The Minimal Bacillus subtilis Nonhomologous End Joining Repair Machinery

It is widely accepted that repair of double-strand breaks in bacteria that either sporulate or that undergo extended periods of stationary phase relies not only on homologous recombination but also on a minimal nonhomologous end joining (NHEJ) system consisting of a dedicated multifunctional ATP-dependent DNA Ligase D (LigD) and the DNA-end-binding protein Ku. Bacillus subtilis is one of the bacterial members with a NHEJ system that contributes to genome stability during the stationary phase and germination of spores, having been characterized exclusively in vivo. Here, the in vitro analysis of the functional properties of the purified B. subtilis LigD (BsuLigD) and Ku (BsuKu) proteins is presented. The results show that the essential biochemical signatures exhibited by BsuLigD agree with its proposed function in NHEJ: i) inherent polymerization activity showing preferential insertion of NMPs, ii) specific recognition of the phosphate group at the downstream 5′ end, iii) intrinsic ligase activity, iv) ability to promote realignments of the template and primer strands during elongation of mispaired 3′ ends, and v) it is recruited to DNA by BsuKu that stimulates the inherent polymerization and ligase activities of the enzyme allowing it to deal with and to hold different and unstable DNA realignments.


Introduction
Among the diverse types of DNA damage, DNA double-strand breaks (DSBs) are the most hazardous, being lethal to dividing cells if they are not repaired in a timely fashion [1][2][3][4]. Two main pathways deal with these lesions: the error-free homologous recombination (HR), in which a second intact double-stranded copy provides the template for DNA synthesis across the break; and the nonhomologous end joining (NHEJ), active mainly when one chromosomal copy is available, and whereby the DNA ends are directly rejoined. This pathway can be either faithful, if the DNA ends are ligated directly, or error-prone, if the ends are processed by nucleases or polymerases before being sealed by a dedicated DNA ligase [reviewed in [5]].
The presence of the NHEJ was first described in eukaryotes being the main DSB repair pathway operating during the G1 phase of the cell cycle [6,7]. Briefly, DNA ends at DSBs are threaded through the open-ring structure of the Ku70/80 heterodimer (Yku70/Yku80 in yeast), allowing their alignment and protection against nucleases [8][9][10]. Once bound to DNA, in higher eukaryotes Ku recruits the DNA-dependent protein kinase catalytic subunit [4,11] and both carry out the synapsis of the broken ends. In yeast, the recruited Mre11/Rad50/Xrs2 complex (MRX), required for NHEJ, has been also proposed to bind and connect two DNA molecules [1]. Those termini are further processed by nucleases (Artemis, Fen1), phosphatases and kinases generating gapped intermediates filled by a DNA polymerase belonging to family X (l, m and yeast Pol IV) [2,3,[12][13][14]. Finally, the complex Ligase IV/XRCC4/XLF (Dnl4/Lif1/Nej1 in yeast) joins the ends together [1,15].
The identification in bacterial genomes of genes homologous to the eukaryotic Ku led to the discovery of a NHEJ pathway in these organisms [16][17][18][19][20][21][22]. Bacterial Ku homologues, typically present as single genes whose product would function as a homodimer, are usually genetically associated to putative ATP-dependent DNA ligase genes (Ligase D, LigD) [16,17,23,24]. In many bacteria, as in Mycobacterium tuberculosis, LigD contains a phosphoesterase, polymerase and ligase domain that could account for the end processing, gap filling and sealing steps during NHEJ [18,20,25] (see scheme in Figure S1). Extensive biochemical characterization of LigD and Ku proteins, mainly those from M. tuberculosis and M. smegmatis, as well as the impact that their absence has in the repair of DSBs, has led to propose recently how the minimal twocomponent NHEJ operates in bacteria [19,21,22,26]. Thus, at first Ku binds to both sides of the DSB and recruits LigD whose polymerization domain recognizes specifically the 59P termini, mediating the synapsis event required for end-joining. If after microhomology pairing non-extendable 39 termini were present, they would be resected by the phosphodiesterase/phosphomonoesterase activity of LigD and further elongated by its polymerization activity. The nicks that result after filling those gaps would be finally sealed by the ligase activity, completing the break repair [26]. Although the bacterial NHEJ was originally envisaged to rely only on Ku and LigD factors, novel bacterial Ku partners, potentially involved in this repair pathway, have been identified, as mycobacterial UvrD1, an ATPase with a Ku-dependent unwinding activity of 39-tailed DNA duplexes [27]. Additionally, mycobacterial Sir-2 protein has been recently shown to interact directly with both, Ku and LigD [28]. This fact, together with the marked sensitivity to ionizing radiation exhibited by Sir-2 deficient cells, provided strong evidence supporting the involvement of this protein in mycobacterial NHEJ together with Ku and LigD. Finally, genetic evidences implicated also mycobacterial LigC in a minor pathway of Ku-dependent NHEJ [21].
