Biochemical basis for dominant mutations in the Saccharomyces cerevisiae MSH6 gene

Abstract

Here, the ATP-binding, ATP hydrolysis, mispair-binding, sliding clamp formation, and Mlh1–Pms1 complex interaction properties of dominant mutant Msh2–Msh6 complexes have been characterized. The results demonstrate two mechanisms for dominance. In one, seen with the Msh6-S1036P and Msh6-G1067D mutant complexes, the mutant complex binds mispaired bases, is defective for ATP-induced sliding clamp formation and assembly of ternary complexes with Mlh1-Pms1, and occludes mispaired bases from other mismatch repair pathways. In the second, seen with the Msh6–G1142D complex, the mutant complex binds mispaired bases and is defective for ATP-induced sliding clamp formation but assembles ternary complexes with Mlh1–Pms1 that either occlude the mispaired base or prevent Mlh1–Pms1 from acting in alternate mismatch repair pathways.

Errors during DNA replication that result in base–base mismatches and small insertion/deletion mismatches can lead to mutations in DNA. The mismatch repair (MMR) system corrects such errors and is highly conserved from bacteria to humans (1–5). MMR defects cause increased mutation rates, and, in humans, inherited and somatic defects in MMR result in increased development of cancer (6–9). In eukaryotic MMR, mispair recognition is performed by two heterodimeric complexes, Msh2–Msh6 and Msh2–Msh3, whose subunits are homologues of the bacterial MutS protein (10–15). The Msh2–Msh6 complex recognizes base–base mispairs and insertion/deletion loops and is the dominant mismatch recognition complex (10, 16–20). The Msh2–Msh3 complex primarily recognizes insertion/deletion loops and can partially substitute for Msh2–Msh6 (10, 13, 16, 21). Similarly, eukaryotic MMR utilizes heterodimeric complexes of proteins related to the bacterial MutL protein of which the Mlh1–Pms1 complex (Pms1 in Saccharomyces cerevisiae is Pms2 in humans) is the dominant complex (22–24). Like their bacterial homologues, the Msh and Mlh complexes form a ternary complex on mispaired bases in DNA, where they presumably interact with downstream effector proteins (for a discussion, see refs. 25 and 26).

The MutS and Msh2–Msh6 proteins form a ring around DNA when they bind to a mispair (27, 28). On ATP binding, in a reaction that does not require ATP hydrolysis, the Msh2–Msh6 complex is converted to a form that slides along DNA (26, 29). These Msh2–Msh6 sliding clamps have two modes of dissociation from DNA: slow direct dissociation and rapid sliding-dependent dissociation off of the ends of linear DNA substrates (26). The Mlh1–Pms1–Msh2–Msh6 ternary complex that forms at mispairs also appears to exhibit sliding behavior (26). However, the role of these sliding protein complexes in the mechanism of MMR is not well understood. Sliding of MMR proteins along DNA plays a critical mechanistic role in only one of the three major MMR models (25, 29), whereas the other two models do not propose a role for sliding (30, 31).

We previously described dominant mutant Msh2–Msh6 complexes that appear to prevent mispaired bases in DNA from being acted on by Msh2–Msh3 (17, 32). Two of the dominant amino acid substitutions, S1036P and G1067D, are located in Msh6 near the γ-phosphate of ATP in the Msh2-binding site, and a third dominant amino acid substitution, G1142D, is in a C-terminal helix–turn–helix of Msh6 that contacts Msh2 near the Msh6 ATP-binding site; S1036P probably contacts ATP in the Msh2 site, whereas G1067D and G1142D probably do not contact bound ATP (17). Biochemical analysis of these three mutant Msh2–Msh6 complexes showed that they exhibited altered ATP-induced dissociation from mispaired bases (17). A fourth Msh6 amino acid substitution, H1096A, that results from a weak dominant msh6 mutation, changes an amino acid predicted to help activate the water that attacks the γ-phosphate of ATP in the Msh6 ATP-binding site; however, this amino acid substitution had little effect on the biochemical properties of Msh2–Msh6. Here, we have performed a detailed biochemical analysis of these mutant Msh2–Msh6 complexes. Our results indicate that the three strong dominant Msh6 amino acid substitutions alter the interaction between ATP binding and downstream conformational changes within Msh2–Msh6 and result in trapping of different intermediates in the process of assembly of the Msh2–Msh6–Mlh1–Pms1 ternary complex.

