Portal de Periódicos da CAPES

The Phe-X-Glu DNA Binding Motif of MutS

2001; Elsevier BV; Volume: 276; Issue: 49 Linguagem: Inglês

10.1074/jbc.c100449200

1083-351X

Mark Schofield, Floyd E. Brownewell, Sunil Nayak, Chunwei Du, Eric T. Kool, Peggy Hsieh,

RNA and protein synthesis mechanisms

The crystal structures of MutS protein fromThermus aquaticus and Escherichia coli in a complex with a mismatch-containing DNA duplex reveal that the Glu residue in a conserved Phe-X-Glu motif participates in a hydrogen-bonded contact with either an unpaired thymidine or the thymidine of a G-T base-base mismatch. Here, the role of hydrogen bonding in mismatch recognition by MutS is assessed. The relative affinities of MutS for DNA duplexes containing nonpolar shape mimics of A and T, 4-methylbenzimidazole (Z), and difluorotoluene (F), respectively, that lack hydrogen bonding donors and acceptors, are determined in gel mobility shift assays. The results provide support for an induced fit mode of mismatch binding in which duplexes destabilized by mismatches are preferred substrates for kinking by MutS. Hydrogen bonding between the Oε2 group of Glu and the mismatched base contributes only marginally to mismatch recognition and is significantly less important than the aromatic ring stack with the conserved Phe residue. A MutS protein in which Ala is substituted for Glu38 is shown to be defective for mismatch repairin vivo. DNA binding studies reveal a novel role for the conserved Glu residue in the establishment of mismatch discrimination by MutS. The crystal structures of MutS protein fromThermus aquaticus and Escherichia coli in a complex with a mismatch-containing DNA duplex reveal that the Glu residue in a conserved Phe-X-Glu motif participates in a hydrogen-bonded contact with either an unpaired thymidine or the thymidine of a G-T base-base mismatch. Here, the role of hydrogen bonding in mismatch recognition by MutS is assessed. The relative affinities of MutS for DNA duplexes containing nonpolar shape mimics of A and T, 4-methylbenzimidazole (Z), and difluorotoluene (F), respectively, that lack hydrogen bonding donors and acceptors, are determined in gel mobility shift assays. The results provide support for an induced fit mode of mismatch binding in which duplexes destabilized by mismatches are preferred substrates for kinking by MutS. Hydrogen bonding between the Oε2 group of Glu and the mismatched base contributes only marginally to mismatch recognition and is significantly less important than the aromatic ring stack with the conserved Phe residue. A MutS protein in which Ala is substituted for Glu38 is shown to be defective for mismatch repairin vivo. DNA binding studies reveal a novel role for the conserved Glu residue in the establishment of mismatch discrimination by MutS. mismatch repair insertion/deletion 4-methylbenzimidazole difluorotoluene DNA mismatch repair (MMR)1 safeguards the integrity of the genome by repairing base-base and insertion/deletion (I/D) mismatches that occur during DNA replication and homologous recombination. Inactivation of this highly conserved repair pathway leads to greatly elevated mutation rates and is the underlying defect in hereditary nonpolyposis colon cancer and familial colorectal cancer (reviewed in Refs. 1Kolodner R.D. Marsischky G.T. Curr. Opin. Genet. Dev. 1999; 9: 89-96Crossref PubMed Scopus (732) Google Scholar and 2Jiricny J. Nystrom-Lahti M. Curr. Opin. Genet. Dev. 2000; 10: 157-161Crossref PubMed Scopus (231) Google Scholar). The methyl-directed MMR pathway ofEscherichia coli has been extensively studied (reviewed in Refs. 3Modrich P. Lahue R. Annu. Rev. Biochem. 1996; 65: 101-133Crossref PubMed Scopus (1340) Google Scholar and 4Harfe B.D. Jinks-Robertson S. Annu. Rev. Genet. 