Nuclear Localization and in Situ DNA Damage by Mycobacterium tuberculosis Nucleoside-diphosphate Kinase*

Nucleoside-diphosphate kinase of Mycobacterium tuberculosis (mNdK) is a secretory protein, but the rationale behind secreting an enzyme involved in the maintenance of cellular pool of nucleoside triphosphates is not clearly understood. To elucidate the biological significance of mNdK secretion, we expressed mNdK fused to green fluorescent protein in HeLa and COS-1 cells. Interestingly, mNdK was detected in the nuclei of HeLa and COS-1 cells. Incubation of mNdK with nuclei isolated from HeLa and COS-1 cells led to in situ damage of chromosomal DNA. Surface plasmon resonance studies demonstrated that mNdK binds supercoiled plasmid DNA lacking apurinic/apyrimidinic sites with a dissociation constant of 30 ± 3.2 μm. Plasmid cleavage by mNdK was found to be dependent on the specific divalent metal ion and inhibited by a metal ion chelator. Moreover, the metal ion-dependent DNA cleavage by mNdK was mediated by superoxide radicals as detected by electron paramagnetic resonance. The cleavage reaction was inhibited under nitrogen atmosphere confirming the necessity of molecular oxygen for DNA cleavage. In view of the findings that mNdK is secreted by intracellular mycobacteria and damages the nuclear DNA, it can be postulated that mNdK may cause cell death that could help in the dissemination of the pathogen.

Nucleoside-diphosphate kinase (NdK) 1 (EC 2.7.4.6) is required for the maintenance of cellular ratios of nucleoside triphosphates and diphosphates. NdK catalyzes phosphoryl transfer from a nucleoside triphosphate to a nucleoside diphosphate and is ubiquitously present in different organisms. There is a wide body of literature available on the diverse array of cellular events and complex regulatory processes brought about by this enzyme in different organisms (1)(2)(3)(4)(5)(6)(7). In this regard, we reported earlier the biochemical characterization and extracellular secretory nature of NdK of Mycobacterium tuberculosis (mNdK) (Rv2445c) (7). Secretory NdKs from other organisms are also well documented (8 -11).
NdKs from different organisms share considerable sequence and structural homology (12). Eight genes of the NdK family (nm23-H1 to nm23-H8) have been discovered in humans; they possess different but specific functions within the cell depending upon their localization (13). One of the major characteristics attributed to NM23-H1 and NM23-H2 is their DNA-associated properties (14). NM23-H1 and NM23-H2 share 88% sequence homology, and both have been shown to bind and nick the c-myc promoter, although they perform distinct biological function (14). NM23-H1 has been shown to have DNase activity, which is regulated by nucleosome assembly protein SET (putative HLA-associated protein II) and induced by granzyme A, leading to apoptosis (15). Postel et al. (16) have reported that DNA cleavage activity of human NM23-H2 involves phosphodiester bond cleavage by a DNA glycosylase/lyase-like mechanism, which entails the formation of a covalent Schiff's base intermediate; they postulated that it is involved in DNA repair function. Additionally, NdK of Escherichia coli (eNdK) has been shown to be involved in the DNA repair function mediated through uracil, which when mispaired with adenine and guanine serves as a specific target for excision and repair (3). Contrary to these observations, a recent report by Bennett et al. (17) shows that eNdK does not act as a uracil-processing DNA repair nuclease. The physiological significance of the DNA-associated properties of bacterial NdK is still not clear.
The current study focuses on understanding the intriguing purpose of NdK secretion by M. tuberculosis. In this report, we show that mycobacterial NdK localizes into the nucleus of mammalian cells and causes in situ DNA nicking. In addition, the study provides conclusive evidence regarding the formation of free radicals, which mediate DNA cleavage by mNdK. The DNA cleavage by mNdK is sequence-specific and requires the presence of specific divalent cations. This is the first report showing that a secreted mycobacterial enzyme is localized into the mammalian cell nucleus and causes superoxide radicalmediated DNA cleavage.

