|Home | About | Journals | Submit | Contact Us | Français|
DNA nucleotide mismatches and lesion arise on chromosomes that are a complex assortment of protein and DNA (chromatin). The fundamental unit of chromatin is a nucleosome that contains ~146 bp DNA wrapped around an H2A, H2B, H3, and H4 histone octamer. We demonstrate that the mismatch recognition heterodimer hMSH2-hMSH6 disassembles a nucleosome. Disassembly requires a mismatch that provokes the formation of hMSH2-hMSH6 hydrolysis-independent sliding clamps, which translocate along the DNA to the nucleosome. The rate of disassembly is enhanced by actual or mimicked acetylation of histone H3 within the nucleosome entry-exit and dyad axis that occurs during replication and repair in vivo and reduces DNA-octamer affinity in vitro. Our results support a passive mechanism for chromatin remodeling where hMSH2-hMSH6 sliding clamps trap localized fluctuations in nucleosome positioning and/or wrapping that ultimately leads to disassembly, and highlights unanticipated strengths of the Molecular Switch Model for mismatch repair (MMR).
Mismatched nucleotides arise in DNA as a result of polymerase misincorporation errors, recombination between heteroallelic parental chromosomes or as a result of chemical and physical damage (Friedberg et al., 2006). MutS homologs (MSH) and MutL homologs (MLH/PMS) are highly conserved proteins, and are essential for the MMR excision reaction that removes mismatches/lesions from DNA (Kolodner et al., 2007). Mutation of hMSH2, hMSH6, hMLH1, and hPMS2 are the causes of a common human cancer predisposition syndrome, Hereditary Non-Polyposis Colorectal Cancer (HNPCC; (Boland and Fishel, 2005). The hMSH2-hMSH6 heterodimer is required for the initial recognition of mismatches during MMR as well as lesion recognition for specific damage-induced signaling pathway(s) (Drummond et al., 1995; Yoshioka et al., 2006). Although MMR occurs in the context of chromatin in vivo, previous biochemical studies have relied exclusively on naked DNA substrates (Constantin et al., 2005; Zhang et al., 2005). The effect of chromatin on MMR is unknown. Moreover, no chromatin remodeling activities have been linked to MMR in spite of numerous cellular and genetic surveys (Ataian and Krebs, 2006; Escargueil et al., 2008).
To determine the effect of nucleosomes on hMSH2-hMSH6 function(s), we have constructed a model DNA substrate containing the Xenopus 5S rDNA nucleosome localization sequence linked to a lac O sequence, mismatch, and terminal biotin on a 3′-tail (Fig. 1A). A single nucleosome was reconstituted on this DNA substrate by salt dialysis, using purified H2A, H2B, H3, and H4 histones that were refolded into histone octamers as previously described (Luger et al., 1999). Nucleosome substrates were formed with three types of histone octamers: those containing no modifications (UN), those containing an acetylation mimic where the H3 lysine-56 is substituted with glutamine [H3(K56Q)], and those containing site-specific acetylation of the histone H3 K115 and K122 residues [H3(K115Ac, K122Ac)]. H3(K56) is located in the nucleosome entry-exit region while H3(K115, K122) are located in the nucleosome dyad beneath the wrapped DNA. All three residues appear important for normal replication, transcription and DNA repair (English et al., 2006; Hyland et al., 2005; Zhang et al., 2003). Site-specific acetylation of histone H3(K115, K122) was accomplished by intein-mediated protein ligation that links a recombinant H3 thioester truncated at L109 with a synthetic peptide containing acetylated K115 and K122; this method generates a native peptide bond and H3 protein sequence (Manohar et al., 2009). The mono-nucleosome DNA substrates were then purified on a 5–30% sucrose gradient (Suppl. Fig. 1A and 1B; (Lowary and Widom, 1998). The nucleosome positions were mapped using an ExoIII protection assay and found to largely occupy the 5S rDNA sequence shielding ~145 bp of DNA, as well as a number of lower frequency positioning sites (Suppl. Fig. 1C). The protection footprint suggests that the nucleosomes are composed of histone octamers and the additional positioning sites appear consistent with the gel migration pattern (Suppl. Fig. 1B).