As other soil organisms, the gram positive bacterium Bacillus subtilis can survive to adverse growth conditions, as nutrient deprivation, low temperature, acidity, dryness or high osmolarity by forming an haploid endospore that will germinate and start a new round of vegetative growth once a favorable environment is reestablished [29,30]. The absence of an intact copy of the bacterial chromosome to act as template for the re-synthesis of DSBs precludes the HR pathway to operate during spore germination [25,31]. As aforementioned, the initial evidence of a NHEJ pathway in B. subtilis was the identification of the gene ykoV [32], predicted to encode for a 35-kDa protein with a significant homology to the eukaryotic Ku proteins (BsuKu) [16,19,24]. This gene exists in an operon including the downstream gene ykoU potentially encoding for a putative 70-kDa bimodular LigD (BsuLigD) with a C-terminal DNA ligase catalytic domain linked to an N-terminal polymerase domain that shares similarity to the catalytic subunit of eukaryal DNA primases [16,23,24] (see scheme in Figure S1). These genes form part of a regulon under the control of both, the RNA-polymerase sigma factor s G and the DNA-binding protein SpoVT, and whose expression is turned on in the forespore [33]. The experimental evidences of the participation of BsuLigD and BsuKu in DSB repair came from the analyses of B. subtilis ykoV and ykoU null mutants. Thus, deletion of those genes sensitized B. subtilis cells to ionizing radiation in the stationary phase [19] and their spores to several DNA-damaging treatments known to cause DSBs, as UV-ray, Xray, ultrahigh vacuum and wet heat [33,34].
Here, an initial and qualitative in vitro characterization of the biochemical properties of BsuLigD and BsuKu proteins is presented. The results show that BsuLigD has intrinsic polymerization and ligase activities that act in concert to fill and finally seal short gaps. In addition, BsuKu recruits BsuLigD stimulating its inherent polymerization and ligase activities allowing the enzyme to deal with and to hold different and unstable DNA realignments.

Nucleotides and proteins
Unlabeled nucleotides, [c-32 P]ATP (3000 Ci mmol 21 ), [a-32 P]dATP (3000 Ci mmol 21 ) and [a-32 P]ATP (3000 Ci mmol 21 ) were obtained from Amersham Pharmacia. Restriction endonucleases and T4 polynucleotide kinase were purchased from New England Biolabs. The independent open reading frames containing the genes ykoU and ykoV from B. subtilis [32] were PCR amplified and further digested with NdeI and BamHI before cloning in the NdeI-BamHI digested pET-28a(+) bacterial expression vector (Novagen), which carries an N-terminal His-Tag configuration to express recombinant proteins as fusions with an N-terminal hexahistidyl for purification on Ni +2 -affinity resins. E. coli BL21(DE3) cells were transformed with the resulting recombinant expression plasmids, named pET28-BsuLigD and pET28-BsuKu, and such constructions were further confirmed by DNA sequencing. Expression of the His-tagged BsuKu protein was carried out in the E. coli strain BL21(DE3), which contains the T7 RNA polymerase gene under the control of the isopropyl b-Dthiogalactopyranoside (IPTG)-inducible lacUV5 promoter [35,36]. Cells, previously transformed with plasmid pET28-BsuKu, were grown overnight in LB medium at 37 uC in the presence of kanamycin. Cells were diluted into the same media and incubated at 30 uC until the DO 600 reached 0.6. Then, IPTG (Sigma) was added to a final concentration of 0.5 mM and incubation was continued for 1 h at 30uC. To express BsuLigD protein, chemically competent E. coli SoluBL21 cells (Genlantis) were transformed with the recombinant plasmid pET28-BsuLigD and grown overnight in M9 minimal media at room temperature in the presence of kanamycin. Cells were diluted into the same media until OD 600 = 0.2 and grown at 37uC until OD 600 = 0.4. After induction of protein expression with 1 mM IPTG, cells were incubated overnight at 20uC. In both cases, cells were collected by centrifugation for 10 min at 6,1436 g.