EXPERIMENTAL PROCEDURES
Materials-All chemicals and reagents were purchased from Sigma. Bacterial culture media were purchased from Difco Laboratories. mNdK and its mutant (H117Q) were expressed as GST fusion proteins in E. coli using the pGEX-2T expression vector (Amersham Biosciences). After thrombin cleavage, proteins were purified by chromatography on a Superdex 200 column as described earlier (18). DNA duplexes were prepared by heating an equimolar mixture of 5Ј-AGTC-TCCTCCCCACCTTCCCCACCCTCC CCACCCTCCCCATAAGC-3Ј and 5Ј-GCTTATGGGGAGGGTGGGGAGGGTGGGGAAGGTGGGGAGGA-GACT-3Ј (complementary sequence) in 10 mM Tris buffer (pH 7.0) containing 200 mM NaCl at 95°C followed by slow cooling. Subsequently, they were labeled with [␥-32 P]ATP in the presence of T4 polynucleotide kinase as described earlier (14). Similarly, a 45-bp ATGC repeat and its complementary oligonucleotides were annealed and labeled.
Construction of Expression Plasmid, Cell Culture, and Transfection-Genomic DNA isolated from M. tuberculosis H 37 Rv was used for PCR amplification of mndk. The nucleotide sequence of primers used was: 5Ј-ATCTAGTGTGAATTCCCATGACCGAACGG-3Ј with an EcoRI site at the 5Ј-end (forward primer) and 5Ј-GCGCACCCGGGATCCGG-CGCCGGG-3Ј with a BamHI site at the 3Ј-end (reverse primer). The amplicons were digested with EcoRI-BamHI and ligated into pEGFP-N1 (Clontech Laboratories, Inc.) to obtain pmNdK-EGFP plasmid.
Fluorescence Microscopy and Immunoblot Analysis-At 24 h posttransfection, the cells were harvested by 0.05% trypsin, and the nuclei were isolated using lysis buffer (0.5% Nonidet P-40, 5 mM MgCl 2 , 25 mM KCl, and 10 mM Tris HCl (pH 7.5)) as described earlier (15). The nuclei were counterstained with the DNA-binding dye Hoechst 33258, and the images were acquired using a fluorescence microscope (Nikon Corp.). Localization of mNdK-GFP in the isolated nuclei was also analyzed by immunoblotting using polyclonal antisera raised against mNdK as described earlier (7).
DNA Nicking Assay-HeLa and COS-1 cells were treated with lysis buffer and centrifuged at 2580 ϫ g to isolate nuclei. The nuclei were incubated with 250 ng of mNdK or 1 unit of DNase I (10 ng) (Amersham Biosciences) for 4 h at 37°C. At the end of incubation, nuclei were pelleted, washed with lysis buffer, and incubated again with 5 units of Klenow fragment of DNA polymerase and 1 Ci of [␣-32 P]dCTP for 30 min at 37°C. Radiolabeled nuclei were pelleted and washed three times with lysis buffer, and DNA was then isolated by proteinase K treatment and ethanol precipitation. The incorporation of [␣-32 P]dCTP into DNA was determined using a liquid scintillation counter (Beckman Instruments). Simultaneously, isolated DNA was also resolved by alkaline agarose gel electrophoresis and analyzed by autoradiography as described earlier (15).
Bioinformatics Analysis-To gain insight into the possible role of mNdK, its amino acid sequence was compared with the "non-redundant data base" using BLAST (19). Some of the homologous sequences retrieved from the program were used for the multiple sequence alignment, and to find conserved motifs by Clustal W (version 1.82). The gap extensions and gap distance penalties were kept as 0.05 and 10, respectively, during the alignment (20).
