The RecA Intein of Mycobacterium tuberculosis Promotes Cleavage of Ectopic DNA Sites IMPLICATIONS FOR THE DISPERSAL OF INTEINS IN NATURAL POPULATIONS*

The RecA intein of Mycobacterium tuberculosis, a novel double-stranded DNA endonuclease, requires both Mn and ATP for efficient cleavage of the inteinless recA allele. In this study, we show that Mg alone was sufficient to stimulate PI-MtuI to cleave doublestranded DNA at ectopic sites. In the absence of Mg , PI-MtuI formed complexes with topologically different forms of DNA containing ectopic recognition sequences with equal affinity but failed to cleave DNA. We observed that PI-MtuI was able to inflict double-strand breaks robustly within the ectopic recognition sequence to generate either a blunt end or 1–2-nucleotide 3 -hydroxyl overhangs. Mutational analyses of the presumptive metal ion-binding ligands (Asp, Asp, and Glu) together with immunoprecipitation assays provided compelling evidence to link both the Mg and Mn and ATP-dependent endonuclease activities to PIMtuI. The kinetic mechanism of PI-MtuI promoted cleavage of ectopic DNA sites proceeded through a sequential mechanism with transient accumulation of nicked circular duplex DNA as an intermediate. Together, these data suggest that PI-MtuI, like group II introns, might mediate ectopic DNA transposition and hence its lateral transfer in natural populations.

Mobile inteins and introns are genetic elements capable of self-propagation by "homing" into host genes in a wide variety of organisms: eubacteria, eukarya, archaea, and viruses (reviewed in Ref. 1). The process is promoted by a homing endonuclease, which is encoded by an open reading frame embedded within the genetic element. The genes for homing endonucleases are found among group I and group II introns, archaeal introns, intein-coding sequences, and free standing open reading frames (reviewed in Refs. [1][2][3][4][5][6]. Inteins are genetic elements present within protein-coding sequences with dual function: protein-splicing and homing endonuclease activities. They are believed to play a central role in rearrangement of organelle as well as nuclear genomes (1)(2)(3)(4)(5)(6). The hallmark of homing endonucleases is their ability to recognize and cleave extended asymmetric sequences (14 -40 bp) that are generally centered on the intein insertion site in inteinless alleles (1,7). Homing endonucleases can be classified into four families based on the presence of conserved motifs: LAGLIDADG, GIY-YIG, His-Cys box, and H-N-H (1,5,6,8). Among these, the LAGLIDADG family is the largest, widespread and much studied class of homing endonucleases. Structural and biochemical studies have demonstrated that homing endonucleases with one LAGLIDADG motif act as homodimers, whereas enzymes with two such motifs function as monomers during catalysis (9 -12). Under standard assay conditions in vitro, these enzymes are extremely specific for their recognition sites. However, recent evidence suggests that self-splicing group II introns transpose into ectopic DNA sites that resemble their natural homing sites (13,14). The transposition of group II introns involves reverse splicing of the intron into the intronless allele, which are then reverse transcribed to give complementary DNA. Following the synthesis of the second strand, the intron is incorporated into the genomic DNA by homologous recombination (1)(2)(3)(4)(5)(6). But equivalent information is not available for any intein endonuclease. The exact sequence of events that lead to recognition and cleavage of DNA by homing endonucleases is poorly understood. In addition, in no case the molecular mechanism underlying cleavage of ectopic DNA sites by an intein endonuclease has been elucidated.
Mycobacterium tuberculosis RecA intein (PI-MtuI) 1 is a member of the LAGLIDADG superfamily of homing endonucleases (15)(16)(17)(18). In previous studies, we showed that PI-MtuI is a novel homing endonuclease, which inflicted a staggered doublestrand break 24 bp upstream of the intein insertion site in the inteinless recA allele (henceforth called cognate site) (18). Typically, Mg 2ϩ is the preferred metal ion and the only cofactor required for cleavage of inteinless alleles by the LAGLIDADG family of homing endonucleases. In contrast, PI-MtuI required both Mn 2ϩ and ATP for cleavage within the inteinless recA allele (18). In this report, we extend our focus on the mode of action of PI-MtuI and define some of the basic conditions essential for optimal cleavage of ectopic DNA sites. Mutational analyses of presumptive metal ion-binding ligands and immunoprecipitation assays provided compelling evidence that Mn 2ϩ and ATP as well as Mg 2ϩ -dependent endonuclease activities are intrinsic to PI-MtuI. Thus, PI-MtuI is the first example of an intein endonuclease demonstrated to have the ability to recognize and cleave cognate as well as ectopic DNA sites in the presence of alternative cofactors. Thus, these data implicate a possible role for PI-MtuI in its lateral transfer in natural populations.

