ABSTRACT
The DNA polymerase I (PolI) from Mycobacterium tuberculosis (Mtb) was overproduced in Escherichia coli as an enzymatically active, recombinant protein with or without an N-terminal His-tag. The proteins catalysed both the DNA polymerisation of homo- and heteropolymer template-primers and the 5'-3' exonucleolytic hydrolysis of gapped and nicked substrates but lacked an associated proofreading activity. In accordance with recent predictions [Tabor,S. and Richardson,C.C. (1995) Proc. Natl. Acad. Sci. USA, 92, 6339-6343], both recombinant forms of the M.tuberculosis enzyme were unable to discriminate against dideoxynucleotide 5'-triphosphates and were thus efficiently inhibited by these chain-terminating nucleotide analogues during DNA synthesis. This unusual property might be potentially exploitable in terms of novel anti-mycobacterial drug design. A mutational analysis of 5' nuclease domain residues allowed the roles of nine invariant acidic residues to be evaluated. Acidic side chain neutralisation resulted in a >= 20-fold reduction in activity, with the most profound reduction ( >= 104-fold) being caused by neutralisation of the Asp125, Asp148 and Asp150 residues.
The DNA polymerase I (PolI) from Escherichia coli and its large proteolytic (Klenow) fragment (KF) have been the subject of detailed studies which have provided a prototypical molecular model of DNA-dependent DNA polymerisation (1 -5 ). The X-ray crystal structures of KF and its binary complexes have yielded important insights into the architecture of the primer and nucleotide binding sites (2 ) and have guided on-going mutational studies aimed at elucidating the roles of individual residues in the substrate binding and catalysis of polymerisation (6 ). The N-terminal 5' nuclease domain of PolI has also been well characterised at the level of mechanism and substrate specificity (7 ,8 ), although the nature of the interaction between the 5' nuclease and DNA polymerase domains during nick translation synthesis remains obscure. The eubacterial PolI-associated 5' nucleases share significant sequence homology with the polymerase-independent 5' nucleases from bacteriophages T4 (designated as T4 RNase H), T5 and T7. Six conserved subdomains containing 14 invariant residues have been identified within this family of prokaryotic nucleases (9 ). Two of the invariant arginines were found to be essential for catalysis by Thermus aquaticus (Taq) PolI (10 ) and of the nine invariant acidic residues, the Asp13, which is located in a region homologous to that containing a catalytically important aspartate in bacterial RNase HI (11 ), was also found to be essential for catalysis by E.coli PolI (2 ). The prokaryotic 5' nucleases are also related to the Rad2/xeroderma pigmentosum complementation group G (XPG) family of proteins (12 ) at the level of amino acid sequence (13 ) and function (14 ). The amino acids essential for substrate binding and for catalysis by a member of the RAD2 family, the human flap endonuclease 1 (FEN-1), have been found to include five acidic residues that are invariant in both 5' nuclease families (15 ).
The recently reported X-ray crystal structures of Taq PolI (16 ), T4 RNase H (17 ) and T5 5'-exonuclease (18 ) have yielded important insights into the structure-function relationships within this class of enzymes. The Taq domain structure revealed a cleft at the bottom of which are clustered seven of the invariant acidic residues within a sphere of 7 Å radius (16 ). Six of these residues were implicated in the coordination of one high-affinity and two low-affinity metal ions (Zn2+ at site I and Mn2+ at sites II and III). The pair at sites I and II were sufficiently close (5 Å) to predict a two-metal-ion hydrolysis mechanism for this enzyme analogous to that postulated for DNA polymerases and reverse transcriptases (RTs; 19 -22 ), polymerase-associated 3'-5' exonucleases (2 , 23 -25 ), RNases H (26 , 27 ) and ribozymes (28 ). The 2.06 Å X-ray crystal structure of T4 RNase H showed a high degree of structural similarity to the Taq 5' nuclease over the region containing the invariant residues, but revealed only two Mg2+ ions coordinated at the bottom of a cleft containing all nine of the invariant acidic residues (17 ). The location and coordination motif of the first Mg2+ ion in this structure (site IT4) corresponded closely to that of the site I Zn2+ in the Taq domain, whereas the site IIT4 Mg2+ ion was located in proximity to the Asp200 carboxyl but considerably further removed from the site I metal ion (7 Å). The T5 5'-exonuclease structure revealed a helical arch ideally suited for the threading of single-stranded (ss) DNA, at the base of which is the active site containing two metal ions separated by an even greater distance (8.1 Å; ref. 18 ). The latter two structures have thus cast doubt on the relevance of a two-metal-ion mechanism for the phage nucleases.
