Chimeric Msh6 protein with an Msh3 mispair-binding domain combines properties of both proteins

Scarlet S. Shell*†‡§, Christopher D. Putnam*, and Richard D. Kolodner*†‡§¶

*Ludwig Institute for Research, Departments of †Medicine and ‡Cellular and Molecular Medicine, and §Cancer Center, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0669

Contributed by Richard D. Kolodner, May 10, 2007 (sent for review March 30, 2007) Msh2–Msh3 and Msh2–Msh6 are two partially redundant mispair- 11). The stacking Phe, a glutamate residue two positions down- recognition complexes that initiate mismatch repair in eukaryotes. stream that hydrogen-bonds with the mispaired bases in the crystal Crystal structures of the prokaryotic homolog MutS suggest the structures, and multiple residues that appear to contact the DNA mechanism by which Msh6 interacts with mispairs because key backbone in the crystal structures are conserved between MutS and mispair-contacting residues are conserved in these two proteins. Msh6 sequences from a number of eukaryotes, suggesting a con- Because Msh3 lacks these conserved residues, we constructed a served mechanism of mispair interaction. Mutation of the stacking series of mutants to investigate the requirements for mispair Phe creates a nonfunctional protein that is unable to bind mispairs interaction by Msh3. We found that a chimeric protein in which the (5–7, 12, 13). In contrast, although several of the putative backbone- mispair-binding domain (MBD) of Msh6 was replaced by the contacting residues are conserved in eukaryotic Msh3 sequences, equivalent domain of Msh3 was functional for mismatch repair. all known Msh3 sequences lack the two conserved residues impli- This chimera possessed the mispair-binding specificity of Msh3 and cated in mispair recognition by MutS and Msh6. Moreover, indi- revealed that communication between the MBD and the ATPase vidual mutations altering some of the conserved backbone- domain is conserved between Msh2–Msh3 and Msh2–Msh6. Fur- contacting residues cause more severe defects in Msh6 than in Msh3 ther, the chimeric protein retained Msh6-like properties with (11, 14). Consequently, the interaction of Msh2–Msh3 with inser- respect to genetic interactions with the MutL homologs and an tions and deletions likely has important differences from that of Msh2 MBD deletion mutant, indicating that Msh3-like behaviors both Msh2–Msh6 and MutS (14). Remarkably, Msh3 is nonfunc- beyond mispair specificity are not features controlled by the MBD. tional when complexed with an Msh2 deletion that lacks the domain homologous to the MutS MBD (Msh2⌬MBD), whereas the mismatch repair ͉ mispair recognition ͉ Msh2 ͉ Mlh1–Pms1 ͉ Mlh1–Mlh3 Msh2⌬MBD–Msh6 heterodimer is almost fully functional (14, 15). Mispair recognition by MutS as well as Msh2–Msh6 and ismatch repair (MMR) is an important mechanism for the Msh2–Msh3 induces exchange of ADP for ATP at the distant Mrepair of base misincorporations and insertions or deletions ATPase domains, likely as a result of coupled conformational of one or more nucleotides that occur during DNA replication. changes within the mispair-recognition proteins (2). ATP bind- MMR has been conserved from bacteria to humans, and MMR ing by MutS is required for interaction with the MutL ho- defects cause increased mutation rates and contribute to cancer modimer, which induces downstream factors to cleave, unwind, susceptibility and development (1–4). Although many of the com- and degrade the nascent DNA strand (2). In eukaryotes, MutL ponents of MMR have been identified, gaps remain in our under- is conserved in the form of multiple heterodimers containing standing of how mispairs are recognized and how this recognition common and unique subunits. The Mlh1–Pms1 complex in is converted into signals for repair. Saccharomyces cerevisiae (Mlh1–Pms2 in higher eukaryotes) is ⌬ Recognition of mispairs by MutS or its eukaryotic homologs required for the majority of MMR because pms1 mutants have ⌬ ⌬ is required for recruitment and activation of downstream factors mutation rates as high as those of or mutants (16). and thereby initiates MMR. In bacteria, both base–base mispairs The Mlh1–Mlh3 complex, in contrast, appears to be required for and insertion/deletions are recognized by the MutS homodimer a subset of Msh3-dependent MMR, as evidenced by the finding ⌬ ⌬ (2). Crystal structures of MutS in complex with mispaired DNA that mlh3 msh6 double mutants have a considerably higher ⌬ ⌬ have shown that the homodimer interacts with DNA in an mutation rate than mlh3 or msh6 single mutants, whereas ⌬ ⌬ ⌬ asymmetric manner, such that only one subunit directly contacts mlh3 mutants have a mutation rate similar to mlh3 or ⌬ the mispair, whereas the other contacts only the backbone of the msh3 single mutants (16). It is not clear whether the ability of DNA (5–7). In eukaryotes, three MutS homologs (MSHs) form Mlh1–Mlh3 to participate in Msh3-dependent MMR depends on two heterodimers that participate in MMR. The Msh2–Msh6 the types of mispairs being recognized or on other factors. complex recognizes both base–base and small insertion/deletion Furthermore, it is not known which domains of MutS or its mispairs, whereas the Msh2–Msh3 complex primarily recognizes eukaryotic homologs interact with the MutL complexes or how both small and large insertion/deletion mispairs (8, 9); the two mispair and ATP binding facilitate the interaction. complexes are thus partially redundant. Specificity in the eu- karyotic heterodimers is mediated by the Msh6 and Msh3 Author contributions: S.S.S., C.D.P., and R.D.K. designed research; S.S.S. and C.D.P. per- subunits (1), which correspond to the mispair-contacting subunit formed research; S.S.S., C.D.P., and R.D.K. analyzed data; and S.S.S., C.D.P., and R.D.K. wrote in the asymmetric MutS homodimer (2). the paper. An N-terminally located mispair-binding domain (MBD) medi- The authors declare no conflict of interest. ates mispair recognition. Structures of MutS complexed with DNA Abbreviations: CSM, complete supplement mixture; MBD, mispair-binding domain; MMR, containing G/T, G/A, A/A, or G/G mispairs or an insertion of a mismatch repair; MSH, MutS homolog; SPR, surface plasma resonance; YPD, yeast extract/ single thymidine demonstrate a single mechanism for mispair peptone/dextrose. recognition by the MBD (5–7). In all cases a highly conserved ¶To whom correspondence should be addressed. E-mail: [email protected]. phenylalanine residue (Phe-36 in Escherichia coli, Phe-39 in Ther- This article contains supporting information online at www.pnas.org/cgi/content/full/ mus aquaticus) stacks with a mispaired base (Fig. 1B), consistent 0704148104/DC1. with the key role indicated for this residue by genetic studies (10, © 2007 by The National Academy of Sciences of the USA

