Evidence That Nucleosomes Inhibit Mismatch Repair in Eukaryotic Cells*

The influence of chromatin structure on DNA metabolic processes, including DNA replication and repair, has been a matter of intensive studies in recent years. Although the human mismatch repair (MMR) reaction has been reconstituted using purified proteins, the influence of chromatin structure on human MMR is unknown. This study examines the interaction between human MutSα and a mismatch located within a nucleosome or between two nucleosomes. The results show that, whereas MutSα specifically recognizes both types of nucleosomal heteroduplexes, the protein bound the mismatch within a nucleosome with much lower efficiency than a naked heteroduplex or a heterology free of histone proteins but between two nucleosomes. Additionally, MutSα displays reduced ATPase- and ADP-binding activity when interacting with nucleosomal heteroduplexes. Interestingly, nucleosomes block ATP-induced MutSα sliding along the DNA helix when the mismatch is in between two nucleosomes. These findings suggest that nucleosomes may inhibit MMR in eukaryotic cells. The implications of these findings for our understanding of eukaryotic MMR are discussed.

DNA mismatch repair (MMR) 2 is an essential genome surveillance pathway that corrects mismatches generated during DNA replication and repair (1)(2)(3). The MMR pathway and its component enzymes are highly conserved from bacteria to humans, and the mechanism of the MMR reaction, although more complex in higher organisms, is related in all organisms studied to date. The steps of the MMR pathway include mismatch recognition by MutS-like proteins, removal of the mispaired nucleotide by nucleases in a manner dependent on MutS-and MutL-like proteins, and DNA repair synthesis by a replicative DNA polymerase in concert with DNA replication factors (3). In comparison with Escherichia coli MMR, human MMR is more complicated, involves more proteins and cofactors, and has more diverse biological roles, including DNA damage signaling (3,4). It is also highly likely that chromatin structure influences the efficiency of MMR in human and other eukaryotic cells and that the mechanism of MMR in eukaryotic cells reflects this additional layer of complexity.
The nucleosome is the basic structural unit of eukaryotic chromatin, and is comprised of an octamer of histone proteins wrapped with a DNA duplex of ϳ147 bp. The histone octamer contains a central (H3-H4) 2 tetramer flanked on either side by a H2A-H2B dimer. During DNA replication, nucleosomes undergo dynamic disassembly/assembly, such that nucleosomes ahead of the replication fork are disrupted and then rapidly re-assembled on nascent DNA behind the replication fork (5). Increasing evidence suggests that MMR is coupled with DNA replication (3, 6 -9) and that active MMR complexes may slide over long distances (i.e. as far as 1000 bp) along DNA during MMR (10 -12). Although human MMR has been reconstituted with purified proteins in vitro using naked DNA heteroduplexes (13,14), the influence of chromatin structure on human MMR has not been analyzed. Therefore, it is not known how human MMR proteins recognize and interact with mismatches in nascent or damaged DNA.
The goal of this study was to analyze the interaction between human MutS␣ (MSH2-MSH6 heterodimer) and a single mismatch in the context of several simple model nucleosome substrates. The results show that MutS␣ specifically recognizes a mismatch located in or adjacent to a nucleosome. However, MutS␣ bound the mismatch with lower efficiency in histone octamer-bound DNA than in histone octamer-free DNA, and ATP-provoked sliding of MutS␣ along DNA helices is blocked when the mismatch is flanked on each side by two nucleosomes. These observations strongly support a notion that chromatin remodeling and/or modification factors are required for MMR in eukaryotic cells.

EXPERIMENTAL PROCEDURES
DNA Preparation-Two complementary 147-mer fragment oligonucleotides (DNA substrate I, Fig. 1A) containing tandem repeats of a high affinity histone octamer-binding DNA sequence, TATAAACGCC (15), were synthesized in The Midland Certified Reagent Company (Midland, TX). After highperformance liquid chromatography purification, the synthetic oligonucleotides were annealed and 5Ј-end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP. The resulting substrate I, which contains a centrally located G-T mismatch flanked by BglII and HindIII restriction enzyme cleavage sites (Fig. 1A, substrate I), was further purified by agarose gel electrophoresis.