We recently reported the cloning and sequence analysis of the polA gene from Mtb (29 ) which is closely related to a homologue occurring at two distinct loci in the Mycobacterium leprae genome (30 ). Although the mycobacterial sequences contained all of the residues essential for the catalysis of DNA synthesis (29 ), a significant sequence deviation was noted in the O helix region of the polymerase domain of the Mtb and M.leprae polypeptides inasmuch as the Phe residue found in all other bacterial PolI sequences examined to date [Phe762 of E.coli PolI (6 )] was replaced by a Tyr [Tyr737 of Mtb (29 ) and Tyr740 of M.leprae PolI (30 )]. Based on the recent study of Tabor and Richardson (31 ), the existence of a Tyr at this position in the O helix was predicted to abolish the ribose ring selectivity of the enzyme and render it highly sensitive to inhibition by chain- terminating nucleotide analogues. Since this prediction has obvious implications for the design of novel anti-mycobacterial agents, we were prompted to analyse the Mtb enzyme, and in this paper we describe the overproduction of two recombinant forms of this protein for use in nucleotide discrimination studies. We also report the use of these enzymes in a mutational study of selected 5' nuclease residues aimed at identifying functionally and/or structurally important side chains in this domain.
Escherichia coli CJ236 was kindly provided by T. Kunkel and E.coli HMS174, BL21(DE3) and BL21(DE3)pLysS, for use in the pET expression system (32 ), were obtained from Novagen (Madison, WI, USA). Restriction enzymes, dNTPs, ddATP and poly(dA) were from Boehringer-Mannheim and dT12-18 was from Pharmacia. Radionucleotides were from New England Nuclear. Oligodeoxyribonucleotides (oligos) were synthesised by Genosys Biotechnologies (The Woodlands, TX, USA). HE19-NB (5'-CGGCTTGGCTGGATCCTCGC), HE19-BX1 (5'-GTCCCGGATACGCCGGCC), HE19-BX2 (5'-AGCTGGATCTGCTAGCCAG), HE19-CB (5'-TCCCGCCCCGGATCCACGCTCGG) and HE19- NCOI (5'-TAGTCACCATGGCTCACAC) were used to create expression cassettes of the polA gene encoding Mtb PolI (29 ). D21N (5'-AATTGCCATTCAGCAACATC), A70V (5'-CGTCGAAAGCCACCGCGATGTG), A71V (5'-ACACGTCGAAAACCGCCGCGAT), AA7071VV (5'-CGTCGAAAACCACCGC- GATGT), D73N (5'-CGGGACACGTTGAAAGCCGC), Q82E (5'-CGGGTAGCGTTCCAAGCGGAA), E123Q (5'-CGTCGGCCTGGAACCCCGG), D125N (5'-TGAGGTCGTTGGCCTCGAA), D126N (5'-CGATGAGGTTGTCGGCCTC), D148N (5'-CGTCACGATTCCCGGTGAC), D150N (5'-GCAGTGCGTTACGATCCCC), D199N (5'-CGCTGGGGTTGCCGCGCAG), D202N (5'-GCAGGTTATTGCTGGGGTC) and Y737F (5'-AGCCAGCCCGAAGGACATCG) were used for mutagenesis of the polA gene. D498N (5'-ATAGTAACAAACTCACAAT), RT6 (5'-GGATTCAGGATTAGAAG) and RT7 (5'-GGATTCAGGATTAGAAGTAAAC) were complementary to M13-RTD443Q/D498N, an M13 subclone of a RNase H domain mutant of HIV-1 reverse transcriptase (33 ; denoted M13-RT herein) and were used to construct the double-stranded (ds) DNA substrates used in the 5' nuclease assays.