10956–10961 ͉ PNAS ͉ June 26, 2007 ͉ vol. 104 ͉ no. 26 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0704148104 Downloaded by guest on September 25, 2021 Msh3 247 A B Msh3 178

α3 α2 α4 β6

β5 α1 β1 β2 β4 Msh3 233 β3 Phe 36 Mispair

Msh3 185

C α1 α2 β1 β2 α3

EcMutS ------MSAIENFDAHTPMMQQYLRLKAQHPEILLFYRMGDFYELFYDDAKRASQLLD 52 TaMutS ------MEGMLKGEGPGPLPPLLQQYVELRDQYPDYLLLFQVGDFYECFGEDAERLARALG 55 hMSH6 -PTVWYHETLEWLKEEKRRDEHRRRPDHPDFDASTLYVPEDFLNSCTPGMRKWWQIKSQ-FDLVICYKVGKFYELYHMDALIGVSELG 441 yMSH6 KFNKQNEERYQWLVDE--RDAQRRPKSDPEYDPRTLYIPSSAWNKFTPFEKQYWEIKSKMWDCIVFFKKGKFFELYEKDALLANALFD 353 hMSH3 ------KSANKRSKSIYTPLELQYIEMKQQHKDAVLCVECGYKYRFFGEDAEIAARELN 262 yMSH3 ------VKKKARKSPTAKLTPLDKQVKDLKMHHRDKVLVIRVGYKYKCFAEDAVTVSRILH 203 * * * β3 β4 α4 β5 β6

EcMutS ISL------TKRGASAGEPIPMAGIPYHAVENYLAKLVNQGESVAICEQIGDPAT------SKGPVERKVVRIVTPG 117 TaMutS LVL------TK--TSKDFTTPMAGIPLRAFEAYAERLLKMGFRLAVADQVEPAEE------AEGLVRREVTQLLTPG 119 hMSH6 LV------FM---KG------NWAHSGFPEIAFGRYSDSLVQKGYK-ARVEQTETPEMMEARCRKMAHISKYDRVVRREICRIITKG 512 yMSH6 LK------IAGGGRA------NMQLAGIPEMSFEYWAAQFIQMGYKVAKVDQRESMLAKEMR------EGSKGIVKRELQCILTSG 421 hMSH3 IYC------HLDHNFMTASIPTHRLFVHVRRLVAKGYKVGVVKQTETAALKAIGDNRSSLF------SRKLTALYTKS 328 yMSH3 IKLVPGKLTIDESNPQDCNHRQFAYCSFPDVRLNVHLERLVHHNLKVAVVEQAETSAIKKHDPGASKSS-----VFERKISNVFTKA 285

Fig. 1. Mispair-binding domain structure and sequence alignments. (A) Ribbon diagrams illustrating the overall fold and recognition of a DNA substrate with

a GT mispair (DNA backbone, black; bases, gray) by E. coli MutS (Protein Data Bank ID 1w7a) (33). The Msh2-like MutS subunit that does not contact the mispair GENETICS is yellow, whereas an Msh6-like mispair-interacting subunit is depicted in orange with the MBD in red. (B) Interactions of the MBD with the DNA mispair. The mispaired thymidine (blue) stacks with Phe-36. Small spheres and changes in ribbon coloring (green at the N terminus, red and the C terminus) indicate C␣ positions of residues at junctions in the subdomain swap constructs. (C) Sequence alignment of E. coli and T. aquaticus MutS and human and yeast Msh6 and Msh3. Secondary structural elements derived from the E. coli MutS structure are shown above the alignment. *, Absolutely conserved position; boxed text, Phe-36; underline, seven residues examined by site-directed mutagenesis; vertical line, beginning and end of the region swapped in Msh6(3-MBD). The figure was generated with PyMOL (34) (A and B) and by SEQUOIA (35) (C).