Purification of BsuLigD. Cells were thawed and ground with alumina at 4uC. The slurry was resuspended in Buffer A (50 mM Tris-HCl, pH 7.5, 0.3 M NaCl, 7 mM b-mercaptoethanol, 5% glycerol) and centrifuged for 5 min at 6506 g, at 4 uC to remove alumina and intact cells. The recombinant BsuLigD protein was soluble under these conditions, since it remained in the supernatant after a new centrifugation for 20 min at 234306 g, to separate insoluble proteins from the soluble extract. The soluble extracts were loaded onto a Ni-NTA column (QIAgen) preequilibrated with Buffer A (0.3 M NaCl). The bound proteins were eluted with 80 mM imidazole in Buffer A (0.3 M NaCl). The eluate was further loaded onto a phosphocellulose column equilibrated with buffer A (0.12 M NaCl) and BsuLigD was eluted with Buffer A (0.2 M NaCl).
Purification of BsuKu. Cells were thawed and ground with alumina at 4uC. The slurry was resuspended in Buffer B (50 mM Tris-HCl, pH 7.5, 0.3 M NaCl, 7 mM b-mercaptoethanol, 1 mM EDTA, 5% glycerol) and centrifuged for 5 min at 6506 g, at 4 uC to remove alumina and intact cells. The recombinant BsuKu protein was soluble under these conditions, since it remained in the supernatant after a new centrifugation for 20 min at 234306 g, to separate insoluble proteins from the soluble extract. The DNA present was removed by stirring for 10 min the soluble extract containing 0.3 % polyethyleneimine (PEI) followed by centrifugation for 10 min at 234306 g. The resulting supernatant was precipitated with ammonium sulphate to 65 % saturation to obtain a PEI-free protein pellet. After centrifugation for 25 min at 234306 g, the pellet was resuspended in Buffer B without NaCl to get a final 0.1 M ammonium sulphate concentration. This fraction was loaded onto a phosphocellulose column equilibrated with buffer B (0.1 M NaCl) and the BsuKu protein was eluted with Buffer A (0.5 M NaCl). The eluate was loaded onto a Ni-NTA column (QIAgen) pre-equilibrated with Buffer A (0.5 M NaCl). The bound protein was eluted by 200 mM imidazole in Buffer A (0.3 M NaCl). Finally, both proteins, BsuLigD and BsuKu were dialyzed against a buffer containing 0.2 M NaCl and 50% glycerol and stored at 270uC. Final purity of the proteins was estimated to be .90 % by SDS-PAGE followed by Coomassie blue staining. The purified proteins were further loaded onto a 4 ml glycerol gradient (15-30%) containing 50 mM Tris-HCl, pH 7.5, 20 mM ammonium sulphate, 180 mM NaCl, 1 mM EDTA, and 7 mM bmercaptoethanol, and centrifuged at 4 uC for 24 h at 3481346 g in a Beckman TST 60.4 Swinging rotor. After centrifugation, 23 (BsuLigD) and 24 (BsuKu) fractions were collected from the bottom of the tube for further analysis.

DNA polymerization on activated DNA
The incubation mixture contained, in 25 ml, 50 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol (DTT), 4 % glycerol, 0.1 mg ml 21 of bovine serum albumin (BSA), 100 nM [a-32 P]dATP and 100 nM of the other three dNTPs, 1.5 mg of activated calf thymus DNA (DNA treated briefly with DNase Ifor preparing substrate used in the study of DNA polymerization reactions) as a substrate and 400 ng of BsuLigD in the presence of the indicated concentration of MnCl 2 or MgCl 2 . After incubation for 20 min at 30uC, the reaction was stopped by adding 10 mM EDTA, 0.1 % SDS, and the samples were filtered through Sephadex G-50 spin columns in the presence of 0.1 % SDS. Nucleotide incorporation (nmol) was calculated by counting the Cerenkov radiation of the excluded volume.