DNA substrate II (200-bp DNA fragment) was constructed by PCR using a plasmid DNA containing the high affinity nucleosome positioning sequence 601 as a template (16), a gift from Timothy Richmond (Eldgenössiche Technische Hochschule, Zürich, Switzerland). Two alternate forward primers with a 5Ј-end-labeled biotin were combined with a unique reverse primer, as follows, with restriction enzyme recognition sites shown in underlined font and the mismatch position in bold font : Forward primer 1,  5Ј-AAGGTCTGCAGCTCGCAGCTAGCCTCGAGATCGAT-TCAAGCAAGCTTGGAATTCCTGGAGAATCCCGGTGC-3Ј; Forward primer 2, 5Ј-AAGGTCTGCAGCTCGCAGCTAGC-TTCGAGATCGATTCAAGCAAGCTTGGAATTCCTGGAG-AATCCCGGTGC-3Ј; and Reverse primer, 5Ј-ACAGGAT-GTATATATCTGACAC-3Ј. The PCR products were purified by agarose gel electrophoresis, mixed together, heat-denatured, re-annealed by slow cooling, digested with XhoI to cleave homoduplex products, and re-purified by agarose gel electrophoresis. The XhoI-resistant DNA, which contains either a G-T mismatch or an A-C mismatch, was extracted from the gel and 5Ј-end-labeled with [␥-32 P]ATP by T4 polynucleotide kinase.
DNA substrate III was derived from PCR amplification of M13mp18-UKY derivatives (17) that contain two pieces of the 601 nucleosome positioning sequence, with one 601 sequence within the EcoRI-KpnI region and another within the XhoI-HindIII (for M13mp18-UKY1) or NsiI-HindIII (for M13mp18-UKY2). The resulting PCR products were purified by agarose gel electrophoresis and mixed together, heat-denatured, re-annealed by slow cooling, and followed by digestion with XhoI and NsiI to cleave homoduplex products. The heteroduplexes, which contain either a G-T or an A-C mismatch, were re-purified by agarose gel electrophoresis and 5Ј-end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase.
Protein Purification-Recombinant human MutS␣ was expressed in insect cells and purified to near homogeneity as described previously (14). Human histone genes (kindly provided by Dr. Jeffrey Parvin, Ohio State University) encoding for H2A, H2B, H3, and H4 were expressed in E. coli, and the purified histone proteins were used to assemble histone octamers as described previously (18).
Nucleosome Assembly and Purification-Nucleosomes were assembled from purified histone octamers and DNA substrates shown in Fig. 1A using the salt-dilution method (19). Reconstituted nucleosomes were separated from free DNA by 5-30% glycerol gradient, and fractions were collected using a Beckman Fraction Recovery System. The assembled nucleosomes were analyzed by electrophoresis in agarose or non-denaturing polyacrylamide gels.
EMSA-Electrophoretic mobility shift assays (EMSAs) were performed by incubating purified MutS␣ and DNA fragments or their corresponding nucleosomal substrates in the presence of 25 mM Tris-HCl (pH 7.6), 100 mM NaCl, 2 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol at room temperature, essentially as described (20). Reactions were terminated by adding 5 l of 50% (w/v) sucrose and analyzed by 4% non-denaturing polyacrylamide gel in buffer containing 50 mM Tris borate (pH 7.6) and 1 mM EDTA. The buffer was re-circulated during electrophoresis. Gels were dried and analyzed using a Storm PhosphorImager (Amersham Biosciences).