2000; 34: 359-399Crossref PubMed Scopus (505) Google Scholar). MutS protein encoded by a single gene specifically recognizes I/D mismatches of 1–3 nucleotides as well as 7 of 8 possible base-base mismatches, C-C mismatches being refractory to repair by this pathway. The binding affinity of MutS for a particular mismatch roughly correlates with the efficiency of repair in vivo (5Su S.S. Lahue R.S. Au K.G. Modrich P. J. Biol. Chem. 1988; 263: 6829-6835Abstract Full Text PDF PubMed Google Scholar). In the presence of ATP, MutL is recruited to the mismatch and a third protein, MutH, a latent endonuclease, is activated. MutH nicks the transiently unmethylated daughter strand at hemimethylated GATC sequences thereby conferring strand specificity. Excision repair spanning the nick and the mismatch completes the process. Eukaryotes possess multiple MutS and MutL homologues. ThreeMutShomologues, MSH2-MSH6 (MutSα) and MSH2-MSH3 (MutSβ), function as heterodimers with the former recognizing base-base and small I/D mismatches and the latter preferentially repairing larger I/D mismatches (reviewed in Ref.4Harfe B.D. Jinks-Robertson S. Annu. Rev. Genet. 2000; 34: 359-399Crossref PubMed Scopus (505) Google Scholar). Recently co-crystal structures of homodimeric Thermus aquaticus (Taq) and E. coli MutS bound to heteroduplex DNA containing an unpaired T or G-T mismatch, respectively, have been determined (6Obmolova G. Ban C. Hsieh P. Yang W. Nature. 2000; 407: 703-710Crossref PubMed Scopus (556) Google Scholar, 7Lamers M.H. Perrakis A. Enzlin J.H. Winterwerp H.H.K. de Wind N. Sixma T.K. Nature. 2000; 407: 711-717Crossref PubMed Scopus (553) Google Scholar). Remarkably, interactions with the I/D and base-base mismatch are quite similar (reviewed in Ref.8Hsieh P. Mutat. Res. 2001; 486: 71-87Crossref PubMed Scopus (152) Google Scholar). The structures reveal the existence of a composite mismatch binding site in which the DNA is kinked at the mismatch by about 60° toward the major groove. Interestingly, a Phe-X-Glu motif at the N terminus of MutS from only one of the two subunits directly contacts the mismatch thereby rendering the protein functionally heterodimeric (see Fig. 1 A). The conserved Phe residue (Phe36in E. coli, Phe39 in Taq) stacks upon the unpaired T or the T in the G-T mispair. The importance of this residue was demonstrated in previous photocross-linking and mutational studies in bacterial MutS proteins (9Malkov V.A. Biswas I. Camerini-Otero R.D. Hsieh P. J. Biol. Chem. 1997; 272: 23811-23817Abstract Full Text Full Text PDF PubMed Scopus (91) Google Scholar, 10Yamamoto A. Schofield M. Biswas I. Hsieh P. Nucleic Acids Res. 2000; 28: 3564-3569Crossref PubMed Scopus (46) Google Scholar) and in eukaryotic MutS homologues (11Bowers J. Sokolsky T. Quach T. Alani E. J. Biol. Chem. 1999; 274: 16115-16125Abstract Full Text Full Text PDF PubMed Scopus (78) Google Scholar, 12Das Gupta R. Kolodner R.D. Nat. Genet. 2000; 24: 53-56Crossref PubMed Scopus (54) Google Scholar, 13Dufner P. Marra G. Raschle M. Jiricny J. J. Biol. Chem. 2000; 275: 36550-36555Abstract Full Text Full Text PDF PubMed Scopus (69) Google Scholar). The structural data reveal for the first time that Glu38 in E. coli and Glu41 inTaq contact the mismatched T via a hydrogen bond between the carboxyl oxygen of the Glu side chain and the N3 position of T. In this study, we probe the importance of hydrogen bonding in mismatch recognition by MutS. Relative binding affinities of E. coliand Taq MutS are reported using DNA duplexes in which hydrogen bonding involving the mismatch base is disrupted via the substitution of Watson-Crick bases with the non-hydrogen-bonding base analogues difluorotoluene, a shape mimic of T, and 4-methylbenzimidazole, a shape mimic of A (reviewed in Ref. 14Kool E.T. Annu. Rev. Biophys. Biomol. Struct. 