To determine the effect of nucleosomes on the initiation of MMR, we examined hMSH2-hMSH6 binding to the nucleosome-DNA substrates (Fig. 1B and 1C). We found little difference in hMSH2-hMSH6 mismatch binding between the free-DNA substrate and the UN and H3(K115Ac, K122Ac) nucleosome-DNA substrates (Fig. 1B and 1C; KD (G/T) = 24nM; KD (G/T•b•UN) = 27 nM; KD (G/T•b•K115Ac/K122Ac) = 22 nM), and in the presence of streptavidin that induces a physical block to one end of the DNA substrate (Fig. 1C; KD (G/T•b-SA•UN) = 26 nM; KD (G/T•b-SA•K115Ac/K122Ac) = 10 nM). The binding of hMSH2-hMSH6 to identical DNA substrates without the mismatch (G/C) was over 50-fold less efficient (Fig. 1B and 1C; KD (G/C) = 1808 nM; KD (G/C•b•K115Ac/K122Ac) = 1342 nM; KD (G/C•b-SA•K115Ac/K122Ac) = 1198 nM). Similar mismatch specific binding was observed for the H3(K56Q) nucleosome-DNA substrate. These results demonstrate that the nucleosome-DNA substrates containing a mismatch outside of the predominant nucleosome localization region are efficiently recognized by hMSH2-hMSH6.
The addition of ATP to hMSH2-hMSH6 bound to a mismatch provokes the formation of a hydrolysis-independent sliding clamp that may diffuse off an open DNA end (Gradia et al., 1999; Mendillo et al., 2005; Selmane et al., 2003). ATP-dependent release of the sliding clamp from the mismatch allows iterative cycles of hMSH2-hMSH6 loading and clamp formation (Acharya et al., 2003; Gradia et al., 2000; Gradia et al., 1999). These iterative cycles can result in multiple ATP-bound hMSH2-hMSH6 clamps that may be trapped on the DNA by blocking the ends with biotin-streptavidin or by using a circular DNA substrate (Acharya et al., 2003; Gradia et al., 1999; Mendillo et al., 2005; Schofield et al., 2001).
Nucleosomes are highly stable protein-DNA complexes that are known to sterically occlude DNA binding proteins from their target sites (Li and Widom, 2004; Polach and Widom, 1995; Utley et al., 1996). The ability of a nucleosome to block the diffusion of hMSH2-hMSH6 sliding clamps and potentially impede MMR is a significant unknown. Consistent with previous work, we found that addition of streptavidin to a free-DNA (F) substrate containing a single biotin on the 3′-tail resulted in a mobility shift (Fig. 2A–C, compare lane 1 and 2; Suppl. Fig. 2A and 2B, compare lane 1 and 2; (Gradia et al., 1999). An additional mismatch-specific shift on this single end blocked DNA was observed with hMSH2-hMSH6 (Fig. 2A–C, lane 3; compared to Suppl. Fig. 2A and 2B, lane 3) that was released from the remaining open-end of the DNA substrate with the addition of ATP (Fig. 2A–C, lane 4). These results are consistent with previous studies that demonstrated a single biotin-streptavidin blocked-end is not sufficient to retain ATP-bound hMSH2-hMSH6 sliding clamps on mismatched DNA (Gradia et al., 1999). We found that the unmodified, H3(K56Q) modification mimic, or H3(K115Ac, K122Ac) modified nucleosome-DNA substrate (N) with an open 3′-tail behaved similarly to the free-DNA (F) substrate containing a single biotin-streptavidin blocked-end (Fig. 2A–C, compare lanes 2–4 with lanes 5–7; Suppl. Fig. 2A and B, lanes 5–7). In this case, the nucleosome-DNA substrate (Fig. 2A–C, lane 5) was bound specifically with hMSH2-hMSH6 (see*, Fig. 2A–C, lane 6; compare Suppl. Fig. 2A and 2B, lane 6), which was then released upon addition of ATP (Fig. 2A–C, lane 7).