MATERIALS AND METHODS
Reagents, Proteins, and DNA-All the chemicals used in this study were of analytical grade. Buffers were prepared using deionized water. Restriction enzymes, nylon N ϩ membrane, and T4 DNA ligase were obtained from Amersham Biosciences. Vent DNA polymerase was obtained from New England Biolabs. Phage T4 polynucleotide kinase and the synthetic oligonucleotides were purchased from Invitrogen. Circular single-stranded and form I DNA from wild-type bacteriophage M13 were prepared as described (19). DNA was dissolved in 10 mM Tris-HCl buffer (pH 7.5) containing 1 mM EDTA and the concentrations were expressed in moles of nucleotide residues.
Bacterial Strains and Media-Escherichia coli strain DH5␣ was used for plasmid manipulations and grown in liquid or solid agar LB media supplemented with appropriate antibiotics. E. coli strain DH5␣ bearing the plasmid pGRI was used for protein overexpression as described (18). PI-MtuI was purified to homogeneity and its concentration was determined as described (18). The same purification method was used to purify the variants of PI-MtuI.
PI-MtuI Cleavage Assay-Endonuclease assays were performed as described (18). Briefly, reaction mixtures (25 l) contained 25 mM Tris-HCl buffer (pH 7.5), 5 mM MgCl 2 , 0.4 mM dithiothreitol, 16 M form I M13 DNA and PI-MtuI at the concentrations indicated in the legend to the figures. After incubation at 37°C, reactions were stopped by the addition of SDS to a final concentration of 0.1%, and the samples were deproteinized by incubation with proteinase K (0.2 mg/ml) for 15 min at 37°C. Three l of gel loading buffer (20% glycerol containing 0.12% (w/v) each of bromphenol blue and xylene cyanol) was added to each sample and separated on a 0.8% agarose gel in 89 mM Tris borate buffer (pH 8.3) containing 2 mM EDTA at 3 V/cm. The gel was stained with ethidium bromide (0.5 g/ml) and DNA was visualized by UV illumination. Subsequently, DNA was transferred to nylon N ϩ membrane and visualized by Southern hybridization (21). The bands were quantified in a UVI-Tech gel documentation station using UVI-BandMap software version 99 and plotted using Graphpad Prism version 2.0.
Immunological Techniques-Polyclonal antibodies against PI-MtuI were generated in rabbits and characterized as described (18). Immunoprecipitation reactions were performed by incubating 20 g of PI-MtuI in the endonuclease assay buffer with preimmune IgG or anti-PI-MtuI IgG tethered to protein A-Sepharose for 2 h with continuous stirring at 4°C. The slurry was centrifuged at 6000 rpm at 4°C for 2 min, and the supernatant was assayed for Mg 2ϩ -dependent PI-MtuI endonuclease activity as described above. Wild-type and variant forms of PI-MtuI were electroblotted onto a nitrocellulose membrane after SDS-PAGE. The membrane was stained with anti-PI-MtuI antibodies and visualized by chemiluminescence as described (22,23).
Site-directed Mutagenesis-PI-MtuI variants were generated by sitedirected mutagenesis by the overlap extension PCR-based method (20). The primers used for this purpose are listed in Table I. Single-site mutations of the conserved amino acid residues Asp 122 , Asp 222 , Glu 220 , Lys 131 , and Lys 195 were generated using the plasmid, pEJ135 (Ref. 5), as a template. PCR products were digested by SmaI and AccI, and the reaction mixture was separated by electrophoresis on a 0.8% agarose gel. The 851-bp fragment isolated from the gel was ligated to pGRI, which had been digested by the same set of enzymes. The variants were screened by digestion with an appropriate restriction endonuclease (21). The mutant sequences were verified by sequencing in ABI PRISM DNA sequencer (PerkinElmer Life Sciences). The recombinant plasmids were maintained and overexpressed in E. coli strain DH5␣. The expression levels of all the PI-MtuI variants were similar to the wildtype enzyme.