Because mispair recognition by Msh2–Msh3 and Msh2–Msh6 conserved among the Msh6 or Msh3 sequences, but are different clearly differs, but the complexes seem to have similar functions between the Msh6 and Msh3 sequences (data not shown). These overall, we investigated whether point mutations or swapping residues likely contribute to mispair specificity and include the portions of the Msh3 sequence homologous to the MutS MBD into Msh6-conserved mispair-contacting Phe residue, as well as a num- Msh6 could produce functional chimeric alleles. We found that a ber of residues nearby in the linear sequence. We therefore made version of Msh6 in which the MBD was replaced by the MBD of MSH3 mutants in which sequence encoding two, three, or seven Msh3 was functional for MMR, and genetic and biochemical conserved residues was changed to encode the equivalent Msh6 characterization demonstrated that the chimeric protein had the mispair-recognition specificities of Msh3. The finding that the residues (Y186K and K187F; Y186K, K187F, and K189E; and chimera was fully functional for MMR indicated that transmission Y186K, K187F, K189E, C190L, F191Y, A192E, and E193K), as of mismatch recognition into signals for downstream repair events well as the reciprocal Msh6 mutants to test the effect of exchanging were not compromised, suggesting that Msh3 and Msh6 retain these homolog-specific residues between Msh6 and Msh3 (Fig. 1). conserved mechanisms mediating communication between do- Patch tests revealed that none of the engineered alleles could mains. Like Msh6, the chimeric protein did not appear to signifi- complement the high mutation rate of an msh3⌬ msh6⌬ strain when cantly interact with Mlh1–Mlh3, indicating that the regions of the present on low copy-number plasmids (data not shown). MSHs dictating specificity toward the MutL homologs lie outside Lack of complementation might have been because of of the MBD, and that recruitment of MutL homologs is not strongly defects altering the stability, affinity, or mispair discrimination influenced by the types of lesions being recognized. Surprisingly, of the resulting chimeric MBD-containing proteins or possible the chimeric allele was functional in combination with the failure to properly signal mispair recognition to induce sub- Msh2⌬MBD mutant, indicating that the MBD of Msh3 is compe- tent to recognize mispairs without the function of the MBD of sequent repair steps. To avoid these potential problems, we Msh2. created an MSH6 allele termed msh6(3-MBD) in which Msh6 residues 305 through 421, which correspond by homology to Results the MBD of MutS, were replaced by the homologous Msh3 Generation of Chimeric MSH6–MSH3 Alleles That Function in Vivo. sequence, residues 155 through 285 (Figs. 1C and 2). Patch Alignments of Msh6 and Msh3 sequences from a variety of organ- tests showed that the plasmid-borne msh6(3-MBD) allele isms show a number of residues in the MBD that are strongly complemented the msh3⌬ msh6⌬ strain to a degree similar to

Shell et al. PNAS ͉ June 26, 2007 ͉ vol. 104 ͉ no. 26 ͉ 10957 Downloaded by guest on September 25, 2021 Allele Plasmid Domain organization Complementation

Msh3 pRDK1088 1-1047 Msh6 pRDK439 1-1242

Msh6(3-185-233) pRDK1097 1-334 372-1242

Msh6(3-155-233) pRDK1092 1-304 372-1242

Msh6(3-185-285) pRDK1093 1-334 422-1242

Msh6(3-178-247) pRDK1112 1-327 386-1242

Msh6(3-MBD) pRDK1087 1-304 422-1242

Msh6(3-1-285) pRDK1126 1-285 422-1242

Msh3(6-305-421) pRDK1103 1-154 286-1047

Msh3(6-1-421) pRDK1119 1-421 286-1047

Msh3(6-1-455) pRDK1128 1-455 311-1047

Msh3(6-1-466) pRDK1151 1-466 322-1047

Msh3(6-1-483) pRDK1153 1-483 339-1047

Msh3(6-1-370) pRDK1175 1-370 233-1047

Msh3(6-1-441a) pRDK1182 1-441 299-1047

Msh3(6-1-441b) pRDK1183 1-441 300-1047

Fig. 2. Properties of point mutants and chimeric alleles. Black bars, sequence derived from Msh3; gray bars, Msh6 sequence; numbering, amino acid positions of the wild-type proteins; complementation, ability (ϩ) or failure (Ϫ) of a plasmid bearing the indicated allele to decrease the high mutation rate of an msh3⌬ msh6⌬ strain in patch tests.

that seen for wild-type MSH3 and MSH6 [supporting infor- of Msh6 and the C-terminal portion of Msh3 [msh3(6–1-455), mation (SI) Fig. 4]. The exchanged sequence could be ex- msh3(6–1-466), msh3(6–1-483), msh3(6–1-370), msh3(6–1-441a), tended successfully into the nonconserved N-terminal region and msh3(6–1-441b)] complemented the msh3⌬ msh6⌬ strain (Fig. by replacing residues 1–421 of Msh6 with residues 1–285 of 2). These results suggest that the chimeras disrupted key aspects of Msh3 in the msh6(3–1-285) allele (Fig. 2). However, minimiz- the Msh3 protein, the MBD of Msh6 failed to make appropriate ing the size of the replaced MBD region by creating several contacts with other domains of Msh3 for signaling, or subtle MSH6 alleles containing portions of the Msh3 MBD did not differences exist in the function of the Msh3 and Msh6 MBDs, such yield chimeric capable of complementing the msh3⌬ that the Msh6 MBD cannot functionally replace the Msh3 MBD. msh6⌬ strain (Fig. 2). Breakpoints for the subdomain swaps were chosen based on analysis of the MutS crystal structure to Mutations Suppressed by the msh6(3-MBD) Allele Are Similar to Those exchange portions of the MBD directly involved in mispair Suppressed by MSH3. To extend the above results, the msh6(3-MBD) recognition with junctions at conserved boundaries between allele was placed at the MSH6 chromosomal locus of wild-type and secondary structural elements within the domain (Fig. 1). We msh3⌬ strains. Fluctuation analysis was performed to determine conclude that either the mode of mispair binding found in the rates of reversion of the hom3–10 and lys2–10A alleles, which Msh3 requires the entire Msh3 MBD or the subdomain swaps primarily result from single-nucleotide deletions, and the rates of disrupted the overall structure or function of the chimeric inactivation of the CAN1 , which reflect the accumulation of MBDs. both frameshift and base substitution mutations (1, 17). In contrast to the successful swap of the MBD of Msh3 into The reversion rates of the msh6(3-MBD) msh3⌬ strain were Msh6, an Msh3 construct encoding the reverse swap (codons similar to those of the msh6⌬ MSH3 strain in the hom3–10 and 155–285 of MSH3 replaced with codons 305–421 of MSH6) failed lys2–10A assays (Table 1), suggesting that the msh6(3-MBD) to complement the msh3⌬ msh6⌬ strain (Fig. 2). Because the allele was similar to wild-type MSH3 in the repair of frameshift msh6(3–1-285) allele that swapped both the N-terminal region and mispairs. Consistent with this finding, the Canr rate of the the MBD was functional, we also tested MSH3 alleles that encoded msh6(3-MBD) msh3⌬ strain was Ϸ13-fold higher than wild type, variants containing both the N-terminal region and MBD of Msh6 similar to both the msh6⌬ MSH3 and the msh6(3-MBD) MSH3 to eliminate the possibility of an incorrect chimeric junction be- strains (Table 1). These Canr rates are consistent with the idea tween the MBD and the N-terminal region. Neither the msh3(6– that the msh6(3-MBD) allele, like MSH3, can support the repair 1-421) allele nor others with different junctions between the MBD of frameshifts in the CAN1 gene, but not the base substitutions