The cloning and sequencing of the Mtb polA gene has been previously described (29 ). The 3.05 kb C-terminal HindIII-EcoRI fragment of pPH1 was subcloned in M13mp19 to produce M13-HE19. Site-directed mutagenesis of M13-HE19 was carried out as previously described (33 ,34 ) using the method of Kunkel et al. (35 ) and Sanger DNA sequencing (36 ) was performed using Sequenase 2.0 (US Biochemical). The coding region of the polA gene was engineered as a BamHI fragment by introducing new sites at the N- and C-termini of the gene and removing the existing BamHI sites within the gene by mutagenesis of M13-HE19 using HE19-NB, HE19-CB, HE19-BX1 and HE19-BX2 to produce M13-HE194M. The HE19-NB primer created a new BamHI site for in-frame fusion of the pET15b vector to the Ala13 residue of the Mtb PolI. This position was chosen on the basis of a multiple PolI amino acid sequence alignment which suggested that the first 12 residues of the Mtb enzyme would not be required for enzymatic activity. The HE19-CB primer created a new BamHI site 11 bp downstream of the stop codon of the gene and the HE19-BX1 and HE19BX2 primers removed the two internal BamHI sites with silent codon changes [Arg662 (CGG -> CGA) and Ala388 (GCG -> GCT), respectively]. The 2697 bp BamHI fragment from M13-HE194M was cloned in pET15b to create pPH3His, which encoded a fusion protein with the N-terminal sequence: MGSS(H)6SSGLVALPRGSHMLEDPAKTPL.., where the italics denote the Mtb PolI sequence, commencing at Ala13. To control for unpredictable effects of this N-terminal His-tag on the catalytic properties on the enzyme, a second construct encoding a non-tagged version of the protein was created by introducing an NcoI site in HE19CB/BX1/BX2 using HE19-NCOI and cloning the resulting 2727 bp NcoI-BamHI fragment in pET15b to create pPH3. This plasmid encoded a protein, V1M, which differed from the native in only the first amino acid (Met in place of Val). Gene expression in E.coli BL21(DE3)pLysS was induced by the addition of isopropyl- [beta]-d-thiogalactopyranoside (IPTG) and incubation at either 30 (pPH3) or 37oC (pPH3His) for a period of 3 h. SDS-PAGE analysis of the induction products confirmed the high-level overproduction of proteins of the predicted size (100 kDa; Fig. 1 ). Furthermore, the soluble fractions recovered post-induction were found to contain >= 103-fold higher DNA polymerase activity than pre-induction controls.
In the case of the 5' nuclease domain mutants, the mutated region was excised as a 1341 bp HindIII-Asp718 fragment and was re-ligated to the 8.5 kb HindIII-Asp718 fragment from wild-type M13-HE194M to re-construct the mutant. For the Y737F mutant, the mutation was excised as a 1700 bp Asp718-EcoRI fragment and was re-ligated to the 8.14 kb Asp718-EcoRI fragment from wild type M13-HE194M. Confirmatory sequence analysis was minimised by including these additional cloning steps.
SDS-PAGE analysis of induced cell extracts suggested that the overall expression levels of the wild-type and mutant proteins directed by pPH3 and by pPH3His were similar. However, the solubility of the non-tagged wild type and D125N proteins was significantly greater than that observed for their His-tagged counterparts (Fig. 1 ). This difference suggested that the presence of the His-tag favoured the formation of inclusion bodies at 37oC in contrast to the preferential partitioning of the non-tagged proteins in the soluble fraction at the lower induction temperature. However, even within the family of wild type and mutant His-tagged proteins, the partitioning between soluble and insoluble fractions varied between extremes of 10-15% for A70V/A71V and 50% for the wild type (data not shown). The His-tagged wild type, mutants and a control from induced cells harbouring vector alone were purified as follows. Induced cell pellets from 500 ml cultures, resuspended in 10 ml Novagen binding buffer [Buffer A; 20 mM Tris-HCl (pH 7.9), 0.5 M NaCl, 5 mM imidazole] were lysed by one freeze-thaw cycle. The soluble fraction recovered by centrifugation (30 min at 25 000 g) was purified by batch-wise metal chelate affinity chromatography using 5 ml resin. The resin was washed twice each with 10 ml