Mapping of the PI-MtuI Cleavage Sites in M13 DNA-To map ectopic DNA sites in the M13 genome, form I DNA was digested with PI-MtuI to generate P1 (3.5 kb) and P2 (2.9 kb) fragments (Fig. 1). The fragments isolated from the gel were digested with appropriate restriction enzymes. The restriction map thus constructed indicated the presence of two PI-MtuI cleavage sites in the wild-type M13 genome: one in the HinfI-HinfI fragment (2498 to 2845 bp), and the second in the NdeI-AccI fragment (6003-6090 bp). Accordingly, we prepared fragments containing the ectopic DNA sites by digestion with appropriate restriction enzymes. The fragments were labeled with [␥-32 P]ATP and polynucleotide kinase. To label at 5Ј-or 3Ј-end, the 2498 -2845-bp fragment was digested by HaeIII, which cleaves once at 2556 bp, thereby generating a 58-bp fragment with label at the 5Ј-end of the upper strand. Similarly, the 2247-2556-bp HaeIII fragment was digested by HinfI, which cleaves once at 2498 bp, generating a 58-bp fragment with label at the 5Ј-end of the lower strand. After electrophoresis on a 8% polyacrylamide gel, the bands corresponding to these fragments were excised from the gel, cut into small pieces, and DNA was eluted with 10 mM Tris-HCl (pH 7.5) buffer containing 1 mM EDTA. DNA was precipitated by ethanol, collected by centrifugation, vacuum-dried, and dissolved in 10 mM Tris-HCl (pH 7.5) buffer containing 1 mM EDTA.
To map the second ectopic DNA site, the 2723-6003-bp NdeI fragment was labeled as described above, digested by AccI to generate the 87-bp fragment with the label at the 5Ј-end of the upper strand. Similarly, form I DNA cut with AccI was end labeled, and digested by NdeI to generate the 87-bp fragment with the label at 5Ј-end of the lower strand. The bands corresponding to these fragments were isolated from acrylamide gels as described above.
Reactions were performed in the cleavage assay buffer containing wild-type or variant forms of PI-MtuI (100 pmol) and the specified DNA fragment with label at the 5Ј-end (100 ng), and stopped as described above. Reaction mixtures were deproteinized by extraction with phenol: chloroform:isoamyl alcohol solution (25:24:1), and the DNA was precipitated by ethanol. Control reactions were subjected to a Maxam-Gilbert chemical sequencing reaction except that the DNA substrates were not incubated with PI-MtuI. Samples were run side by side on 12 or 14% polyacrylamide gels in the presence of 8 M urea at 1800 V for 2.5 h (24). The gel was dried onto a Whatman 3MM filter paper and visualized as described (18).

Purification of Wild-type and Variants of PI-MtuI-Wild-
type and variants of PI-MtuI were expressed in E. coli and purified to homogeneity as described (18). Gel filtration on Superdex 75 column in a buffer containing 20 mM Tris-HCl (pH 7.5), 0.15 M NaCl, and 10% glycerol showed that PI-MtuI coeluted with molecular mass corresponding to 47 kDa, indicating that the quaternary structure of PI-MtuI is that of a monomer (data not shown). The data are consistent with that of a monomer, provided that the protein behaves as an average globular protein in solution. The identity of the purified protein was verified by sequencing 10 amino acid residues at the N-terminal end. Purified PI-MtuI and its variants were devoid of both 5Ј to 3Ј and 3Ј to 5Ј exonuclease activities.
