The Vsr mismatch endonuclease recognises the sequence CTWGG (W = A or T) in which the underlined thymine is paired with guanine and nicks the DNA backbone on the 5′-side of the mispaired thymine. By using base analogues of G and T we have explored the functional groups on the mismatch pair which are recognised by the enzyme. Removal of the thymine 5-methyl group causes a 60% reduction in activity, while removing the 2-amino group of guanine reduces cleavage by 90%. Placing 2-aminopurine or nebularine opposite T generates mismatches which are cut at a much lower rate (0.1%). When either base is removed, generating a pseudoabasic site (1′,2′-dideoxyribose), the enzyme still produces site-specific cleavage, but at only 1% of the original rate. Although TT and CT mismatches at this position are cleaved at a low rate (~1%), mismatches with other bases (such as GA and AC) and Watson–Crick base pairs are not cleaved by the enzyme. There is also no cleavage when the mismatched T is replaced with difluorotoluene.
Received March 13, 2000; Revised and Accepted May 12, 2000.
In Escherichia coli the dcm protein catalyses methylation of the underlined cytosine residue in the sequence CCWGG (W = A or T) (1–3). Since 5-methylcytosine is prone to hydrolytic deamination producing thymine, GT mismatches can be formed at this position, which if not repaired cause transition mutations. GT mismatches arising by this mechanism are corrected by the VSP repair pathway (very short patch DNA repair synthesis), which depends on the genes vsr and polA (4,5) and is strongly stimulated by MutL and MutS (6–8). The vsr gene product (Vsr, Mr 18 000) is a DNA mismatch endonuclease which nicks double-stranded DNA within the sequence CTWGG (9). Cleavage occurs on the 5′-side of the underlined thymidine residue which is mismatched to 2′-deoxyguanosine, leaving a 5′-phosphate group on the thymidine. The incision is sequence-dependent, mismatch-dependent and strand-specific (9). The GT mismatch repair activity of the E.coli Vsr endonuclease has been investigated in a variety of sequence contexts differing in one or two positions from the canonical pentanucleotide sequence, such as CTWGN and NTWGG (10). All were processed by the enzyme, though at lower rates. A crystal structure of the Vsr protein has been determined (11), as too has the structure of the enzyme bound to a cleaved oligonucleotide (12). This crystal structure reveals a novel recognition mechanism in which three aromatic residues intercalate from the DNA major groove deforming the local base pair stacking. The mismatched guanine is contacted by Lys89 in the major groove and by the main chain carbonyl of Met14 in the minor groove, while the only specific contact to the thymine which could differentiate it from cytosine is made by Asn93. Although there is little specific recognition of the mismatched base pair there is a good steric match with the protein surface. Thr19, which is positioned in the minor groove, prevents interaction with a normal Watson–Crick base pair but allows binding to GT since the thymine is displaced towards the major groove in this wobble base pair. Phe67 stacks directly onto the exposed portion of the mismatched thymine in the major groove.
Since the GT mismatch consists of two natural DNA constituents it is devoid of markers for identifying the damaged DNA strand. In this study we have used several base analogues to examine the role of different functional groups on the GT mismatch for recognition by Vsr endonuclease. There have been several structural studies on GT mismatches (13–15), which are known to form a wobble base pair in which the guanine 2-amino group is unpaired (Fig. 1A). We have examined the importance of this group by replacing guanine with inosine (Fig. 1B), which is able to form a similar base pair, but lacks the exposed amino group. In contrast, 2-aminopurine (2AP) (Fig. 1C) retains this amino group but is unable to form a wobble pair with thymine. Nebularine (deoxypurine, Fig. 1D) lacks any of these functional groups. The role of various groups of the thymine residue has been examined using deoxyuridine (lacking the 5-methyl) and 8-oxoguanine (Fig. 1E), which in the syn configuration presents the same arrangement of hydrogen bond donors and acceptors as thymine and has the potential for forming a similar wobble base pair with G. It has previously been shown that VSP repair can correct GU as well as GT mismatches (16). We have also examined the importance of either base by replacing G or T with an abasic site (1′,2′-dideoxyribose, ϕ).