Plasmid Cleavage Assay-Supercoiled pUC19myc plasmid (having a nuclease hypersensitive element (NHE) of the c-myc promoter) was incubated with mNdK or mNdK-H117Q in the presence of varying concentrations of ATP (0.25-0.5 mM), ADP (2 mM), or AMP (2 mM) in reaction buffer I (50 mM Tris-HCl (pH 7.9) containing 100 mM KCl, 1.5 mM MgCl 2 , 50 g/ml bovine serum albumin, and 2% glycerol) for 30 min at 37°C. The reaction was terminated by incubating the reaction mixture with stop solution (1% SDS, 10 mM EDTA, and proteinase K (200 g/ml)) for 30 min at 55°C. Samples were resolved on a 1% agarose gel. Moreover, the pUC19myc plasmid used for the study was examined for the presence of apurinic/apyrimidinic (AP) or abasic sites using 1,2ethylenediamine (21).
DNA Binding Assays-An electrophoretic mobility shift assay was used to determine the binding of mNdK with duplex DNA. Briefly, 5Ј-32 P-labeled 45-bp duplex DNA containing an NHE sequence (5 ng) was incubated with mNdK or mNdK-H117Q in the presence of EDTA and ATP in reaction buffer I for 30 min at 37°C. After incubation, samples were resolved on a 5% native PAGE, vacuum-dried, and autoradiographed.
Surface Plasmon Resonance Analysis-The interaction of mNdK with plasmid DNA was studied using a BIAcore 2000 instrument (Biacore AB, Uppsala, Sweden) as described earlier (22). Anti-GST antibody (2 l/min) was chemically coupled (density ϳ3000 RU) on the CM5 sensor chip activated by treatment with 1-ethyl-3-(3-dimethyaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide, and then the unoccupied chip surface was deactivated using ethanolamine. Thereafter, GST-tagged mNdK (250 ng/l) was passed over the chip for 10 min at 2 l/min. Various concentrations of pUC19myc plasmid DNA (1.85-99.60 M) dissolved in filtered and degassed reaction buffer I containing 0.005% surfactant P20 (Biacore AB) were injected (10 l/min for 3 min) to study its interaction with chip-bound mNdK. After each binding cycle, glycine (10 mM, pH 2.0) was injected (10 l/min for 30 s) to remove both GST-mNdK and plasmid, allowing the base line to return back to that of the captured antibody. A blank-flow cell (without attached mNdK) was used to account for any bulk or solvent effect not specific to the interaction. Data analysis was done by plotting the signal (in response units) obtained against various concentrations of pUC19myc plasmid. The data were fitted into the equation: where RU eq is the measured response, RU max is the maximum response, and K D is the equilibrium dissociation constant.
Free Radical Detection Assays-Plasmid cleavage was studied under both aerobic and anaerobic (nitrogen atmosphere) conditions as described by Crumbliss and Basolo (23). The effect of superoxide dismutase on plasmid cleavage was investigated as described earlier by Hertzberg and Dervan (24).
The electron paramagnetic resonance (EPR) spectra of plasmid cleavage assays were recorded on X band (ϳ9 GHz) using a Varian E-109 spectrometer with 100 kHz field modulations. A quartz flat liquid cell was used in a dual sample rectangular cavity TE104. The effects of anaerobic conditions and superoxide dismutase on free radical generation were also examined. The EPR settings were: microwave power, 1.50 milliwatts; modulation amplitude, 10 G; scan time, 240 s; time constant, 0.016 s; and scan range, 100 G. The g value was determined by comparison with diphenylpicrylhydrazyl sample (g value, 2.0036) as described earlier (25).

RESULTS
Sequence Analysis of mNdK-Sequence alignment of mNdK with its homologs indicated that the critical residues (lysine 12, involved in the covalent interaction of NM23-H2 with duplex DNA (14), and histidine117, involved in autophosphorylation (7)) are phylogenetically conserved (Fig. 1). mNdK and NM23-H2 have 45% sequence identity over 135 common residues (26). X-ray structure of mNdK has revealed that unlike several bacterial NdKs, which are tetrameric in nature, the mycobacterial NdK forms a hexamer similar to human NdK (NM23-H2 and NM23-H1) (26), which suggests that mNdK may be a more evolved protein and may have an important role in the pathogenesis of M. tuberculosis.