Reconstitution and Characterization of Mismatch-contain-
ing Nucleosomes-Histone octamers were assembled from purified recombinant human histone proteins and purified by a fast-protein liquid chromatography Superdex 200 column (data not shown) and verified by SDS-gel electrophoresis (Fig.  1B). Nucleosomes were assembled using the salt dilution method (19) from purified histone octamers and DNA substrates shown in Fig. 1A. Substrates I (147 bp) and II (200 bp) form mononucleosomes, and substrate III (428 bp) forms dinucleosomes. The resulting nucleosome substrates were purified by glycerol gradient sedimentation and verified by restriction enzyme digestion and gel electrophoresis. As shown in Fig. 2, nucleosomes migrated more slowly than their corresponding DNA substrates as if the latter had gained ϳ200 bp per histone octamer (Fig. 2, A-C), which has a mass approximately equal to that of 200-bp DNA (23,24). Restriction enzyme susceptibility of individual DNA substrates and their corresponding mononucleosome were also examined. As expected, the substrate I naked DNA (Fig. 2D, lane 2) and its nucleosome (Fig. 2D, lane 5) were resistant to cleavage by XhoI, because the G-T mismatch is in the XhoI recognition site. Naked substrate I DNA but not its nucleosome was sensitive to cleavage by BglII and HindIII (Fig. 2D, lanes 3 and 4 compared with lanes 6 and 7), indicating that the histone octamer blocks access of the restriction enzymes to the DNA (i.e. steric hindrance). Both substrates II and III contain the high affinity nucleosome positioning sequence 601 (16). Indeed, nucleosomes assembled with these two DNA substrates are located in the 601 sequence, as judged by the fact that substrate II nucleosome was sensitive to digestions by HindIII (Fig. 2E); and substrate III nucleosome was sensitive to cleavage by KpnI, which converts dinucleosomal substrate III into two mononucleosomes (Fig. 2F).
Nucleosome Structure Inhibits Binding of MutS␣ to a Mismatch-EMSAs were used to examine interactions between human MutS␣ and the 147-bp DNA heteroduplex (i.e. substrate I in Fig. 1A) or its corresponding nucleosome. As shown in Fig. 3A, MutS␣ bound the heteroduplex nucleosome (lanes 8 -14) but with lower efficiency than the naked heteroduplex (lanes 1-7), because a higher minimum amount of MutS␣ was required to supershift the nucleosome than the 147-bp DNA fragment. This binding reflects a specific interaction with the mismatch, because the heteroduplex nucleosome (Fig. 3A, lanes 19 -22) but not the 147-bp homoduplex nucleosome (Fig. 3A, lanes 15-18) was supershifted in the presence of 20-fold excess homoduplex competitor DNA. This result shows that, although a histone octamer-bound mismatch is specifically recognized by MutS␣, the binding affinity is reduced by the presence of the histone octamer. It is noted that, although the naked DNA substrate and its nucleosome differ obviously in their electrophoresis mobility, it is not obvious to distinguish their complexes with MutS␣ (Fig. 3A). This is likely due to binding of multiple molecules of MutS␣ to the naked DNA substrate.
Previous studies showed that ATP and non-hydrolyzable ATP analogs inhibit MutS␣ binding to heteroduplexes (25), likely by inducing a conformational change in MutS proteins, followed by DNA end-dependent dissociation of MutS proteins from DNA substrates (11,26). We therefore tested the ATP effect on MutS␣ binding to nucleosomal heteroduplex. As shown in Fig. 3B, addition of ATP (lanes 6 -9) or ATP analogy AMP-PNP (lanes 10 -13) to EMSA reactions induced rapid dissociation of the MutS␣-nucleosome complex, as free nucleosomes appeared in early time reactions. These results suggest  32 P-end-labeled DNA substrate I (0.5 pmol) or its corresponding nucleosome (1 pmol) were incubated with the indicated restriction endonucleases, and the resulting products were digested with proteinase K, followed by ethanol precipitation and 6% polyacrylamide gel electrophoresis. E, analysis of nucleosomal substrate II. The 32 P-labeled nucleosomal substrate II (1 pmol) was treated with or without HindIII, as indicated, and the samples were directly loaded onto 6% polyacrylamide gel for electrophoresis. F, analysis of nucleosomal substrate III. The 32 P-labeled dinucleosomal substrate III (1 pmol) was digested with or without KpnI, followed immediately by 6% polyacrylamide gel electrophoresis. The nucleosomal substrate II was used as a mononucleosome (Mono) control for cleavage products of the dinucleosome (Di). DNA or nucleosomes in D-F were detected by using a PhosphorImager.  1-7) only, whereas those on the right are for nucleosome reactions (lanes 8 -22). B, inhibition of MutS␣-nucleosome interaction by ATP or AMP-PNP. EMSA experiments were performed with the 147-bp heteronucleosome and MutS␣ (5 nM) for the indicated times as described in A but in the absence (lanes 1-5) or presence (lanes 6 -9) of 4 mM ATP or 4 mM AMP-PNP (lanes 10 -13).