DNA polymerization assays on defined DNA molecules DNA-dependent polymerization was assayed on different primer-template structures, obtained by hybridization of the indicated 59-labeled primer (sp1, sp1p and sp1CGG) to the specified template oligonucleotide. To obtain ''gapped'' structures of defined length, a third oligonucleotide was also hybridized to the corresponding template oligonucleotide, as described above. The incubation mixture contained, in 12.5 ml, 50 mM Tris-HCl, pH 7.5, the indicated catalytic metal ion, 1 mM DTT, 4% (v/v) glycerol, 0.1 mg ml 21 BSA, 1.5 nM of the DNA molecule indicated in each case, the indicated concentrations of BsuLigD and BsuKu and the specified nucleotides. After incubation for the different times at 30 uC, the reactions were stopped by adding EDTA up to 10 mM. Samples were analyzed by 8 M urea-20% PAGE and autoradiography.
Processivity assays. The processivity of BsuLigD was analyzed at different enzyme/DNA ratios. The incubation mixture contained, in 12.5 ml, 50 mM Tris-HCl, pH 7.5, 5 mM MnCl 2 , 1 mM DTT, 4% (v/v) glycerol, 0.1 mg ml 21 BSA, 1.5 nM of the DNA substrate indicated, 10 mM of the specified type of nucleotide, and the indicated decreasing amounts of enzyme. When stated, 40 ng of BsuKu was also added to the reaction. After incubation for 20 minutes at 30uC, the reactions were stopped by adding EDTA up to a final concentration of 10 mM. Samples were analyzed by 8 M urea-20% PAGE and autoradiography. Processivity of polymerization was assessed by analysis of the length of replication products under decreasing DNA polymerase/DNA ratios.

Enzyme-AMP complex formation
Reactions were performed as described [37]. The incubation mixture contained, in a final volume of 20 ml, 12 mM Tris-HCl (pH 7.5), 1 mM EDTA, 20 mM ammonium sulphate, 0.1 mg ml 21 BSA, 5 mM [a-32 P]ATP (2.5 mCi), 5 mM MnCl 2 and the indicated amount of BsuLigD. After incubation for 10 min at 30uC the reactions were stopped by adding 10 mM EDTA and 0.1% SDS. The samples were then filtered through Sephadex G-50 spin columns to remove the non-incoporated ATP, and further analyzed by 12% SDS-PAGE. Label transfer to the 70-kDa BsuLigD polypeptide was visualized by autoradiography of the dried gel.

Results and Discussion
BsuLigD is endowed with a polymerization activity The purified BsuLigD was first assayed for a potential DNA polymerase activity on activated calf-thymus DNA, in the presence of increasing concentrations of either Mg 2+ or Mn 2+ ions. As observed in Figure 1A, although BsuLigD catalyzed DNA synthesis under both conditions, it showed a preferential use of Mn 2+ over Mg 2+ , as described for other LigDs [25,31,38,39].
As shown in Figure 1B, BsuLigD catalyzed the template directed addition of both dNTPs and NTPs on a template/primer structure. Although the enzyme displayed a higher efficiency in the use of NTPs respect to dNTPs, a hallmark of bacterial LigDs [18,20,21,25,31,[38][39][40], NMP incorporation was mainly limited to 1-5 nucleotides. Only a very low fraction of the primers were extended beyond the n+5 template position, strongly suggesting that BsuLigD prefers a DNA to an RNA primer-terminus, as described for Pseudomonas aeruginosa LigD (PaeLigD) whose polymerization activity decreased progressively after each serial NMP incorporation [38,41]. The explanation for the preferential utilization of NTPs came from the crystallographic structures of the polymerization domain of both, PaeLigD and MtuLigD bound to Mn 2+ and incoming NTP [25,31,42,43]. In those complexes the 29-OH group of the ribose sugar was hydrogen bonded to the side chain hydroxyl and backbone carboxyl oxygen of MtuLigD residue T236 (PaeLigD residue S768) and to the dN of MtuLigD residue H111 (PaeLigD residue H651). Those residues, conserved among the polymerization domain of bacterial LigDs [26], that would correspond to BsuLigD residues T539 and H422 (see Figure 1C), could account for the preferential NMP additions exhibited by LigDs [43]. This fact has been related with the necessity to repair DSBs when the intracellular pool of dNTPs is depleted, as it occurs during the stationary phase or cellular quiescent states [18,25]. To determine the polymerization processivity limit of BsuLigD, chain length distribution was analyzed as a function of enzyme/DNA ratio. As shown in Figure 1D, the length of the products synthesized by BsuLigD decreased with the enzyme/DNA ratio, in agreement with a distributive polymerization behavior. From these data it can be inferred that the enzyme dissociates from the DNA after each nucleotide incorporation step.