PI-MtuI Binds Topologically Different Forms of DNA-In light of the evidence that negatively supercoiled DNA influences the activities of DNA-binding proteins, we explored the ability of PI-MtuI to interact with topologically different forms of M13 DNA. We observed that PI-MtuI bound form I, form II, or circular single-stranded DNA in a sequence nonspecific manner, and with similar affinities (data not shown). Further evidence in favor of sequence nonspecific binding of PI-MtuI emerged from DNase I protection assays. In this assay, a 16 M single-stranded, form I or linear duplex [ 3 H]DNA was first incubated with varying amounts of PI-MtuI (0.5-7.5 M), and then digested by DNase I (10 g/ml) for 10 min. Under the conditions used in this assay, the amount of protection rendered by PI-MtuI at Ն5 M was Ͼ75% in all cases, indicating that binding is independent of single-or double-stranded nature of DNA or its topological state (data not shown).
PI-MtuI Cleaves Ectopic DNA Sites in the Presence of Mg 2ϩ -To explore whether interaction of PI-MtuI with ectopic DNA elicits endonuclease activity, we used form I M13 DNA as the substrate. The rationale for using form I DNA rather than  (Fig. 1, lane 2), incubation of form I DNA with PI-MtuI produced form III DNA and two products of size 3.5 and 2.9 kb ( Fig. 1, marked P1 and P2). The combined sizes of P1 and P2 products equal the size of form III DNA, indicating that the latter was further cleaved to produce two products. Thus, form I M13 DNA contained two cleavage sites for PI-MtuI. Interestingly, cleavage proceeded to a similar extent in the presence of ATP or its analogues ( In previous studies, we showed that PI-MtuI required both Mn 2ϩ and ATP for optimal cleavage of the inteinless recA allele (18). Thus, PI-MtuI shows dual target specificity depending, first, on whether the substrate contains a cognate or ectopic site and, second, on whether the cofactor is Mg 2ϩ or Mn 2ϩ and ATP. One possible reason for this dual target specificity could be because of a contaminating endonuclease in the purified PI-MtuI preparation. Although the identity of PI-MtuI was established by sequencing 10 amino acid residues at the Nterminal end, it is necessary to validate the dual target specificity of PI-MtuI. Accordingly, immunoprecipitation assay was performed using polyclonal antibodies raised against PI-MtuI and the resulting immunosupernatant was assayed for endonuclease activity. As shown in Fig. 2, cleavage assay performed with immunosupernatant from the preimmune serum displayed Mg 2ϩ -dependent endonuclease activity. However, the same from the reaction carried out with anti-PI-MtuI failed to cleave M13 DNA, indicating that anti-PI-MtuI antibodies had precipitated Mg 2ϩ -dependent endonuclease activity. In gel filtration chromatography, both Mg 2ϩ -and Mn 2ϩ and ATP-dependent endonuclease activities were coincident with that of PI-MtuI elution profile (data not shown). Together, these observations suggest that both Mg 2ϩ -and Mn 2ϩ and ATP-dependent activities are intrinsic to PI-MtuI.
Specificity of DNA Cleavage by PI-MtuI-The finding that PI-MtuI bound various topologically different forms of M13 DNA with equal affinity raised the pertinent question of whether PI-MtuI can distinguish these substrates during catalysis. To this end, we incubated the same molar concentrations of topologically different forms of M13 DNA separately with increasing concentrations of PI-MtuI in the presence of Mg 2ϩ . After incubation, reaction mixtures were deproteinized and the samples were analyzed by agarose gel electrophoresis. The results show that PI-MtuI failed to cleave single-stranded DNA (Fig. 3, lanes 2-5), whereas form I and form II (Fig. 3,  lanes 7-10), or form III DNA (Fig. 3, lanes 12-15) were converted to P1 and P2 products. Although PI-MtuI bound form III as well as single-stranded DNA with similar binding pattern and affinity, it was able to distinguish between single-and double-stranded DNA during catalysis.