Nuclear Localization of mNdK-Because M. tuberculosis is an intracellular pathogen, the extracellular secretion of mNdK (7), like the NdKs of other species (8 -11), encouraged us to study its possible role in the host pathogen interaction. To ascertain the subcellular localization of mNdK in host cells, we cloned it in a mammalian expression vector (pEGFP-N1) and transfected it into HeLa and COS-1 cells. Fluorescence microscopy of nuclei isolated from these cells transfected with pmNdK-EGFP plasmid revealed that the mNdK protein localizes in the nucleus, whereas nuclei isolated from cells transfected with pEGFP-N1 did not exhibit green fluorescent protein (GFP)-associated fluorescence ( Fig. 2A). Nuclear localization of mNdK was further confirmed by probing the nuclear lysate of HeLa cells transfected with pEGFP-N1 and pmNdK-EGFP, using mNdK-specific polyclonal antibodies. A 40-kDa protein (corresponding to the predicted molecular mass of NdK-GFP fusion protein) reacted with anti-NdK antibody in the nuclear lysate of HeLa cells transfected with pmNdK-EGFP plasmid, whereas this protein was absent in the nuclear lysate of untransfected and pEGFP-N1-transfected cells (Fig. 2B). Similar results were obtained with the COS-1 cell line (data not shown).
DNA Nicking by mNdK-NM23-H1, the human homolog of mNdK, has been shown to localize in the nucleus and to have DNase I-like activity (15). Considering the homology of mNdK with human NdK, we investigated its effect on chromosomal DNA using isolated nuclei of HeLa and COS-1 cells. mNdK incubated with isolated nuclei showed no sign of oligonucleosomal DNA fragmentation when analyzed by 1% agarose gel electrophoresis (data not shown). To assess more subtle forms of DNA damage, single-stranded breaks were identified by radiolabeling the nicked ends with Klenow and resolving them by denaturing agarose gel electrophoresis. Alkaline agarose gel electrophoresis and scintillation counting showed high incorporation of [ 32 P]dCTP in DNA isolated from mNdK and DNase I-treated nuclei (1220 cpm/ng in mNdK-treated nuclei and 38273 cpm/ng in DNase I-treated nuclei) (Fig. 3). On comparison, mNdK showed ϳ3% of the DNA cleavage activity obtained with bovine pancreas DNase I.
mNdK Cleaves Supercoiled pUC19myc Plasmid-NM23-H1 and NM23-H2 were shown earlier to bind and nick the NHE region of the c-myc promoter (14). Incubation of mNdK with pUC19myc plasmid showed magnesium-dependent nicking and cleavage of the plasmid (Fig. 4A, lane 3). The requirement of metal ions for mNdK activity was further confirmed by the observation that cleavage was inhibited in the presence of EDTA (Fig. 4B), whereas some other divalent ions used had varied effect on the activity of mNdK. It was found that Mn 2ϩ and Ni 2ϩ could replace Mg 2ϩ (Fig 4A, lanes 4 and 7), whereas Zn 2ϩ and Ca 2ϩ failed to complement Mg 2ϩ (Fig. 4A, lanes 5 and  6). mNdK did not cleave pUC19 plasmid, suggesting that the cleavage was sequence-specific (Fig. 4C, lanes 2 and 4). Similarly, mNdK did not cleave pGEX-5X-3 (Amersham Biosciences) lacking the NHE region of the c-myc promoter (data not shown), which revealed that DNA cleavage by mNdK is sequence-specific. The presence of abasic sites in the plasmid used to assess the DNA-associated properties of mNdK was ruled out by treating the plasmid with 1,2-ethylenediamine (21). 1,2-Ethylenediamine was unable to cleave the pUC19myc plasmid, which indicates that plasmid used for the cleavage assays lack AP sites (Fig. 4C, lane 6). The DNA cleavage activity of mNdK was catalytic, as increasing the amount of DNA resulted in increased cleavage when plasmid was incubated with equal amount of mNdK (data not shown).
