HDAC6 regulates DNA damage response via deacetylating MLH1

MutL homolog 1 (MLH1) is a key DNA mismatch repair protein, which plays an important role in maintenance of genomic stability and the DNA damage response. Here, we report that MLH1 is a novel substrate of histone deacetylase 6 (HDAC6). HDAC6 interacts with and deacetylates MLH1 both in vitro and in vivo. Interestingly, deacetylation of MLH1 blocks the assembly of the MutSα–MutLα complex. Moreover, we have identified four novel acetylation sites in MLH1 by MS analysis. The deacetylation mimetic mutant, but not the WT and the acetylation mimetic mutant, of MLH1 confers resistance to 6-thioguanine. Overall, our findings suggest that the MutSα–MutLα complex serves as a sensor for DNA damage response and that HDAC6 disrupts the MutSα–MutLα complex by deacetylation of MLH1, leading to the tolerance of DNA damage.

MutL homolog 1 (MLH1) is a key DNA mismatch repair protein, which plays an important role in maintenance of genomic stability and the DNA damage response. Here, we report that MLH1 is a novel substrate of histone deacetylase 6 (HDAC6). HDAC6 interacts with and deacetylates MLH1 both in vitro and in vivo. Interestingly, deacetylation of MLH1 blocks the assembly of the MutS␣-MutL␣ complex. Moreover, we have identified four novel acetylation sites in MLH1 by MS analysis. The deacetylation mimetic mutant, but not the WT and the acetylation mimetic mutant, of MLH1 confers resistance to 6-thioguanine. Overall, our findings suggest that the MutS␣-MutL␣ complex serves as a sensor for DNA damage response and that HDAC6 disrupts the MutS␣-MutL␣ complex by deacetylation of MLH1, leading to the tolerance of DNA damage.
Mismatch repair is a mutation avoidance mechanism that corrects DNA replication errors, including small insertions, deletions, and mis-incorporated bases, as well as some forms of DNA damage (1)(2)(3). This process begins with mismatch recognition, which is mainly carried out by the MSH2-MSH6 heterodimer (MutS␣). Next, the DNA-bound MutS␣ recruits MLH1-PMS2 heterodimer (MutL␣) to increase the footprint on the DNA. Eukaryotes have three different forms of heterodimers that include MLH1, designated as MLH1-PMS2 (MutL␣), MLH1-PMS1 (MutL␤), and MLH1-MLH3 (MutL␥). Among these heterodimers, MutL␣ acts as the matchmaker and facilitator, coordinating events in mismatch repair (4,5). This step may serve as an entry point for the exonuclease activity that removes mismatched DNA in the presence of other required proteins, including MutS␣ and downstream proteinproliferating cell nuclear antigen. Because of its critical role in MMR, 3 deletion or mutation of MLH1 causes genomic instability in hereditary nonpolyposis colorectal cancer (HNPCC) (6,7). More than 240 mutations have been described in HNPCC, and about 60% of these mutations occurred in hMLH1. From the analyses of HNPCC, more than 25% of gene alterations in hMLH1 were identified as minor variants such as amino acid replacement or small in-frame deletions. These mutations are scattered throughout the entire coding region, indicating the importance of every single domain of the MLH1 protein in its full function. To date, the UMD-MLH1 mutation database shows a total of 3,063 recorded mutations, and 568 different variants from 2,527 samples. Among all the mutations, 47 nonsense and 198 missense and synonymous variants in hMLH1 are reported. Aside from cancer susceptibility, MLH1 also affects fertility, as reported in a knockout mouse study. Both males and females exhibit normal mating behavior; however, both are sterile (8). In addition to mutations on MLH1 cDNA, hypermethylation of its promoter is another main cause of hMLH1 gene silencing involved in sporadic cancers (9).
Antimetabolites (e.g. 6-thioguanine (6-TG)), alkylating agents (e.g. N-methyl-NЈ-nitro-N-nitrosoguanidine (MNNG) and N-methyl-N-nitrosourea (MNU)), and platinum-containing drugs (e.g. cisplatin) can trigger cell cycle arrest and apoptosis in cells via MMR proteins (10). Cell lines, such as 2008/A and HCT116 that are deficient in hMLH1, became more resistant to MNNG and cisplatin, respectively (11), suggesting the involvement of MLH1 in DNA damage response. It has been reported that MLH1 plays a critical role in apoptosis in either the p53dependent or -independent mechanism. In the presence of p53, MLH1 mediates MNNG-and MNU-induced cell death by increasing phosphorylation of Ser-15 in p53, leading to apoptosis (12). In the absence of p53, MLH1 is associated with the c-Abl-p73-apoptosis pathway in response to cisplatin-in-duced DNA damage (13). Overall, the above evidence has shown that MLH1 indeed plays an important role in DNA damage response. However, how MLH1 is regulated at the posttranslational level is understudied.
Histone deacetylase 6 (HDAC6) was cloned from mice and humans as a mammalian homolog of yeast HDA1 (14,15). It contains two deacetylase domains, termed DAC1 and DAC2, as well as a ZnF-UBP domain in the C terminus. Our previous studies have shown that DAC2 has full deacetylase activity, whereas DAC1 possesses intrinsic E3 ligase activity both in vitro and in vivo (16). HDAC6 is now considered to be a master regulator of cellular response to cytotoxic assaults (17)(18)(19) and plays a role in genotoxic stress responses (16,20,21). Here, we have identified the MMR protein MLH1 as a factor for HDAC6mediated DNA damage response function. HDAC6 interacts with and deacetylates MLH1 both in vivo and in vitro. We also identified four new acetylation lysine sites in the MLH1 protein, and HDAC6 may regulate the acetylation status of these sites to disrupt the assembly of the MutS␣-MutL␣ complex, leading to 6-TG tolerance. These findings suggest that as a downstream target of HDAC6, MLH plays an important role in HDAC6mediated DNA damage response.

