MutS stimulates the endonuclease activity of MutL in an ATP-hydrolysis-dependent manner

Authors


Abstract

In the initial steps of DNA mismatch repair, MutS recognizes a mismatched base and recruits the latent endonuclease MutL onto the mismatch-containing DNA in concert with other proteins. MutL then cleaves the error-containing strand to introduce an entry point for the downstream excision reaction. Because MutL has no intrinsic ability to recognize a mismatch and discriminate between newly synthesized and template strands, the endonuclease activity of MutL is strictly regulated by ATP-binding in order to avoid nonspecific degradation of the genomic DNA. However, the activation mechanism for its endonuclease activity remains unclear. In this study, we found that the coexistence of a mismatch, ATP and MutS unlocks the ATP-binding-dependent suppression of MutL endonuclease activity. Interestingly, ATPase-deficient mutants of MutS were unable to activate MutL. Furthermore, wild-type MutS activated ATPase-deficient mutants of MutL less efficiently than wild-type MutL. We concluded that ATP hydrolysis by MutS and MutL is involved in the mismatch-dependent activation of MutL endonuclease activity.

Abbreviations
CCC

covalently closed circular

CTD

C-terminal domain

L

linear

MMR

DNA mismatch repair

NTD

N-terminal domain

OC

open circular

ttMutL

Thermus thermophilus MutL

ttMutS

Thermus thermophilus MutS

Introduction

During DNA replication, DNA polymerases generate misincorporation, deletion and insertion errors [1-3]. To prevent these errors from being fixed as mutations, DNA mismatch repair (MMR), a highly conserved DNA repair system [4-6], recognizes and repairs mispairs and small insertion/deletion loops. In eukaryotic MMR, MutSα (MSH2/MSH6) and MutLα (MLH1/PMS2 and MLH1/PMS1 heterodimers in Homo sapiens and Saccharomyces cerevisiae, respectively) function during the initial steps of MMR. Mutations or epigenetic silencing of genes encoding MutSα and MutLα homologs can cause Lynch syndrome (also known as hereditary nonpolyposis colorectal cancer) in humans [7-9]. MutSα binds DNA and searches for a replication error [10-12], and recruits MutLα onto the DNA in cooperation with other MMR proteins. MutS and other MMR proteins direct MutLα to cleave the error-containing strand [13, 14]. MutS and MutL homodimers whose fundamental properties are similar to those of eukaryotic MutSα and MutLα are conserved in the majority of bacteria except for a few species, including Escherichia coli [15]. The E. coli-type MMR is characterized by a lack of endonuclease activity in MutL. In E. coli-type MMR, the nicking endonuclease activity is provided by MutH [16].

Structural and biochemical studies revealed that MutS homologs belong to the ABC ATPase superfamily [17-20]. Mismatch binding induces ATP uptake by MutS [21, 22] and stabilization of the ATP-bound form [23-25]. ATP binding induces a MutS conformational change, resulting in the formation of a clamp-like structure to slide along the DNA strand [26, 27]. In E. coli-type MMR, ATP hydrolysis of MutS is necessary for activation of MutH [28]. However, in eukaryotic and general bacterial-type MMR, the role of ATP hydrolysis by MutS is unclear.

MutL endonucleases possess nonspecific endonuclease activity and degrade lesion-less DNA in vitro [13, 14, 29-32], suggesting that MMR requires the sequence- or structure-nonspecific endonuclease activity to generate an excision entry point wherever it is needed. The regulatory mechanism for this apparently nonspecific endonuclease activity has been argued. MutL homologs belong to the GHL ATPase superfamily [33, 34] and contain the Bergerat ATP-binding fold in their N-terminal domain (NTD). The N-terminal ATPase domain of MutL tightly binds ATP, and ATP binding provokes the conformational change and dimerization of the NTD of MutL [34]. It has been reported that the addition of ATP causes stimulation or inhibition of MutL endonuclease activity in vitro [13, 14, 29, 35]. ATP-binding-dependent suppression of the endonuclease activity may protect lesion-less DNA from nonspecific degradation in the cell. However, ATP-dependent enhancement of the MutL endonuclease activity may be required when mismatch recognition and matchmaking by other MMR proteins are completed. In order to achieve mismatch-specific DNA incision, it is expected that interactions with other MMR proteins unlock the ATP-binding-dependent suppression of MutL endonuclease activity and then, ATP enhances the nick introduction.

MutL interacts with other MMR proteins such as the β clamp, clamp loader, and MutS [36-41]. In general bacterial-type MMR, interactions between the β clamp, MutS and MutL are important to recruit and stabilize the MutS–MutL complex [41, 42]. Furthermore, in eukaryotic-type MMR, the interaction between MutLα and PCNA (an eukaryotic counterpart of β clamp) is thought to be necessary for discrimination between newly synthesized and template strands [43]. Because these MMR proteins are involved in the upstream of nick introduction by MutL (Fig. 1), it can be expected that the ATP-bound form of MutL is activated by these proteins. In this study, we examined the effects of these MMR proteins on the activation of MutL, and investigated the roles of ATP binding and hydrolysis by MutS and MutL proteins in regulation of MutL endonuclease activity.