Results

Dominant Mutant Msh2–Msh6 Complexes Show Altered Kinetics of ATP Hydrolysis. Previous studies showed that three of the dominant mutant Msh2–Msh6 complexes had reduced ATPase activity in the presence of some mispaired bases, in contrast to stimulation of wild-type Msh2–Msh6 (17). To extend these results, a kinetic analysis of ATP hydrolysis was performed, and the kinetic constants are presented in Table 1. Compared with wild-type Msh2–Msh6, the strong dominant Msh6–S1036P and Msh6–G1067D complexes showed a lower k_cat under most reaction conditions, which was most pronounced in the presence of the GT mispair and +A substrates. This reduced rate of ATP hydrolysis likely reflects a defect in turnover from mispaired bases rather than a general inability to bind ATP, given the nucleotide-binding data presented below and that these two mutant complexes show reduced ATP-induced dissociation from mispaired bases (17). Compared with wild-type Msh2–Msh6, the strong dominant Msh6–G1142D complex generally showed a significantly reduced k_cat in the presence of DNA that was most pronounced in the presence of the GT mispair and +A substrates. This finding, combined with the nucleotide-binding data presented below, suggests that an ATP-binding defect may underlie the ATP hydrolysis defect of the Msh6–G1142D complex. Finally, compared with wild–type Msh2–Msh6, the weakly dominant Msh6–H1096A complex showed a generally reduced k_cat that was somewhat more pronounced in the presence of the GT mispair and +A substrates. These results suggest that a general ATP hydrolysis defect results in the weakly dominant Msh6–H1096A complex.

Table 1. Kinetic parameters of ATP hydrolysis by wild-type and mutant Msh2–Msh6 complexes.

	Km, μM				kcat, min-1				kcat/Km, μM-1·min-1
Protein	No DNA	GC	GT	+A	No DNA	GC	GT	+A	No DNA	GC	GT	+A
WT	14 ± 1.4	31 ± 5.6	81 ± 7.9	52 ± 6.6	4.2 ± 0.8	9.2 ± 0.9	16.2 ± 2.3	18.4 ± 1.6	0.30 ± 0.06	0.30 ± 0.07	0.20 ± 0.04	0.36 ± 0.07
S1036P	4 ± 2.4	10 ± 1.2	18 ± 3.1	17 ± 0.5	2.0 ± 0.3	3.5 ± 0.9	0.9 ± 0.3	5.9 ± 1.7	0.54 ± 0.28	0.37 ± 0.13	0.05 ± 0.01	0.36 ± 0.11
G1067D	18 (17-19)	25 (25-25)	32 (26-39)	31 (25-36)	5.1 (4.9-5.3)	6.1 (5.9-6.3)	1.4 (1.3-1.6)	7.8 (6.7-9.0)	0.28 (0.28-0.29)	0.24 (0.24-0.25)	0.05 (0.04-0.05)	0.26 (0.25-0.26)
H1096A	8 (7-10)	22 (18-26)	84 (83-85)	35 (28-42)	2.4 (2.3-2.5)	4.1 (3.3-4.8)	6.6 (6.1-7.1)	5.7 (4.7-6.7)	0.30 (0.24-0.35)	0.19 (0.19-0.19)	0.08 (0.07-0.09)	0.16 (0.16-0.17)
G1142D	90 ± 7.5	85 ± 11.2	357 ± 104	122 ± 17.3	7.2 ± 2.6	5.5 ± 1.9	6.9 ± 2.7	5.5 ± 0.9	0.08 ± 0.03	0.06 ± 0.02	0.02 ± 0.01	0.05 ± 0.01

dsDNA substrates (17) are as follows: GC, base pair; GT, base mispair; and +A, insertion (see Methods). The values presented are the mean ± SD of the values obtained in independent experiment when three or more experiments were performed or, in parentheses, the range of values when only two experiments were performed.

Filter-Binding Assays Reveal That Dominant Mutant Msh2–Msh6 Complexes Have Reduced ATP Binding. Previous studies have shown that the S. cerevisiae Msh2–Msh6 complex has a high-affinity ATP-binding site with a K_d for binding adenosine 5′-[γ-thio]triphosphate (ATPγS) of ≈3.7 μM and a low-affinity ATP-binding site with a K_d for binding ATPγS of ≈17 μM (33). We have used filter binding to evaluate the ability of the mutant Msh2–Msh6 complexes to bind ATP in the absence of DNA and in the presence of 5-fold excess of DNA to protein of a base-paired DNA, a GT mispair DNA, or a +A insertion substrate. Two basic conditions were studied: binding to [α-³²P]ATP in the presence of EDTA to eliminate ATP hydrolysis and binding of ATP[γ-³⁵S] in the presence of Mg²⁺. We also evaluated the effect of the order of addition of ATP and DNA and the time course of binding, including binding times as short as 10 sec. However, none of the modified binding conditions yielded results that were significantly different from those results obtained with the representative conditions presented in Fig. 1.