MATERIALS AND METHODS
Preparation of Vsr endonuclease
Full details of the cloning and purification of the Vsr endonuclease will be published elsewhere. Briefly, the Vsr gene was prepared by PCR synthesis from 50mer oligonucleotides according to the method of Stemmer et al. (17). This PCR fragment was cloned between the NdeI and BamHI sites of plasmid pET15b (Novagen), generating plasmid pET-VSR. Vsr endonuclease was expressed as a His-tagged protein and purified on Ni–NTA resin (Qiagen). The purified protein was estimated to be >95% pure as judged by SDS–PAGE and stained with Coomassie blue. The protein concentration was estimated from the A280, using an extinction coefficient of 31 010 M–1 cm–1 determined by the method of Gill and von Hippel (18). This was stored at –20°C in 50 mM HEPES/NaOH buffer pH 7.6 containing 500 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA and 15% glycerol at a concentration of 25 µM.
A 50 bp DNA fragment was used for these cleavage studies, containing the Vsr recognition sequence (CTAGG/CCTGG) in the centre (Fig. 1F). DNA oligonucleotides were synthesised on an Applied Biosystems 394 solid phase DNA/RNA synthesiser on a 0.2 or 0.04 µmol scale. The unmodified phosphoramidite monomers, DNA solid supports and other reagents were purchased from Applied Biosystems. Inosine, deoxyuridine, 2AP, nebularine and 8-oxoguanine phosphoramidites were purchased from Glen Research. Difluorotoluene β-cyanoethylphosphoramidite (19,20) and 1′,2′-dideoxyribose phosphoramidite (21) were prepared as previously described. Oligonucleotides were purified by reversed phase HPLC.
Approximately 40 pmol of the T-containing oligonucleotide strand (upper sequence of Fig. 1) was radiolabelled with 20 µCi (6.7 pmol) [γ-32P]ATP (3000 Ci/mmol; Amersham) using polynucleotide kinase. The radiolabelled DNA was purified on a 10% polyacrylamide gel containing 8 M urea. This was then annealed with an excess (20 pmol) of the unlabelled complementary strand in 10 mM Tris–HCl pH 7.5 containing 100 mM NaCl and the resulting duplex was then further purified on a non-denaturing 10% polyacrylamide gel. This lengthy purification procedure ensures that there is no single-stranded DNA (labelled or unlabelled) in the reaction mixtures. The purified DNA was dissolved in 400 µl of 10 mM Tris–HCl containing 0.1 mM EDTA in which we estimate the final strand concentration to be <40 nM. For the single turnover experiments described in this paper the absolute DNA concentration is not important, so long as it is much lower than that of the enzyme.
Vsr endonuclease was diluted to working concentrations in 10 mM HEPES/NaOH pH 7.5 containing 100 mM NaCl, 10 mM MgCl2 and 100 µg/ml acetylated bovine serum albumin (BSA) (Promega). Cleavage was initiated by adding 20 µl of this enzyme (typically at a concentration of 250 nM) to 10 µl of radiolabelled DNA. Samples (3.5 µl) were removed from the reaction at various times (typically up to 40 min) and added to 4 µl of 80% formamide containing 10 mM EDTA and 0.1% (w/v) bromophenol blue. When examining the cleavage of oligonucleotides containing different mismatches or base modifications the fragment containing the GT mismatch was always included as an internal control. Samples were denatured by heating at 100°C for 3 min, before cooling on ice and loading onto a 13% polyacrylamide gel containing 8 M urea. Gels were run at 1500 V for 2 h before fixing in 10% acetic acid, transferring to Whatman 3MM paper and drying under vacuum at 80°C. Dried gels were exposed to phosphorimaging using a STORM 860 phosphorimager.
The intensity of bands in each digest was estimated using ImageQuant software. The amount of digested material at each time point was expressed as the fraction of the total radioactivity (cut and uncut species). A simple exponential curve was fitted to these data using FigP for Windows. The cleavage rate constant for each modified oligonucleotide was expressed as a percentage of that measured with the GT mismatch. In early experiments, in which BSA was not included in the reaction mixture, we found that although the reaction profiles were described by an exponential curve, this extrapolated to less than 20% cleavage and the extent of cleavage was lower for poorer substrates. In these instances the cleavage of different substrates was compared by estimating the initial reaction rates. The values for the relative cleavage of different substrates obtained in this way were identical to those obtained from analysis of full cleavage in the presence of BSA. A similar procedure using the initial reaction rates was used to compare poor Vsr substrates for which the total cleavage was less than 1% of the total material.