Figure 1.

The effect of MMR proteins on ttMutL endonuclease activity. (A–H) pUC119-mis was used as a substrate. (A) Endonuclease assay without ttMutL. Reaction performed in the presence of MnCl2 and ATP. (B) The percentages of reaction products in (A) are shown in a bar graph. (C) The endonuclease assay was carried out in the presence of MnCl2 and ATP. (D) The percentages of reaction products in (C) are shown. (E) The endonuclease assay was carried out without ATP in the presence of MnCl2. (F) The percentages of reaction products in (E) are shown. (G) The endonuclease assay was carried out in the absence of MnCl2 and in the presence of ATP. (H) The percentages of reaction products in (G) are shown. (I–L) pUC119 was used as a substrate. (I) The reaction was carried out in the presence of ATP. (J) The percentages of reaction products in (I) are shown in a bar graph. (K) The reaction was carried out in the absence of ATP. (L) The percentages of reaction products in (K) are shown. L, CL, B and S indicate ttMutL, clamp loader, β clamp, and ttMutS, respectively. Error bars indicate standard deviations. The concentrations of ATP, MgCl2, and MnCl2 used in this study were 750 μm, 5 mm and 5 mm, respectively.

Results

The cellular concentrations of MutL and MutS

It has been reported that the endonuclease activity of relatively low concentrations (500 nm) of MutL from Thermus thermophilus (ttMutL) and Aquifex aeolicus (aaMutL) was inhibited in the presence of ATP [29]. However, it was reported that higher concentrations (6.5 μm) of Aquifex aeolicus MutL was activated by the addition of ATP [35]. We showed that high concentrations of ttMutL were also activated by ATP (Fig. S2). Thus, these differences in the response of MutL to ATP was attributed to differences in MutL concentrations. Therefore, we first determined the concentration of ttMutL in T. thermophilus HB8.

T. thermophilus HB8 was cultured and harvested at each point in the growth curve (Fig. S3A). The cell lysates were western blotted to detect endogenous MutS from T. thermophilus (ttMutS) and ttMutL (Fig. S3B,C). Referencing the standard curve generated by the same experiment with recombinant ttMutS and ttMutL proteins (Fig. S3B,C), the cellular concentrations of ttMutS and ttMutL were determined to be ~ 500 and 300 nm at the logarithmic growth phase, respectively. Protein concentrations gradually decreased as growth progressed (Fig. S3B,C). Previously, it was demonstrated that the cellular concentrations of E. coli MutS and MutL are ~ 1000 and 600 nm, respectively. It was also reported that the concentrations of E. coli MutS and MutL decreased gradually as growth progressed [44]. Therefore the concentrations of ttMutL and ttMutS were comparable with those of E. coli MutS and MutL. Because DNA replication and MMR frequently occur during logarithmic growth, the concentration of ttMutL at the logarithmic phase is adequate for the endonuclease assay. Therefore, in this study, we conducted endonuclease assays with 500 nm ttMutS and ttMutL.

MutL endonuclease activity was inhibited by ATP

The endonuclease activity of ttMutL was analyzed by monitoring the conversion of a covalently closed circular (CCC) form of plasmid DNA to its open circular (OC) or linear (L) form. To examine the ability of ttMutL to distinguish homo- and heteroduplexes, we used perfectly matched plasmid DNA (pUC119) and the GT mismatched plasmid DNA (pUC119-mis) as substrates. The construction of pUC119-mis is described in Materials and Methods (Fig. S4A). To determine the electrophoretic mobility of the OC form of plasmid DNA, the recognition site for nickase Nb.Bpu10I was introduced into pUC119-mis. The purified CCC form of pUC119-mis was treated with EcoRI, PstI, or Bpu10I, and then electrophoresed on 1% agarose gel (Fig. S4B). Because pUC119-mis has a GT mismatch in the PstI site, the CCC form of pUC119-mis was resistant to digestion by PstI. The CCC form of pUC119-mis treated with EcoRI and Nb.Bpu10I was converted into the L and OC forms, respectively. The CCC, L and OC forms exhibited high, moderate and low mobility, respectively.

ATP tightly inhibited the endonuclease activity of ttMutL (Fig. S5A,B). The inhibition was observed in the presence of various concentrations of ATP and MnCl2 (Fig. S6A,B). The nicked and linearized products generated by ttMutL were approximately fivefold higher in the absence of ATP than in the presence of ATP (Fig. S5C). In Fig. S5C, plasmid DNA containing a GT mismatch was used as a substrate. With the perfectly matched pUC119 substrate, the endonuclease activity of ttMutL was also inhibited in the presence of ATP (Fig. S5D–F).

Addition of ATP did not suppress the endonuclease activity of ttMutL when MnCl2 was substituted with MgCl2 (Fig. S5G–I). The necessity of manganese ions for ATP regulations suggests that manganese may be a physiological cofactor for the endonuclease activity of ttMutL. This is also supported by previous findings that MutL homologs from Homo sapiens, Saccharomyces cerevisiae, Bacillus subtilis, A. aeolicus and Neisseria gonorrhoeae exhibit maximal endonuclease activity in the presence of MnCl2 [13, 14, 30-32].