Fig. 1.

Filter-binding analysis of ATP binding to mutant Msh2–Msh6 complexes. In the first series of experiments, the indicated Msh2–Msh6 complexes (4 pmol/200 mM) were incubated in the absence of DNA (A) or in the presence of GC (B), GT (C), or +A(D) DNA substrate in EDTA-containing buffer, followed by the addition of [α-³²P]ATP and the measurement of the amount of ATP bound. In the second series of experiments, the indicated Msh2–Msh6 complexes were incubated in Mg²⁺-containing buffer and either 5 μM(E) or 25 μM (F) ATP[γ-³⁵S], followed by the addition of the indicated DNA and the measurement of the amount of ATP bound.

First, binding to [α-³²P]ATP in the presence of EDTA was evaluated as a function of ATP concentration in the presence or absence of different DNA substrates with the DNA added before the addition of the nucleotide (Fig. 1 A–D). The highest levels of ATP binding were seen for the wild-type protein, and there was little effect of the addition of GC base pair, GT mispair, or +A insertion/deletion mispair DNA on ATP binding compared with no DNA. In all reaction conditions with the wild-type Msh2–Msh6, the ATP binding detected is probably a measure of ATP bound to protein that is not bound to DNA because, under conditions where ATP cannot be hydrolyzed, ATP binding causes Msh2–Msh6 to dissociate from both base pairs and mispairs, and ATP-bound Msh2–Msh6 cannot rebind to the DNA substrates (26). In the case of the Msh6–S1036P and Msh6–G1067D complexes, ATP binding in the absence of DNA was reduced by ≈50% compared with wild-type Msh2–Msh6, and each of the DNAs further reduced the level of ATP binding observed to ≈25% of that seen for the wild-type Msh2–Msh6. These results suggest that the Msh6–S1036P and Msh6–G1067D complexes have some sort of ATP-binding defect but can still bind ATP to a significant extent. In addition, because the concentrations of proteins and DNA in these reactions are 10- and 50-fold higher, respectively, than the K_d of binding of the Msh6–S1036P complex to a GT mispair in the presence of ATP (note that the Msh6–G1067D complex showed GT binding kinetics that are virtually identical to those of the Msh6–S1036P complex) and because these two protein complexes show stable binding to mispaired bases in the presence of ATP (17), these results suggest that they can likely bind ATP while bound to mispaired bases. In the case of the Msh6–H1096A complex, ATP binding in the absence of DNA was similar to that of wild-type Msh2–Msh6, whereas each of the DNAs significantly reduced ATP binding. This finding suggests that the Msh6–H1096A exhibits an ATP-binding defect in the presence of DNA, which parallels the reduced catalytic efficiency of the Msh6–H1096A complex in the presence of the three DNA substrates (Table 1). Finally, the Msh6–G1142D complex exhibited very low ATP binding under all four conditions. However, the addition of ATP both decreases the binding of the Msh6–G1142D complex to mispaired bases and increases the dissociation of prebound Msh6–G1142D complex so the Msh6–G1142D complex cannot be completely defective for ATP binding (17).

Second, binding of ATP[γ-³⁵S] in the presence of Mg²⁺ was evaluated at two different ATPγS concentrations, a low concentration (5 μM) that was slightly above the reported K_d of the high-affinity ATP-binding site and a high concentration (25 μM) that was slightly above the reported K_d of the low-affinity ATP-binding site (33). In the examples shown (Fig. 1 E and F), the ATPγS was added to the protein before the DNA was added; however, similar results were obtained when the DNA was added before the ATPγS (data not shown). In the case of wild-type Msh2–Msh6, the same level of ATPγS binding was observed in the absence of DNA or in the presence of the three different DNA substrates, and possibly slightly higher binding was observed at 25 μMATPγS than at 5 μMATPγS. In the case of the Msh6–S1036P, Msh6–G1067D, and Msh6–H1096A complexes, the level of ATPγS binding observed was the same: ≈50% lower and ≈50% higher, respectively, than seen for wild-type Msh2–Msh6. Similar levels of binding were seen at 25 μMATPγS and 5 μMATPγS, and the addition of the three different DNA substrates had no effect on ATPγS binding. In the case of the Msh6–G1142D complex, there was significantly reduced ATPγS binding in the absence of DNA compared with wild-type Msh2–Msh6, and the addition of the three DNA substrates further reduced the binding of ATPγS, with the mispaired DNAs having the greatest effect. Again, similar results were obtained with 25 μMATPγS and 5 μMATPγS. This finding shows that the Msh6–G1142D complex has a reduced ability to bind ATP, particularly in the presence of DNA.