Table 1. Mutation rate analysis of msh6(3-MBD) Mutation rate (fold increase)

Relevant genotype Strain no. Thrϩ Lysϩ CanR

MSH3 MSH6 3686 2.0 [1.2–3.7] ϫ 10Ϫ9 (1) 1.5 [1.0–3.6] ϫ 10Ϫ8 (1) 7.4 [4.6–13.5] ϫ 10Ϫ8 (1) MSH3 msh6⌬ 4151 3.0 [1.9–3.9] ϫ 10Ϫ8 (15) 9.8 [8.4–15.3] ϫ 10Ϫ7 (67) 9.4 [6.1–14.9] ϫ 10Ϫ7 (13) msh3⌬ MSH6 4149 2.2 [1.6–3.5] ϫ 10Ϫ8 (11) 1.3 [0.87–1.9] ϫ 10Ϫ7 (9) 1.1 [0.55–1.3] ϫ 10Ϫ7 (2) msh3⌬ msh6⌬ 4234 3.1 [1.9–4.8] ϫ 10Ϫ6 (1,548) 4.8 [3.9–10.6] ϫ 10Ϫ5 (3,272) 3.1 [1.7–4.6] ϫ 10Ϫ6 (42) MSH3 msh6(3-MBD) 5153 2.7 [1.7–4.0] ϫ 10Ϫ9 (1) 3.7 [1.8–5.9] ϫ 10Ϫ8 (3) 9.9 [8.1–17.4] ϫ 10Ϫ7 (13) msh3⌬ msh6(3-MBD) 5249 6.6 [5.7–8.0] ϫ 10Ϫ8 (33) 1.0 [0.88–1.3] ϫ 10Ϫ6 (70) 9.8 [7.9–14.6] ϫ 10Ϫ7 (13)

Data are median rates of 18–35 cultures per strain, with 95% confidence intervals in brackets and fold increase in rate relative to the wild-type strain in parentheses.

10958 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0704148104 Shell et al. Downloaded by guest on September 25, 2021 Table 2. CAN1 mutation spectra Base substitutions Ϯ1 Frameshifts Larger deletions or duplications

Strain Percent Rate (fold increase) Percent Rate (fold increase) Percent Rate (fold increase)

Wild type 80 3.9 ϫ 10Ϫ8 (1) 14 6.9 ϫ 10Ϫ9 (1) 6 2.9 ϫ 10Ϫ9 (1) msh6⌬ 93 6.2 ϫ 10Ϫ7 (16) 5 3.4 ϫ 10Ϫ8 (5) 0 N/A msh3⌬ 51 3.2 ϫ 10Ϫ8 (1) 24 1.5 ϫ 10Ϫ8 (2) 24 1.5 ϫ 10Ϫ8 (5) msh6⌬ msh3⌬ 33 6.9 ϫ 10Ϫ7 (18) 65 1.4 ϫ 10Ϫ6 (203) 2 4.2 ϫ 10Ϫ8 (14) msh6(3-MBD) 98 8.9 ϫ 10Ϫ7 (24) 2 1.8 ϫ 10Ϫ8 (3) 0 N/A msh6(3-MBD) msh3⌬ 92 8.4 ϫ 10Ϫ7 (22) 4 3.6 ϫ 10Ϫ8 (5) 4 3.6 ϫ 10Ϫ8 (12)

Data are percentages of total mutations that fall into the category indicated (n ϭ 49–59 per strain), and rates of CAN1 inactivation via mutation type are indicated (fold increase over wild-type rate in parentheses).