The effect of MMR proteins on MutL endonuclease activity

To characterize the activation mechanism for ATP-bound ttMutL, we tested the effects of other MMR proteins on ttMutL endonuclease activity. We selected clamp loader (DNA polymerase III γ, τ, δ and δ′ subunits), β clamp and ttMutS because these proteins are expected to function upstream of nick introduction. Although the full clamp loader complex in a bacterial cell is composed of DNA polymerase III γ, δ, δ′, ψ and χ subunits [45], it was reported that a complex of γ, τ, δ, and δ′ subunits was able to load the β clamp onto DNA [46].

Our far-western blotting experiments implied that, in T. thermophilus, DNA polymerase III γ and τ subunits, β clamp and ttMutS interact with ttMutL (Fig. S7). These interactions were confirmed by coimmunoprecipitation analyses (Fig. S8). The far-western analyses support the interaction of the recombinant DNA polymerase III γ and τ subunits with the recombinant δ and δ′ subunits (Fig. S7A,B), indicating that the clamp loader complex was successfully reconstituted from recombinant subunits.

To investigate the effect of MMR proteins on the endonuclease activity of ttMutL, each MMR protein was preincubated with substrate DNA, and then ttMutL was added to the reaction mixture. Without ttMutL, the other MMR proteins showed no detectable activity to generate nicked or linearized plasmid DNA (Fig. 1A,B). When the reaction mixture contained a mismatch, ATP and MnCl2, ttMutS stimulated ttMutL endonuclease activity: the rate of accumulation of the products generated by ttMutL was sixfold higher in the presence of ttMutS than in its absence (Fig. 1C,D). No stimulation by ttMutS was observed when the reaction was performed without ATP (Fig. 1E,F). The stimulating effect of ttMutS was also abolished by substituting MnCl2 with MgCl2 (Fig. 1G,H). Furthermore, ttMutS did not induce ttMutL endonuclease activity when mismatch-free plasmid DNA was used as a substrate (Fig. 1I–L), indicating that the ttMutS-dependent stimulation of ttMutL requires a mismatch. Unlike ttMutS, DNA polymerase III subunits (γ, τ, δ, δ′ subunits and β clamp) showed no effect on the endonuclease activity of ttMutL (Fig. 1C–H).

It has been shown that the clamp and clamp loader exhibit an enhancing effect on eukaryotic MutL endonuclease when the substrate DNA contains a discontinuity such as a nick or gap. Therefore, we tried to detect the effect of the clamp and clamp loader on the endonuclease activity of ttMutL by using the nicked pUC119-mis plasmid DNA as a substrate. Curiously, the β clamp and clamp loader neither enhanced nor conferred any strand specificity to the nicking nuclease activity of ttMutL (Fig. S9). To confirm that the clamp loader and the β clamp were active under the experimental condition, we reconstituted the loading reaction of the β clamp on immobilized circular DNA (Fig. S10) [47]. Importantly, the β clamp was preferentially recovered with DNA when the clamp loader was included in the reaction, demonstrating that both the β clamp and the clamp loader are active. Although the loading efficiency was not stimulated by the presence of a nick on the DNA, the eukaryotic PCNA clamp is also shown to be loaded on a closed circular DNA [43]. Consistent with the clamp loading assay, the ATPase activity of the clamp loader complex, which is an indirect readout of the β clamp loading reaction [48-50], was equally stimulated by closed circular and OC DNAs (Fig. S11). Correctively, these data suggest that, although they interact with ttMutL, the clamp loader and β clamp do not significantly alter either the activity or specificity of the ttMutL endonuclease (see Discussion).

Activation of ttMutL required ATP hydrolysis by ttMutS

To determine whether ATP hydrolysis is required for the stimulation of ttMutL endonuclease activity, the endonuclease assay was performed in the presence of ADP or AMPPNP, a nonhydrolyzable analog of ATP. Only slight stimulation by ttMutS was observed in the presence of ADP (Fig. 2A,G) or AMPPNP (Fig. 2B,G), suggesting that the activation requires ATP-hydrolysis energy generated by ttMutS and/or ttMutL.

Figure 2.

The effect of ATP hydrolysis by ttMutS and MutL on the endonuclease activity of ttMutL. (A) The reaction was performed in the presence of ADP. (B) The reaction was performed in the presence of AMPPNP. (C–F) The reaction was performed in the presence of ATP. (C) ttMutS K597M was used for the reaction instead of ttMutS WT. (D) ttMutS E671A was used for the reaction instead of ttMutS WT. (E) ttMutL E28A was used for the reaction instead of ttMutL WT. (F) ttMutL D57A was used for the reaction instead of ttMutL WT. (G) The percentages of reaction products in (A–F) and ttMutL WT–ttMutS WT in the presence of ATP are shown. (H) The effect of ttMutS on the endonuclease activity of ttMutL. The velocity of generating OC form of plasmid DNA was plotted against concentrations of ttMutL WT or E28A mutant. Red, blue, yellow, and green plots are the results of the endonuclease assay using ttMutL WT without ttMutS, ttMutL WT with 500 nm ttMutS WT, ttMutL E28A with 500 nm ttMutS WT, and ttMutL WT with 500 nm ttMutS K597M, respectively. The data were fitted to the Michaelis–Menten equation using igor pro 4.03, and the theoretical curves are shown. Because our method does not detect the incision reaction on the OC form of plasmid DNA, accurate parameters were not determined.