BsuLigD is well suited to act on short gapped DNA substrates. Specific recognition of the phosphate group at the downstream 59 end LigDs are proposed to act on DNA gapped substrates that can arise during the limited base pairing realignment of the DNA ends produced by DSBs [18,25]. To analyze the BsuLigD filling-in ability, DNA molecules harboring gaps of 1-5 nucleotides were used as substrates of the polymerization activity. In addition, the influence of a phosphate group at the 59 end of the downstream strand in the BsuLigD gap-filling efficiency was also evaluated. As shown in Figure 2 (left panels), in the presence of a downstream 59-OH end first ribonucleotide addition was higher than that of dNMPs, regardless of the length of the gap. Notwithstanding BsuLigD accomplished efficient filling of the 2-nt gap with NTPs, it mainly inserted only one NMP when acting on longer gaps. These results would be pointing to gaps of 1-2 nt as the preferred substrates of the enzyme, suggesting a simultaneous binding of LigD to the upstream and downstream regions of the DNA substrate. Therefore, the longer the gap, the lower the insertion efficiency, maybe due to the loss of downstream contacts. In this sense, the binding of LigD to the downstream strand in the 2-nt gap molecule could be aiding the second NMP addition observed to occur with this molecule. The presence of a 59-phosphate at the distal margin of the gap stimulated insertion of dNMPs principally in the 2-nt gap molecules (Figure 2, right panel) pointing to a specific recognition of this group by the enzyme. This result would  [55][56][57], using as template the crystallographic structure of the MtuLigD preternary complex [PDB code 3PKY [42]]. BsuLigD residues K331 and K341, potentially involved in the specific recognition of the downstream 59-P group are shown as lime and green space-filling spheres, respectively. The 29-OH group of the incoming UTP is proposed to make hydrogen bonds (black lines) with BsuLigD residues H422 and T539, represented as space-filling spheres. Figure  represent an extension to the rest of LigDs of previous data obtained with Mycobacterium tuberculosis LigD (MtuLigD) whose templated polymerization reaction was greatly stimulated by a 59-P group [31]. The crystal structure of the polymerase domain of MtuLigD mediating the synapsis of two non-complementary DNA ends showed the enzyme making intimate contact with the upstream primer portion and with the 59-P on the downstream strand through two Lys residues absolutely conserved in NHEJ LigDs [26,42] and whose homologs in BsuLigD would correspond to Lys331 and Lys341 (see Figure 1C).
BsuKu interacts functionally with BsuLigD enhancing its gap-filling efficiency Both, eukaryotic and prokaryotic Ku have been described to bind dsDNA with high affinity [8,19,44]. In the current working model of bacterial NHEJ, Ku recognizes and binds to the broken ends, recruiting LigD that further catalyzes the sealing step of break repair [18,25,26].
To study whether BsuKu exerts a functional influence on BsuLigD activity, the gap-filling efficiency of the latter was examined in the presence or absence of BsuKu (see Figure 3A). As it can be observed, BsuKu slightly impaired the primer usage by BsuLigD when acting on the 1-nt gapped molecule. This result could indicate that both proteins cannot accommodate simultaneously in such a short gap. Conversely, BsuKu enhanced BsuLigD polymerization efficiency on gaps longer than one nucleotide, irrespective of the identity of the 59-group of the downstream DNA, increasing the proportion of products longer than 1 nucleotide on the 2-and 3-nt gaps. Interestingly, in the presence of BsuKu the NMP insertion catalyzed by BsuLigD was not halted after the first nucleotide addition and the enzyme performed gap filling to completion. Similar results were observed with the 4-and 5-nt gap molecules on which BsuKu enhanced gap filling efficiency of BsuLigD. Interestingly, although BsuKu caused a nearly 3-fold improvement of the primer usage by BsuLigD ( Figure 3B), it did not prevent dissociation of the enzyme after each NMP insertion, as the length of the polymerization products decreased with the enzyme/DNA ratio (see Figure 3C). These results led to propose that BsuKu promotes a fast recruitment of BsuLigD to the 39 ends increasing its polymerization efficiency, and suggest that, as proposed for P. aeruginosa, utilization of ribonucleotides for DNA synthesis during NHEJ in B. subtilis could be restricted to short repair tracts [38,41].