There were significant differences in the results obtained with [α-³²P]ATP in the presence of EDTA and ATP[γ-³⁵S] in the presence of Mg²⁺, which suggests either that the absence of Mg²⁺ destabilizes ATP binding or that ATPγS and ATP have different binding characteristics. Regardless, the filter-binding experiments reveal that each of the four mutant Msh2–Msh6 complexes has defects in ATP binding.

UV Crosslinking Reveals That Dominant Mutant Msh2–Msh6 Complexes Have Different ATP-Binding Defects. UV crosslinking was used to analyze the interaction of the mutant Msh2–Msh6 proteins with ATP under conditions (–Mg²⁺) where ATP hydrolysis cannot occur (Fig. 2). The concentration of ATP used, 100 μM, is well above the K_d of binding to the Msh2 and Msh6 ATP-binding sites and induces both mispair-dependent sliding of Msh2–Msh6 and ternary complex formation with Mlh1–Pms1 (D.J.M., M.L.M., and R.D.K., unpublished data). When the wild-type protein was analyzed, both the Msh2 and Msh6 subunits were labeled; ATP titration studies (data not shown) indicated that this level of labeling represents 80% binding saturation for Msh2 and 97% binding saturation for Msh6. In addition, the individual Walker A amino acid substitutions Msh2–K694M and Msh6–K988M that change critical residues within the ATP binding pockets of the Msh2 and Msh6 ATP sites, respectively, reduced crosslinking to Msh2 and Msh6, respectively (D.J.M. and R.D.K., unpublished data). These latter results support the view that these ATP crosslinking experiments provide a useful method for characterization of the nucleotide-binding properties of Msh2 and Msh6.

Fig. 2.

UV crosslinking of ATP to mutant Msh2–Msh6 complexes. Crosslinking reaction mixtures (20 μl) containing 4 pmol (200 nM) of the indicated Msh2–Msh6 complexes and 100 μM[γ-³²P]ATP were prepared, UV irradiated, and analyzed by SDS/PAGE as described in Methods.

The S1036P and G1067D Msh6 mutants contain amino acid substitutions in a region of Msh6 that is adjacent to the Msh2 ATP-binding site. These two amino acid substitutions eliminated ATP crosslinking to Msh2 and had no apparent effect on ATP crosslinking to Msh6. The Msh6 amino acid substitution H1096A changes a residue predicted to help coordinate the γ-phosphate of ATP in the Msh6 ATP-binding site. Under the UV-crosslinking conditions, this amino acid substitution had little effect on ATP crosslinking to either subunit. The Msh6 amino acid substitution G1142D is adjacent to the Msh6 ATP-binding site. This amino acid substitution almost completely eliminated ATP crosslinking to both Msh2 and Msh6; the level of ATP crosslinking observed was 20% and 14% of that seen for Msh2 and Msh6 in the wild-type protein complex, respectively. These results show that the three strong dominant msh6 mutations affect ATP binding by Msh2–Msh6 and define at least two classes of ATP-binding defects.

Dominant Mutant Msh2–Msh6 Complexes Are Defective in Sliding Clamp Formation. In a previous study, we developed a method using an IAsys biosensor to evaluate binding Msh2–Msh6 to a mispaired base and conversion to a sliding form that undergoes end-dependent dissociation on binding ATP or ATP analogs (26). DNA substrates are attached at one end to the surface of an IAsys cuvette, and the free end of the DNA, depending on the experiment, is blocked with LacI protein, which can be dissociated (t_1/2 = 1.6 sec) on addition of isopropyl β-d-thiogalactoside (IPTG). When Msh2–Msh6 is incubated with either GT or +A mispair-containing substrates in the presence of ATP, bound ATP is hydrolyzed to ADP, allowing Msh2–Msh6 to bind to the mispair. When ATP then binds to the Msh2–Msh6, it is converted to a form that dissociates from the mispair by sliding off of the end of the DNA. Under these conditions, a steady-state level of binding is reached that is higher on mispair-containing DNA (Fig. 3 A and F) than base pair-containing DNA (Fig. 3K). In the presence of LacI protein, a higher level of Msh2–Msh6 binding is seen on the mispair substrates (Fig. 3 A and F), because while the Msh2–Msh6 slides away from the mispair, allowing additional Msh2–Msh6 to bind, the end block prevents its rapid dissociation off of the ends (note that on end-blocked substrates, Msh2–Msh6 still undergoes a slow direct mode of dissociation). In contrast, blocking the ends of base pair-containing substrates did not lead to increased binding (Fig. 3K), because dissociation by sliding is mispair-dependent. Finally, when Msh2–Msh6 was bound to end-blocked, mispair-containing substrates, and IPTG was added to dissociate the LacI, rapid dissociation of Msh2–Msh6 by sliding off of the ends was seen (Fig. 3 A and F); no IPTG-induced dissociation of Msh2–Msh6 was seen on base pair-containing substrates, because dissociation by sliding is mispair-dependent (Fig. 3K).