2001; 30: 1-22Crossref PubMed Scopus (423) Google Scholar). In a second approach, Glu38 of E. coli MutS and Glu41 of Taq MutS are mutated to Ala, and the abilities of these proteins to carry out MMR in vivo and bind DNA in vitro are examined. Taq MutS protein encoded by pETMutS and E. coli His6-tagged MutS protein encoded by pTX412 were purified as described previously (15Schofield M.J. Nayak S. Scott T.H. Du C. Hsieh P. J. Biol. Chem. 2001; 276: 28291-28299Abstract Full Text Full Text PDF PubMed Scopus (119) Google Scholar). The TaqMutS-E41A and E. coli E38A mutants were constructed by converting the Glu GAG codon to an Ala GCG codon using QuikChange (Stratagene) and verified by DNA sequencing. Deoxyoligonucleotides were synthesized, purified by denaturing gel electrophoresis, and 5′-32P-end-labeled as described previously (15Schofield M.J. Nayak S. Scott T.H. Du C. Hsieh P. J. Biol. Chem. 2001; 276: 28291-28299Abstract Full Text Full Text PDF PubMed Scopus (119) Google Scholar). The 2,4-difluorotoluene (F) and 4-methylbenzimidazole (Z) phorphoramidites were synthesized as described previously (16Ren R.X.-F. Chaudhuri N.C. Paris P.L. Rumney IV S. Kool E.T. J. Am. Chem. Soc. 1996; 118: 7671-7678Crossref PubMed Scopus (221) Google Scholar, 17Morales J.C. Guckian K.M. Sheils C.J. Kool E.T. J. Org. Chem. 1998; 63: 9652-9656Crossref PubMed Scopus (88) Google Scholar). The homoduplex A-T substrate was made by annealing top and bottom strands. Substrates containing mismatches and non-hydrogen-bonding analogues (Z and F) were created by making the appropriate changes to the central T on the top strand or to the central A on the bottom strand, both shown in bold. Top strand 5′TCA CTC ATT AGG ATA CCG TCG ACG CTA GCGTGC GGC TCG TCG TCG ACC TCA CCC CAG GCT T 3′. Bottom strand: 3′AGT GAG TAA TCC TAT GGC AGC TGC GAT CGC ACG CCG AGC AGC AGC TGG AGT GGG GTC CGA A 5′. DNA binding was assessed at 37 °C forE. coli MutS and 21 °C for Taq MutS in gel mobility shift assays as described previously (15Schofield M.J. Nayak S. Scott T.H. Du C. Hsieh P. J. Biol. Chem. 2001; 276: 28291-28299Abstract Full Text Full Text PDF PubMed Scopus (119) Google Scholar). Error bars represent S.D. values of triplicate measurements. pET-15b (Novagen), pTX412, or pE38A were transformed into TX2929 (CC106 mutS 201:: Tn5; Kanr) (18Feng G. Winkler E.W. BioTechniques. 1995; 19: 956-965PubMed Google Scholar). Eleven independent transformants were assessed for rifampicin resistance as described previously (19Biswas I. Obmolova G. Takahashi M. Herr A. Newman M.A. Yang W. Hsieh P. J. Mol. Biol. 2001; 305: 805-816Crossref PubMed Scopus (41) Google Scholar). The relative affinities of E. coli andTaq MutS for a series of "homoduplex" DNAs containing normal Watson-Crick bases or nonpolar shape analogues of adenine, 4-methylbenzimidazole (Z), and thymidine, difluorotoluene (F) (see Fig.1, B and C) were measured in DNA gel mobility shift assays (Fig.2, TableI). The mismatch specificity of E. coli MutS was poor. The apparent K D for homoduplex DNA (A-T) was 20-fold higher than that for the best mismatch (ΔT), but only severalfold higher than the K D values of most other mismatches, a result consistent with earlier DNase I footprinting experiments (5Su S.S. Lahue R.S. Au K.G. Modrich P. J. Biol. Chem. 1988; 263: 6829-6835Abstract Full Text PDF PubMed Google Scholar). Studies of eukaryotic MutS homologues also indicated relatively poor mismatch specificity (20Alani E. Mol. Cell. Biol. 1996; 16: 5604-5615Crossref PubMed Scopus (143) Google Scholar, 21Marsischky G.T. Kolodner R.D. J. Biol. Chem. 1999; 274: 26668-26682Abstract Full Text Full Text PDF PubMed Scopus (113) Google Scholar, 22Gradia S. Acharya S. Fishel R. Cell. 1997; 91: 995-1005Abstract Full Text Full Text PDF PubMed Scopus (303) Google Scholar, 23Blackwell L.J. Martik D. Bjornson K.P. Bjornson E.S. Modrich P. J. Biol. Chem. 1998; 273: 32055-32062Abstract Full Text Full Text PDF PubMed Scopus (161) Google Scholar). In contrast, Taq MutS exhibited a roughly 1000-fold greater affinity for I/D mismatch (e.g. ΔT) heteroduplexes than for homoduplexes (see below). However, the very large preference ofTaq MutS for I/D mismatches was not apparent for base-base (e.g. G-T) mismatches as reflected in the much lower affinities for base-base mismatches. Thus, discrimination between homoduplex and base-base mismatches was, as in the case of E. coli, relatively poor (this study). 