In the experiments described in this paper the enzyme concentration was always in excess of the substrate so as to achieve single turnover kinetics. Under these conditions the cleavage rate should be independent of enzyme concentration. However, if the interaction is weak then all the DNA may not be bound by the enzyme at the start of the experiment, resulting in multiple turnover kinetics. We therefore examined the rate of the reaction as a function of enzyme concentration. We found that for the substrate containing the GT mismatch, changing the enzyme concentration from 165 nM to 1.65 µM increased the reaction rate by less than 15%, while decreasing the enzyme concentration to 40 nM decreased the rate by 50%. Since most of the experiments described in the paper with different DNA substrates used 165 nM enzyme, this appears to be sufficient to ensure single turnover kinetics with the GT-containing substrate. However, it is possible that some of the poorer substrates may bind the enzyme less well so that single turnover is not being measured. In this case the measured reaction rates will overestimate the true single turnover rate and so should be regarded as an upper limit.
Figure 2 shows the time course of Vsr cleavage of the fragment containing a single GT mismatch in the sequence CTAGG. It can be seen that, as expected, the DNA is cleaved at a single position, corresponding to the 5′-side of the mismatched thymine. The reaction requires the presence of a divalent metal ion (Mg2+) but is not affected by the presence of dithiothreitol. However, we find that it is necessary to include BSA in the reaction mixture in order to achieve complete digestion. Although the initial reaction rate is the same in the presence and absence of BSA, the extent of cleavage is improved on its addition. This effect is especially noticeable at low Vsr concentrations or when working with poor substrates (see below). A full consideration of the kinetic properties of the enzyme will be presented elsewhere. A quantitative analysis of the reaction is presented in Figure 3. Under these single turnover conditions, in which the enzyme concentration (165 nM) is greater than the substrate, the first order rate constant is 0.23 ± 0.04 min–1. This value is much lower than that typically observed with restriction enzymes such as EcoRV, which have turnover rates of ~200 min–1 (22). For all the experiments described below, in which we compare the efficiency of cleavage of different DNA substrates by Vsr, the reaction rates are compared to that of the GT-containing substrate which was always run in parallel with the same sample of diluted enzyme.
The most informative method for comparing the ability of enzymes to discriminate between different substrates is to calculate specificity constants, kcat/Kd, where kcat is the single turnover rate constant and Kd is the dissociation constant (which may not be the same for all substrates). We attempted to estimate the Kd by band shift analysis but found that enzyme concentrations in excess of 1 µM were required to produce a retarded DNA species with the GT-containing oligonucleotide. This is consistent with the weak binding observed in a previous study using band shift (8), which in any case will measure the dissociation constant for the cleaved complex. The analysis described below therefore only compares the relative kcat values of the different substrates. This limitation is considered further in the Discussion.
Comparison of GT, GU, IT and IU
In order to assess the importance of various substituents on the GT mismatch which might be recognised by Vsr, we have compared the cleavage of four fragments of identical sequence except for the identity of the mismatched base pair. Representative Vsr cleavage patterns of these fragments, which contain GT, IT, GU and IU pairs, are presented in Figure 2, while the quantitative plots are shown in Figure 3. The relative cleavage efficiencies of these different sites are summarised in Table 1. It can be seen that each of these fragments is cut by the enzyme at a single site on the 5′-side of the mismatched base. GU, IT and IU are cleaved less efficiently than GT, suggesting that both the 5-methyl group of thymine and the 2-amino group of guanine are recognised by the enzyme, though neither is essential for activity. Changing T to U causes a 60% decrease in activity (compare GT with GU and IT with IU), while changing G to I causes a 90% reduction in activity (compare GT with IT and GU with IU). Cleavage of GU is not surprising since it has previously been shown that GU mispairs are repaired by VSP, albeit at a lower efficiency than GT (16).