Because AMPPNP is not identical to ATP, we could not rule out the possibility that the difference observed with ATP and AMPPNP was due to the different binding modes of the each nucleotide to ttMutS and/or ttMutL. Then, we created the ATPase-deficient mutants of ttMutS and ttMutL by mutagenizing putative catalytic amino acid residues in order to further examine the effect of ATP hydrolysis on the stimulation of ttMutL endonuclease. In E. coli MutS and MutL, several amino acid residues have been shown to be involved in ATP hydrolysis [18, 20, 34]. The amino acid residues participating in ATP hydrolysis in ttMutS and ttMutL were predicted based on an alignment of ATPase motifs (Fig. S12) and mutagenized. The ttMutS mutant derivatives K597M and E671A showed kcat values 5- and 18-fold smaller than WT, respectively (Table 1). It is known that the conserved lysine and glutamate residues in Walker A and B motifs are crucial for the ATPase activity; however, it is also reported that mutagenizing these residues to others results in only partial abolishment of the ATPase activity [51-53]. The Km value of the K597M mutant was similar to that of WT and was sufficiently smaller than the ATP concentration (750 μm) used for the endonuclease assay. Therefore, it can be expected that the majority of K597M mutant molecules were in the ATP-binding form under the reaction conditions. As shown in Table 1, ttMutL E28A exhibited no ATP-hydrolysis activity. It was reported that the ATP-binding ability of the corresponding mutant E29A from E. coli MutL was 10-fold lower than that of WT [34]. By contrast, the kcat value of D57A was hardly distinguishable from that of WT (Table 1). Because the Km value of D57A was comparable with the ATP concentration used for the endonuclease assay, it can be expected that the half of D57A mutant molecules were in the ATP-binding form and were only capable of half-maximum velocity of ATP hydrolysis in the assay. The kcat values of ttMutS and ttMutL were seven- and twofold higher at 70 than at 55 °C, respectively (data not shown). It is known that the enzymes from T. thermophilus often exhibit maximum activity at ~ 70 °C. Therefore, it is indicated that ATPase activities observed here were derived from ttMutS and ttMutL and not from the contaminated proteins.

Table 1. Kinetics of ATPase activity. The ATPase assays of MutS and MutL were performed at 70 °C for 10 and 30 min, respectively. n.d., not detected
ProteinKmm)kcat (min−1)
ttMutS WT50.7 ± 9.879.19 ± 1.81
K597M53.9 ± 3.551.92 ± 0.120
E671A457 ± 2670.522 ± 0.0912
ttMutL WT461 ± 1380.614 ± 0.123
E28An.d.n.d.
D57A785 ± 2020.564 ± 0.125

The stimulation of ttMutL endonuclease activity by K597M (Fig. 2C,G) and E671A (Fig. 2D,G) mutants of ttMutS was at the lower limit of detection. These results strongly indicate that ATP hydrolysis by ttMutS is essential for ttMutL activation. Although the ATPase-deficient ttMutL mutants were slightly activated by ttMutS WT (Fig. 2E–G), the efficiencies of ttMutS-dependent activation of the ttMutL mutants were twofold lower than that of ttMutL WT. The kapp value determined using WT ttMutS and ttMutL was seven times higher than that determined using mutants of ttMutS and ttMutL (Fig. 2H). Thus, ATP hydrolysis by ttMutS and ttMutL was involved in the activation of ttMutL. ATP hydrolysis by ttMutS more significantly contributed to the activation of ttMutL than the ATP hydrolysis by ttMutL under this experimental condition.

No stimulation was detected in the NTD-deleted mutant of ttMutL

To analyze the function of the N-terminal ATPase domain of ttMutL, we constructed an NTD deletion mutant, leaving only the C-terminal domain (CTD) of ttMutL. It has been reported that the NTD has MutS-interacting region [54], and the CTD contains endonuclease motifs and is sufficient for the endonuclease activity [2, 13, 15, 29-31] (Fig. S13). We demonstrated that the ATP-dependent inhibition of ttMutL endonuclease was not observed with the ttMutL CTD (Figs 3 and S13). Interestingly, ttMutS was unable to activate ttMutL CTD even in the presence of a mismatch, ATP and MnCl2, indicating that the NTD of ttMutL is required for the activation of ttMutL by ttMutS (Fig. 3A,B). This result may indicate that ttMutS-dependent activation of ttMutL involves the interdomain interaction within ttMutL or that ttMutL NTD is required for interaction with ttMutS. We next investigated the interaction of ttMutS with the full-length and CTD of ttMutL by surface plasmon resonance. The results demonstrated that ttMutS was capable of interacting with full-length ttMutL and the CTD (Fig. S14), although ttMutS interacted less stably with ttMutL CTD than with full-length ttMutL. These results suggest that interdomain interaction within ttMutL plays a significant role in ttMutS-dependent activation of the endonuclease. This notion has been proposed by the previous report regarding the interdomain interaction in A. aeolicus MutL [55].