ATP Inhibits DNA Cleavage Activity of mNdK-The relationship between kinase activity and the DNA cleavage property of mNdK was examined by ascertaining the effect of ATP on DNA cleavage and using the kinase mutant of mNdK (H117Q). The presence of ATP in the reaction mixture abrogated the cleavage of plasmid by mNdK (Fig. 5A). ATP-mediated inhibition was not abrogated by the increased concentration of MgCl 2 (20 mM), suggesting that inhibition of plasmid cleavage by ATP was not because of scavenging of magnesium ions (Fig. 5A, lane 4). However, ADP and AMP had no effect on the plasmid cleavage activity of mNdK (Fig. 5B). To further examine ATP-induced inhibition, we used a mNdK mutant, H117Q, which is deficient in ATP binding, hydrolysis, and autophosphorylation activities (7). This mutant of mNdK also showed plasmid cleavage activity (Fig. 5C). It was observed that ATP also inhibited the plasmid cleavage activity of mutant mNdK, as observed for wild type mNdK (Fig. 5D), indicating that the kinase and DNA cleavage activities of mNdK are independent.
mNdK Binds to Double-stranded DNA-The DNA binding and cleavage properties of mNdK and its kinase mutant were further probed by electrophoretic mobility shift assay using a radiolabeled 45-bp duplex DNA fragment. Electrophoretic mobility shift assay studies revealed that mNdK and its mutant, H117Q, bind to radiolabeled duplex DNA. The DNA binding activity of mNdK was also found to be magnesium ion-dependent (Fig. 6A) and to be inhibited by EDTA and ATP (Fig. 6A,  lanes 6 and 7). The molecular interaction between mNdK and pUC19myc plasmid was characterized by surface plasmon resonance (27), and typical surface plasmon resonance sensograms were generated, where the plasmid in solution was allowed to interact with immobilized mNdK (Fig. 6B). Equilibrium response obtained from these curves was used to determine the dissociation constant as described under "Experimental Procedures." A dissociation constant of 30 Ϯ 3.2 M was obtained from the binding isotherms (Fig. 6B, inset). The rate of dissociation was independent of plasmid concentration in the solution phase, suggesting that the rebinding phenomenon was not significant under the reaction conditions employed. As control, no interaction was observed between pUC19myc and GST or pUC19 and GST-mNdK (data not shown).
Oxygen-dependent Free Radical-mediated DNA Cleavage by mNdK-The contribution of metal ions in the DNA binding and cleavage activities of mNdK prompted us to investigate the role of molecular oxygen in the reaction. Inhibition of plasmid cleavage by mNdK under nitrogen atmosphere showed a requirement for molecular oxygen (Fig 7A, lane 3). These observations indicated that DNA cleavage by mNdK may involve a Fenton-  like reaction. To validate this hypothesis, EPR experiments were carried out, and indeed superoxide formation was detected during the course of the reaction (Fig. 7C, I), which was not observed in the absence of magnesium ion (Fig. 7C, IV). To identify the nature of free radical, the g value of the cleavage reaction mixture was determined by comparison with diphenylpicrylhydrazyl. The g value of the reaction was 2.0039, suggesting the presence of superoxide. The formation of superoxide was found to be associated with the molecular oxygen, as no EPR signal was observed under nitrogen atmosphere (Fig.  7C, V). superoxide dismutase had no effect on the generation of free radicals (Fig. 7C, II) or the plasmid cleavage activity of mNdK (Fig. 7B), suggesting that the electrons on superoxo species are resonating between an mNdK-metal-oxygen complex and hence are not accessible to superoxide dismutase (23,24,28). DISCUSSION Intracellular pathogens secrete several proteins to gain advantage for their survival and pathogenesis. Recently, it has been shown that M. tuberculosis secretes protein