Effect of Ionic Strength on Cleavage of Ectopic DNA Sites by PI-MtuI-It has been reported that altering the ionic strength in the assay buffer influences the ability of homing endonucleases to recognize and cleave cognate DNA (1)(2)(3)(4)(5)(6). In accord, we found that 50 mM NaCl or potassium glutamate was sufficient to abolish both cleavage and nicking of cognate DNA by PI-MtuI (18). Interestingly, ectopic DNA was cleaved equally well in the absence or presence of 50 mM salt. However, cleavage of  ectopic DNA decreased progressively whereas nicking increased across a range of NaCl concentrations from 50 to 250 mM (data not shown). In contrast, cleavage of ectopic DNA by PI-MtuI was not attenuated even in the presence of 250 mM potassium glutamate (data not shown). The basis for these differences in relative salt profiles is unclear.
Mechanism of Cleavage of Ectopic DNA by PI-MtuI-We next investigated the mechanism of Mg 2ϩ -dependent cleavage of ectopic DNA by PI-MtuI using form I DNA. Given form I DNA containing the recognition sequence, PI-MtuI can linearize the substrate by two alternative mechanisms. In a sequential mechanism, nicking of form I DNA would lead to the formation of form II, and cutting the second strand in the vicinity of the first would generate form III DNA. In the concerted mechanism, form I DNA is directly converted to form III by cutting both strands simultaneously. The kinetics of the cleavage reaction showed a relatively slow phase during the first 10 min, at the same time the amount of form I DNA progressively decreased with a gradual increase in the levels of form II, followed by form III DNA (Fig. 4A). Quantification of cleavage products suggest that all of form I DNA was converted to form III and P1 and P2 products by 60 min (Fig. 4B). Together, these results indicate that Mg 2ϩ -dependent cleavage of ectopic DNA by PI-MtuI proceeds through a sequential mechanism. Whereas this mechanism is not unique, it is one of the simplest mechanisms that can fit the data presented in this report.
To further characterize the kinetics of cleavage, initial velocities of cleavage reactions catalyzed by PI-MtuI were measured using form I DNA in the presence of Mg 2ϩ . The velocities were determined from the initial linear phase of the cleavage reaction across a range of DNA concentrations from 2.5 to 30 M. A plot representing velocity versus substrate concentration displayed a linear increase at lower substrate concentrations, and then plateaued at higher concentrations. The reaction velocities followed a conventional Michaelis-Menten curve when plotted against substrate concentration (Fig. 5A). The linear phase of the curve was fitted into Lineweaver-Burk plot (  (Fig. 6). The prominent products in the Mg 2ϩ -dependent reaction were two expected cleavage products (marked P1 and P2) resulting from cleavage at two ectopic sites. As for the Mn 2ϩ and ATP-dependent cleavage of cognate DNA, the rate of product formation was much slower (Fig. 6B). The kinetics of cleavage of ectopic DNA (this study) or cognate DNA (18) in separate reactions were not significantly different from that of mixed reactions.
Mutation of Acidic Residues Alter the Catalytic Activity of PI-MtuI-Based on the crystal structure of PI-SceI, and multiple sequence alignments of the LAGLIDADG family of homing endonucleases, we identified residues Asp 122 , Asp 222 , Glu 220 , and Lys 195 in PI-MtuI as important for the recognition and cleavage of DNA (1,25,26). Asp 122 , Asp 222 , and Glu 220 are highly conserved ligands probably required for binding of a divalent cation and hence catalysis. In addition to Lys 195 , Lys 131 was chosen to investigate its role in catalysis. These residues were subjected to site-directed mutagenesis, and the variants were overexpressed in E. coli and purified to homogeneity (Fig. 7A). Each variant protein exhibited the same purification behavior as the wild-type enzyme, indicating proper folding of proteins (data not shown). The identity of PI-MtuI variants were further confirmed by Western blot analysis using antibodies raised against the wild-type PI-MtuI (data not shown).
We examined the ability of PI-MtuI variants to cleave ectopic DNA in the presence of Mg 2ϩ (Fig. 7B). In comparison with the endonuclease activity of the wild-type enzyme, the variant enzymes can be classified into three categories. First, the pattern and the extent of cleavage products generated by the D122Y variant was quite similar to the wild-type enzyme. Second, the E220A and K195F variant enzymes produced a significant amount of form II DNA together with a trace amount of form III DNA. Third, D222T and K131L variant enzymes generated nearly equivalent amounts of form II and form III DNA, but failed to convert the latter to P1 and P2 products. Although nicking activity was displayed by all the variant enzymes, nicking was not at the same positions as created by the wild-type enzyme (see below). These results indicate that the penultimate residue in the first LAGLIDADG motif, Asp 122 , is dispensable, and the second LAGLIDADG motif in PI-MtuI is necessary for Mg 2ϩ -dependent cleavage of ectopic DNA sites.