that ATP also inhibits interactions between MutS␣ and nucleosomal heteroduplexes.
Nucleosome Acts as a Barrier for MutS␣ Sliding-The moving model of MMR (3) suggests that MutS and its homologs move bi-directionally away from the mismatch and dissociate from the DNA when it reaches a DNA terminus (11,26). However, this process has never been examined on a chromatin-like template, and the impact of nucleosomes on the ability of MutS␣ to slide and dissociate from heteroduplex DNA has not been explored. To examine this question, mononucleosomes were formed with a 5Ј-biotin-labeled 200-bp fragment with a unique G-T or A-C mismatch and HindIII restriction site upstream of the 601 nucleosome positioning sequence (Fig. 1A, substrate II). Because the resulting nucleosome (Fig. 2B) remained sensitive to cleavage by HindIII (Fig. 2E), the octamer bound stably to and was positioned by the 601 sequence, and the mismatch remained nucleosome-free. This nucleosomal heteroduplex was used to examine DNA sliding by MutS␣ in the presence or absence of streptavidin and ATP or AMP-PNP (Fig. 4). Because the streptavidin-biotin complex blocks dissociation of MutS␣ from DNA termini (27)(28)(29), this experiment allowed us to test whether the stably bound histone octamer downstream of the mismatch inhibits dissociation of MutS␣ from the terminus.
The results show that MutS␣ bound to the nucleosomal heteroduplex with or without streptavidin as efficiently as it bound to the corresponding naked heteroduplex (data not shown). This is not surprising, because the mismatch is in a histone octamer-free region of the substrate. In the presence of ATP, ϳ50% of MutS␣ dissociated from the heteroduplex nucleosome without streptavidin (Fig. 4A, lanes  9 -12); however, in the presence of streptavidin, the MutS␣-nucleosome complex was stable in the presence or absence of ATP (Fig. 4A, lanes 3-7) or AMP-PNP (Fig. 4B). These data strongly suggest that ATP-or AMP-PNP-induced sliding/dissociation of MutS␣ can be also blocked by a stably bound histone octamer.
We further tested this idea using a dinucleosome derived from DNA substrate III (Fig. 1A), in which a mismatch is located in a histone octamer-free region between two mononucleosomes as the nucleosomal substrate is sensitive to cleavage by KpnI (see Figs. 1A and 2F). MutS␣ bound efficiently to nucleosome-free substrate III, and this interaction was inhibited in the presence of ATP (Fig. 5, lanes 7-10), as expected for naked DNA. In contrast, MutS␣ formed a stable complex with dinucleosomal substrate III, which persisted in the presence of ATP (Fig. 5, lanes 2-5). This result is consistent with the notion that stably bound histone octamers flanking a mismatch inhibit MutS␣ sliding and subsequent dissociation of the MutS␣-DNA complex.