Buffer A and salt-free wash buffer [Buffer B; 60 mM imidazole, 20 mM Tris-HCl (pH 7.9)] before eluting the bound protein with 10 ml salt-free elute buffer [Buffer C; 1 M imidazole, 20 mM Tris-HCl (pH 7.9)]. This procedure resulted in the recovery, in a significantly purified form (~50%), of full-length protein in all cases except D73N, E123Q and D126N, where the eluate was found to consist instead of a wide range of smaller polypeptides. However, the extent of degradation during purification was markedly reduced by including CompleteTM protease inhibitor cocktail (Boehringer Mannheim) in all buffers, implying that these three mutants were particularly prone to proteolysis. The inhibitors were therefore included in all buffers for the purification of all of the enzymes. The co-purification of a minor E.coli nuclease contaminant with the His-tagged recombinant (37 ) necessitated further purification by anion exchange chromatography. The affinity column eluate was therefore loaded on a 5 ml Q Sepharose (Fast Flow, Sigma) column which was washed with 20 ml Buffer C containing 0.25 M NaCl and eluted with an 80 ml linear gradient from 0.25 to 0.5 M NaCl. The Mtb PolI eluting at 0.3-0.35 M NaCl was 80-95% homogeneous. Peak fractions were concentrated 10-fold by centrifugation over Ultrafree MC filters (Millipore) and were stored at 0.05-0.1 mg/ml in a storage buffer [10 mM Tris-HCl (pH 7.9), 1 mM dithiothreitol (DTT), 0.15 M NaCl, 0.5 M imidazole, 50% glycerol]. The activity loss and protein degradation were negligible after 3 months storage at -20oC for all of the proteins except for D126N which degraded rapidly (t0.5 = 5 d).
The non-tagged wild-type and D125N mutant proteins were purified as follows. Induced cell pellets from 2.5 l cultures, resuspended in 100 ml salt-free Buffer A and containing protease inhibitors, were lysed as described above and clarified by ultracentrifugation (30 min at 110 000 g). The high-speed supernatant was loaded on a 100 ml phospho-cellulose (Whatman P11) column equilibrated in salt-free Buffer B. The column was washed with 40 ml Buffer B and was eluted with a 100 ml linear gradient from 0 to 0.5 M NaCl in Buffer C. The peak Mtb PolI-containing fractions eluting at 0.2-0.3 M NaCl were pooled and further purified to >95% homogeneity on a 5 ml Q Sepharose column eluting with an 80 ml linear gradient from 0.25 to 0.5 M NaCl. Peak fractions were concentrated as described above. In all cases, proteins were diluted in 50 mM Tris-HCl (pH 7.0), 150 mM NaCl, 1 mM DTT and 50% glycerol immediately before use in enzyme assays and their concentrations were determined using the BioRad protein assay (Kit I, bovine gamma globulin standard).
Assays (11 [mu]l) containing 700 nM 3'-OH [poly(dA)[middot]oligo (dT)12-18; A:T nucleotide ratio = 860:1], 33 [mu]M [3H]TTP (2500 c.p.m./pmol), 8 mM MgCl2, 5 mM DTT, 8 mM KCl, 15 mM NaCl, 5% glycerol and 16-33 nM PolI in 50 mM Tris-HCl (pH 8.3) were incubated at 37oC for 1-20 min. Samples were quenched and processed by a DE81 filter binding assay (34 ). For polymerase assays of E.coli PolI and KF, reactions were carried out at 37oC in 50 mM Tris-HCl (pH 7.4) containing 2 mM MgCl2.
5'-[32P]-D498N was annealed to ss M13-RT at a molar ratio of 1:3. Polymerisation was followed by the gel electrophoretic separation of products generated using a nucleotide mix optimised for T7 DNA polymerase [ddNTP/dNTP ratio of 0.1 (8 [mu]M ddNTP, 80 [mu]M dNTPs); Sequenase version 2.0 kit (US Biochemical)]. Assays containing 25 mM Tris-HCl (pH 8.0), 5 mM MgCl2, 5 mM DTT, template-primer (33 nM 3'-OH), nucleotide mix and enzyme (30-150 nM) were incubated at 37oC for 15 min. Aliquots (1.5 [mu]l) were quenched with 4 [mu]l sequencing gel sample loading buffer and fractionated by electrophoresis in 6% polyacrylamide, 7 M urea, 1* TBE [90 mM Tris, 64.6 mM boric acid, 2.5 mM EDTA, pH 8.3] sequencing gels which were exposed at -70oC for 18 h.
Assays (5 [mu]l) containing 90 nM 5'-[32P]-D498N in 5 mM MgCl2, 2 mM DTT and 50 mM Tris-HCl (pH 8.0) were initiated by the addition of enzyme and incubation at 37oC for 0-40 min. Aliquots (1.5 [mu]l) were quenched and fractionated in 16% sequencing gels, as described above.