To determine whether nucleosomes blocked the sliding of hMSH2-hMSH6 clamps we examined the nucleosome-DNA substrates containing biotin-streptavidin blocked 3′-tails (Fig. 2A–C, lanes 8–15; Suppl. Fig. 2A and 2B, lanes 8–15). Nucleosome stability may be calculated from data with the nucleosome-DNAs containing a G/C duplex, where hMSH2-hMSH6 displays insignificant binding activity (Suppl. Fig. 2A and 2B, lane 8–15; t1/2 (G/C•UN) = 578 min, t1/2 (G/C•K115Ac/K122Ac) = 347 min; Fig. 2D). These results demonstrate that unmodified and H3(K115Ac, K122Ac) modified nucleosomes are stable for 10–20 hrs under our experimental conditions. A similar stability is observed with H3(K56Q) nucleosomes (data not shown). In contrast, we found that incubation of the biotin-streptavidin blocked G/T mismatch nucleosome DNA substrates with hMSH2-hMSH6 and ATP resulted in the eviction of the histone octamer (Fig. 2A–C, lanes 8–15, red arrow; quantified in Fig. 2D). These results suggest that a nucleosome does not block ATP-bound hMSH2-hMSH6 sliding clamps and that the nucleosome appeared to be disassembled by hMSH2-hMSH6. Moreover, there was a significant difference in the ability of hMSH2-hMSH6 to disassemble unmodified versus the H3(K56Q) mimic or H3(K115Ac, K122Ac) modified nucleosomes (Fig. 2D; t1/2 (G/T•UN) = 117 min, t1/2 (G/T•K56Q) = 53 min, t1/2 (G/T•K115Ac/K122Ac) = 23 min, respectively). Our previous work has demonstrated that H3(K115Ac, K122Ac) increases the rate of thermal repositioning and reduces the DNA-histone binding free-energy compared to unmodified nucleosomes (Manohar et al., 2009). These observations are consistent with the conclusion that nucleosomes containing H3 acetylation mimics and/or modifications that reduce their intrinsic DNA affinity may be disassembled more efficiently by hMSH2-hMSH6.
ATP-dependent chromatin remodeling is required for numerous cellular DNA transactions including transcription, replication and repair (Groth et al., 2007). Disassembly of a nucleosome from a localized region on DNA suggests that hMSH2-hMSH6 performs a chromatin remodeling reaction. To explore the mechanism behind this new hMSH2-hMSH6 function, we examined the ATP requirement for chromatin remodeling (Fig. 3; Suppl. Fig. 3). The hMSH2(K675A)-hMSH6(K1140A) mutant heterodimer binds mismatched DNA similar to the wild type heterodimer, but is incapable of ATP binding and/or hydrolysis (N.P, S.J. and R.F. in preparation; (Haber and Walker, 1991). We found that in spite of a normal mismatch binding activity, hMSH2(K675A)-hMSH6(K1140A) was incapable of catalyzing the disassembly of unmodified or H3(K115Ac, K122Ac) modified nucleosomes (Fig. 3A and 3C; Suppl. Fig. 3A–C). These results suggest that ATP binding and/or hydrolysis are required for hMSH2-hMSH6 catalyzed chromatin remodeling. Since the hMSH2(K675A)-hMSH6(K1140A) protein was purified by an identical method to the wild type protein, these results also imply that preparation contaminants are unlikely to be responsible for the chromatin remodeling activity.
To examine the role of ATP hydrolysis on hMSH2-hMSH6 chromatin remodeling activity we performed nucleosome disassembly studies with the ATP analog adenosine 5′-[γ-thio]-triphosphate (ATPγS). We determined that the rate of ATPγS hydrolysis (kcat) by hMSH2-hMSH6 in the absence of DNA (0.04 ± 0.02 min−1) or in the presence of mismatched DNA (0.06 ± 0.05 min−1; Suppl. Fig. 4A), and compared it to the well-known rate of ATP hydrolysis in the absence of DNA (1 ± 0.5 min−1) or in the presence of mismatched DNA (22 ± 1.2 min−1; (Mazurek et al., 2009). These results clearly demonstrate that hMSH2-hMSH6 is more than 350-fold less capable of hydrolyzing ATPγS compared to ATP when a mismatch is present, and that repeated rounds of mismatch-dependent hydrolysis are dramatically suppressed by ATPγS. Perhaps more importantly, ATPγS is the only analog of ATP that appears to bind hMSH2-hMSH6 and provoke the formation of a sliding clamp similar to ATP; although the kinetics of sliding clamp formation appear slower than ATP (Suppl. Fig. 4B; (Gradia et al., 1997).