Figure 3.

ttMutL CTD was not activated by ttMutS. The endonuclease assay was performed using ttMutL CTD. (A) The reaction was carried out in the presence of ATP. (B) The percentages of reaction products in (A) are shown in a bar graph. (C) The reaction was carried out in the absence of ATP. (D) The percentages of reaction products in (C) are shown.

Trans-activation by ttMutS

In E. coli MMR studies, cis- and trans-activation assays have been employed to investigate the activation mechanism for MutH endonuclease [20]. In the cis-activation assay system, a mismatch and cleavable site reside on the same DNA molecule, whereas in the trans-activation assay they reside on separate DNA molecules. In previous cis experiments, we verified that ttMutS cis-activates the endonuclease activity of ttMutL (Fig. 1). To further explore the activation mechanism for ttMutL endonuclease, we performed the trans-activation assay, in which the mismatch is not on the cleavable plasmid DNA but on a separate double-stranded linear DNA molecule. In this experiment, we sought to determine whether ttMutS stimulates the endonuclease activity of ttMutL bound on a different DNA molecule.

To trap ttMutS on the linear mismatched DNA, we prepared a 120-bp GT mismatched DNA on which both 5′-ends were blocked with biotin-streptavidin (Materials and Methods). In a DNase I footprinting experiment, we confirmed that ttMutS was trapped on the 120-bp heteroduplex DNA (Fig. S15A) but not on the 120-bp homoduplex DNA (Fig. S15B). Then, ttMutS trapped by the 120-bp GT mismatched or perfectly matched double-stranded DNA was added to the reaction mixture containing ttMutL and plasmid DNA substrate, and incision of the plasmid DNA was monitored (Fig. 4A).

Figure 4.

Trans-activation analysis using homoduplex plasmid DNA. (A) Scheme for trans-activation analysis. Briefly, ttMutS and ttMutL were separately preincubated with 120-bp double-stranded DNA and CCC form of homoduplex plasmid DNA (pUC119) at 55 °C for 10 min, respectively. Then, preincubated ttMutS and ttMutL solutions were mixed and incubated at 55 °C for 10 min. (B) Preincubation of ttMutS was carried out with 120-bp GT mismatched DNA. (C) The percentages of reaction products in (B) are shown. (D) Preincubation of ttMutS was carried out with 120-bp homoduplex DNA. (E) The percentages of reaction products in (D) are shown.

We performed the trans-activation assay with perfectly matched plasmid DNA and found that ttMutS trapped on the 120-bp GT mismatched DNA was unable to stimulate ttMutL endonuclease activity against the homoduplex plasmid DNA (Fig. 4B,C). As a control, 120-bp homoduplex DNA was used instead of heteroduplex DNA, and no activation was observed (Fig. 4D,E). These results show that ttMutS is capable of activating ttMutL endonuclease activity in cis, but not in trans.

Discussion

We found that intracellular concentration of ttMutL is ~ 300 nm (Fig. S3C). Then, we employed 500 nm ttMutL for in vitro experiments in this study. Our experiments revealed that the endonuclease activity of 500 nm ttMutL was suppressed by the addition of ATP in the absence of mismatch and other MMR proteins, although much higher concentrations of ttMutL were activated by ATP under the same reaction conditions (Fig. S2). Thus, our results suggest that the endonuclease activity of a physiological concentration of ttMutL is regulated to prevent nonspecific nick introduction by binding of ATP until mismatch recognition and matchmaking by other MMR proteins are completed. There might be two possible interpretations for the ATP-dependent inhibition of ttMutL endonuclease activity. First is that the association of ttMutL NTD caused a conformational change to form an ATP-binding state inadequate to exhibit the activity. Second, it might be also possible that the chelating activity of ATP against Mn2+ interfered the activity. However, no ATP-dependent inhibition was observed for the endonuclease activity of ttMutL CTD, which lacks the ATP-binding domain (Fig. 3A–D). Therefore, we concluded that the ATP-dependent inhibition of the endonuclease activity was caused by ATP-binding to ttMutL. As shown in Table 1, Km of ttMutL to ATP is much smaller than the intracellular concentration of ATP (~ 10 mm in a bacterial cell [56]), and the kcat value is extremely small. Therefore, the majority of intracellular ttMutL molecules likely exist in an ATP-bound form, in which the endonuclease activity is kept inactive.