Fig. 3.

Analysis of mutant Msh2–Msh6 complexes for sliding clamp formation. Biosensor analysis of the association and dissociation of the indicated Msh2–Msh6 complexes with a DNA substrate containing a GT mispair (A–E), +A insertion (F–J), or a GC base pair (K–O) is shown. The dashed red line indicates association with DNA substrates with a free end; the solid red line indicates association with DNA substrates with a LacI-blocked end; and the solid green line indicates dissociation on addition of IPTG to complexes formed on LacI-blocked DNA substrates. Note the small IPTG-induced decrease (B, C, E, G, H, and J–O) is due to LacI dissociation, whereas the larger decrease (A, D, F, and I) is mainly due to rapid Msh2–Msh6 dissociation from newly freed ends.

The Msh6–S1036P (Fig. 3 B, G, and L), Msh6–G1067D (Fig. 3 C, H, and M), and Msh6–G1142D (Fig. 3 E, J, and O) complexes all showed higher binding to the unblocked GT and +A mispair-containing substrates in the presence of ATP compared with the wild-type protein, which is consistent with previous observations that ATP does not induce rapid dissociation of these mutant complexes from mispairs (17). In contrast to wild-type Msh2–Msh6, blocking the ends of the mispair-containing substrates did not increase the steady-state levels of binding of the Msh6–S1036P, Msh6–G1067D, and Msh6–G1142D complexes relative to binding to the unblocked substrates, and the addition of IPTG to dissociate the LacI did not induce dissociation of these three mutant complexes. The binding behavior of Msh6–S1036P, Msh6–G1067D, and Msh6–G1142D complexes on the base pair-containing substrate was the same as the wild-type complex. These results show that although the Msh6–S1036P, Msh6–G1067D, and Msh6–G1142D complexes recognize mispairs in DNA, they are defective for ATP-induced conversion to the sliding configuration. Analysis of the Msh6–H1096A complex (Fig. 3 D, I, and N) revealed a binding behavior that was similar to wild-type Msh2–Msh6 (Fig. 3 A, F, and K), indicating that this mutant complex was proficient for ATP-induced sliding. Interestingly, the Msh6–H1096A complex showed lower binding to unblocked GT or +A mispair-containing substrates and slower but increased binding to the blocked mispair-containing substrates than wild-type Msh2–Msh6. These results indicate that the relatively minor ATP-binding/hydrolysis defect of the Msh6–H1096A complex alters its mispair-binding and -dissociation dynamics.

Dominant Mutant Msh2–Msh6 Complexes Exhibit Different Defects in Assembly of Ternary Complexes with Mlh1–Pms1. To examine the ability of mutant Msh2–Msh6 complexes to form ternary complexes with Mlh1–Pms1, the Msh2–Msh6 complexes were incubated with immobilized LacI-blocked substrate in the presence of ADP or ADP + Mlh1–Pms1 to allow Msh2–Msh6 mispair binding to occur (Fig. 4); ternary complexes do not form in the absence of ATP (26). Then the Msh2–Msh6–DNA complexes were challenged by the addition of ATP to allow assembly with Mlh1–Pms1; the level of Mlh1–Pms1 binding observed was determined by subtracting the Msh2–Msh6-only curve from the Msh2–Msh6 + Mlh1–Pms1 curve (Fig. 4 Insets). The LacI end-block was necessary to prevent assembly of Msh2–Msh6–Mlh1–Pms1 complexes on the ends of DNA. In the case of wild-type Msh2–Msh6, robust ternary complex formation was observed on GT and +A mispair-containing substrates (Fig. 4 A and F) compared with lower level ternary complex formation on the base pair substrate (Fig. 4K); prior studies have shown that the ternary complex formed on base pair substrates exhibits distinctly different dissociation behavior than ternary complexes formed on mispair DNA (26). The significant increase in ternary complex formation above that predicted for loading of one Mlh1–Pms1 complex per prebound wild-type Msh2–Msh6 complex (Fig. 4 A and F) has been attributed to sliding of the wild-type ternary complex off of the mispair, allowing additional Msh2–Msh6 complexes and hence ternary complexes to assemble. Ternary complex formation with the Msh6–H1096A complex (Fig. 4 D, I, and N) was essentially the same as the wild-type Msh2–Msh6 complex.