Control reactions with free-DNA demonstrated streptavidin binding (Fig. 3B, compare lane 1 and 2), specific mismatch binding by hMSH2-hMSH6 (Fig. 3B, lane 3), and the release of hMSH2-hMSH6 upon addition of ATPγS (Fig. 3B, lane 4). These results are similar to previous studies and are consistent with the conclusion that ATP-binding by hMSH2-hMSH6 results in the formation of a hydrolysis-independent sliding clamp (Gradia et al., 1999; Mendillo et al., 2005; Selmane et al., 2003). In addition, the single nucleosome substrate DNA containing a G/T mismatch (Fig. 3B, lane 5) specifically binds hMSH2-hMSH6 (see *, Fig. 3B, lane 6) that is largely released upon the addition of ATPγS (Fig. 3B, lane 7). We note that for both the free-DNA and the nucleosome-DNA substrates the efficiency of ATPγS-induced release appears reduced compared to ATP. These observations are consistent with kinetic analysis (Suppl. Fig. 4B; (Gradia et al., 1997) and suggest that the nucleosome-DNA substrates provoke hMSH2-hMSH6 to form a sliding clamp in the presence of ATPγS, which although modestly slower appears nearly identical to single biotin-streptavidin blocked-end free DNA (Fig. 3B, compare lanes 1–4 with lane 5–7). The addition of ATPγS to the pre-bound hMSH2-hMSH6 in a chromatin remodeling reaction suggests reduced but significant nucleosome disassembly (Fig. 3B, lanes 8–15; t1/2 (G/T•K115Ac/K122Ac) = 108 min; Fig. 3D; Suppl. Fig. 3D–F). Contrasting the ~ 4-fold slower rate for nucleosome disassembly in the presence of ATPγS, to the ~ 350-fold slower rate of ATPγS hydrolysis compared to ATP (Suppl. Fig. 4C), and assuming that the rate-limiting step(s) of the disassembly reaction remain similar, these observations support the notion that γ-phosphate hydrolysis is unlikely to be a significant contributor to the disassembly process. It is important to note that these studies are complicated by a competitive ATPγS pre-binding reaction that inactivates hMSH2-hMSH6 mismatch binding and freezes iterative mismatch-dependent loading of sliding clamps, which may ultimately contribute to the reduced rate of ATPγS-induced nucleosome disassembly (Acharya et al., 2003; Gradia et al., 1999). Taken as a whole these observations are consistent with the conclusion that ATP binding and not hydrolysis is the most significant contributor to hMSH2-hMSH6 chromatin remodeling, and that iterative ATP binding likely sustains an efficient reaction.
In the absence of DNA, hMSH2-hMSH6 displays an intrinsic low-level ATP hydrolysis (ATPase) activity (Fig. 3E, bar 1) that is stimulated by mismatched DNA (Fig. 3E, lane 2). This mismatch-dependent hMSH2-hMSH6 ATPase activity (Fig. 3E, compare bar 2 with bar 5) may be progressively reduced to the background level in the absence of DNA (red line) when one and then both of the DNA ends are blocked with biotin-streptavidin (Fig. 3A, compare bar 2 with bars 3 and 4 or bar 5 with bars 6 and 7). These results are consistent with previous studies that have demonstrated the hMSH2-hMSH6 ATPase is accelerated by mismatch provoked ADP→ATP exchange and hydrolysis only occurs when hMSH2-hMSH6 translocates off a DNA end (Fig. 3E, cycle A; (Gradia et al., 1999). We examined the hMSH2-hMSH6 ATPase activity with the unmodified and H3(K115Ac, K122Ac) modified single nucleosome substrates containing a biotin-streptavidin blocked 3′-tail (Fig. 3E, bar 8–11). Unlike traditional chromatin remodelers that display an increased ATPase activity with nucleosome substrates (Gangaraju and Bartholomew, 2007), we found that the ATPase activity of hMSH2-hMSH6 with the biotin-streptavidin blocked G/T mismatch nucleosome substrates was reduced compared to G/T mismatch free-DNA containing a single biotin-streptavidin blocked-end (Fig. 3E; see blue bars, compare bar 3 with bar 8 and 10). As expected, the hMSH2-hMSH6 ATPase activity with the biotin-streptavidin blocked G/T mismatch nucleosome substrates