5'-[32P]-D498N was 3'-ddA-terminated in a 120 [mu]l reaction containing ss-M13-RT/[5'-32P]-D498N (29 nM 3'-OH, ss DNA:primer = 4.3:1), 100 [mu]M ddATP, 6.7 mM DTT, 33 mM NaCl, 13 mM MgCl2 and 10 U Sequenase version 2.0 in 25 mM Tris-HCl (pH 7.5) incubated at 37oC for 1.3 h. The reaction was stopped by PhOH/CHCl3 extraction and was diluted with 680 [mu]l water. Aliquots (100 [mu]l) were purified by centrifugation over 1 ml Sephadex G25 spin columns. The hybrid, M13-RT/(3'-ddA/ 5'-[32P])-D498N, was resuspended in 60 [mu]l water (47 nM hybrid; 80% recovery) and partitioned into three equal aliquots. The substrates, Gap-[infinity], Gap-5 and Gap-0, were prepared by annealing either water (Gap-[infinity]) or a 17-fold molar excess of the appropriate unlabelled upstream primer (RT6 or RT7) to this hybrid by heating to 70oC for 1 min and cooling over 1.5 h to 37oC in a buffer containing 48 mM Tris-HCl (pH 8.0) and 4.8 mM MgCl2. Since equivalent concentrations of the various upstream primers were annealed to equal aliquots of the precursor hybrid, the final concentrations of the DNA substrates were the same in all cases. The gap sizes denote the spacing in nucleotides (nt) between the 3'-terminus of the upstream primer ([infinity], 5 or 0 nt) and the scissile 5'-end of the downstream 3'-ddA-terminated D498N primer.
The mutant complementation vector, pMTBPOLAV1M, was constructed with the 3.05 kb HindIII-EcoRI fragment from HE19NCOI/CB/BX1/BX2 (harbouring the V1M mutation) being used in place of the corresponding fragment from HE19CB that had been used to create the wild type counterpart, pMTBPOLA (38 ). The ability of the mutant vector to complement the plasmid replication defect of the polA755::aph mutant of M.smegmatis (39 ) was analysed by comparing its transformation frequency to that of the pMTBPOLA (wild type) control (38 ). All experimental procedures were carried out as previously described (38 ).
The plasmid pPH3His directed the overproduction of a His-tagged protein of limited solubility in E.coli. However, the protein released in the soluble fraction of the lysate was associated with DNA-dependent DNA polymerase activity and could be purified to homogeneity by a combination of affinity and ion exchange chromatography (Fig. 1 A). Since the N-terminal His-tag extension might conceivably affect the enzymatic properties of the protein, a second recombinant form of the enzyme, which differed from that predicted from the gene sequence in only the first amino acid, was also over-expressed. This recombinant form was highly soluble at an induction temperature of 30oC and was purified to >95% homogeneity by ion exchange chromatography (Fig. 1 B).
The effect of the V1M mutation on the DNA polymerase activity of Mtb PolI, expressed in a heterologous mycobacterial host under the control of its own promoter, was analysed by studying the ability of the V1M polA gene to complement the plasmid replication defect of a polA mutant of M.smegmatis insertionally inactivated within its DNA polymerase domain (38 ). The Mtb V1M polA gene and flanking sequences were cloned in pOLYG (a shuttle plasmid harbouring the pAL5000 origin of replication and a hygromycin resistance gene) to yield the V1M mutant form of the previously described wild-type complementation vector, pMTBPOLA (38 ). The vector, pMTBPOLAV1M, transformed M.smegmatis polA755::aph to hygromycin resistance with ~30% of the frequency observed for its wild type counterpart (0.3 * 104 c.f.u./[mu]g; ref. 38 ), confirming that it retained the ability to direct the expression of functional PolI from the M.tuberculosis polA promoter in vivo. Since the DNA polymerase activity of PolI plays a well established role in the initiation of both continuous- and discontinuous-strand synthesis in the replication of plasmids such as ColE1 (40 ) and pAL5000 (39 ), we propose that these data support the validity of using the recombinant V1M protein as a surrogate for studying the biochemical properties of the DNA polymerase activity of the native enzyme. However, we emphasise that this argument is speculative, since it rests on the assumption that the translational initiation site of the Mtb polA gene in both M.tuberculosis and M.smegmatis is indeed at this position [Val in the wild type (29 ) versus Met in the mutant].