Fig. 4.

Analysis of mutant Msh2–Msh6 complexes for assembly of ternary complexes with Mlh1–Pms1. Biosensor analysis of the association of the indicated Msh2–Msh6 complexes with a DNA substrate containing a GT mispair (A–E), +A insertion (F–J), or a GC base pair (K–O) end-blocked with 30 nM LacI is shown. Mlh1–Pms1 was present in the association (solid black line) buffer, but the Mlh1–Pms1–Msh2–Msh6–DNA ternary complex formed only after addition of ATP (solid red line). A control lacking Mlh1–Pms1 was included for comparison (dashed black line) to determine the effect of adding ATP on Msh2–Msh6 (dashed red line). The level of Mlh1–Pms1 binding was determined by subtracting the Msh2–Msh6-only curve from the Msh2–Msh6 + Mlh1–Pms1 curve (Insets).

The strong dominant mutant Msh2–Msh6 complexes exhibited two distinctly different ternary complex formation behaviors. The Msh6–S1036P (Fig. 4 B, G, and L) and Msh6–G1067D (Fig. C, H, and M) complexes did not support ternary complex formation with Mlh1–Pms1 on any DNA substrate. Thus, these mutant complexes appear to be defective for one or more ATP-induced conformational changes required for sliding clamp formation and ternary complex formation. In contrast, the Msh6–G1142D complex was proficient for ternary complex formation on the GT and +A mispair-containing substrates (Fig. 4 E and J) and exhibited ≈25% of wild type levels of ternary complex formation on the base pair substrate (Fig. 4O). However, in comparison with ternary complex formation with wild-type Msh2–Msh6, mispair-dependent ternary complex formation with the Msh6–G1142D complex reached a plateau at a lower level consistent with loading of one Mlh1–Pms1 complex per prebound Msh2–Msh6 complex. This result is predicted if the Msh6–G1142D complex, in contrast to the wild-type Msh2–Msh6 complex (see above), is defective for sliding away from the mispair, preventing the binding of more Msh2–Msh6 and hence preventing the assembly of additional ternary complexes. These results indicate that the Msh6–G1142D substitution allows ternary complex formation even though the mutant Msh2–Msh6 complex is defective for ATP binding and sliding clamp formation.

Discussion

Here, we have performed a detailed biochemical analysis of four dominant mutant Msh2–Msh6 complexes. The strongly dominant mutant Msh6–S1036P and Msh6–G1067D complexes had a lower k_cat for ATP hydrolysis in the presence of mispaired bases, bound ATP only in the Msh6 ATP binding site, bound to mispaired bases but were defective for conversion to sliding clamps, and were defective for assembly of ternary complexes with Mlh1–Pms1. These results indicate that these two mutant complexes remain bound to mispairs in the presence of ATP because they are defective for ATP binding in the Msh2 site and as a consequence do not undergo some type of ATP-induced conformational change. The strongly dominant Msh6–G1142D complex had a reduced k_cat for ATP hydrolysis, particularly in the presence of mispaired bases. The Msh6–G1142D complex only weakly bound ATP, which was further reduced by DNA containing mispaired bases, and bound to mispaired bases, but it was defective for conversion to rapidly dissociating sliding clamps. These results show this mutant complex remains bound to mispairs in the presence of ATP because it is defective for ATP binding. Remarkably, the Msh6–G1142D complex did support mispair-dependent, ATP-dependent assembly of ternary complexes with Mlh1–Pms1. These results demonstrate at least two mechanisms for dominance: one in which the mutant Msh2–Msh6 complex binds and occludes mispaired bases and the other in which the mutant Msh2–Msh6 complex binds mispaired bases and assembles a Msh2–Msh6–Mlh1–Pms1 ternary complex that also occludes mispaired bases. The Msh6–H1096A complex had a relatively minor ATP-binding/hydrolysis defect and a subtle alteration of its mispair binding and dissociation dynamics; however, additional experimentation will be required to determine how this results in a weak dominant mutator phenotype.

