and RecA may suggest that RecO is not needed for recruitment of RecA to SSB-coated ssDNA. We predict that if RecO functions as a mediator for mycobacterial RecA, then the role of RecO would not be to recruit RecA to SSB-coated DNA, but rather to facilitate RecA/SSB exchange. Alternatively, the direct recruitment of RecA by the SSB-Ct may reflect a fundamentally different mechanism of RecA loading in mycobacteria that may not require RecO.

Multiple other components of the two parallel pathways of recombination in mycobacteria remain to be identified. For the RecO pathway, the nuclease that resects the DSB before RecO action is unknown. In the RecFOR systems of other bacteria, RecJ performs this function (16,39). However, no RecJ exonuclease has been identified in mycobacteria, leaving the identity of the resection nuclease in the mycobacterial RecO pathway obscure.

Our prior results indicated that mycobacteria have an SSA pathway that depends on RecBCD (3). In this work we implicate the RecO protein in the SSA pathway. RecO proteins are known to exhibit single-strand annealing activity. However, the in vivo function of this annealing activity has not been demonstrated. It has been suggested that this annealing activity may catalyse the annealing of the second resected end onto the displaced D-loop created by strand invasion during RecA-dependent HR (28). However, this model was based on bacteria that do not apparently encode an SSA pathway of DSB repair. Our data suggest an additional in vivo function for the DNA annealing activity of RecO in the annealing of single-stranded DNA during repair by the SSA pathway. This function of RecO in mycobacteria echoes that of yeast Rad52, which also anneals single-stranded DNA and is genetically required for both the gene conversion and SSA pathways (25).

Our biochemical characterization of mycobacterial RecO indicates that the protein is adept at DNA annealing. Mycobacterial RecO binds single-stranded DNA more avidly than E. coli RecO. This high-affinity DNA binding might allow mycobacterial RecO to anneal SSB-coated single-stranded DNA without interacting with the SSB C-terminal tail, an interaction that is required for the annealing activity of E. coli RecO. MsRecO DNA binding and annealing activities are dependent on added Zn2+, a requirement that is likely due to the tetracysteine zinc-binding motif found in some RecO orthologs but not E. coli RecO. It is tempting to speculate that there is a correlation between the presence of a tetracysteine zinc-binding motif in RecO and a lack of binding to the SSB-Ct, because MsRecO and DrRecO share these features. B. subtilis RecO possesses a similar 4xCys motif, can compete for DNA binding with SsbA and SsbB (40), and can anneal ssDNA in the presence of SsbA (41), which contains a conserved SSB-Ct. However, the requirement for the SsbA C-terminus in this process has not been examined (41,42). An alternate possibility is that mycobacterial RecO may interact with a region of SSB distinct from the C-terminus, such that its DNA annealing activity still relies on interaction with SSB but via a distinct mechanism.

In summary, our data document a dual function of mycobacterial RecO in two pathways of homology-directed DNA repair. Further biochemical characterization of the RecO–SSB interaction and investigation of the ability of RecO to load RecA onto ssDNA will elucidate the biochemical function of RecO during HR. In addition, separation of function mutations that ablate RecO annealing activity will be useful to understand the novel requirement for RecO in the SSA pathway documented here.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online: Supplementary Figures 1–4.

FUNDING

National Institutes of Health (NIH) [AI064693 to M.S.G. and GM073837 to S.K.]. American Cancer Society Research Professorship (to S.S.). Funding for open access charge: NIH [AI 64693].

Conflict of interest statement. None declared.