The effect of O6-methylguanine DNA adducts on the adenosine nucleotide switch functions of hMSH2-hMSH6 and hMSH2-hMSH3.

The human homologs of prokaryotic mismatch repair have been shown to mediate the toxicity of certain DNA damaging agents; cells deficient in the mismatch repair pathway exhibit resistance to the killing effects of several of these agents. Although previous studies have suggested that the human MutS homologs, hMSH2-hMSH6, bind to DNA containing a variety of DNA adducts, as well as mispaired nucleotides, a number of studies have suggested that DNA binding does not correlate with repair activity. In contrast, the ability to process adenosine nucleotides by MutS homologs appears to be fundamentally linked to repair activity. In this study, oligonucleotides containing a single well defined O(6)-methylguanine adduct were used to examine the extent of lesion-provoked DNA binding, single-step ADP --> ATP exchange, and steady-state ATPase activity by hMSH2-hMSH3 and hMSH2-hMSH6 heterodimers. Interestingly, O(6)-methylguanine lesions when paired with either a C or T were found to stimulate ADP --> ATP exchange, as well as the ATPase activity of purified hMSH2-hMSH6, whereas there was no significant stimulation of hMSH2-hMSH3. These results suggest that O(6)-methylguanine uniquely activates the molecular switch functions of hMSH2-hMSH6.


INTRODUCTION
and expression of a wild type MMR gene has been shown to restore the sensitivity to DNA damaging agents (21-23). These and other findings suggest a direct role for the MMR system in potentiating the toxicity of DNA damaging drugs such as MNU and MNNG.
The mechanism of MMR potentiated toxicity to O 6 -methylguanine lesions remains enigmatic although a number of models have been proposed. One model is based on the observation that replication past O 6 -methylguanine lesions results in the misincorporation of a nucleotide opposite the lesion (24). The mismatched nucleotide, often a thymidine, then serves as a substrate for repair directed to the newly replicated strand by hMSH2-hMSH6. This newly synthesized region of DNA continues to misincorporate a nucleotide opposite the O 6methylguanine lesion, which is subsequently excised and re-synthesized. Repeated cycles of binding, excision and re-synthesis is suggested to result in the accumulation of strand breaks, cell cycle arrest and apoptosis: termed the "futile-cycle model" (17,25,26). Cells lacking MMR are thought to escape such futile cycles of repair and display a damage tolerance (resistant) phenotype. A second model proposes that the binding of mismatch repair proteins to damaged bases may shield the adduct from cellular repair processes allowing lesions to persist in the genome: termed "binding-occlusion model" (27). Yet a third model has suggested that MMR proficient cells are capable of inhibiting replication past certain DNA adducts leading to cell cycle arrest and a cytotoxic response (28). Finally, an alternative to these models suggests that specific types of DNA damage are recognized by the MutS homologs, which provokes the exchange of ADP→ATP resulting in the formation of a sliding clamp associated with the DNA by guest on April 27, 2019 http://www.jbc.org/ Downloaded from adjacent to the lesion. In the absence of a targeted repair event, a threshold number of DNAbound sliding clamps are proposed to ultimately signal apoptosis (29).
Biochemical analysis has demonstrated that mismatched nucleotides stimulate the intrinsic ATPase activity of both hMSH2-hMSH3 and hMSH2-hMSH6 (10,(30)(31)(32). This intrinsic ATPase is controlled by mismatch-provoked ADP→ATP exchange which results in a conformational transition and the formation of a hydrolysis-independent sliding clamp on the DNA. However, it is not clear whether the same biochemical process(es) operate on DNA containing adducts such as O 6 -methylguanine. Here we have explored the biochemical interaction of purified hMSH2-hMSH3 and hMSH2-hMSH6 heterodimeric complexes with oligonucleotides containing a single O 6 -methylguanine adduct at a defined position. It was our goal to understand the distinctions between mispair and lesion recognition by the human MutS homologs toward clarifying the models which have been proposed to explain the influence of MMR on the toxicity of such lesions. It would appear that the defining biochemical feature that distinguishes these models is the ability of the MutS homologs to recognize O 6 -methylguanine lesions in the context of a homoduplex substrate and to be displaced by ADP→ATP exchange.
Our results appear most consistent, but not uniquely confined to, the direct apoptosis signaling model.

MATERIALS AND METHODS
Protein Purification: Recombinant hMSH2-hMSH3 and hMSH2-hMSH6 protein complexes were overexpressed in SF9 cells using the pFastBac dual expression vector (GIBCO-BRL) and purified as previously described (10,30). Protein concentration was determined by the Bradford assay using globular protein standards (Biorad). glycerol. Samples were applied to pre-wet nitrocellulose filters (Millipore HAWP02500, 25mm, 0.45um) and washed with 4 ml of wash buffer. Filters were dried and Cerenkov counted. Protein concentrations were set to 10-20% total binding in the absence of any unlabelled DNA competitor which was 12.5 nM hMSH2-hMSH6 for radiolabelled G/T and 25 nM hMSH2-hMSH3 for radiolabelled +(CA) DNA substrates.