The Msh6–S1036P and Msh6–G1067D mutants have amino acid changes near the Msh2 ATP-binding site and the Msh6–G1142D mutant has an amino acid change near the Msh6 ATP-binding site. Based on studies of proteins like Rad50 (34–36), these amino acid substitutions would likely affect ATP binding-induced or ADP–ATP exchange-induced conformational changes across the Msh2–Msh6 interface. These amino acid changes may also affect mispair-induced conformational changes transmitted to this region of the protein (27, 28, 37, 38). The observation that the Msh6–S1036P and Msh6–G1067D amino acid substitutions prevent ATP binding in the Msh2 site suggests that either they alter the structure of the Msh2 ATP-binding site or they block a conformational change at the Msh2–Msh6 interface that allows ATP binding in the Msh2 site. These results further indicate that this conformational change and/or ATP binding in the Msh2 site are required for both sliding clamp formation and ternary complex formation with Mlh1–Pms1. Finally, these results indicate that the interaction with ATP in the Msh2 site is critical for inhibition of mispair binding by prebound ATP (26). The observation that the Msh6–G1142D amino acid substitution inhibits binding in both sites, which is further diminished by mispair binding, suggests that this amino acid substitution stabilizes a conformation that cannot bind ATP but is still competent for mispair binding and induced conformational changes; these results further suggest that ATP binding in the Msh6 site affects ATP binding in the Msh2 site.

Remarkably, the Msh6–G1142D complex can assemble mispair-dependent ternary complexes with Mlh1–Pms1 even though the Msh6–G1142D complex is significantly defective for binding ATP at the Msh2 and Msh6 sites. This observation supports several hypotheses. First, it is possible that a critical mispair-induced conformational change at an interface between Msh2–Msh6 and Mlh1–Pms1 is required for ternary complex formation rather than a mispair-induced alteration of ATP-bound states or ATP hydrolysis. Second, it is possible that either ATP binding by Msh2–Msh6 is not required for ternary complex formation, or the Msh6–G1142D amino acid substitution stabilizes a conformational state in the absence of ATP binding that is normally induced by ATP binding, thus preventing the mutant complexes from binding ATP but allowing assembly of Mlh1–Pms1; in this regard, it should be noted that the addition of DNA substrates significantly inhibits the already reduced ATP binding exhibited by the Msh6–G1142D complex (Fig. 1). However, because ternary complex formation with the Msh6–G1142D complex requires ATP, these first two possibilities would place the ATP requirement on the Mlh1–Pms1 complex. Finally, because Msh2–Msh6 in the ternary complex formation experiments is in large excess relative to DNA, it is possible that the fraction of Msh6–G1142D complex that binds mispairs and forms ternary complexes is the fraction of Msh6–G1142D complex that binds ATP, but that this ATP-bound form does not support sliding. In the context of each hypothesis, it is tempting to speculate that msh6–G1142D is a separation-of-function mutation resulting in a sliding clamp-defective but otherwise MMR-competent form of Msh2–Msh6, supporting the view that sliding clamp formation is needed for MMR (29). In this regard, the sliding defect of the Msh6–G1142D and related reduced assembly of Msh6–G1142D–Mlh1–Pms1 complexes can be partially overcome at very high ATP concentrations (data not shown).

In this study we have performed a detailed biochemical characterization of three strong dominant mutant Msh2–Msh6 complexes and one weak dominant mutant Msh2–Msh6 complex. This analysis has defined two different biochemical defects that underlie dominant mutations in MSH6. In addition, these mutations have allowed further dissection of the early biochemical steps in mismatch recognition and MMR. This basic strategy may be applicable to the analysis of other biochemical steps in MMR.

Methods

Proteins. Wild-type and Msh6 mutant Msh2–Msh6 complexes and wild-type Mlh1–Pms1 complex were purified as described in refs. 17 and 26. The LacI protein was provided by Kathleen Matthews (Rice University).

Nucleotide-Binding and ATP-Hydrolysis Assays. Nucleotide-binding assays were performed in 20 μl volumes containing 4 pmol of wild-type or mutant Msh2–Msh6, 25 mM Hepes buffer at pH 7.5, 100 mM NaCl, 1 mM dithiothreitol, 100 μg/ml BSA, 1 mM EDTA, 1–100 μM [α-³²P]ATP (0.6–30 Ci/mmol, Amersham Pharmacia Biosciences) (1 Ci = 37 GBq), 15% glycerol, and the indicated previously described dsDNA substrates (17) at 1 μM. Proteins were mixed with DNA substrates for 5 min at room temperature, and then ATP was added for 12 min. In some experiments, the 12-min incubation with ATP preceded the 5-min incubation with DNA. The reactions were then transferred to ice, diluted with 3 ml of ice-cold stop buffer consisting of 25 mM Hepes buffer pH 7.5, 100 mM NaCl, and 1 mM EDTA, and immediately filtered over 25-mm nitrocellulose filters (HAWPO2500, Millipore). The filters were washed two times with 5 ml of stop buffer, and the bound radioactivity was determined by liquid scintillation counting. In experiments containing ATP[γ-³⁵S] (2.5–12.5 Ci/mmol, PerkinElmer) instead of [α-³²P]ATP, the EDTA in all buffers was replaced with 10 mM MgCl₂.