kinase G and tyrosine phosphatases that modulate host defense mechanisms and help in the survival of the organism (29,30). Our earlier studies established that mycobacterial NdK, like the NdKs of other pathogenic organisms (7,9,11), is a secretory protein and is cytotoxic to macrophage cells in the presence of ATP (7). The human homologs of mNdK, NM23-H1 and NM23-H2, have been shown to bind and nick the NHE region of the c-myc promoter (14). These observations prompted us to investigate the significance of mNdK secretion and its DNA-associated properties. In the present study, we show that mNdK localizes into the nucleus and causes DNA damage. Nuclear localization of mNdK was shown by expression of mNdK as a GFP fusion protein in HeLa and COS-1 cells using fluorescence microscopy and immunoblotting (Fig. 2). mNdK induces a type of DNA damage (single-stranded nicks) that leads to larger DNA fragments not detected by conventional nondenaturing agarose gels. The in situ DNA damage by mNdK was observed by alkaline agarose gel electrophoresis (Fig. 3). Incorporation of radiolabeled dCTP at the site of nicks with Klenow fragment of DNA polymerase confirmed that mNdk causes in situ nicking of chromosomal DNA.
DNA nicking was further studied by using pUC19 plasmid containing the NHE sequence of the c-myc promoter (pUC19myc). It was observed that cleavage of pUC19myc plasmid mediated by mNdK resulted in relaxation and subsequent linearization of the plasmid due to DNA nicking (Fig. 4). It has been reported by Freifelder and Trumbo (31) that single strand breaks can be produced in double-stranded DNA in a variety of ways and that some of these breaks in opposite strands that are in close proximity lead to double strand breaks. Proteinase K treatment after completion of plasmid cleavage reaction results in a band corresponding to the linearized form of the plasmid (Fig. 4), which excludes the possibility of band shift due to a DNA-mNdK complex. This fact was further corroborated by the observation that pUC19myc linear DNA obtained from plasmid digested with BamHI (unique site in the multiple cloning site of the plasmid) migrated at the same position as mNdK-cleaved linearized plasmid (Fig. 4C, lane 5). The plasmid cleavage mediated by mNdK was sequence-specific because it cleaved pUC19 plasmid containing a 45-bp NHE sequence of c-myc promoter but failed to cleave pUC19 plasmid (Fig. 4C, lanes 2  and 4). A similar finding has been reported for human NdK, NM23-H2 (14). The binding of mNdK to DNA was also found to be sequence-specific as it was unable to bind nonspecific 45-bp double-stranded DNA (data not shown). These results suggest that the cleavage activity is mediated by mNdK and not caused by any kind of DNase contamination.
Recently, Bennet et al. (17) have given compelling evidence regarding eNdK lacking detectable uracil-DNA glycosylase activity, which was in contradiction to the earlier report by Postel et al. (3). Therefore, we also investigated the magnesium iondependent AP lyase activity of mNdK. The sequence comparison of mNdK with uracil-DNA glycosylase of M. tuberculosis (Swiss Protein accession number P95119) and E. coli (Swiss Protein accession number BAA10923.1) showed that five highly conserved structural motifs (QDPYH, AIPPS, LLLN, GS, and HPSPLSAHR), which serve as a signature for the uracil-DNA glycosylases, are absent in mNdK. 1,2-Ethylenediamine has been shown to cleave the abasic/AP sites in doublestranded DNA (21). It was found that 1,2-ethylenediamine did not cleave the pUC19myc (Fig. 4C, lane 6), thereby proving that this plasmid does not have AP sites. Finally, mNdK cleaved pUC19myc plasmid specifically and had no effect on other plasmids such as pUC19 and pGEX-5X-3. These observations collectively suggest that DNA cleavage activity by mNdK is not related to AP lyase activity.