was greater than the corresponding biotin-streptavidin blocked G/C duplex nucleosome substrates (Fig. 3E, compare grey bars with blue bars or bar 8 and 10 with bar 9 and 11). Moreover, we found that the ATPase activity was greater with the H3(K115Ac, K122Ac) modified nucleosome substrate compared to the unmodified nucleosome substrate (Fig. 3E, compare bar 8 with 10). These results mirror the hMSH2-hMSH6 catalyzed chromatin remodeling studies and suggest an intimate connection between ATPase activity, a mismatch, and the ability to disassemble a nucleosome. Taken together with our previous studies (Gradia et al., 2000; Gradia et al., 1999), we consider it likely that the ATPase activity with nucleosome substrates results from a combination of two ATPase cycles, since the product of nucleosome disassembly is a single end-blocked free-DNA substrate (Fig. 3E, cycle B to cycle A via dashed blue arrow). This would explain the reduced ATPase activity since efficient hydrolysis with nucleosome-free DNA (cycle A) would be delayed until the nucleosome was disassembled (cycle B). Alternatively, nucleosomes might enhance hMSH2-hMSH6 ATPase cycling on the DNA. However, it is hard to reconcile the catalytic enhancement of ATPase cycling by histone modifications like H3(K115Ac, K122Ac) that are buried in the nucleosome dyad.
Chromatin remodeling proteins typically interact directly with nucleosomes (Gangaraju and Bartholomew, 2007). To determine whether hMSH2-hMSH6 catalyzed chromatin remodeling requires a mismatch to load hMSH2-hMSH6 sliding clamps that must translocate along the DNA (cis) or interacts directly with nucleosomes (trans), we placed a lac O sequence between the mismatch and the nucleosome (Fig. 1A). The addition of Lac I protein to a lac O sequence has been previously shown to provide a high-affinity block to the diffusion of MSH2-MSH6 sliding clamps (Mendillo et al., 2005). We found that the addition of Lac I to the biotin-streptavidin blocked G/T mismatch nucleosome substrate induces a near complete inhibition of hMSH2-hMSH6 catalyzed nucleosome disassembly (Fig. 4; Suppl. Fig. 5; t1/2 (G/T•UN) = 791 min, t1/2 (G/T•K115Ac/K122Ac) = 198 min). These results strongly suggest that the hMSH2-hMSH6 chromatin remodeling activity requires a mismatch in cis with the nucleosome and that hMSH2-hMSH6 must translocate from the mismatch to the nucleosome for disassembly.
Nucleosomes are disassembled in front of and reassembled behind a replication fork (Groth et al., 2007). The first fully formed nucleosome may be found approximately 250 bp behind the replication fork with intermediates in the assembly process occurring in the intervening region (Jackson, 1988; Sogo et al., 1986). Post-replication MMR is likely to be initiated in vivo shortly after a mismatch escapes the replication machinery and has been shown to form excision tracts that encompass 100–1000 bp in vitro (Fang and Modrich, 1993). These observations suggest that the human MMR machinery may encounter both fully formed nucleosomes as well as nucleosome assembly intermediates.
Here we have demonstrated a new chromatin remodeling function for the MMR initiation heterodimer hMSH2-hMSH6. Chromatin remodeling by hMSH2-hMSH6 requires a cis-mismatch and translocation of the heterodimer along the DNA, ATP binding but not ATP hydrolysis, and it is enhanced by histone post-translational modifications that increase thermal repositioning and/or reduce histone-DNA affinity. We used the 5S rDNA positioning sequence, which strongly localizes nucleosomes compared to native DNA (Thastrom et al., 1999). These observations suggest that genome wide nucleosome disassembly by hMSH2-hMSH6 may be significantly more efficient. Moreover, artificially high affinity nucleosome positioning sequences, such as the non-physiological 601 positioning sequence, may mask the hMSH2-hMSH6 nucleosome disassembly process (Thastrom et al., 1999).