RESULTS
Recognition of DNA Substrates -Purified hMSH2-hMSH6 and hMSH2-hMSH3 were assessed for their ability to bind to duplex DNA oligonucleotides containing a single O 6 -methylguanine adduct at a defined position paired with either cytosine (mG/C) or thymine (mG/T) nucleotides ( Figure 1). A number of reports have demonstrated that hMSH2-hMSH6 binds primarily to single base mispairs in DNA with the highest apparent binding affinity for G/T mismatches (33)(34)(35)(36). In addition, hMSH2-hMSH6 has been shown to bind DNA containing a variety of adducts including O 6 -methylguanine (11). In contrast, hMSH2-hMSH3 recognizes primarily larger insertion deletion loop-type (IDL) mismatched DNA (34). While biochemical studies of hMSH2-hMSH3 on IDL's have been reported (10), its affinity for DNA adducts is not known.
We examined the relative gel-shift binding of hMSH2-hMSH6 and hMSH2-hMSH3 to the five model DNA substrates (Figure 1 and 2). A comparison of hMSH2-hMSH6 binding affinity suggests G/T >> +CA > mG/T > mG/C ≈ G/C. In contrast, the hMSH2-hMSH3 binding affinity suggests +(CA) >> mG/T ≈ mG/C ≈ G/T ≈ G/C. However, these results should be tempered by the lack of equivalent binding saturation by the DNA substrates which suggests complex binding functions (see discussion).
We have used competition studies to overcome such complex binding activities and to gauge the relative binding specificity of the human MutS homologs for DNA (33). Our previous studies demonstrated that hMSH2-hMSH3 exhibited a substantial affinity for DNA containing a +(CA) mismatch (10) while hMSH2-hMSH6 exhibited a substantial affinity for DNA containing a G/T mismatch substrates (33). Thus, we assessed the relative affinity of O 6 -methylguanine containing DNA substrates for hMSH2-hMSH3 and hMSH2-hMSH6 relative to their binding to 20 nM of the 32 P-labelled model high affinity substrates, respectively ( Figure 3). We determined that the amount of unlabelled test DNA competitor required to reduce binding of hMSH2-hMSH6 to 32  and mG/C (375 nM). These observations suggest that hMSH2-hMSH6 displays a significant affinity for both mG/C and mG/T containing DNA substrates while hMSH2-hMSH3 does not appear to exhibit any significant affinity for the O 6 -methylguanine DNA substrates.
O 6 -methylguanine-DNA Stimulation of the ATPase Activity -DNA containing mispaired nucleotides has been shown to stimulate the intrinsic ATPase activity associated with both hMSH2-hMSH3 and hMSH2-hMSH6 (10,30). However, it is unclear whether DNA containing nucleotide damage stimulates the intrinsic ATPase activity of the heterodimeric complexes in a similar manner. We determined the ability of the mG/C and mG/T substrates to stimulate the ATPase activity of hMSH2-hMSH3 and hMSH2-hMSH6 ( Figure 4 and Table 1). The steadystate ATPase activity of hMSH2-hMSH6, relative to that of homoduplex (G/C) DNA, was increased for both mG/C and mG/T substrates. It is interesting to note that when O 6methylguanine was paired with T (mG/T), the steady-state ATPase was dramatically reduced compared to the G/T substrate, while the mG/C substrate increases steady-state ATPase compared to the G/C substrate. Surprisingly, we find the k cat value for hMSH2-hMSH6 in the presence of the +CA substrate was approximately half of that reported by Gradia (33) while the values for the G/T mismatch substrate and homoduplex DNA were similar. Comparison of the oligonucleotides used in these two studies suggested that the sequence context surrounding the +CA mismatched base pairs was different and likely to account for the differences between experiments (data not shown). Studies are in progress to understand the sequence context effects associated with ATPase activation of the hMSH2-hMSH3 and hMSH2-hMSH6 molecular switch. Importantly, we found that mG/C and mG/T only slightly elevated the ATPase activity of hMSH2-hMSH3 over that of the G/C DNA substrate.
We find a continuous discrimination of O 6 -methylguanine-containing substrates versus homoduplex DNA by hMSH2-hMSH6 over a wide variety of DNA concentrations ( Figure 5 and  (10,30). We examined the ability of O 6methylguanine containing oligonucleotides to provoke ADP→ATP exchange ( Figure 6).
Relative to the homoduplex DNA, the presence of the O 6 -methylguanine group resulted in a significantly faster rate-of-exchange with both mG/C and mG/T with hMSH2-hMSH6.
However, the rate of ADP→ATP exchange provoked by mG/C and mG/T was nearly identical to that of homoduplex DNA with hMSH2-hMSH3. These results are consistent with the steady- In order to compare the efficiency that these substrates induce the steady-state ATPase of the human MutS homologs we calculated the k cat /K m (Table I). Although the k cat for the methylated substrates relative to the homoduplex DNA appeared to increase with hMSH2-hMSH3, little change was observed in the k cat /K m values (even for the +CA which is a known substrate for hMSH2-hMSH3). These results contrast the ADP→ATP exchange data which clearly demonstrate that the +CA substrate provokes ADP→ATP exchange, while the G/T, mG/T, mG/C, and G/C substrates appeared nearly identical. Moreover, while we find the k cat and ADP→ATP exchange data are concordant, the K m for the +CA substrate increases significantly relative to the G/T, mG/T, mG/C, and G/C substrates. These results suggest that there is likely to be little or no discrimination of O 6 -methylguanine containing DNA over homoduplex DNA by the hMSH2-hMSH3 heterodimer, and that the ATPase efficiency (as measured by the k cat /K m ) of hMSH2-hMSH3 for a recognized substrate may indicate an alternative feature distinct from hMSH2-hMSH6. In contrast, the k cat /K m and ADP→ATP exchange for both mG/C and mG/T increased relative to that of the homoduplex G/C for hMSH2-hMSH6.  repair has been somewhat controversial. It has been proposed that ATP hydrolysis is necessary for the translocation of the hMSH2-hMSH6 heterodimer along the DNA backbone following recognition of mispaired nucleotides (38). An alternative model suggests that mismatched or lesion-containing DNA provokes the exchange of ADP→ATP resulting in a conformational transition and in the formation a hydrolysis independent sliding clamp on DNA (31). In this model, it is the increased local concentration of sliding clamps associated DNA adjacent to the lesion/mismatch that serve as a first step for downstream signaling events (29). While O 6methylguanine lesions appear to stimulate hMSH2-hMSH6 adenosine nucleotide processing activity, the modest stimulation of cognate hMSH2-hMSH3 activities suggests only a weak or modest overlapping and redundant function; similar to mispair recognition (39). However, it is worth noting that the ratio of hMSH2-hMSH6 to hMSH2-hMSH3 in mismatch proficient HeLa cells has been estimated to be 10:1 (34). Together with the observation that hMSH2-hMSH3 displays a near order-of-magnitude weaker ATP processing activity than hMSH2-hMSH6, these results suggest that processing of O 6 -methylguanine lesions by hMSH2-hMSH3 is unlikely to contribute to the cytotoxicity of these lesions.