MLH1 interacts with HDAC6
We have previously shown that HDAC6 forms a complex in the nucleus with several DNA mismatch repair proteins, including MSH2, MSH6, and MLH1, in HeLa cells (16). Although we have shown that HDAC6 sequentially deacetylates and ubiquitinates MSH2, it is completely unknown whether HDAC6 also regulates MLH1. To this end, we first verified that HDAC6 interacts with MLH1. We assessed the endogenous interaction between HDAC6 and MLH1 via coimmunoprecipitation. As shown in Fig. 1A, lung cancer A549 cell extracts were immunoprecipitated with either anti-rabbit IgG or anti-HDAC6 antibody. MLH1 could only be detected in the immunoprecipitates from the anti-HDAC6 antibody but not from the control anti-IgG. In the reciprocal experiment, anti-MLH1 antibody, but not the anti-rabbit IgG control, specifically pulled down HDAC6 protein (Fig. 1B). Thus, HDAC6 and MLH1 are indeed associated with each other in vivo.
To determine whether HDAC6 directly binds to MLH1 or through other associated proteins, we performed in vitro GST pulldown assays with bacterially-purified GST-HDAC6 and His-MLH1. As shown in Fig. 1C, GST-HDAC6, but not GST or GSH-agarose, efficiently pulled down His-MLH1. This result indicates a direct binding between HDAC6 and MLH1.
We next attempted to map which region of HDAC6 interacts with MLH1. HDAC6 has two deacetylation domains in the N terminus, termed DAC1 and DAC2, and a ZnF-UBP domain in the C terminus. As shown in Fig. 1, D and E, both DAC1 (1-503 aa) and DAC2 (488 -839 aa), but not the C terminus (835-1215 aa), bind to MLH1. We then examined which region of MLH1 binds to HDAC6. MLH1 has an ATPase domain (1-147 aa) in the N terminus. The middle region (147-320 aa) is for MutS␣ binding, and the very C-terminal domain is for PMS2/MLH3/ PSM1 binding (491-756 aa). We found that both N-and C-ter-minal domains can bind to HDAC6, except a portion of the middle region (320 -491 aa) that binds to the MutS␣ complex ( Fig. 1, F and G). We previously showed that HDAC6 may interact with MLH1 in HeLa nuclear extracts (16). To confirm whether HDAC6 binds to MLH1 in the nucleus, we performed the co-immunoprecipitation experiments using anti-IgG or anti-HDAC6 to pull down MLH1 in T29 nuclear extracts. The method of preparing nuclear extracts was described in Zhang et al. (16). As shown in Fig. 2A, endogenous HDAC6 indeed binds to MLH1 in the nucleus. Next, we performed immunofluorescence staining of MLH1 and HDAC6. As shown in Fig. 2, B and C, HDAC6 is co-localized with MLH1 in H1299 cells, an effect that is enhanced upon etoposide treatment. The result suggests that HDAC6 may be translocated into the nucleus upon DNA damage.
To examine the portion of HDAC6 in the nucleus, we performed nuclear and cytoplasmic fractionation of four nonsmall cell lung cancer cell lines, A549, H460, H292, and H125. PARP1 and Hsp60 were used as nuclear-and cytoplasmic-specific markers, respectively. As shown in Fig. 2D, it appears that HDAC6 is almost equally distributed in nuclear and cytoplasmic fractions in all four cell lines except H460, in which there is at least twice as much HDAC6 protein in the cytoplasm compared with the nucleus. Because equal amounts of protein were loaded in each lane, and our prior experience has indicated that there is 3-4-fold more protein in the cytosol than that in the nucleus, we therefore estimated that ϳ20% of HDAC6 is localized in the nucleus of A549, H292, and H125 cell lines, and 10% of HDAC6 is localized in the nucleus of H460 cell line.