While we have demonstrated that the H3(K56Q) mimic of the replication-associated acetylation modification H3(K56Ac) clearly enhances nucleosome disassembly by hMSH2-hMSH6, there is growing evidence that bona fide histone acetylations additionally accelerate nucleosome thermal repositioning, which may substantially enhance hMSH2-hMSH6 dependent chromatin remodeling (Manohar et al., 2009). Moreover, SWI/SNF-independent (SIN) histone mutations that are located in the nucleosome dyad near H3(K115) and H3(K122) appear to increase the rate of nucleosome repositioning following thermal heating (Flaus et al., 2004; Muthurajan et al., 2004) and reduce DNA-histone interactions (Kurumizaka and Wolffe, 1997); thus reducing or eliminating the requirement for these chromatin remodeling factors in several DNA transactions (Kruger et al., 1995).
The rate of nucleosome disassembly (t1/2 = 23 min) appears well within the window of MMR in vitro (Constantin et al., 2005; Zhang et al., 2005), although the rate of MMR may be somewhat reduced in the presence of nucleosomes compared to naked DNA. Our results are consistent with the conclusion that hMSH2-hMSH6 performs two important functions for MMR: 1) it specifically recognizes mismatched nucleotides to initiate repair, and 2) it creates a nucleosome-free and perhaps protein-free environment surrounding the mismatch for the excision reaction. A requirement for translocation and the lack of any detectable interaction(s) with histone components or nucleosomes strongly suggests that hMSH2-hMSH6 chromatin remodeling functions are uniquely linked to its ability to form sliding clamps. A related reaction has been considered for RAD51 polymerization-dependent chromatin remodeling (Dupaigne et al., 2008). Because chromosomes throughout phylogeny contain complex mixtures of protein-DNA, our observations might be generalized to suggest that all MutS homologs that form sliding clamps function similarly. Several mechanisms for MMR have been proposed and remain controversial (for review see ref. (Kolodner et al., 2007). The Molecular Switch Model posits the mismatch-dependent loading of multiple MSH hydrolysis-independent sliding clamps that recruit MLH/PMS proteins, and connect mismatch recognition to an iterative dynamic and redundant strand excision process (Acharya et al., 2003; Gradia et al., 1997). Our observations appear to highlight an unanticipated strength of the Molecular Switch Model by suggesting that the iterative MSH hydrolysis-independent sliding clamps also perform chromatin remodeling.
Examining the role of hMSH2-hMSH6 in chromatin remodeling in vivo is complicated by the overlapping requirement for sliding clamps in both MMR and nucleosome disassembly. Thus, dissociating the hMSH2-hMSH6 chromatin remodeling activity from MMR activity has been impracticable. One prediction of our studies is that there may be a synergistic phenotype when partially defective alterations of the MMR machinery and chromatin modifying machinery are combined. While these studies are underway, they are technically challenging and may be subtle as a result of the significant redundancies associated with histone modification enzymes.
The absence of an energetic component associated with the translocation of hMSH2-hMSH6 sliding clamps suggests a unique passive mechanism for chromatin remodeling. We consider two models in which hMSH2-hMSH6 sliding clamps might trap inherent structural fluctuations in nucleosomes leading to disassembly (Fig. 5). One model proposes that the formation of iterative sliding clamps may capture thermally induced position-shifts of the nucleosome away from the mismatch; ultimately “nudging” the nucleosome off the open-end of our model DNA substrates (Fig. 5). Since free DNA ends are rare in vivo, such a Nudging Model would be envisioned to detain nucleosomes away from the mismatch along the DNA. A second model considers thermal fluctuations (breathing) by the nucleosome DNA (Li and Widom, 2004; Polach and Widom, 1995), which might be irreversibly captured in the open-state by hMSH2-hMSH6 sliding clamps (Fig. 5). In this Unwrapping Model, hMSH2-hMSH6 sliding clamps would iteratively occupy the DNA of a breathing nucleosome, beginning at the entry-exit region, until a critical DNA length is engaged and the nucleosome spontaneously disassembles. Both models do not appear to be mutually exclusive and may occur in concert. Passive chromatin remodeling has been considered for transcription factors where binding sites are occluded by nucleosomes (Polach and Widom, 1995). However, this would be the first case in which stable translocating DNA clamps, with no DNA sequence specificity, provoke the disassembly of nucleosomes. Regardless of the detailed mechanics, it appears that hMSH2-hMSH6 typifies an entirely new class of passive DNA lesion-dependent chromatin remodeling factors.