normally recognized by the Msh2–Msh6 complex. Thus, substi- DNA substrates with a GT mispair or a ϩ1 insertion, but interacted tution of the Msh3 MBD for the Msh6 MBD in the msh6(3- more weakly with ϩ3 and ϩ4 insertions. In contrast, the Msh2– MBD) allele likely changed the mispair-recognition specificity Msh6(3-MBD) chimera did not bind the GT mispair better than the from that of MSH6 to that of MSH3. Intriguingly, the msh6(3- fully base-paired DNA, and bound to the ϩ1, ϩ3, and ϩ4 insertions MBD) MSH3 strain had hom3–10 and lys2–10A reversion rates more strongly than wild-type Msh2–Msh6, with the ϩ3 insertion similar to those of the wild-type strain, indicating that a func- showing the strongest relative interaction (Fig. 3). Thus, the tional copy of MSH3 combined with the chimeric allele was Msh6(3-MBD) chimera was compromised relative to Msh6 in sufficient for wild-type levels of repair of frameshift mispairs. recognizing base–base mispairs, but had an increased ability to This result suggests that the small defect seen in msh3⌬ and recognize the larger insertion mispairs bound by Msh3 (2). msh6⌬ single mutants in these assays may be because of de- creased gene dosage, rather than loss of nonredundant activities. The msh6(3-MBD) Allele Functions with an msh2 MBD Deletion Mutant. To confirm the results of the mutation rate analysis, we analyzed Recent reports have indicated that Msh3, but not Msh6, requires the spectra of mutations in the CAN1 gene in the same six strains. the Msh2 domain homologous to the MutS MBD for in vivo MMR Base substitutions made up the majority of mutations found in the activity (14, 15). To determine whether this requirement is intrinsic wild-type strain, and the rate at which base substitutions accumu- to the Msh3 MBD or because of other features of the Msh2–Msh3 lated increased Ϸ20-fold in all strains lacking wild-type MSH6 heterodimer, we assessed complementation of an msh2⌬ msh3⌬ (Table 2). The msh6(3-MBD) allele, like MSH3, was unable to msh6⌬ strain by plasmid-encoded MSH3, MSH6, and msh6(3- suppress these high base substitution rates, consistent with the MBD) alleles in combination with plasmids encoding either wild- absence of normal MSH6-like recognition of base–base mispairs type Msh2 or Msh2 missing residues 2–133 (msh2⌬MBD), which (Table 2). In contrast, MSH3, MSH6, and msh6(3-MBD) all sup- correspond to the entire MBD-like domain (14, 15). Consistent GENETICS pressed the accumulation of Ϯ1 frameshift mutations to a similar with previous reports, coexpression of MSH3 and msh2⌬MBD did extent. We did not observe sufficient numbers of larger insertion/ not complement the triple mutant, but coexpression of MSH6 and deletion events to reliably discern differences among MSH3, MSH6, msh2⌬MBD complemented well (Table 3). Surprisingly, msh6(3- and msh6(3-MBD) with respect to repair of these types of mispairs. MBD) in combination with msh2⌬MBD complemented the triple Together the mutation rate and spectrum data suggest that the mutant just as well as msh6(3-MBD) in combination with wild-type mispair-recognition properties of the MBD encoded by MSH3 are MSH2, suggesting that the MBD-like domain of Msh2 is not maintained when this domain is transferred into Msh6. required for mispair recognition by the MBD of Msh3 per se, but rather that the MBD-like domain of Msh2 may be required for Msh2-Msh6(3-MBD) Chimera Shows Specificity for Insertion/Deletion some other aspect of Msh3 function. Loops in Vitro. To assess the ability of the Msh2–Msh6(3-MBD) chimera complex to recognize larger insertion/deletion mispairs, we Unlike Msh3, Msh6(3-MBD) Does Not Mediate Significant Repair purified the chimeric Msh6(3-MBD) protein as a heterodimer with Through the Mlh1–Mlh3 Complex. The basis of the specificity of Msh2 for biochemical analysis. Surface plasmon resonance (SPR) Mlh1–Mlh3 for Msh2–Msh3-bound mispairs is not known. In was used to assess the interaction of Msh2–Msh6(3-MBD) and wild-type Msh2–Msh6 with 236-bp DNA harboring centrally lo- cated mispairs (Fig. 3). The wild-type complex bound strongly to Table 3. Complementation of msh2⌬ msh3⌬ msh6⌬ strain by MSH3/6 alleles and msh2⌬MBD Mutation rate (fold increase) WT Msh2-Msh6 Msh2-Msh6(3-MBD) ϩ ϫ 7 ϩ ϫ 6 +3 Plasmid alleles* Thr rate 10 Lys rate 10 150 GT +4 None 55 [32–120] (103) 130 [86–220] (338) +1 100 +1 MSH2 MSH6 0.54 [0.34–1.3] (1) 0.38 [0.17–0.68] (1) MSH2 MSH3 0.29 [0.16–0.56] (1) 1.5 [0.92–2.3] (4) +3 GT 50 MSH2 msh6(3-MBD) 1.7 [1.2–5.3] (3) 2.6 [1.8–4.2] (7) +4 GC ⌬ GC msh2 MBD MSH6 1.4 [0.90–2.3] (3) 0.62 [0.35–0.97] (2) response (units) 0 msh2⌬MBD MSH3 95 [72–220] (176) 140 [99–190] (377) -10 01020304050 -10 01020304050 msh2⌬MBD msh6(3-MBD) 1.7 [1.2–4.8] (3) 2.3 [1.7–2.9] (6) time (seconds) time (seconds) Data are median rates of 14–20 cultures, with 95% confidence intervals Fig. 3. SPR analysis of DNA binding, in which 10 nM wild-type Msh2–Msh6 in brackets and fold increase over MSH2 MSH6-complemented strain in or Msh2–Msh6(3-MBD) complexes were flowed over sensor chips preconju- parentheses. gated with 75 to 85 response units of 236-bp DNA containing central GT, ϩ1, *On low copy-number plasmids in the msh2⌬ msh3⌬ msh6⌬ strain (RDKY ϩ3, or ϩ 4 mispairs as indicated, or no mispairs (GC). 5963).

Shell et al. PNAS ͉ June 26, 2007 ͉ vol. 104 ͉ no. 26 ͉ 10959 Downloaded by guest on September 25, 2021 Table 4. Effect of MLH3 genotype on mutation rate Mutation rate (fold increase)