As a control to ensure proper assignment of the polymerization activity to BsuLigD and its stimulation by BsuKu, the purified proteins were individually sedimented through a glycerol gradient, and the collected fractions assayed accordingly (see Materials and Methods). Thus, BsuLigD fractions were evaluated for DNA polymerase activity on a primer/template substrate in the presence of BsuKu (see Materials and Methods). As shown in Figure 4A, the single activity peak observed coincided with the protein peak, identified by Coomassie Blue staining after SDS-PAGE analysis of each gradient fraction. These results allowed to ascribe the DNA polymerization activity to BsuLigD, ruling out the presence of a contaminant DNA polymerase from E. coli. Similarly, each BsuKu fraction was analyzed in its ability to stimulate BsuLigD polymerization activity on the same substrate (see Figure 4B). As it is shown, stimulation of the polymerization reaction was observed with those fractions containing the maximal amount of protein, a result that led us to conclude that the improvement of BsuLigD activity is attributable to BsuKu.
Altogether these experimental evidences suggest i) that the complex BsuKu/BsuLigD is well suited and specialized to fill short gaps 2-3 nucleotides long, mainly those bearing a P group at the 59 end of the downstream strand, and ii) the existence of a functional interaction between BsuKu and BsuLigD.
Ligase activity of BsuLigD enables the enzyme to catalyze complete repair of a gapped DNA substrate As it can be observed in Figure 3A, with the 2 nt gap-59/P substrate and in the presence of BsuKu, BsuLigD yielded a faint band (marked with an asterisk) that would correspond to a small proportion of the +2 product ligated to the downstream strand, supporting the prediction of a ligase activity in this enzyme [24]. Accordingly, BsuLigD catalyzed the metal dependent formation of a radiolabeled covalent ligase-adenylate adduct in the presence of [a-32 P]ATP, the initial step in DNA ligation, which comigrated with the BsuLigD in a SDS-PAGE (see Figure S2). The coincidence of the maximal ligase activity, assayed on a nicked DNA molecule, with the mass peak fractions obtained from the glycerol gradient sedimentation of the purified protein shown in Figure 4A, corroborated the proper assignment of the ligase activity to BsuLigD ( Figure S3). The low proportion of the ratio between the ligated and the +2 products aforementioned led to analyze the optimal metal requirement to couple the polymerization and ligation activities. Therefore, BsuLigD/BsuKu complex was incubated with the 5-nt gapped molecule in the presence of increasing concentrations of Mn 2+ as metal activator and a higher amount of BsuLigD to augment the quantity of elongated +5 products. As it can be observed in Figure 5A, BsuLigD filled the gap and sealed the resulting nick to the 59-P terminus of the downstream strand, giving rise to the 33-mer repaired product, the MnCl 2 concentration required to get the maximal ligation/+5 product ratio ranging between 0.16 and 1.2 mM, in good correlation with the in vivo levels of Mn 2+ within the cell (0.3 mM; [45]). At 5 mM MnCl 2 the amount of ligation products was negligible, a result that would account for the low proportion of repaired products observed with the 2-nt gapped molecule previously mentioned. As shown in Figure 5B the presence of a downstream 59-P group and BsuKu was absolutely required to allow BsuLigD to repair totally the 5-nt gap, with either dNTPs or NTPs.

Mismatched 39 termini elongation on gapped molecules
BsuLigD/Ku complex should deal in vivo with potential mispairings resulting during the rejoining of non-compatible DNA ends. Therefore, it was of interest to test the capability of BsuLigD/Ku complex to elongate mismatched 39 termini. As it can be observed in Figure 6A (left panel), in the presence of the next correct ATP, BsuLigD could elongate the dG:dG mispair by adding two AMP residues opposite the two consecutive dTMPs of the template. Interestingly, the enzyme could also incorporate the "incorrect" CMP residue to the mispaired 39 end, a result that could be indicative of an intrinsic ability of BsuLigD to slip the primer-terminus back to catalyze CMP insertion opposite the dGMP of the mispair. Accordingly, BsuLigD could extend the dG:dA mispair by adding either AMP or UMP at the same extent ( Figure 6A, right panel). As it is observed in Figure 6B, the presence of BsuKu did not prevent the primer slippage. Interestingly, previous studies of PaeLigD showed that the polymerization domain could insert two consecutive TMPs opposite a template dA-dT dinucleotide [40]. This result led to conclude that PaeLigD accomplished misincorporation of TMP directed by the templating dT. However, based on the results presented here, an alternative and non-exclusive possibility would be that the templating dA directed the iterative addition of two TMP residues, a fact that would substantiate the ability to slip the primer terminus back as a general signature of bacterial LigDs.