hMSH2–hMSH6 and the hMSH2(K675A)-hMSH6(K1140A) were purified as previously described (Gradia et al., 1997). Lac I protein was a generous gift from Dr. Kathleen Matthews (Rice University). The G/C and G/T oligonucleotides (5′-GCT TAG GAT CAT CGA GGA TCG AGC TCG GTG CAA TTC AGC GGG-3′with the complementary strand 5′-T CGA CCC GCT GAA TTG CAC CGA GCT (T/C)GA TCC TCG ATG ATC CTA AGC-3′ containing a 3′-biotin moiety) were synthesized (Midland Certified Research Company), annealed and purified by HPLC using a Waters Gen-Pak column (Gradia et al., 1997). The site of the mismatch is indicated in bold. The Xenopus 5S rDNA nucleosome localization sequence containing the lac O sequence (5′-TGG AAT TGT GAG CGG ATA ACA ATT-3′) on the 3′-end was amplified by PCR from a pBluescript (SK-) plasmid containing the Xenopus 5S rDNA sequence using tailed primers [5′-GCC CGG GGG ATC CAC TAG TTC - 3′; 5′-ACC GCC TGG GCC TGG TAC AAT TGT TAT CCG CTC ACA ATT CCA CTC GAG CGA -3′]. The PCR product (5S rDNA plus the lac O sequence) was digested with XhoI on the 3′-end and SmaI on the 5′-end. The annealed synthetic oligonucleotide containing a G/C duplex or G/T mismatch was ligated to the PCR product, purified by native PAGE and verified by restriction analysis.
Histone H3 acetylated at K115 and K122 was prepared by expressed protein ligation (Manohar et al., 2009). A peptide containing amino acids 110–135 was synthesized manually on Boc-Ala-PAM resin (Novabiochem) using standard Boc-Nα protection strategies and HBTU activation protocols. K115 and K122 were acetylated prior to HF cleavage from the resin and purified by RP-HPLC. Truncated histone H3 (residues 1–109) was cloned as a fusion protein with the GyrA intein into the pTXB1 vector (New England Biolabs). The H3-intein fusion protein was expressed in E. coli BL21 (DE3) cells and purified from inclusion bodies by ion exchange and gel filtration chromatography. The purified protein was refolded by dialysis into a high-salt buffer. Thiolysis was then initiated by addition of 100 mM MESNA (mercaptoethanesulfonic acid) and allowed to continue for 24 hours at 4°C. The buffer components were then adjusted to generate protein-ligation buffer I: 50 mM HEPES (pH 7.5), 6 M urea, 1 M NaCl, 1 mM EDTA, 50 mM MESNA and the protein concentrated to > 1 mg/mL of the thioester and stored at −80°C. Expressed protein ligation was done with ten molar equivalents of the acetylated H3(110–135) peptide to the H3(1–109) thioester in protein ligation buffer II: 50 mM HEPES pH 7.5, 6 M urea, 1 M NaCl, 1 mM EDTA, 20 mM TCEP, which proceeded overnight at room temperature with gentle agitation. Full-length semisynthetic H3 was then purified by ion exchange chromatography over a TSKgel SP-5PW column (TOSOH Bioscience).
Recombinant unmodified histones: H2A, H2B, H3 and H4 were expressed and purified as previously described (Luger et al., 1999). The unmodified, H3(K56Q) and H3(K115Ac, K122Ac) histones were unfolded separately in: 7 M guanidine, 20 mM Tris (pH 7.5) and 10 mM DTT for 1 to 3 hours and then spun to remove aggregates. The four core histones were combined at equal molar ratio with total histone concentration adjusted to 5 mg/ml in 200 ul. The octamer was refolded by double dialysis in: 2 M NaCl, 10 mM Tris-HCl (pH 7.5), 1 mM EDTA and 5 mM BME. The recovered refolded octamer was centrifuged to remove large aggregates and then purified over a Superdex 200 (GE healthcare) column. The purity of each octamer was confirmed by SDS-PAGE and mass spectrometry.