2T. H. Scott and P. Hsieh, unpublished results.Table IEffect of nonpolar DNA base-shape mimics on the binding affinity of MutSSubstrateE. coli K DTaq K DSubstrateE. coli K DTaq K DT-A34 ± 6.13800 ± 360F-A9.3 ± 1.53000 ± 1300T-Z16 ± 2.5780 ± 39F-Z10 ± .62100 ± 250ΔT1.4 ± 0.52.2 ± 2.1ΔF8.3 ± 2.116 ± 2.8ΔA5.7 ± 2.175 ± 26ΔZ11 ± 0.683 ± 14T-G5.0 ± 1.7310 ± 73F-G8 ± 3.6960 ± 370T-C9.3 ± 3.2450 ± 77F-C9.7 ± 0.6830 ± 210T-T9.0 ± 1.72300 ± 1100F-T12 ± 2.1800 ± 550A-A6.3 ± 1.53400 ± 570A-Z16 ± 1.02200 ± 58G-A7.7 ± 2.95100 ± 210G-Z12 ± 3.12100 ± 440C-A4.3 ± 0.61900 ± 780C-Z19 ± 2.12300 ± 610K D values in nm were measured using a polyacrylamide gel retardation assay as described under "Experimental Procedures." Errors are calculated by S.D. values of triplicate measurements. Open table in a new tab K D values in nm were measured using a polyacrylamide gel retardation assay as described under "Experimental Procedures." Errors are calculated by S.D. values of triplicate measurements. Interestingly, both E. coli and Taq MutS exhibited a small but reproducible preference for homoduplexes containing the nonpolar mimics F and Z as compared with homoduplexes containing Watson-Crick (A-T) base pairs. Duplexes containing F-A, T-Z, and F-Z base pairs were bound by E. coli MutS with apparent affinities (K D = 9.3, 16, and 10 nm, respectively) reflecting a 2–3.5-fold increase over the affinity for the corresponding Watson-Crick homoduplex (K D = 34 nm). For Taq MutS, the affinity for the F-A duplex (apparent K D = 3000 nm) was essentially unchanged from that of the normal homoduplex (apparent K D = 3800 nm), whereas the affinity for the F-Z duplex was marginally improved (apparent K D = 2100 nm). However, binding to the T-Z duplex was significantly enhanced (apparentK D = 780 nm). Previous studies of duplex stabilities as determined by thermal denaturation experiments revealed that duplexes in which T is replaced by F have lowered stability compared with normal duplexes on the order of 3.6 kcal/mol due to loss of hydrogen bonding and the cost of desolvation of the A opposite F (24Moran S. Ren R.X.-F. Kool E.T. Proc. Natl. Acad. Sci. U. S. A. 1997; 94: 10506-10511Crossref PubMed Scopus (302) Google Scholar). The observed preference of MutS for F- and Z-containing duplexes is consistent with an induced fit mode of mismatch binding in which intrinsically destabilized duplexes containing mismatches (or nonpolar base mimics) are preferred substrates for MutS-mediated kinking. To assess the role of hydrogen bonding in mismatch recognition, we determined the relative affinities of E. coli and Taq MutS for I/D heteroduplex DNAs in which the single, unpaired base was T or A or, alternatively, the F and Z nonpolar analogues (Fig. 2; Table I). The F analogue lacks hydrogen bond donors and acceptors, i.e. N3 and O4 of T (see Fig. 1), implicated in hydrogen bonding with the Oε2 of Glu38 or Glu41 in crystal structures of E. coli andTaq MutS-mismatch complexes. Although we cannot prove that the F- and Z-containing heteroduplexes are recognized via the same Phe-X-Glu binding site, the data presented below are most consistent with an interaction involving the canonical mismatch binding site of MutS. The unpaired ΔT substrate and the unpaired ΔF substrate were bound by E. coli MutS with relative affinities corresponding toK D = 1.4 nm and 8.3 nm, respectively. This reduced affinity for the ΔF substrate, corresponding to a ΔΔG°bind of ∼1.0 kcal mol−1, was consistent with the loss of hydrogen bonding between Glu38 and the unpaired F analogue and any accompanying desolvation of Glu38. Interestingly, there was only a 2-fold difference in the relative affinities of MutS for ΔA compared with ΔZ, K D = 5.7 nm versus 11 nm. Thus, the relative contribution of the hydrogen bond between Glu38 and an I/D mismatch was not as significant for ΔA as for ΔT. A possible caveat is one of sequence context; the A/Z mismatches are on the C-rich bottom strand, whereas the T/F mismatches are on