The DNA polymerase activities of the recombinant Mtb PolI enzymes were studied using a homopolymer template-primer. The pH and [Mg2+] optima for the His-tagged protein were found to be 8.4 and 8 mM, respectively (data not shown). The kcat values of the His-tagged and non-tagged forms were 0.4 s-1 and 0.6 s-1, respectively, compared to 1.5 s-1 for E.coli PolI. Since the extent of contamination of the recombinant Mtb PolI preparations with inactive protein is unknown, the significance of these relatively low values is unclear. However, the similar kinetics of synthesis by the two recombinant forms suggested that the artificial N-terminal structure of the His-tagged recombinant [27 amino acid (His)6-bearing peptide fused to Ala13] had a negligible effect on the polymerase activity of the enzyme. This observation is in accordance with the well established fact that the 5' nuclease and DNA polymerase domains of the E.coli and Taq PolI enzymes are distinct and separate folding entities (5 ,41 ). We therefore concluded that the DNA polymerase properties of both recombinant Mtb PolI enzymes would reflect, equally well, those of the native enzyme at a qualitative (but not necessarily quantitative) level.
An interesting feature of the PolI enzymes from Mtb (29 ), M.leprae (30 ) and M.smegmatis (V. Mizrahi and A.R. De Meyer, unpublished) is the presence of a Tyr residue at a position occupied by a Phe in other bacterial homologues in a region involved in dNTP binding (O helix; Phe762 in E.coli PolI; ref. 6 ). We therefore surmised that DNA synthesis catalysed by Mtb PolI might be highly susceptible to inhibition by chain-terminators. This hypothesis was tested by monitoring the effect of ddNTPs on the extension of a 5'-end labelled primer annealed to a ss M13 DNA template. To control for potential complications arising from 5'- nuclease-catalysed product degradation, the 5'-nuclease-deficient D199N mutant was analysed in parallel with the wild-type His-tagged enzyme in comparative assays with KF and T7 DNA polymerase using nucleotide mixes optimised for DNA sequencing with the latter enzyme (Fig. 2 ). In contrast to the high discrimination against ddNMP incorporation demonstrated by KF, the His-tagged Mtb PolI was even more susceptible to inhibition of synthesis by ddNMP incorporation than was T7 DNA polymerase. Furthermore, ribose ring discrimination was predictably introduced into the nucleotide binding site of Mtb PolI by Tyr737 -> Phe mutagenesis. Interestingly, the Y737F and Y737F/D199N mutations were poorly tolerated at the level of DNA polymerase activity (6-fold lower than wild type; data not shown), but for the reasons outlined above, the significance of this observation is unclear. The lack of effect of the His-tag on the dideoxynucleotide susceptibility of the DNA polymerase activity of the Mtb enzyme was confirmed by analysing the effect of chain terminators on DNA synthesis catalysed by the non-tagged recombinant. A comparison of panels C and H suggested that both forms of the enzyme were indeed equally susceptible to inhibition by ddNTPs.
Exonuclease activity analysis of both recombinant forms of the Mtb PolI enzyme using a 5'-end-labelled ss oligo substrate showed negligible 3'-5' degradation, suggesting that the mycobacterial enzyme lacked an associated proofreading activity (Fig. 3 ). This property contrasts that of E.coli PolI and KF, both of which showed the expected progressive accumulation of 3'-derived hydrolysis products (Fig. 3 ; ref. 5 ). The lack of hydrolysis by the Mtb enzyme is consistent with the absence of conserved ExoI-III subdomains containing the four critical acidic residues required for 3'-5' exonuclease activity (42 ).
The His-tagged mutant enzymes were purified according to the same procedure as that developed for the corresponding wild type protein (data not shown). As shown in Table 1 , the mutations had little or no effect on the specific activity of the DNA polymerase. The 5' nuclease activities of the mutants were then compared to the wild-type enzyme using the nicked (Gap-0) substrate. Although the A70V, A71V, A70V/ A71V and Q82E mutations had no detectable effect on activity, neutralisation of each of the nine invariant acidic residues reduced the activity by a factor of at least 20-fold with the D125N, D148N and D150N mutations resulting in the most profound reduction in rate ( >= 104-fold; Table 1 ). The non-tagged form of the D125N mutant was also purified and analysed and as shown in Table 1 , the two forms of this mutant were enzymatically indistinguishable. This observation strongly supported the validity of the remaining mutagenesis data that were obtained using the His-tag expression system alone.