T. thermophilus DNA polymerase III subunits and ttMutS were chosen as the candidate activator of ATP-bound ttMutL because far-western and coimmunoprecipitation experiments revealed an interaction between these proteins and ttMutL (Figs S7 and S8). We investigated the effects of these proteins on the endonuclease activity of ttMutL in the presence of ATP. The endonuclease activity of ttMutL was stimulated by ttMutS in the presence of a mispair and ATP (Fig. 1C,D), suggesting that the signal of mismatch binding by ttMutS was transmitted to ttMutL in an ATP-dependent manner. However, DNA polymerase III subunits were unable to activate ttMutL under our experimental conditions (Fig. 1 and Fig. S9); therefore, the function of the interaction between the DNA polymerase III subunits and ttMutL remains unknown. It has been reported that in E. coli clamp loaders, DNA polymerase III γ, τ, δ and δ′ subunits, interact with MutL, but the function of these interactions is unclear [36, 57]. However, in eukaryotes, PCNA stimulates the endonuclease activity of MutLα, and the interaction between PCNA and MutLα is reported to be essential for MMR to distinguish between template and newly synthesized strands [43]. In this study, β clamp did not stimulate the endonuclease activity of ttMutL although the intracellular interaction between β clamp and ttMutL was implied by the immunoprecipitation analysis (Fig. S8). Because the β clamp-interaction motif is hardly found in the amino acid sequence of ttMutL [15], it can be suspected that the observed coimmunoprecipitation of β clamp with ttMutL reflects their indirect interaction that is mediated by ttMutS. It is known that MutS interacts with both MutL and β clamp, therefore anti-ttMutL Ig may catch the ttMutS–β clamp complex. Similarly, ttMutS–ttMutL complex may be coimmunoprecipitated with anti-β clamp Ig. We postulate that the interactions between bacterial MutL and DNA polymerase III subunits mediate the loading of MutL onto DNA for strand discrimination, or loading of the DNA polymerase III complex for DNA resynthesis. Further experiments are necessary to test this hypothesis.

Intriguingly, the ttMutS-dependent activation of ttMutL was observed in the presence of ATP but not ADP and AMPPNP (Fig. 2A,B,G), strongly indicating that the activation requires ATP hydrolysis. Analyses of ATPase-deficient mutants of ttMutS and ttMutL revealed that the activation of ttMutL required the ATP-hydrolysis energy generated by both ttMutS and ttMutL (Fig. 2C–H), although ATP hydrolysis by ttMutS was more critical for the activation than that by ttMutL. Several previous experiments have suggested that ATP hydrolysis by MutL is needed for the mismatch-specific endonuclease activity of MutL. Even in the absence of MutS and mismatch, ATP (but not AMPPNP) promotes the endonuclease activity of relatively high concentrations of MutL [35], suggesting the potential of ATP hydrolysis by MutL to enhance its endonuclease activity. Our experiments verified that the ATPase activity is involved in the endonuclease activity of physiological concentrations of ttMutL in the presence of a mismatch and ttMutS. By contrast, there has been no evidence support the necessity of ATP hydrolysis by MutS for the activation of MutL endonuclease activity. In this study, we demonstrated that ATP hydrolysis by ttMutS is required for mismatch-specific incision by ttMutL.

The ATPase and nuclease catalytic sites reside in the NTD and CTD of MutL, respectively, and an ATPase cycle-dependent interdomain interaction has been proposed [55, 58]. ATP-hydrolysis energy generated by the MutL NTD may be transmitted to the CTD via interdomain interaction. In our experiments, the difference between WT and ATPase-deficient mutants of ttMutL was observed only in the presence of ttMutS and a mismatch, suggesting ATP hydrolysis by ttMutL was activated by interaction with a ttMutS–mismatch complex. Furthermore, the ATPase-deficient mutant of ttMutS did not affect the endonuclease activity of ttMutL. These results suggest ATP hydrolysis by the ttMutS–mismatch complex enhances the ATPase activity of ttMutL. Subsequent ATP hydrolysis by the ttMutL NTD would cause a conformational change in the ttMutL CTD via interdomain interaction.

SPR experiments revealed that ttMutS interacts with the ttMutL CTD (Fig. S14). Therefore, the stimulation of the endonuclease activity would involve direct interaction between ttMutS and the endonuclease domain of ttMutL. However, our results also clarified that unlike full-length ttMutL, the CTD of ttMutL was not activated by ttMutS (Fig. 3). The ATP-hydrolysis- and interdomain-interaction-dependent conformational change in the ttMutL CTD may be necessary to enable activation by ttMutS.

In E. coli, MutS forms a complex with MutL, and then stimulates MutH endonuclease activity. To determine whether translocation of E. coli MutS is required for activation of MutH, a trans-activation assay was employed, wherein the mismatch and MutH-recognizing sequence are place on separate DNA molecules [20]. Previous studies have revealed that E. coli MutS can activate MutH in trans, indicating that translocation of E. coli MutS is necessary for activation of the downstream incision event. In this study, we demonstrated that ttMutS enhanced the endonuclease activity of ttMutL in cis (Fig. 1) but not in trans (Fig. 4). This result strongly suggests that the interaction between ttMutS and ttMutL requires both proteins to be on the same DNA molecule. This raises the following two hypotheses. The first is that the translocation of ttMutS along the DNA from the mismatch is essential for activation of ttMutL. The second is that mismatch-bound ttMutS does not translocate but recruits ttMutL onto the mismatch to introduce a nick. ttMutL has no sequence specificity, unlike E. coli MutH, and the incision sites for ttMutL were not identified in our experiments. Therefore, we could not exclude either hypothesis. However, previous studies have revealed that, in eukaryotic MMR, MutSα travels along the DNA after mismatch recognition [12], suggesting ttMutS translocation is necessary for activation of ttMutL endonuclease activity. This may be correlated with the strand discrimination mechanism. In E. coli MMR, the strand discrimination signal is included in the MutH recognition sequence (GATC site) as a hemimethylated site. Therefore, E. coli MutS or other MMR proteins need not convey the strand discrimination signal to MutH. By contrast, in MMR of eukaryotes and most bacteria, MutL incision sites contain no strand discrimination signal, and MutS or other MMR proteins must transmit the signal to MutL. Translocation of ttMutS may have a role in conveying the strand discrimination signal.