the G-rich top strand. Qualitatively similar results were obtained with Taq MutS. Substitution of ΔF for ΔT resulted in a 7-fold decrease in the apparent binding affinity, K D = 16 nm versus 2.2 nm. Again, loss of hydrogen bonding resulting from the substitution of ΔZ for ΔA had no effect on relative affinities, K D = 83 nm for ΔZversus 75 nm for ΔA. This was not unexpected given the relatively poor intrinsic affinity of Taq MutS for ΔA. These results suggest that hydrogen bonding may play a small role in binding to the high affinity ΔT substrate, but is unlikely to be a factor in binding to ΔA. This is in sharp contrast to the role of the aromatic ring stack involving Phe36/39 as mutation of Phe to Ala reduced mismatch binding affinities for both Taq andE. coli by a 1000-fold (9Malkov V.A. Biswas I. Camerini-Otero R.D. Hsieh P. J. Biol. Chem. 1997; 272: 23811-23817Abstract Full Text Full Text PDF PubMed Scopus (91) Google Scholar, 10Yamamoto A. Schofield M. Biswas I. Hsieh P. Nucleic Acids Res. 2000; 28: 3564-3569Crossref PubMed Scopus (46) Google Scholar). Consistent with the small effect of hydrogen bonding on binding to I/D mismatches, substitution of Watson-Crick bases with Z or F analogues in base-base mismatches did not reveal striking effects on mismatch binding by E. coli or Taq MutS. The largest change resulted from alteration of a C-A mismatch to a C-Z mismatch yielding a 4-fold decrease in relative affinities for E. coli MutS, K D = 4.3 nm versus 19 nm. In the case of TaqMutS, changing a T-G mismatch to an F-G mismatch resulted in a 3-fold decrease in relative affinity, from a K D = 310 nm to 960 nm. However a change from T-T to F-T resulted in a 3-fold increase in affinity, from a K D = 2.3 μm to 800 nm. Although no clear trends were apparent, in most cases, substitutions of nonpolar analogues for normal bases in base-base mismatches resulted in only small changes in binding affinity. While these observations are most simply interpreted as confirming the absence of a major role for hydrogen bonding in mismatch recognition, we cannot rule out the possibility that the result is attributable to two relatively modest opposing forces. One is the destabilization of DNA duplexes by the F and Z analogues that would favor binding by MutS. Opposing this is the loss of hydrogen bonding by Glu38 (E. coli) or Glu41(Taq) that would destabilize mismatch binding. Although determination of binding affinities is imprecise at protein concentrations in the micromolar range, the data clearly reveal the relatively poor binding of base-base mismatches by Taq MutS compared with E. coli MutS. The relatively minor role of hydrogen bonding involving Glu and the mismatch base is at odds with the highly conserved nature of the Glu38/41 residue. To further probe the role of this amino acid residue in mismatch recognition, we replaced the Glu residue with Ala. The relative binding affinities of E. coli E38A and Taq E41A MutS for the high affinity ΔT substrate were unchanged (Table II). Thus, consistent with results presented above, hydrogen bonding by Glu38/41is not required for heteroduplex DNA binding, suggesting that the essential role of Glu lies elsewhere. Information regarding the critical nature of this amino acid was revealed when binding to several other DNA substrates by the mutant proteins was examined. The replacement of the acidic Glu with a neutral Ala restored high affinity binding to the ΔF heteroduplex such that it was now indistinguishable from the affinity of wild type MutS for the favored ΔT mismatch,K D values in the 1–2 nm range. In fact, the E38A and E41A mutant proteins exhibited enhanced binding for several mismatch and homoduplex substrates relative to wild type MutS protein (Table II). Strikingly, the E38A mutation in E. coliMutS substantially narrowed the gap in affinities for a mismatchversus a perfectly paired homoduplex.Table IIEffect on DNA binding of a glutamic acid to alanine mutationSubstrateE. coliTaqWild typeK DE38A K DWild typeK