The two recombinant proteins described in this study have been used to analyse selected properties of the native Mtb PolI enzyme. As described above, the ability of the non-tagged, V1M recombinant to complement the plasmid replication defect of a polA mutant of M.smegmatis strongly supported the validity of using this protein as a surrogate for studying the DNA polymerase activity of the native enzyme. Since in vitro studies suggested that the DNA polymerase activity of the His-tagged recombinant was indistinguishable from that of its non-tagged counterpart, we speculate that this form of the enzyme would also function as a DNA polymerase in vivo.
Table 1
These recombinants were used to confirm the prediction the native enzyme would be unable to discriminate against ddNTP analogues during DNA synthesis (31 ). The Mtb enzyme is therefore the first bacterial PolI described to date that possesses this unusual property. This observation is analogous to that of Eriksson et al. (43 ) which confirmed that the yeast mitochondrial DNA polymerase utilised the chain-terminating anti-HIV agents, ddTTP, 3'-fluoro-TTP and ddCTP almost as efficiently as natural dNTPs. This finding is in turn, consistent with the fact that the mitochondrial polymerase, which is a member of the type I DNA polymerase family, also bears a Tyr residue at the ribose ring discrimination position (6 ,31 ). The high degree of primary structural conservation within bacterial PolI enzymes suggests that they may have evolved from a common progenitor which contained a Tyr at the discrimination position. Exposure of rapidly-growing bacteria to naturally occurring chain terminators might have provided the selective pressure necessary for the organisms to evolve a way of discerning the ribose ring structure of the incoming nucleotide. Since the indiscriminate incorporation of chain terminators has potentially catastrophic consequences for bacterial survival, it is reasonable to speculate that the retention of a Tyr residue at this position in the PolI enzymes from both slow- and fast-growing mycobacteria is more likely due to the formidable barrier to entry of naturally occurring chain terminators provided by the unusual mycobacterial cell wall (44 ) rather than to the intrinsically slow growth rate of these organisms, as has been suggested [ref. 31 ; the doubling times of M.smegmatis, Mtb and M.leprae are 2-3 h, 18-24 h and 13 d, respectively (45 )].
The unusual DNA polymerase enzymology of mycobacterial PolI has obvious implications for anti-mycobacterial drug design. PolI is a repair enzyme required during the growth phase of an organism, since the repair of damage-induced lesions must occur between successive rounds of chromosomal DNA replication. In light of the potential lethality of persistent, unrepaired ss breaks, it is possible that this enzyme is also expressed during the dormant phase of Mtb. If this hypothesis is correct, then PolI may be a useful target for chemotherapeutic intervention against dormant tubercle bacilli that are not susceptible to conventional mycobactericidal drugs since these all act specifically against dividing cells. The wide variety of chain-terminating nucleoside analogues developed for anti-viral chemotherapy may therefore include compounds with potential anti-mycobacterial activity. However, the efficacy of such molecules in vivo may well be limited by the extent of drug uptake and subsequent metabolism to the triphosphate form.