Elongation of the mispaired 39 terminus was also evaluated in a 3-nt gapped molecule, as these substrates would mimic NHEJ intermediates, namely the bridging step of an end joining reaction, as described [18]. As it can be observed in Figure 6C BsuLigD recognized specifically the downstream 59-P group also when acting on these substrates. In addition, irrespectively of the 59 group, the presence of a downstream strand did not prevent the primer slippage described above, as inferred from the CMP insertion. Additionally, on this substrate BsuLigD also inserted a UMP residue very proficiently, a proportion of the UMP-extended primers being further sealed to the downstream 59P group to give rise to a 29 mer ligation product. This result suggests that, as it occurs with other bacterial LigDs [31,40], BsuLigD can use as template of its polymerization reaction the marginal dA by flipping out of the catalytic site the two intermediate dTs of the template. The resulting matched base pair is further ligated to the downstream 59P end causing a two nucleotides deletion mutation (see scheme on the right). It was significant that on this substrate the presence of the four NTPs did not prevent the preferential insertion and further ligation of UMP opposite the marginal dA of the template. However, BsuKu compelled BsuLigD to fill the gap to completion following the 39-59 template instructions, guaranteeing a minimal lost of the original sequence and giving rise to the final 31 mer ligation product (see Figure 6D).
Similar experiments were also carried out in the presence of a 3 nucleotides mismatched 39 terminus (see Figure 6E). As it is shown, when acting on the 5-nt gapped molecule, and in the absence of nucleotides, BsuLigD gave rise to a 31 mer band. The formation of this product would indicate a direct pairing of the 39 G to the marginal dC of the template and further ligation to the downstream strand, yielding a two nucleotides deletion (see left panel Figure 6E). Additionally, when the nucleotides were supplied independently, the enzyme inserted specifically G opposite the marginal dC, producing a final 32 mer ligated product. As expected, substitution of the marginal dC of the template by a dG precluded BsuLigD to promote direct ligation of the 39 and 59 ends (see right panel and scheme in Figure 6E), and now C was the nucleotide mainly inserted. Thus, it seems that to get final ligation, BsuLigD polymerization active site would be designed to accommodate the "bubble" formed by a double or a triple mispairing to juxtapose the upstream 39 and the downstream 59 termini. Under these conditions a low proportion of misinserted and further ligated A and U was also observed suggesting that, despite its potential mutagenicity, the enzyme would cope with mispaired nucleotides at the joining site to restore the integrity of the broken DNA strand.

Conclusions
Here, it has been demonstrated that B. subtilis gene ykoU encodes for a LigD that shares with other well documented members of this family the essential biochemical features to repair DNA breaks [18,25,31,40]. Thus, BsuLigD i) has a polymerization activity, ii) shows a preferential insertion of rNMPs, iii) recognizes specifically a phosphate group at the downstream 59 end of gapped DNA molecules, iv) contains an intrinsic ligase activity and v) it can promote dislocations of the template and primer strands, suggesting the high flexibility of the catalytic site to deal with and hold different and unstable DNA realignments. These biochemical characteristics would have evolved to allow the enzyme to play the crucial joining of the two broken strands, despite of its potential mutagenicity.