In a parallel set of experiments, we sought to relate the site of cleavage by restriction analysis. In particular, we wished to determine which of the two ectopic sites in M13 DNA was first cleaved by PI-MtuI. Restriction mapping revealed that PI-MtuI inflicted a double-strand break first at 2529 bp resulting in linearization of circular DNA, followed by a second doublestrand break at 6042 bp to generate P1 and P2 products (data not shown).

D122Y Variant of PI-MtuI Displays Gain of Function
Effect-The above results identified presumptive metal-binding ligands that are necessary for catalysis. Asp 122 is a highly conserved residue, therefore, likely to be critical for metal binding and catalysis. However, with ectopic DNA as the substrate, the Mg 2ϩ -dependent cleavage activity of the D122Y variant was similar to that of the wild-type enzyme (Fig. 7B). To re-evaluate the role of Asp 122 , we compared the cleavage activity of the D122Y variant using cognate and ectopic DNA substrates at varying concentrations of divalent metal ions. The cleavage activity of the D122Y variant increased gradually with increasing concentrations of MnCl 2 to reach its peak at ϳ3 mM similar to that seen previously for the wild-type enzyme (Fig. 8, lower panels). Surprisingly, unlike the wild-type enzyme (Fig. 8A, upper panel), the D122Y variant was most active  Fig. 1. At the specified time intervals as indicated above each lane, aliquots were withdrawn and the reaction was immediately stopped by the addition of SDS followed by proteinase K. Samples were analyzed on an agarose gel by electrophoresis. Panel B, the amount of form I DNA cleaved during the reaction was determined by quantitation of the gel shown in A, and represented as a function of time. In panel A, separation of DNA fragment P2 and form I DNA (last band) was not complete and migrate as a single band at the bottom in this gel system. However, loading them separately indicated that they are closely spaced bands. First and last lanes represent M13 or pEJ244 DNA in the absence of PI-MtuI. on the cognate substrate in the presence of MgCl 2 (Fig. 8B,  upper panel). The optimal Mg 2ϩ concentration for cleavage activity was ϳ18 mM, and gradually decreased at higher Mg 2ϩ concentrations. In this regard, the D122Y mutation in PI-MtuI resulted in a gain of function effect by acquiring the ability to catalyze cleavage of cognate DNA in the presence of Mg 2ϩ . The metal-mediated inhibition of cleavage activity at high Mg 2ϩ has been noted for several Mg 2ϩ -dependent nucleases (6). What is the mechanism of metal ion-dependent cleavage of DNA? It is known that Asp can coordinate with only Mn 2ϩ , whereas Tyr can coordinate with oxophilic Mg 2ϩ , as well as thiophilic Mn 2ϩ , during catalysis. This partly explains the altered specificity of the D122Y variant to act on cognate DNA in the presence of Mg 2ϩ .
Mapping of PI-MtuI Ectopic Cleavage Sites-To map the cleavage sites to the nucleotide level, we used 58-or 87-bp double-stranded DNA derived from form I M13 DNA. The 32 P-labeled DNA substrates were incubated with wild-type or variant enzymes of PI-MtuI in the presence of Mg 2ϩ as described above. The cleavage products, together with the Maxam-Gilbert chemical sequencing ladder, were separated by electrophoresis on polyacrylamide gels in the presence of urea. Interestingly, wild-type and D122Y variant cleaved both 58and 87-bp substrates at very similar positions (compare Fig. 9,  A and B versus C and D). Overall, the efficiency of cleavage was comparable. On the other hand, D222T and K131L variant enzymes displayed residual amounts of nicking activity in both the upper and lower strands of the 87-bp DNA substrate at several positions upstream of the ectopic site. E220A and K195F variant enzymes showed very weak nicking activity with the 87-bp substrate on both strands. However, except D122Y, the variant enzymes had no observable nucleolytic activity on the 58-bp DNA substrate (Fig. 9, A and B). The base sequence encompassing the cleavage site corresponding to P1   Fig. 9.