ATPase activity was measured, as previously described, in 20-μl reactions consisting of 25 mM Hepes buffer at pH 7.8, 10 mM MgCl₂, 0.1 mM EDTA, 100 mM NaCl, 1 mM dithiothreitol, 10% glycerol, 0.001% IGEPAL (Nonidet P-40, Sigma), 75 μg/ml acetylated BSA (Promega), 10–250 μM [γ-³²P]ATP, and the indicated previously described dsDNA substrates (17) at 200 nM. The reactions were started by adding wild-type or mutant Msh2–Msh6 to 60 nM, incubated at 30°C, and processed as described in ref. 17. All values were determined by subtracting the value obtained for a no-protein control.

UV Crosslinking Experiments. Reactions were performed in 20-μl volumes containing 4 pmol of wild-type or mutant Msh2–Msh6, 50 mM Tris at pH 8.0, 110 mM NaCl, 2 mM dithiothreitol, 200 μg/ml BSA, 0.5 mM EDTA, and 5% glycerol. Proteins were mixed with 100 μM [γ-³²P]ATP (25 Ci/mmol, Amersham Pharmacia Biosciences) and incubated on ice for 10 min, followed by 20 min of crosslinking by using a Stratalinker (Stratagene). Reaction mixtures were then fractionated by SDS/PAGE, and radiolabeled bands were detected and quantified by using a PhosphorImager (Molecular Dynamics). Crosslinking to BSA (present in all samples) was not detected.

Total Internal Reflectance Measurements. A 236-bp DNA substrate that was biotinylated at one end, contained either a central GT mispair or GC base pair and had the lacO₁ operator sequence incorporated at the other end was made as previously described in ref. 26. In the case of the GT substrate, the G is located at nucleotide 103 of the top strand, and the T is located in the complementary position of the bottom strand, where the C would be located in the GC control substrate. An additional substrate containing a +A insertion was constructed by the same method. In this case, the biotinylated top strand was made by PCR amplification using plasmid RDK4720 as template DNA, and the bottom strand was made by PCR amplification using the plasmid RDK4719 as template. The top strand contains the +A insertion at nucleotide 104; in addition, nucleotide 103 of the top strand has been changed to a C, and a G is present as its complement in the bottom strand.

Experiments analyzing either Msh2–Msh6 binding or the Msh2–Msh6–Mlh1–Pms1 ternary complex formation on the GT mispair, +A insertion, or the GC base pair substrates were performed by using an IAsys Auto Plus instrument using a running buffer consisting of 25 mM Tris at pH 8.0, 110 mM NaCl, 4 mM MgCl₂ 0.5 mM dithiothreitol, 2% glycerol, 0.05% IGEPAL CA-630, 25 μM ADP, and 250 μM ATP as described in ref. 26. For Msh2–Msh6 binding to unblocked DNA substrates, the equilibration buffer was removed and replaced with 50 μl of the same buffer but containing 50 nM wild-type or mutant Msh2–Msh6. Incubation was carried out for ≈3.5 min or until equilibrium was reached. For experiments on end-blocked DNA substrates, 30 nM LacI was included in the running buffer, during both the equilibration and Msh2–Msh6-binding phases. The baseline of the association curves of MMR proteins depicted was taken after LacI was bound to the DNA substrates, so the binding curves reflect association of MMR proteins only. The dissociation of Msh2–Msh6 off of the end of the DNA substrate was monitored by adding 5 μl of running buffer containing 10 mM IPTG to the 50 μl of protein-containing buffer in the cuvette. This technique allowed dissociation of LacI bound (t_1/2 ≈ 1.6 sec) at the ends of the DNA substrate without significantly altering the concentrations of protein in solution. All experiments were performed at 25°C.

Analysis of Mlh1–Pms1 binding to the Msh2–Msh6–DNA complex was performed as follows. The cuvette containing the indicated bound DNA substrate was first equilibrated in running buffer lacking ATP and containing 30 nM LacI. This buffer was then replaced with 50 μl of running buffer lacking ATP and containing 20 nM of wild-type or mutant Msh2–Msh6, 30 nM LacI, and 40 nM Mlh1–Pms1, as indicated. Incubation was carried out for ≈3.5 min or until equilibrium was reached; Mlh1–Pms1 does not interact with Msh2–Msh6 and DNA in the presence of ADP alone (26). Next, ternary complex formation was initiated by adding 5 μl of running buffer containing 2.75 mM ATP, which resulted in a final concentration of 250 μM ATP. Binding was then monitored for ≈8 min. All experiments were performed at 25°C.