Mtuand Pae-LigDs contain, besides a polymerization and ligase domain, a 39-phosphoesterase (PE) domain with a nucleolytic activity that would remodel the 39 modified ends prior to their elongation and further ligation [18,25]. Thus, PE domain of PaeLigD has been shown to hydrolyze 39-phosphates from the DSB to produce a 39-OH primer for the polymerization activity. In addition, the enzyme can also degrade the RNA tract, synthesized by the polymerization activity, to yield a 39-NMP, the preferred substrate of its ligase activity [41,46,47]. The PE domain of MtuLigD can also resect nucleotide flaps often produced by microhomology-mediated synapsis [20]. The potentiality of bacterial Ku protein to recruit proteins other than LigD has been recently demonstrated with the identification of the DNA repair helicase UvrD1 and Sir2-like protein as novel interaction partners for mycobacterial Ku [27,28]. This finding, together with the absence of a PE domain in BsuLigD opens the possibility of DNA end-cleaning protein recruitment by BsuKu to the DSB. In this sense, several B. subtilis proteins have been proposed to play a role in basal end processing [48] as B. subtilis SdcD whose ortholog in Escherichia coli has a nucleolytic activity [49] and forms a complex with the B.subtilis SbcC protein shown to participate in the repair of DSB [50]. Moreover, Mre11 protein from the MRX complex required for NHEJ in yeast is a nuclease of the SbcCD family shown to interact with Yku80 [1]. Another protein that could take part in processing 39 ends during NHEJ is B. subtilis polynucleotide phosphorylase (PNPase) as it achieves Mn 2+ dependent 39-59 phosphorolytic degradation on 39-tailed duplex DNAs [51], a null B. subtilis pnpA mutation being epistatic to the Dku mutation. Finally, an attractive candidate would be B. subtilis polymerase X (PolX) as this DNA repair protein is endowed with a 39-59 After incubation for 20 min at 30uC the elongation and ligation products were analyzed by 8 M urea-20% PAGE and autoradiography. Position of the unextended primer is indicated on the right. Asterisks indicate the 32 P 59-labeled end of the primer strand. (C) and (D) Activity on a mispaired 39 end in the 3 nucleotides gapped molecule. The assay was performed essentially as described in Materials and Methods in the presence of 1 mM of the indicated nucleotide (4N: the four nucleotides), 0.6 mM MnCl 2 , 1.5 nM of the depicted gapped substrate and 50 ng of BsuLigD, in the absence (C) or presence (D) of 50 ng of BsuKu. After incubation for 20 min at 30uC the elongation and ligation products were analyzed by 8 M urea-20% PAGE and autoradiography. Position of the unextended primer and ligation products is indicated on the right. (E) Activity on a 3 nucleotides mismatched 39 terminus present in 5 nucleotides gapped molecules. The assay was performed essentially as described in Materials and Methods in the presence of 1 mM of the indicated nucleotide, 0.6 mM MnCl 2 , 1.5 nM of the depicted gapped substrate, 50 ng of BsuLigD and 50 ng of BsuKu. After incubation for 20 min at 30 uC the elongation and ligation products were analyzed by 8 M urea-20% PAGE and autoradiography. Position of the unextended primer and ligation products is indicated on the right. The two DNA substrate molecules only differ in the marginal nucleotide of the template region. Schematic representations of the proposed realignments of the primer and template strands are drawn on the right. BsuKu homodimer and BsuLigD are represented by orange arches and a green rectangle, respectively. doi:10.1371/journal.pone.0064232.g006 exonuclease activity specialized in resecting 39 flap structures [52] and it can sanitize 39 damaged ends [53]. The potential involvement of bacterial PolX in NHEJ would agree with the increased sensitivity against DSBs provoked by c-rays observed in the bacterium D. radiodurans upon deletion of the gene that codes for its own PolX [54], and with the certainty that in eukaryotes the gap filling step during NHEJ is carried out by a family X DNA polymerase [2,3,12]. Thus, the probable recruitment and involvement of a trans-acting nucleolytic activity to remodel 39 ends before the repair of DSBs would imply that bacterial NHEJ machinery could be more complex than currently thought. Figure S1 Organization of the different enzymatic activities of BsuLigD and representatives of the two LigD subfamilies described in [26]. The polymerization (Pol), phosphoesterase (PE) and ligase domains are represented as green, yellow and orange cilinders, respectively. (TIF) Figure S2 BsuLigD-adenylate complex formation. Reactions were performed as described in Materials and Methods by incubating the indicated amount of purified BsuLigD with 5 mM [a-32 P]ATP (2.5 mCi) in the presence of 5 mM MnCl 2 . After incubation for 10 min at 30uC the reaction was stopped by adding 10 mM EDTA and 0.1% SDS. The samples were then filtered through Sephadex G-50 spin columns to remove the non-incoporated ATP, and further analyzed by 12% SDS-PAGE. Label transfer to the 70-kDa BsuLigD polypeptide was visualized by autoradiography of the dried gel. The position of the BsuLigD-AMP complex is indicated on the left. (TIF) Figure S3 BsuLigD has an inherent ligase activity. The assay was performed by incubating 4 ml of each fraction collected after sedimentation of the purified BsuLigD on a 15-30% glycerol gradient (see Materials and Methods) with 1.5 nM of the nicked DNA depicted (asterisk indicates the 59 32 P-labeled end of the primer strand) in the presence of 20 mM MnCl 2 and 50 ng of BsuKu. After incubation for 10 min at 30uC the ligation products were analyzed by 8 M urea-20% PAGE and autoradiography. (TIF)