As mentioned earlier, the MMR reaction requires the mismatch- and daughter-strand-specific nicking endonuclease activity of MutL. In this study, we succeeded in unraveling the activation mechanism that renders the endonuclease activity of ttMutL mismatch specific. It has been reported that, in the presence of clamp and clamp loader, eukaryotic MutL endonuclease specifically incises the discontinuity-containing strand of the duplex [43]. By contrast, the β clamp and clamp loader exhibited no effect on the strand specificity of the endonuclease activity of ttMutL (Fig. S9). A caveat in our experiments is that additional factor(s) might be required for the β clamp and clamp loader to confer strand specificity to the MutL endonuclease. Although such a possibility needs to be carefully examined in the future, at this point we favor the possibility that ttMutL is not activated by the β clamp, because ttMutL lacks the β-clamp-interacting motif that is conserved among most bacterial MutL homologs [15]. If this is the case, an as yet unidentified mechanism might be involved in determining the strand to be repaired in T. thermophilus. Complete understanding of the strand discrimination mechanism in this organism must await full reconstitution of the strand specific MMR reaction in vitro.

In conclusion, we elucidated the fundamental mechanism for the activation of MutL endonuclease in MMR, which is expected to be conserved among most bacteria. Based on the results of this and previous studies, we propose a model for the initial steps of the MMR pathway in most bacteria (Fig. 5). ADP-bound MutS binds DNA and searches for a mismatch along the DNA [12, 59]. MutS recognizes a mismatch and a conformational change is induced by the exchange of ADP with ATP, triggering the sliding clamp mode of MutS. Mismatch binding allows MutS to form the active complex with MutL, and then MutS hydrolyzes ATP and transmits the ATP-hydrolysis energy to the MutL NTD. The MutL NTD hydrolyzes ATP and causes a conformational change in the CTD via interdomain interaction, which allows the CTD to be stimulated by MutS. The activated CTD of MutL introduces a nick into the DNA.

Figure 5.

The model of MutL activation. ADP-bound MutS binds DNA and searches for a mismatch. Mismatch-bound MutS exchanges ADP with ATP, resulting in formation of the clamp-like structure. The ATP-bound form of MutS interacts with the ATP-bound form of MutL. Mismatch-bound MutS hydrolyzes ATP upon interaction with MutL. The MutL NTD hydrolyzes ATP, and the ATP-hydrolysis energy is utilized to stimulate the endonuclease activity of MutL CTD. A nick is introduced by activated MutL.

Experimental procedures

Construction of MutS and MutL mutant

The pET3a/ttha1324 (ttMutS) and pET11a/ttha1323 (ttMutL) plasmids were obtained from RIKEN BioResource Center (Tsukuba, Japan). pET3a/ttha1324 was used as a template to generate the overexpression plasmids for ttMutS K597M and E671A by QuickChange mutagenesis (Stratagene, La Jolla, CA, USA). The primers used to generate the expression plasmid for ttMutS K597M were 5′-CCCAACATGGCGGGGATGTCCACCTTCCTCCGC-3′ and 5′-GCGGAGGAAGGTGGACATCCCCGCCATGTTGGG-3′, and the primers for the expression plasmid of ttMutS E671A were 5′-GTCCTCCTGGACGCGGTGGGCCGGGGC-3′ and 5′-GCCCCGGCCCACCGCGTCCAGGAGGAC-3′. The pET11a/ttha1323 was used as a template to generate the overexpression plasmids for MutL E28A and D57A by QuickChange mutagenesis (Stratagene). The primer sets used to generate the expression plasmids for MutL E28A and D57A mutants were 5′-GGACGCCGTGCGGGCGCTTCTGGAAAACGCC-3′ and 5′-GGCGTTTTCCAGAAGCGCCCGCACGGCGTCC-3′, and 5′-GCTTGTGGTGGAGGCCGACGGGGAGGGGATC-3′ and 5′-GATCCCCTCCCCGTCGGCCTCCACCACAAGC-3′, respectively. A DNA fragment expressing the T. thermophilus MutL CTD was generated by PCR using pET-11a/ttha1323 as a template. The forward and reverse primers used for the amplification were 5′-ATATCATATGGCCCTCCCCGAGCCCAAGCCCCTC-3′ and 5′-ATATAGATCTTTAAGGTTCTCGGGGTAGAGGTG-3′, respectively. The forward and reverse primers contained NdeI and BglII sites, respectively (underlined). The amplified T. thermophilus MutL CTD gene fragment was ligated into the NdeI and BamHI site of pET–HisTEV (Novagen, Madison, WI, USA) to obtain the expression plasmid for histidine-tagged T. thermophilus MutL CTD.