Nucleosomes were reconstituted with [32P]-labeled nucleosome-DNA substrate (Fig. 1A) and with octamer containing unmodified, H3(K56Q), or H3(K115Ac, K122Ac) histones by salt double dialysis as previously described (Thastrom et al., 2004). The reconstituted nucleosomes were purified by ultracentrifugation on a 5–30% sucrose gradient. Fractions corresponding to the peak of reconstituted nucleosomes were pooled and concentrated in a Centricon 30 concentrator (Amicon) and washed twice with 0.5 X TE. The nucleosome purity was verified with a 5% native polyacrylamide gel containing 1/3 X TBE.
Reactions were performed in: 25 mM HEPES (pH 7.8), 15% glycerol, 100 mM NaCl, 1 mM DTT, 2 mM MgCl2, containing 20 ng/μL poly dI-dC, 200 μg/mL acetylated BSA (Promega), and approximately 5 fmols of [32P]-labeled mono-nucleosome or the 265 bp free-DNA substrate in a final volume of 20 μl. hMSH2-hMSH6 (at the indicated concentration) was preincubated with the nucleosome-DNA on ice for 10 min. Reactions were separated on a 5% native polyacrylamide/5% glycerol in 1/3 X TBE at 4°C for 3 hours. Gels were dried, quantified by phosphorimager (Molecular Dynamics), and represented as percent substrate shifted. Standard deviation was calculated from at least three separate experiments. The ATPase activity was determined in: 25 mM HEPES (pH 7.8), 100 mM NaCl, 10 mM MgCl2, 1 mM DTT, 0.01 mM EDTA, 15% glycerol, 200 μg/mL acetylated BSA (Promega), 500 μM unlabeled ATP and 16.5 nM [γ-32P]-ATP in a final volume of 20 μl. Steady-state reactions were performed using 25 nM hMSH2-hMSH6 and 25 nM free-DNA, nucleosome-DNA or without DNA as indicated. We determined that ATP hydrolysis was linear under these conditions for at least 2 hr. Reactions were incubated at 37°C for 60 min and processed as described previously (Gradia et al., 1997).
Reactions were performed in: 25 mM HEPES (pH 7.8), 15% glycerol, 100 mM NaCl, 1 mM DTT, 5 mM MgCl2, 20 ng/μL poly dI-dC, 200 μg/mL acetylated BSA, and approximately 5 fmols of [32P]-labeled nucleosome-DNA in a 20 μl reaction volume. Where indicated, 900 nM of streptavidin was included for 5 min on ice prior to the addition of hMSH2-hMSH6 or hMSH2(K675A)-hMSH6(K1140A). Reactions were incubated with hMSH2-hMSH6 (250 nM) or hMSH2(K675A)-hMSH6(K1140A) (250 nM) on ice for 10 minutes. Where indicated, 4 nM Lac I was incubated with the nucleosome-DNA for 10 min on ice prior to the addition of hMSH2-hMSH6. Dissociation with 1 mM ATP (or ATPγS) was performed where indicated by addition of nucleotide and a further incubation from 10–60 min at 37°C. The reactions were separated on a 5% native polyacrlyamide/5% glycerol gel in 1/3 X TBE at 4°C for 3 hr. Gels were dried and quantified by phosphorimager (Molecular dynamics). Standard deviations were calculated from at least three independent experiments.
The authors wish to thank Michael Smerdon and Ravindra Amunugama for 5S rDNA plasmids and constructs; Kathleen Matthews for Lac I protein; Justin North and Robin Nakkula for help in octamer preparation; Thomas Haver for technical assistance; and Kristine Yoder and Jessica Tyler for helpful discussions. This work was funded by NIH/NCI grants CA067007 and GM062556 (R.F.); GM083055 (M.G.P. and J.J.O.) and a Career Award in Basic Biomedical Sciences from the Burroughs Welcome (M.G.P.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.