We considered the possibility that the enzyme merely recognises an unpaired base (G or T) in the sequence by examining the digestion of fragments containing a pseudoabasic site (1′,2′-dideoxyribose, ϕ) on either or both strands at the mismatch site, generating Gϕ, ϕT and ϕϕ pairs. The results of these experiments are presented in Figure 4. It can be seen that replacing either the G or T of the GT mismatch with an abasic site causes a drastic reduction in cleavage, with a further reduction on replacing both bases. The relative cleavage efficiencies, averaged over five experiments, are presented in Table 1 and show that the activity is reduced by over 98%. It should, however, be noted that despite the low activity, cleavage is still located exclusively at this site.
Other purine analogues
Figure 5 shows the results of further experiments examining the cleavage of fragments containing 2AP and nebularine in place of the G in the GT mismatch, alongside the patterns for GT and IT for comparison. Since removal of the 2-amino group of guanine caused a large reduction in activity we anticipated that the 2AP pair might be efficiently cleaved. However, it can be seen that this is a very poor substrate for Vsr cleavage, worse even than the abasic site ϕT (Fig. 4), though once again the low level of cleavage is still restricted to the single site. The nebularine·T base pair was cleaved with even lower efficiency. These cleavage efficiencies are summarised in Table 1.
We have examined the ability of Vsr to cleave several other mismatches consisting of natural bases when these are located at the same position. The results are summarised in Table 1. Both TT and CT mismatches are cut with low efficiency, similar to or lower than that seen with the abasic sites. In contrast, we could not detect any cleavage of AC, A5MeC or GA mismatch pairs. As expected, the fully Watson–Crick duplex, containing an AT base pair in this position, was not cut by the enzyme. We also examined the activity of the enzyme against a base pair in which 8-oxoguanine replaced thymine. In the syn configuration 8-oxoguanine presents the same arrangement of hydrogen bond donors and acceptors as thymine and has the potential for forming a wobble base pair with G which is similar to the GT mismatch. We found that the fragment containing the G8oxoG base pair was cut with low efficiency, but was a poorer substrate than Gϕ.
It has been reported that difluorotoluene (F) can form stable base pairs with adenine (19,20). The AF base pair could involve either shape-specific recognition between the bases, as F is isostructural with T, or the formation of hydrogen bonds with the fluorine atoms. Since F lacks the pyrimidine N3-H it is unlikely to form a wobble base pair with guanine. We found that a fragment containing the GF mismatch was not cleaved by the Vsr endonuclease (not shown).
The experiments described in this paper have explored the importance of various functional groups on the GT mismatch for cleavage by Vsr mismatch endonuclease. It should be remembered that, since we were unable to estimate the dissociation constant (Kd) for each substrate, the relative cleavage rates that we have determined are not true specificity constants (kcat/Kd). If the enzyme binds less well to substrates with modified bases (i.e. with higher Kd) then the discrimination between different substrates will be even greater than we have estimated. If, however, Vsr binds more tightly to any of the poorer substrates then the discrimination will be less pronounced.
At first sight the 2-amino group of guanine is implicated in the recognition process since its removal, generating an IT mismatch, reduces the activity by 90%. This substituent is located in the DNA minor groove, whereas it has previously been suggested that the enzyme binds from the major groove (11). The crystal structure reveals one specific contact at this location from the main chain carbonyl of Met14. However the 2AP·T pair, in which the 2-amino group is retained, is a very poor substrate (0.15% activity), demonstrating that the presence of this substituent is not sufficient for recognition. However, as noted below, 2AP·T is not a wobble base pair and has a different structure to the GT mismatch. The poor cleavage of 2AP·T seems to implicate the purine 6-keto group for recognition. When this is replaced by an amino group, generating an AT base pair, cleavage is totally abolished. We might therefore suggest that efficient GT mismatch cleavage requires the presence of a hydrogen bond acceptor at the purine 6 position. Removing this group (as in 2AP) leads to a drastic reduction in activity, while replacing it with a hydrogen bond donor (as in A) abolishes enzyme cleavage. However, it is clear that neither the 2-amino nor 6-keto substituents on the purine ring are sufficient for cleavage, as a Watson–Crick GC pair at this position is not a substrate. It therefore appears that the enzyme must also be sensing the shape of the GT wobble base pair. The observation that removing the 5′-methyl group of thymine leads to a 60% decrease in activity also suggests that recognition occurs from the major groove.