DE41A K DΔT1.4 ± 0.50.9 ± 0.52.2 ± 2.12.8 ± 0.6ΔF8.3 ± 2.11.5 ± 0.116 ± 2.82.0 ± 0.06ΔA5.7 ± 2.16.0 ± 1.075 ± 266.6 ± 0.6ΔZ11 ± 0.65.9 ± 0.583 ± 1429 ± 4.6T-G5.0 ± 1.73.5 ± 0.8310 ± 73240 ± 30T-A34 ± 6.19.2 ± 4.23800 ± 3602200 ± 290K D values in nm were measured using a polyacrylamide gel retardation assay as described under "Experimental Procedures." Errors are calculated by S.D. values of triplicate measurements. Open table in a new tab K D values in nm were measured using a polyacrylamide gel retardation assay as described under "Experimental Procedures." Errors are calculated by S.D. values of triplicate measurements. Taken together, these results indicate that Glu38/41functions in the discrimination between mismatch-containingversus perfectly paired DNA as opposed to making a critical hydrogen bond to the mismatch base. As a corollary, the data suggest that charge neutralization of an acidic residue (or removal of an unfavorable contact) in the mismatch binding site can have a significant effect on mismatch discrimination by MutS. This finding is consistent with the structural data of Taq and E. coli MutS-mismatch complexes that predict electrostatic repulsion of Glu with the phosphate backbone of a normal DNA duplex (6Obmolova G. Ban C. Hsieh P. Yang W. Nature. 2000; 407: 703-710Crossref PubMed Scopus (556) Google Scholar, 7Lamers M.H. Perrakis A. Enzlin J.H. Winterwerp H.H.K. de Wind N. Sixma T.K. Nature. 2000; 407: 711-717Crossref PubMed Scopus (553) Google Scholar). Supporting evidence for the role of acidic residues in enhancing substrate selectivity stems from studies of Glu and Asp residues in the tRNA and DNA binding sites of E. coli methionyl-tRNA synthetase and E. coli RuvA, respectively (25Schmitt E. Meinnel T. Panvert M. Mechulam Y. Blanquet S. J. Mol. Biol. 1993; 233: 615-628Crossref PubMed Scopus (46) Google Scholar, 26Ingleston S.M. Sharples G.J. Lloyd R.G. EMBO J. 2000; 19: 6266-6274Crossref PubMed Scopus (23) Google Scholar). Alteration of the charge by amino acid substitution resulted in increased binding of noncognate tRNAs by methionyl-tRNA synthetase and duplex DNA by RuvA. The ability of the E. coli E38A MutS protein to carry out mismatch repair was assessed in vivo in a general mutator assay measuring the frequency of mutation to rifampicin resistance.E. coli TX2929 (CC106 mutS 201:: Tn5; Kanr) (18Feng G. Winkler E.W. BioTechniques. 1995; 19: 956-965PubMed Google Scholar) was transformed with pTX412 encoding wild typemutS, pE38A mutS, or pET15b vector alone, and relative frequencies of rifampicin-resistant colonies were scored. The mutation frequency of pE38A transformants was similar to that of pET15b vector controls (mutation frequencies of 340 ± 180 × 10−9 and 150 ± 80 × 10−9, respectively) and 2 orders of magnitude higher than transformants harboring pTX412 (0.7 ± 1.1 × 10−9). The DNA binding studies described above suggest that the mismatch repair defect of the E38A mutant protein is attributable at least in part to the loss of mismatch specificity and not to the loss of DNA binding per se. The following conclusions stem from the findings presented here. First, the data support an induced fit mode of mismatch binding in which duplexes destabilized by mismatches are preferred substrates for deformation by MutS. Second, although hydrogen bonding between Glu38/41 of the Phe-X-Glu motif and a mismatched base may contribute in some cases to binding, its relative contribution is marginal and significantly smaller than that of the aromatic ring stack involving Phe. Third, Glu38/41contributes to mismatch discrimination. Charge neutralization of this conserved acidic residue leads to an increased affinity for homoduplex DNA and loss of mismatch discrimination that may account for the mismatch repair defect of the E. coli E38A mutant protein. Our findings are consistent with the idea that Glu38/41promotes DNA kinking by MutS, because kinking reduces or eliminates unfavorable interactions between the Glu side chain and DNA.