Earlier, it was hypothesized that ATP-mediated inhibition of DNA cleavage activity of NM23-H2 could be because of (a) the kinase activity of NdK, (b) competition by Mg 2ϩ with ATP for the active site, (c) quenching of Mg 2ϩ by ATP directly, or (d) the inhibitory effect of phosphorylation of His-118 (14). We also elucidated the effect of ATP on the DNA cleavage and binding properties of mNdK. Abrogation of the DNA binding and cleavage properties of mNdK by ATP was not restored even after increasing the magnesium concentration to 15-fold (Fig. 5A,  lane 4), suggesting that inhibition of DNA cleavage by ATP is not due to magnesium ion scavenging. Histidine 117 of mNdK, a highly conserved residue among other NdKs, was shown to be essential for its autophosphorylation (7,(32)(33)(34), and H117Q mutant failed to complement an ndk-disrupted strain of Pseudomonas aeruginosa (35). Interestingly, H117Q mutant of mNdK was found to have both DNA binding and cleavage activities (Fig. 5C, lane 3, and Fig. 6A, lane 4). The DNA cleavage property of H117Q mutant was also found to be inhibited by ATP, indicating that plasmid cleavage occurs by a mechanism that is independent of the ATP binding/hydrolysis property of NdK. Earlier reports indicated that tightening of the erythrocytic NdK conformation was observed in the presence of ATP, which led to the changes in its activity toward p-chloromercuricbenzoate (36). Accordingly, it can be hypothe- sized that ATP may cause structural changes in the DNA binding pocket of mNdK, leading to the abrogation of DNA binding and cleavage activity. Collectively, these observations suggest that kinase and DNA binding/cleavage activities of mNdK are not linked and appear to be mediated by different residues.
While examining the role of metal ions in the plasmid cleavage activity of mNdK, we found that certain divalent cations such as Mn 2ϩ or Ni 2ϩ could mediate this activity, whereas Zn 2ϩ and Ca 2ϩ failed to complement Mg 2ϩ (Fig 4A). These results were substantiated by the fact that in the presence of EDTA, the DNA cleavage activity was inhibited (Fig. 4B). Ca 2ϩ and Zn 2ϩ have a low propensity for coordinate bond formation, whereas Mn 2ϩ and Ni 2ϩ can easily form coordinate bonds because of their partially filled outermost orbitals where they can accept electrons (37,38). On the other hand, Mg 2ϩ does not have a vacant outer orbital; yet it is able to mediate DNA cleavage, presumably because it has a low radius (0.66 Å) to charge ratio, which leads to an increase in the Lewis acidity allowing this ion to coordinate with multiple negatively charged ligands (37)(38)(39). Hence it can be proposed that a suit-able metal ion forms a coordination complex with mNdK, and subsequently mNdK delivers this metal ion to the site of DNA helix where activation of molecular oxygen occurs through electron transfer mechanism resulting in DNA cleavage. The proposed hypothesis was strengthened by the observation that mNdK-mediated cleavage of DNA was completely inhibited under N 2 atmosphere, suggesting the necessity of molecular oxygen for the reaction (Fig. 7A, lane 3). The involvement of oxygen and metal ions in the DNA cleavage prompted us to investigate the nature of species involved in this electron transfer reaction. The presence of superoxide species in the mNdKmediated plasmid cleavage reaction was detected by EPR studies (Fig. 7C), and the calculated g value of 2.0039 was similar to the g value of the chemical system containing potassium superoxide in dimethyl sulfoxide (40). An EPR signal was not observed in the absence of Mg 2ϩ , molecular oxygen, DNA, or mNdK (Fig. 7C), implying that all these components are part of a redox system responsible for the free radical generation.
The nuclear localization and DNase-like activity of mNdK, similar to NM23-H1 (15), suggest that mNdK can be a potent virulence factor. Secretion of NdK by M. tuberculosis might help in the dissemination of this intracellular pathogen, and subsequent establishment of the disease, by killing host cells via its DNase activity. Our experiments mimic in vivo infection, and it is likely that the same phenomenon may occur during mycobacterial infection.