The data presented here show that although the flanking sequences around the ectopic sites are very different, the cleavage pattern generated by the wild-type enzyme and D122Y variant with both the substrates is strikingly similar. The influence of flanking sequences on the rate of cleavage of P1 and P2 sites remains to be elucidated. Intriguingly, the hexameric sequence in the recognition site of PI-MtuI is similar to the cleavage site determined for ClaI (27). We searched sequence data bases for sites in phage , pBR322, and pET11d, and the potential ClaI recognition sites were identified. Comparison of the profiles of reaction products generated by PI-MtuI with that of ClaI indicated that the former cleaved at canonical ClaI sites to varying degrees of efficiency, whereas the latter cleaved at equal efficiency (data not shown). As a further test for specificity of their action, we examined the ends generated by these enzymes. Consistent with the previous studies (27), ClaI generated 5Ј-ended termini, whereas PI-MtuI cleaved within its recognition sequences to leave 1-2-bp 3Јhydroxyl overhangs. The DNA ends in the P1 and P2 products generated by PI-MtuI possessed 5Ј-phosphate and 3Ј-hydroxyl groups (data not shown).

DISCUSSION
In previous studies, we showed that PI-MtuI is a novel Mn 2ϩ and ATP-dependent homing endonuclease, which was able to cleave the inteinless recA allele at 24 and 33/43 bases upstream of the intein insertion site, in the upper and lower strands, respectively (18). Here, we show that PI-MtuI possesses an intrinsic ability to cleave ectopic DNA sites within the sequence 5Ј-ATCGAT-3Ј, in the presence of Mg 2ϩ , resulting either in a blunt end or 1-2-nucleotide 3Ј-hydroxyl overhangs. However, the reaction promoted by PI-MtuI with these two substrates are seemingly different. PI-MtuI cleaved only a fraction of cognate substrate even after prolonged incubation, whereas all the substrate was digested in the case of ectopic DNA. The rate at which PI-MtuI cleaved the ectopic DNA sites was much faster than the cognate substrate (this study and Ref. 18). The cleavage reaction profile with ectopic DNA sites indicated a sequential mechanism similar to the substrate bearing the cognate site. It is intriguing how PI-MtuI could recognize and cleave two different sequences using alternative cofactors. Although the crystal structure of PI-MtuI has not yet been determined, the crystal structure of PI-SceI, a LAGLIDADG homing endonuclease, showed that it contains two active sites per monomer (4). If so, each LAGLIDADG motif of PI-MtuI might be involved in recognition and cleavage of cognate or ectopic DNA sites using alternative cofactors.
In the absence of divalent metal ions, PI-MtuI was able to form stable protein-DNA complexes with topologically different forms of M13 DNA under physiological ionic strength. However, binding to single-stranded DNA failed to elicit cleavage in the absence or presence of metal ion, indicating that the singlestranded DNA⅐PI-MtuI complex is not in a favorable conformation for cleavage. The key to our findings is the sensitive assay with which we were also able to decipher the cleavage mechanism promoted by PI-MtuI with ectopic DNA. There are other examples from the LAGLIDADG family of endonucleases, such as PI-SceI (28), PI-PfuI (29), and group II introns (13,14), that promote cleavage of both cognate as well as ectopic DNA sites. Furthermore, group II introns have been shown to promote transposition events into ectopic DNA sites in vitro and in vivo. However, in all these examples the ectopic DNA site closely resembles the natural homing site. In contrast, the ectopic DNA site of PI-MtuI is distinct in its nucleotide sequence, and its cleavage requires a different cofactor.