DISCUSSION
The mechanism by which loss of DNA mismatch repair results in resistance O 6methylguanine and cisplatin lesions is not completely understood. However, it is becoming increasingly clear that methylation damage (likely O 6 -methylguanine lesions) induce apoptosis through a MMR-dependent signaling pathway (19,20). A recent study has shown that cells treated with MNU induce the phosphorylation and stabilization of the p53 tumor suppressor, which appears to require a functional MMR system (40). However, other studies have clearly shown that apoptosis induced by methylating agents is dependent on MMR but independent of p53 (18,19). Interestingly, a p53 related gene, p73, appears to be over-expressed and directly linked to the MMR-dependent apoptotic response induced by cisplatin (41). These studies are complicated by the observation that, initiation of apoptosis appears to be lesion specific, since cells treated with ionizing radiation undergo apoptosis regardless of MMR status (18).
The observation that hMSH2-hMSH6 adenosine nucleotide processing is stimulated by mG/C reduces the likelihood of a model for methylation sensitivity in mammalian cells that uniquely depends on replication misincorporation followed by a "futile-cycle" of repair.
Moreover, the significant steady-state ATPase implies cycles of lesion recognition, ADP→ATP exchange, and the formation of an hMSH2-hMSH6 sliding clamp that diffuses away from the lesion containing oligonucleotide; thus reducing the likelihood of a "binding-occlusion" model.
These observations are consistent with a model in which a portion of the sensitivity to O 6methylguanine occurs via recognition of mG/C by the MMR machinery and direct transduction of this lesion recognition to the apoptotic machinery (although we can not completely eliminate an intertwined lesion-dependent replication-block model). The importance of MMR complexes in the direct initiation of apoptosis has been underlined by the observation that over-expression of hMSH2 or hMLH1, but not hMSH3, hMSH6 or hPMS2, in primary mouse embryo fibroblasts provokes apoptosis in both repair proficient and deficient lines (19). The precise mechanism of MMR signaling to downstream effectors, such as the repair or apoptotic processes, remains to be established. It is tempting to speculate that two separate signal transduction pathways exist for the MMR machinery, one to signal DNA mismatch/lesion repair and one, in the case of high lesion loads, to signal apoptosis.    Table 1 (see materials and methods).