Nicking endonuclease assay

A 150 ng CCC sample of DNA in 10 μL buffer containing 50 mm Hepes/KOH, 1 mm dithiothreitol, 40 μg·mL−1 BSA, 4% glycerol, 0.1 mm EDTA, 750 μm ATP, 5 mm MgCl2 and 5 mm MnCl2, pH 7.5 was preincubated with 750 nm ttMutS, 40 nm DNA polymerase III δ and δ′ subunits, 60 nm γ and τ subunits, and/or 300 nm β clamp at 55 °C for 10 min. Subsequently, 5 μL 1.5 μm ttMutL in the same buffer containing 180 mm KCl was added to the reaction mixture and incubated at 55 °C for 10 min. Reaction conditions are detailed in each figure legend. In the assay using various concentrations of ttMutL, 100 ng CCC DNA in 10 μL buffer containing 50 mm Hepes/KOH, 1 mm dithiothreitol, 40 μg·mL−1 BSA, 4% glycerol, 0.1 mm EDTA, 750 μm ATP, 5 mm MgCl2 and 5 mm MnCl2, pH 7.5 was preincubated with or without 750 nm WT or K597M mutant of ttMutS at 55 °C for 10 min. Subsequently, 5 μL of various concentrations of WT or E28A mutant of ttMutL in the same buffer was added to the reaction mixture and incubated at 55 °C for 3 or 10 min. The final concentrations of WT and E28A mutant of ttMutL are indicated in the figure legends. To observe initial velocity, we stopped the reaction before linear form of plasmid DNA was generated. The reaction was terminated by the addition of 5 μL of stop solution containing 0.35% SDS, 0.3 mg·mL−1 proteinase K, 400 mm KCl and 5 mm MgCl2, and the mixture was incubated at 25 °C for 15 min. The samples were electrophoresed on a 1% agarose gel in buffer containing 25 mm Tris/HCl, 25 mm boric acid, 0.5 mm EDTA and 0.5 μg·mL−1 ethidium bromide. The DNA bands were visualized by UV irradiation at 254 nm and quantified using imagej software (http://rsb.info.nih.gov/ij/).

ATPase assay

ATPase activity was assayed in 10 μL reaction mixture containing 50 mm Hepes/KOH, 5 mm MgCl2, 1 mm dithiothreitol, 4% glycerol, 40 μg·mL−1 BSA, 500 nm MutS or MutL, 10 μm to 2 mm ATP and 22 nm [32P]ATP[γP], pH 7.5. Reactions were initiated by addition of hot and cold ATP, incubated at 70 °C for 10 or 30 min, and terminated by addition of an equal volume of phenol/chloroform/isoamylalcohol (25 : 24 : 1) and 1 μL of 10 mm EDTA. Aqueous solutions were recovered by centrifugation, and then 0.5 μL of each samples were spotted onto a polyethyleneimine cellulose plate (Merck KGaA, Darmstadt, Germany) and analyzed by thin-layer chromatography. Intact [32P]ATP[γP] and released radiolabeled phosphate were separated by developing the thin-layer chromatography plate in the buffer containing 0.5 m formic acid and 0.25 m LiCl at room temperature for 5 min. The thin-layer chromatography plate was dried and placed in contact with an imaging plate to visualize and analyze the spots using BAS2500 imaging analyzer (Fuji Photo Film Co., Kanagawa, Japan). The initial rate was calculated by quantifying the proportion of released phosphate to unreacted ATP, and kinetic parameters were determined from the Michaelis–Menten equation.

Trans-activation analysis

The 120-bp double-stranded DNA with a GT mismatch was obtained by hybridizing the single-stranded DNA, 120-2T with the complementary single-stranded DNA 120-1G (Table S1). The GT mismatched double-stranded DNA (2.25 pmol, 120 bp) was preincubated with 750 nm MutS in reaction buffer (10 μL) containing 50 mm Hepes/KOH, 1 mm dithiothreitol, 40 μg·mL−1 BSA, 4% glycerol, 0.1 mm EDTA, 750 μm ATP, 5 mm MgCl2 and 5 mm MnCl2, pH 7.5, at 55 °C for 10 min. A 5-μL aliquot of 1.5 μm MutL in the same buffer containing 180 mm KCl and 48 fmol CCC DNA with or without a GT mismatch was then added to the reaction mixture and incubated at 55 °C for 10 min. The reaction was terminated by the addition of 5 μL of termination mixture containing 0.35% SDS, 0.3 mg·mL−1 proteinase K, 400 mm NaCl and 5 mm MgCl2. The samples were electrophoresed on a 1% agarose gel in buffer containing 25 mm Tris/HCl, 25 mm boric acid, 0.5 mm EDTA and 0.5 μg·mL−1 ethidium bromide. The DNA bands were visualized by UV irradiation at 254 nm.

Acknowledgements

This work was supported by JSPS KAKENHI (Series of single-year grants) Grant Number 23‧2064. We thank Dr Hitoshi Iino and Dr Noriko Nakagawa for their valuable discussions on this study.

Ancillary