Relevant genotype Strain no. Thrϩ Lysϩ CanR

MSH3 MSH6 MLH3 3686 2.0 [1.2–3.7] ϫ 10Ϫ9 (1) 1.5 [1.0–3.6] ϫ 10Ϫ8 (1) 7.4 [4.6–13.5] ϫ 10Ϫ8 (1) MSH3 MSH6 mlh3⌬ 5296 1.7 [1.0–2.7] ϫ 10Ϫ8 (9) 1.7 [0.72–2.9] ϫ 10Ϫ7 (11) 1.2 [1.0–1.6] ϫ 10Ϫ7 (2) MSH3 msh6⌬ MLH3 4151 3.0 [1.9–3.9] ϫ 10Ϫ8 (15) 9.8 [8.4–15.3] ϫ 10Ϫ7 (67) 9.4 [6.1–14.9] ϫ 10Ϫ7 (13) MSH3 msh6⌬ mlh3⌬ 5298 1.9 [1.4–2.8] ϫ 10Ϫ7 (97) 7.9 [6.5–9.4] ϫ 10Ϫ6 (540) 1.7 [1.3–2.2] ϫ 10Ϫ6 (22) msh3⌬ MSH6 MLH3 4149 2.2 [1.6–3.5] ϫ 10Ϫ8 (11) 1.3 [0.87–1.9] ϫ 10Ϫ7 (9) 1.1 [0.55–1.3] ϫ 10Ϫ7 (2) msh3⌬ MSH6 mlh3⌬ 5974 3.8 [2.8–5.1] ϫ 10Ϫ8 (19) 1.9 [1.7–2.8] ϫ 10Ϫ7 (13) 2.5 [1.9–4.6] ϫ 10Ϫ7 (3) msh3⌬ msh6(3-MBD) MLH3 5249 6.6 [5.7–8.5] ϫ 10Ϫ8 (33) 1.0 [0.88–1.3] ϫ 10Ϫ6 (70) 9.8 [7.9–14.6] ϫ 10Ϫ7 (13) msh3⌬ msh6(3-MBD) mlh3⌬ 5882 6.8 [5.5–8.6] ϫ 10Ϫ8 (34) 1.3 [1.1–1.8] ϫ 10Ϫ6 (86) 1.4 [1.1–1.6] ϫ 10Ϫ6 (19)

Data are median rates of 18–35 cultures per strain, with 95% confidence intervals in brackets and fold increase in rate relative to the wild-type strain in parentheses.

principle, both protein–protein interactions and protein-mispair ATPase domain, but this result suggests that the underlying mo- interactions could influence which MutL-related complex is re- lecular mechanisms are highly conserved. cruited. The Msh6(3-MBD) chimera contains the MBD of Msh3 By recognizing the same types of substrates as Msh3, Msh6(3- placed within a scaffold of Msh6 and thus recognizes the same types MBD) allowed us to distinguish between aspects of Msh3 function of substrates as Msh3. Therefore, we used the msh6(3-MBD) allele that are driven by interactions of the Msh3 MBD with mispaired to test the roles of the MBD and mispair recognition in the DNA and protein partners and aspects of Msh3 function that are recruitment of Mlh1–Mlh3. Consistent with previous studies (16), directed by the protein scaffold in which the Msh3 MBD is inserted. deletion of MLH3 in an msh6⌬ background, in which MMR is In addition to a preference for recognizing large insertions and initiated only by Msh2–Msh3, caused a substantial increase in deletions, MSH3 differs from MSH6 with respect to interaction with mutation rate, whereas deletion of MLH3 in an msh3⌬ background the msh2⌬MBD mutant (14, 15), interaction with MutL homologs caused only a small increase in mutation rate, indicating that (16), suppression of interchromosomal crossover recombination Mlh1–Mlh3 is not important in repair events initiated by Msh2– between homeologous substrates (18), and removal of nonhomolo- Msh6 (Table 4). Interestingly, the small increase in mutation rate in gous ends during gene conversion (19) and single-strand annealing the msh3⌬ mlh3⌬ double mutant relative to the msh3⌬ mutant was (20). In the cases of both gene conversion and single-strand significant in all three assays (P Ͻ 0.01, two-tailed Mann–Whitney annealing, specific recognition of double- and single-stranded junc- U test), although the mutation rate of the msh3⌬ mlh3⌬ mutant was tions of branched DNA structures by Msh2–Msh3 has been impli- only significantly higher than that of the mlh3⌬ mutant in the cated as the basis of MSH3 specificity (21). ⌬ hom3–10 and CAN1 assays. These data suggest that an Mlh3- The inability of Msh3 to function in complex with Msh2 MBD dependent pathway could be involved in a small fraction of Msh2– (14, 15) could be because of defects in recognition of mispairs or defects in downstream steps. Our results showed that the combi- Msh6-mediated repair events. However, the increases in the mu- ⌬ tation rates were small when compared to the effect of an MLH3 nation of the msh6(3-MBD) and msh2 MBD alleles was functional deletion in an msh6⌬ background. As in the msh3⌬ background, for MMR, indicating that the MBD of Msh3 lacks an inherent requirement for the MBD of Msh2 for mispair recognition. This deletion of MLH3 in the msh6(3-MBD) msh3⌬ background only finding suggests that the defects in the Msh2⌬MBD–Msh3 complex resulted in small increases in mutation rate in the lys2–10A and cause reduced mispair affinity (14) as a result of something more CAN1 assays, whereas the hom3–10 reversion rate was not signif- subtle about the interaction between the two subunits. We note that icantly changed (Table 4). Thus, msh6(3-MBD) behaved like MSH6 insertions of more than one nucleotide modeled into the known with respect to genetic interaction with MLH3 rather than like MutS structures (5, 6) would probably be positioned such that the MSH3. This finding suggests that interactions between the MutS unpaired, extrahelical nucleotides could make physical contact with and MutL homologs are not specified by the types of mispairs the MBD of the Msh2-like subunit. If this potential interaction were recognized or by physical contacts involving the MBD, but rather important in activating the Msh2–Msh3 complex, it would explain by other features of the MSHs. both the defect of the Msh2⌬MBD–Msh3 complex and the failure Discussion of the reverse swap constructs where the Msh6 MBD was placed into the Msh3 protein. Here we show that a chimera in which the MBD of Msh6 was The msh6(3-MBD) allele also allowed us to investigate the replaced by that of Msh3 was functional for MMR and had interactions involved in the recruitment of specific MutL homolog Msh3-like mispair-recognition properties in vivo and in vitro. How- complexes during MMR. Deletions of, and mutations in, the ever, the chimera retained MSH6-like genetic interactions with N-terminal domain of Mlh1 disrupt the Mlh1–Pms1 interaction MLH3 and retained the ability of Msh6 to function in complex with with Msh2–Msh3/6. However, the interfaces of the MutL homologs an Msh2 mutant lacking its MBD. with Msh2, Msh3, and Msh6 have not been defined (22–24). The ability of the Msh2–Msh6(3-MBD) complex to bind larger Previous work indicated that the Mlh1–Mlh3 complex functions insertions in vitro and of msh6(3-MBD) to complement an msh3⌬ primarily in the Msh2–Msh3 pathway, whereas the Mlh1–Pms1 msh6⌬ strain in vivo leads to two interesting conclusions about complex functions through both Msh2–Msh3 and Msh2–Msh6 (16). Msh3 and Msh6. First, these proteins are modular, with a discrete We found that the msh6(3-MBD) allele is equivalent to MSH6, not MBD conferring mispair-recognition properties without apparent MSH3, with regard to utilization of the Mlh1–Mlh3 complex, contribution by the other domains of the protein. Second, the despite the ability of the Msh2–Msh6(3-MBD) complex to recog- mechanisms by which Msh3 and Msh6 communicate mispair bind- nize the types of mispairs recognized by Msh2–Msh3. These ing at the MBD to the C-terminal ATPase domain must be similar, differences suggest that neither the lesion being recognized nor the such that the Msh3 MBD can replace functions normally per- MBD is sufficient to impart MutL homolog complex specificity, and formed by the Msh6 MBD within the Msh2–Msh6 complex. It is not that the interactions between the MSHs and the MutL homologs known how mispair binding leads to conformational changes in the are mediated by other domains of the MSHs.