Effects of Nucleosomal Heteroduplexes on ATPase and Nucleotide Binding Activities of MutS␣-MutS␣ possesses an ATPase activity essential for MMR. Previous studies have shown that DNA heteroduplexes stimulate MutS␣ ATPase activity (27,30). To determine the effect of nucleosome on MutS␣ ATPase, we examined ATP hydrolysis by MutS␣ during its interactions with individual DNA heteroduplexes or their corresponding nucleosomes. As expected, all three naked DNA heteroduplexes stimulated MutS␣ ATPase activity to at least 2.5-fold (Fig. 6, lanes 3-5) as compared with MutS␣ alone (Fig.  6, lane 2). However, the stimulation by nucleosomal heteroduplexes differs from naked DNA substrates. First, there was essentially no stimulation of the MutS␣ ATPase activity by substrate I nucleosome (lane 6), which might be due to the weak interaction between MutS␣ and the nucleosome (see Fig. 3). Second, nucleosomes derived from DNA substrates II and III did enhance ATP hydrolysis by MutS␣, but the stimulation was weaker compared with naked DNA substrates (compare lanes 7 and 8 with lanes 4 and 5). Because nucleosome substrate I has a lower affinity for MutS␣, and because nucleosomal substrates II and III, although bound normally by MutS␣, possess limited  length for MutS␣ sliding, our data appear to suggest that both the sliding and heteroduplex binding are important factors stimulating MutS␣ ATPase activity.
It is also known that MutS␣ binds both ATP and ADP, with the MSH6 subunit preferentially binding ATP and the MSH2 subunit preferentially binding ADP (21,22). We examined the effect of nucleosomal heteroduplexes on MutS␣ nucleotide binding by performing nucleotide UV cross-linking experiments (21,22). As shown in Fig. 7, nucleosomes essentially had little influence on MSH6 ATP binding, as judged by the fact that in the absence of Mg 2ϩ (i.e. no ATP hydrolysis), almost equal amounts of the MSH6 subunit were labeled with [␥-32 P]ATP (Fig. 7A) or [␣-32 P]ATP (Fig. 7C) in all reactions. However, in the presence of Mg 2ϩ , which supports ATP hydrolysis, more labeled MSH2 (by [␣-32 P]ADP) was observed in reactions with naked DNA substrates than those with nucleosomal substrates, with the least labeling in the reaction containing nucleosomal substrate I (Fig. 7D). Thus, the ADP binding pattern is almost identical to the ATP hydrolysis profile shown in Fig. 6. To determine if the differences in ADP binding among reactions reflect changes in the ADP-binding ability of MSH2 or simply the binding of available ADP (from ATP hydrolysis) to MSH2 in individual reactions, cross-linking experiments were conducted in the presence of [␣-32 P]ADP (Fig. 7, E and F). In the presence of Mg 2ϩ (Fig. 7F), we did observe an ADP binding pattern similar to the one shown in Fig. 7D. Despite the fact that all reactions contained the same amounts of ADP, much less MSH2 labeling was detected in the nucleosomal substrate I reaction (Fig. 7F), indicative of inhibition of MSH2 ADP binding by the nucleosomal heteroduplex. As shown in Table 1, this conclusion was also supported by quantitative analysis of the data shown in Figs. 6, 7D, and 7F. It is clear that both types of substrates (except nucleosomal substrates I) stimulate MutS␣ ATPase-and ADP-binding activities, but the nucleosomal heteroduplexes are weaker stimulator than the naked DNA heteroduplexes; there was little (if any) stimulation generated by nucleosomal substrate I (Table 1). Given the importance of the ATPase-and nucleotide-binding activities of MutS␣ in MMR, we conclude that nucleosomes inhibit or partially inhibit MMR.