The 5' nuclease domain of Mtb PolI provided a useful system for analysing structure-function relationships in 5' nucleases in general. As described above, the efficient 5'-3' hydrolysis of a nicked substrate catalysed by the His-tagged Mtb enzyme provided an appropriate assay for this study. The fact that neutralisation of invariant 5' nuclease domain acidic residues had a negligible effect on the DNA polymerase activity of Mtb PolI argued strongly in favour of independently folded polymerase and 5' nuclease domains, as has been found for the E.coli (5 ) and Taq PolI counterparts (41 ). However, this same property negated the value of using the DNA polymerase activity as a surrogate marker for mutagenesis-induced global folding defects and hence limited the conclusions that could be drawn from this study to the following:
(i) The effects of neutralisation of the Asp21, Asp73, Asp125 and Asp148 residues were consistent with their involvement in the coordination of a catalytic site I metal ion, on the following grounds: (a) the requirement of the Asp21 residue for catalysis is in agreement with the functional importance of E.coli PolI Asp13 (2 ) and FEN-1 Asp34 (15 ), as established by mutagenesis. The Taq PolI Asp18 is within coordinating distance of the site I Zn2+ ion (16 ) and the T4 RNase H Asp19 is bridged to the site IT4 Mg2+ ion through a water molecule (17 ). Furthermore, this invariant acidic residue is located in a sequence context related to that found in other nucleases (2 ), suggesting a functional homology between this residue and the E.coli RNase HI Asp10 (11 ) and HIV-1 RNase H Asp443 (33 , 34 ) which participate in coordinating the catalytic metal ion in those enzymes. (b) The role of the T4 RNase H Asp73 residue in site I metal ion coordination has been established (17 ) and since the FEN-1 Asp86 is also implicated in catalysis (15 ), the severely reduced activity of the D73N mutant is thus consistent with weakened or loss of site I metal ion binding. (c) The Taq PolI and T4 RNase H counterparts of Asp125 and Asp148 are in close proximity to the site I metal ion, suggesting that these carboxylates participate as ligands for its coordination. However, for reasons that are unclear, the functional homology in the case of the Asp148 does not extend to the FEN-1 protein since its counterpart (Asp179) could be neutralised without effect (15 ).
(ii) The high degree of functional conservation of carboxylates involved in site I metal ion coordination in T4 RNase H and Mtb PolI did not extend to the putative site II metal ion ligands, Asp150 and Asp202, since the T4 RNase H D157N and D200N mutants retained detectable and wild type activity respectively (17 ), whereas the Mtb PolI D155N and D202N mutants and the corresponding FEN-1 D181A and D233A mutants (15 ) all had sharply reduced activity. The reasons underlying these differences are presently unclear. However, by analogy with the proposed role of the FEN-1 Glu158 and Asp233 residues (15 ), we speculate that the Glu123, Asp150 and Asp202 residues of Mtb PolI are involved in the positioning of a second metal ion which may play an indirect role in substrate binding.
(iii) The extensive degradation of certain mutants during purification in the absence of protease inhibitors suggested that conformational perturbation may have been caused by neutralisation of the Asp126, Asp73 and Glu123 residues, resulting in increased susceptibility of the corresponding polypeptides to degradation by E.coli proteases. By analogy with T4 RNase H, the Asp126 residue might participate in a H-bonding interaction with the invariant Arg80 residue, and since the T4 RNase H Asp71 and Glu130 are involved in the positioning of water molecules around the Mg2+ ions, neutralisation of Asp73 and Glu123 may have resulted in structural destabilisation by weakening the metal ion binding and exposing uncoordinated carboxylates to intramolecular electrostatic repulsion.
(iv) The lack of effect of mutations of the conserved Ala70, Ala71 and Gln82 Mtb PolI residues confirmed that they are neither functionally nor structurally important. `Subdomain B' should therefore be re-defined as a single residue corresponding to Asp63 of E.coli PolI. Moreover, since the conserved acidic residue in subdomain C [Glu72 of E.coli PolI (46 )] can be neutralised without loss of function, a carboxylate is obviously not required in this subdomain.
The authors gratefully acknowledge the financial support of the Glaxo International Tuberculosis Research Initiative, the South African Medical Research Council, the South African Institute for Medical Research and the Foundation for Research Development. We thank B. Gordhan for performing the complementation experiments. We are deeply indebted to T. Mueser, C. Hyde, N. Nossal and M. Bhagwat for providing us with data prior to publication and for helpful structural insights. We also thank C. Joyce and K. Duncan for advice and encouragement and anonymous reviewers for constructive comments.
*To whom correspondence should be addressed. Tel: +27 11 489 9370; Fax: +27 11 489 9001; Email: 075val@chiron.wits.ac.za
Enzymea
5' Nuclease (%)b
DNA polymerase (%)c
His-tagged
Wild type
100
100
D21N
0.1
90
A70V
100
100
A71V
100
100
A70V/ A71V
100
100
D73N
0.3
90
Q82E
100
100
E123Q
0.5
90
D125N
<= 0.01
90
D126N
5
90
D148N
<= 0.01
100
D150N
<= 0.01
100
D199N
5
100
D202N
0.1
100
Non-tagged
Wild type
100
100
D125N
<= 0.01
100
REFERENCES
Return