10960 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0704148104 Shell et al. Downloaded by guest on September 25, 2021 Materials and Methods 3 days before counting colonies. Amino acids were omitted from all Strains, Media, and Plasmids. S. cerevisiae was grown on standard media used for plasmid-bearing strains as necessary to maintain media, either yeast extract/peptone/dextrose (YPD) or complete plasmid selection in both patch-testing and fluctuation analysis. supplement mixture (CSM) (US Biological, Swampscott, MA), Strains were grown at 30°C in all experiments. Rates of hom3–10 lacking specific amino acids. Then 60 mg/liter canavanine was reversion, lys2–10A reversion, and CAN1 inactivation were deter- added to CSM Ϫarg medium to select for canavanine resistance. mined by the method of the median (1, 17, 29). P values were All yeast strains were isogenic derivatives of S288C. The wild- calculated by using the two-tailed Mann–Whitney U test (www.fac- type strain RDKY 3686 (25) has the genotype MAT␣ ura3–52 ulty.vassar.edu/lowry/utest.html). CAN1 mutation spectra were de- leu2⌬1trp1⌬63 his3⌬200 hom3–10 lys2::InsE-A10. Derivatives termined by sequencing the CAN1 gene in 49 to 59 independent RDKY 4149, 4151, 4234, and 5249 have msh3::hisG and/or isolates per strain as described (1). msh6::hisG mutations (Table 1). In RDKY 5153, 5249, and 5882, MSH6 was replaced by msh6(3-MBD) by a pop-in/pop-out strategy Protein Purification. Msh2 was coexpressed with wild-type Msh6 or (26). Strains RDKY 5296, 5298, 5882, and 5974 were made from Msh6(3-MBD) in the E. coli strain BL21 CodonPlus (DE3)-RIL 3686, 4151, 5249, and 4149, respectively, by replacing nucleotides (Stratagene, La Jolla, CA) from the plasmids PLANT-2-B-MSH2 ϩ26 through ϩ2094 of MLH3 with a PCR fragment of pRS303 (30), pET11a-Msh6 (30), and pRDK1231, which was derived from containing the HIS3 gene. RDKY 5963 was made from 4234 by pET11a-Msh6. Lysis was performed as described (30), and purifi- replacement of nucleotides ϩ10 through ϩ2885 of MSH2 with cation was performed by sequential chromatography over SP HIS3. Sepharose, ssDNA cellulose, heparin, and high performance Q Wild-type MSH2 was present on the URA3 plasmid pII-2 (27), (Amersham Pharmacia Biotech, Piscataway, NJ) columns (see SI and msh2⌬MBD was present on a derivative of this plasmid missing Materials and Methods). codons 2–133, pRDK1250. Other plasmids used for complemen- tation assays were based on the ARS CEN LEU2 vector pRS315 SPR Analysis. SPR experiments were performed with a BIAcore (28). In pRDK439, MSH6 is present on a genomic fragment (BIAcore AB, Uppsala, Sweden) 3000 instrument. DNA substrates extending from the BamHI site at position Ϫ806 relative to the (236 bp) with a biotin conjugated to one end and a centrally located MSH6 start site to the HindIII site at position ϩ4626, 897 bp GT mispair or no mispair were constructed as described (31, 32). downstream of the stop codon, cloned into pRS315. In pRDK1088, DNAs harboring insertions of one, three, or four nucleotides were MSH3 is present on a genomic fragment bordered by the same two constructed in the same way (see SI Materials and Methods for sites (positions Ϫ1632 to ϩ4633) and cloned into pRS315. Other details). Between 12 and 14 ng of DNA (75–85 response units) was plasmids were made from pRDK439 and pRDK1088 by site- conjugated to each flow cell of an avidin-coated SA BIAcore chip. ␮ directed mutagenesis or PCR and recombination (Fig. 2). Those Then 10 nM protein was flowed over the chip at 30 l/min in with the N-terminal domain of MSH6 had the native MSH6 running buffer (25 mM Tris, pH 8, 2% glycerol, 4 mM MgCl2, 110 , whereas in pRDK1103 and pRDK1126 the native MSH3 mM NaCl, 0.01% igepal, 2 mM DTT). The DNA-coated surface ␮ promoter was present. was regenerated with a 15- l pulse of 3 M NaCl. Data were analyzed with BIAevaluation 3.1 software. Reference subtraction