DISCUSSION
Growing evidence supports a role for chromatin structure in MMR. Genetic studies in yeast suggest that local chromatin structure influences MMR efficiency, because mutation rates of an identical polyGT tract varied significantly in different locations of the yeast genome (31). A recent biochemical study revealed that regulatory factor X, a chromatin remodeling factor that regulates gene expression, stimulates MMR in vitro (32). In this study, we directly tested possible effects of nucleosomes on mismatch recognition. The data presented here suggest that (i) human MutS␣ displays reduced ATPase-and ADP-binding activities when interacting with nucleosomal heteroduplexes; (ii) MutS␣ binds the mismatch within a nucleosome with lower efficiency than a naked heteroduplex or a heterology free of histone proteins but between two nucleosomes; and (iii) nucleosomes flanking a mismatch prevent MutS␣ from sliding along the DNA helix.
These findings may have significant implications for our understanding of the mechanism of MMR in eukaryotic cells. First, reduced binding for nucleosome-bound heteroduplexes by MutS␣ will certainly lead to reduced MMR activity. Second, because ATPase-and nucleotide-binding activities of MutS␣ are essential for MMR, inhibition of MutS␣ ATPase-and ADPbinding activities by nucleosomes will inhibit or partially inhibit MMR. Third, depending on two current MMR models, nucleosomes flanking a mismatch could completely or partially block the repair of the heterology. The moving model (11,12,26) proposes that MutS proteins slide away from the mismatch along the DNA helix toward the strand discrimination signal

TABLE 1 Relative ATPase and ADP binding activities of MutS␣
The data shown in this table were obtained from Figs. 6 (ATPase), 7D (ADP* binding), and 7F (ADP binding) by dividing radioactive intensity in individual reactions of a given assay with the intensity of their corresponding control reaction containing MutS␣ alone (i.e., no DNA or nucleosomal substrate). (i.e. a single strand nick), whereas the stationary model (33,34) suggests that MutS proteins remain bound at the mismatch during the course of MMR. If the moving model is correct, nucleosomes would inhibit MMR, because they block MutS␣ sliding (Figs. 4 and 5). Therefore, chromatin-remodeling or modification activities that disrupt nucleosomes would be required for efficient MMR. In contrast, if the stationary model is correct, nucleosomes may not inhibit MMR, although higher order chromatin structures still could reduce its efficiency. The results presented here are a first step toward unraveling the detailed interaction between the MMR machinery and chromatin substrates, albeit in a simplified in vitro system. There is precedent for the idea that chromatin-remodeling/modification activities could be required during MMR in eukaryotic cells, because ample evidence demonstrates the importance of chromatin-remodeling or modification factors in the context of eukaryotic gene transcription and DNA replication (5,35). In fact, chromatin-remodeling factors and histone-modifying enzymes alter higher order chromatin structure by affecting contacts between nucleosomes or interactions between histones and DNA (for a review see Ref. 35). Recent studies identified at least 30 modifications (e.g. acetylation and phosphorylation) in histone proteins, with the majority of the modifications occurring in H3 and H4 (36), which are the core histone proteins responsible for contacting the phosphate backbone of the wrapped DNA. Thus, lysine acetylation of H3 and H4 residues located in the DNA-histone interface may reduce DNA-histone interaction(s) and facilitate nucleosome disassembly. Using a chemical ligation strategy to acetylate recombinant histones at specific residues, several groups have shown that acetylation of H3K115, H3K122, or H4K16 reduces DNA binding and inhibits the formation of higher order chromatin structure (37,38). The influence of these specific modifications on MMR remains to be examined.

MutS␣ activity
It should be noted that the experiments described here were performed in a simplified in vitro system, including only human MutS␣ and mono-or dinucleosome substrates. In addition, histone octamers were formed from unmodified histone proteins, and nucleosomes were positioned using a high affinity nucleosome binding sequence (e.g. the 601 positioning sequence). Therefore, it remains possible that MutS proteins (e.g. MutS␣) and/or other MMR proteins may possess chromatin-remodeling activity by displacing modified or more natural histone octamers on DNA. Therefore, future studies should evaluate the extent to which higher order chromatin structure and modified histone or non-histone chromosomal proteins, including MMR proteins, promote, inhibit, or modulate MMR in eukaryotic cells.