However, hydrogen bond recognition of the mismatched base pair alone is not sufficient to explain the cleavage observed at sites containing abasic residues. Gϕ is cut at a much lower rate than GT, suggesting that the enzyme normally makes some contacts with the mismatched thymine. In addition, ϕT and ϕϕ, which lack the purine base, are also cleaved by the enzyme. At least part of the recognition must arise from contacts with the rest of the dcm sequence C_WGG, though other sequences, such as CTWGN and NTWGG, are substrates for the enzyme (10). Comparison of the activities measured at variants of the canonical sequence, in which significant cleavage is observed at some sites which differ by two bases from the canonical sequence (10), suggests that this part of the recognition process is affected by subtle changes in the local conformation. It therefore seems that, in addition to forming some essential hydrogen bond contacts, the Vsr endonuclease recognises the shape of the mismatched sequence and hence the conformation of the phosphodiester backbone. It should be noted that A·T, 2AP·T and nebularine·T are not wobble base pairs and contain a hydrogen bond from the thymine 3-NH to purine N1. In the wobble base pairs GT and IT the position of the pyrimidine ring is shifted so that 3-NH contacts the group in the 6 position of the purine base. The importance of this shift in the position of the pyrimidine ring can be seen in the crystal structure of the cleaved complex for which a normal Watson–Crick base pair would push against Thr19, which is positioned in the DNA minor groove (12). A model involving shape-specific recognition of the mismatch also accounts for the appreciable (though small) cleavage observed at abasic sites, since these will be less restrained and might be able to adopt the required structure. TT and CT mismatches may also be able to adopt this structure, but do so less favourably than the abasic sites, since the backbone of the pyrimidine·pyrimidine base pair may not be in the correct position or orientation to be cleaved. Nonetheless, shape-specific recognition alone cannot explain the selectivity, since A5MeC and AC, which are isostructural with GT, are not substrates. Although it is not clear what structure is adopted by the GF mismatch, this can involve (at most) one hydrogen bond to fluorine and it is unlikely to adopt the wobble configuration required for enzyme cleavage.
In conclusion, we propose that Vsr endonuclease recognises the GT mismatch by sensing the shape of the wobble base pair, as suggested by the crystal structure (12). The reduced activity at IT suggests that optimum binding involves contacts with the purine 2-amino group in the DNA minor groove. The observation that the A5MeC wobble pair is not cleaved suggests a role for the guanine 6-keto and thymine O4 groups in the major groove. The lower activity at GU compared with GT also indicates some contacts with the DNA major groove.
This work was supported by a grant from the BBSRC.
To whom correspondence should be addressed. Tel: +44 2380 594374; Fax: +44 2380 594459; Email: email@example.com The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors
|Mismatch/sequence||Relative cutting efficiency (%)|
|GU||40.6 ± 3.3|
|IT||9.3 ± 0.6|
|IU||3.1 ± 0.3|
|Gϕ||1.5 ± 0.5|
|ϕT||1.8 ± 0.7|
|ϕϕ||0.6 ± 0.2|
|2AP·T||0.14 ± 0.04|
|Neb·T||0.08 ± 0.04|
|TT||1.1 ± 0.6|
|G8oxoG||0.8 ± 0.2|
|Mismatch/sequence||Relative cutting efficiency (%)|
|GU||40.6 ± 3.3|
|IT||9.3 ± 0.6|
|IU||3.1 ± 0.3|
|Gϕ||1.5 ± 0.5|
|ϕT||1.8 ± 0.7|
|ϕϕ||0.6 ± 0.2|
|2AP·T||0.14 ± 0.04|
|Neb·T||0.08 ± 0.04|
|TT||1.1 ± 0.6|
|G8oxoG||0.8 ± 0.2|
In all experiments cleavage of the fragment containing the GT mismatch was examined alongside the various mismatches and used for internal comparison. The values were determined from analysis of phosphorimages of the gels as described in Materials and Methods. Each value is the average of at least three determinations. In each case the base in the radiolabelled strand is underlined. All mismatches were in the sequence CTAGG. n.d., no cleavage detected.