Qualitative and Quantitative Evaluation of Mutator Phenotypes. was made from an unmodified flow cell. GENETICS Patch tests were performed by patching colonies onto YPD or CSM Ϫleucine plates and replica-plating onto Ϫthreonine, Ϫlysine, We thank Dan Mazur for assistance in protein purification, Manju Ϫarginine ϩcanavanine, and YPD or Ϫleucine media after 23 days Hingorani for expression plasmids, and Jill Harrington and Marc Men- dillo for comments on the manuscript. This work was supported by of growth. Fluctuation analysis was performed on 14 to 35 inde- National Institutes of Health Grants GM50006 and CA92584 (to pendent cultures for each strain as follows. Then 10-ml YPD or R.D.K.), National Institutes of Health/National Institute of General CSM Ϫleucine cultures were inoculated with whole colonies, grown Medical Sciences-funded University of California at San Diego Genetics 12 h (24 h for drop-out media), serially diluted, plated on CSM and Training Program Grant T32 GM08666 (to S.S.S.), and a Damon CSM Ϫthreonine, Ϫlysine, and Ϫarginine ϩcanavanine, and grown Runyon Cancer Research Foundation Robert Black fellowship (C.D.P.).

1. Marsischky GT, Filosi N, Kane MF, Kolodner R (1996) Genes Dev 10:407–420. 18. Nicholson A, Fabbri RM, Reeves JW, Crouse GF (2006) Genetics 173:647–659. 2. Iyer RR, Pluciennik A, Burdett V, Modrich PL (2006) Chem Rev 106:302–323. 19. Sugawara N, Paques F, Colaiacovo M, Haber JE (1997) Proc Natl Acad Sci USA 3. Fishel R, Lescoe MK, Rao MR, Copeland NG, Jenkins NA, Garber J, Kane 94:9214–9219. M, Kolodner R (1993) Cell 75:1027–1038. 20. Sugawara N, Ira G, Haber JE (2000) Mol Cell Biol 20:5300–5309. 4. Kane MF, Loda M, Gaida GM, Lipman J, Mishra R, Goldman H, Jessup JM, 21. Surtees JA, Alani E (2006) J Mol Biol 360:523–536. Kolodner R (1997) Cancer Res 57:808–811. 22. Kadyrov FA, Dzantiev L, Constantin N, Modrich P (2006) Cell 126:297–308. 5. Obmolova G, Ban C, Hsieh P, Yang W (2000) Nature 407:703–710. 23. Plotz G, Welsch C, Giron-Monzon L, Friedhoff P, Albrecht M, Piiper A, Biondi 6. Lamers MH, Perrakis A, Enzlin JH, Winterwerp HH, de Wind N, Sixma TK RM, Lengauer T, Zeuzem S, Raedle J (2006) Nucleic Acids Res 34:6574–6586. (2000) Nature 407:711–717. 24. Plotz G, Raedle J, Brieger A, Trojan J, Zeuzem S (2003) Nucleic Acids Res 7. Natrajan G, Lamers MH, Enzlin JH, Winterwerp HH, Perrakis A, Sixma TK 31:3217–3226. (2003) Nucleic Acids Res 31:4814–4821. 25. Amin NS, Nguyen MN, Oh S, Kolodner RD (2001) Mol Cell Biol 21:5142–5155. 8. Acharya S, Wilson T, Gradia S, Kane MF, Guerrette S, Marsischky GT, 26. Scherer S, Davis RW (1979) Proc Natl Acad Sci USA 76:4951–4955. Kolodner R, Fishel R (1996) Proc Natl Acad Sci USA 93:13629–13634. 27. Reenan RA, Kolodner RD (1992) Genetics 132:963–973. 9. Marsischky GT, Kolodner RD (1999) J Biol Chem 274:26668–26682. 28. Sikorski RS, Hieter P (1989) Genetics 122:19–27. 10. Schofield MJ, Brownewell FE, Nayak S, Du C, Kool ET, Hsieh P (2001) J Biol 29. Lea DE, Coulson CA (1948) J Genet 49:264–285. Chem 276:45505–45508. 11. Drotschmann K, Yang W, Brownewell FE, Kool ET, Kunkel TA (2001) J Biol 30. Antony E, Hingorani MM (2003) Biochemistry 42:7682–7693. Chem 276:46225–46229. 31. Mendillo ML, Mazur DJ, Kolodner RD (2005) J Biol Chem 280:22245–22257. 12. Bowers J, Sokolsky T, Quach T, Alani E (1999) J Biol Chem 274:16115–16125. 32. Lin YL, Shivji MK, Chen C, Kolodner R, Wood RD, Dutta A (1998) J Biol 13. Dufner P, Marra G, Raschle M, Jiricny J (2000) J Biol Chem 275:36550–36555. Chem 273:1453–1461. 14. Lee SD, Surtees JA, Alani E (2007) J Mol Biol 366:53–66. 33. Lamers MH, Georgijevic D, Lebbink JH, Winterwerp HH, Agianian B, de 15. Goldfarb T, Alani E (2005) Genetics 169:563–574. Wind N, Sixma TK (2004) J Biol Chem 279:43879–43885. 16. Flores-Rozas H, Kolodner RD (1998) Proc Natl Acad Sci USA 95:12404–12409. 34. DeLano WL (2002) PyMOL (DeLano Scientific, San Carlos, CA). 17. Tran HT, Keen JD, Kricker M, Resnick MA, Gordenin DA (1997) Mol Cell Biol 35. Bruns CM, Hubatsch I, Ridderstrom M, Mannervik B, Tainer JA (1999) J Mol 17:2859–2865. Biol 288:427–439.

Shell et al. PNAS ͉ June 26, 2007 ͉ vol. 104 ͉ no. 26 ͉ 10961 Downloaded by guest on September 25, 2021