The observation that PI-MtuI promotes DNA cleavage at ectopic sites in the presence of Mg 2ϩ was unexpected. Consequently, it remained plausible that the display of such an activity could arise from a contaminating endonuclease. Although preincubation of PI-MtuI with anti-PI-MtuI antibodies prior to the endonuclease assay indicated that both the activities are intrinsic to the same polypeptide, the evidence was less direct. We therefore sought a mutagenesis approach to formally resolve this issue. Sequence comparison of PI-MtuI with PI-SceI revealed conserved residues at the active site region. In addition, the availability of the crystal structure of PI-SceI (25) abetted mutagenesis experiments. Based on these analyses, we assumed that PI-MtuI is similar to PI-SceI with an analogous catalytic domain.
In the co-crystal structure of I-CreI the two LAGLIDADG motifs are a part of two catalytic centers (10). Structural and mutational evidence suggests a role for this motif in DNAbinding and phosphodiester bond cleavage. The penultimate Asp residue of the first LAGLIDADG motif is one of the ligands for coordination of metal ion. The role(s) of conserved acidic amino acid residues in DNA cleavage has been ascertained from crystal structure and mutational analysis of I-CreI (30), PI-SceI (31), and PI-PfuI (32). In PI-MtuI, the first dodecapeptide motif bears the sequence: 115 LLGYLIGD*G 123 and, the second, 214 LLFGLFE*SD*G 223 (asterisks denote conserved acidic amino acid residues). We performed mutational analysis of these presumptive metal ion-binding ligands (D122Y, E220A, and D222T) in PI-MtuI. In addition, K195F and K131L variant enzymes were also generated to test their influence on catalysis. We measured the catalytic activity of these variant enzymes in comparison with the wild-type enzyme in the presence of Mg 2ϩ . The PI-MtuI variant enzymes can be grouped according to the severity of the defect: E220A and K195F were inactive in DNA cleavage. Whereas D222T and K131L displayed partial activity by generating form II and form III DNA thus confirming the essential role of these residues in catalysis (Fig. 7B). There is, however, an exception, the D122Y mutation did not affect the cleavage activity thus indicating plasticity of the catalytic center. We further show that the D122Y mutation in PI-MtuI resulted in a gain of function effect by acquiring the ability to catalyze cleavage of cognate DNA in the presence of Mg 2ϩ . On the other hand, E220A and K195F variants were inactive in endonuclease activity implicating these residues in cleavage of ectopic sites in the presence of Mg 2ϩ . Nevertheless, these data suggest that Asp 222 and Glu 220 are likely to be the ligands involved in the Mg 2ϩ -dependent cleavage activity of PI-MtuI. How do these mutations inhibit DNA cleavage? One possibility is that they might promote a particular inefficient conformation in PI-MtuI. The results of our mutational analysis of the active site residues of PI-MtuI closely mimic those observed for PI-SceI. Furthermore, both mutational and crystal structure data implicate the corresponding residues (Asp 218 and Asp 326 ) of PI-SceI to be the residues involved in metal ion binding and hence nucleophile activation (31)(32)(33). Thus, these data provide compelling evidence that Mg 2ϩ , as well as ATP and Mn 2ϩ -dependent endonuclease activities, are intrinsic to PI-MtuI.
Our understanding is less complete regarding the molecular mechanism of lateral transfer of inteins in natural populations. Equally, our knowledge is limited by the paucity of factors thus far shown to be involved in this process. Therefore, the recent identification that bacterial group II introns effectively transpose into ectopic DNA sites (13), and yeast group II intron into sites that resemble the natural homing sites (14) has generated much excitement. The finding that PI-MtuI was able to cleave the ectopic DNA site efficiently has raised the possibility that it might use such an activity to spread through natural populations. Inteins are now seen to be widespread among mycobacteria where they might play an important evolutionary role in restructuring the genome. It is also possible that the ability of PI-MtuI to cleave at ectopic sites may provide defense against intruding DNA molecules. Although the occurrence of such a phenomenon is unclear in regard to M. tuberculosis, it is interesting to note that some eubacteria have evolved to do just that. Because overproduction of PI-MtuI in E. coli had no deleterious effect on growth, it is possible that the sites are masked by the interaction of sequence nonspecific proteins such as HU, IHF, and other DNA-binding proteins. However, it is also possible that activation of PI-MtuI in the presence of Mg 2ϩ under in vitro conditions might only reflect the potential, rather than the primary biological activity under in vivo conditions.