HDAC6 deacetylates MLH1
Given the fact that HDAC6 is a deacetylase, we hypothesized that MLH1 could be a substrate of HDAC6. To this end, we tested whether MLH1 is an acetylated protein. We treated 293T cells with trichostatin A (TSA), a class I, II, and IV HDAC inhibitor, and we immunoprecipitated MLH1 with an anti-MLH1 antibody. As shown in Fig. 3A, the level of acetylated MLH1 is increased upon TSA treatment, suggesting that HDACs regulate MLH1 acetylation. To determine the involvement of HDAC6 in MLH1 deacetylation, we treated 293T cells with tubastatin A, an HDAC6-specific inhibitor. As shown in Fig. 3B, the level of acetylated MLH1 is increased upon tubastatin A treatment, suggesting that HDAC6 deacetylates MLH1. Next, we tested whether histone acetyltransferases (HATs) could elevate MLH1 acetylation. As shown in Fig. 3C, p300, a HAT, enhances the acetylation of MLH1 in 293T cells. To determine whether MLH1 is a substrate of HDAC6, we then performed a deacetylation assay. As shown in Fig. 3D, acetylation of MLH1 was reduced in the cells overexpressing HDAC6 compared with the control cells, indicating that HDAC6 deacetylates MLH1 in vivo. To exclude the possibility that HDAC6 deacetylates MLH1 through its associated proteins, we performed an in vitro deacetylation assay. HDAC6 was purified from 293T cells and incubated with acetylated MLH1. As shown in Fig. 3E, the acetylation level of MLH1 was significantly decreased in the reaction with purified HDAC6. Thus, HDAC6 can deacetylate MLH1 both in vivo and in vitro.

Lysines 33, 241, 361, and 377 of MLH1 are targeted for acetylation
To detect the acetylation sites present in MLH1, we overexpressed FLAG-MLH1 in 293T cells followed by treatment with TSA and subjected the samples to MS analyses. As shown in Fig. 4, Lys-33, Lys-241, Lys-361, and Lys-377 were identified as acetylation sites, which spread across the ATPase and MutS␣ interaction domains. Lysine 33 is located in the ATPase domain and conserved among mammals, zebrafish (Danio rerio), clawed frog (Xenopus tropicalis), fruit fly (Drosophila melanogaster), Caenorhabditis (Caenorhabditis elegans), and even in plant (Arabidopsis thaliana), yeast (Saccharomyces cerevisiae and Schizosaccharomyces pombe), and Escherichia coli (Fig.  5A), indicating that lysine 33 plays an important role in MLH1 function. Lysine 241 is located in the MutS homologs' interaction domain and is conserved among mammals, zebrafish, and clawed frog, indicating their potential role in MMR complex formation between Muts␣ and MutL␣ (Fig. 5B). Lysine 361 is also located in the MutS homologs' interaction domain and is conserved from mammals to clawed frog, with the exception of mouse (Fig. 5B). The Lys-377 site shows a less conserved pattern compared with the other three sites, indicating a distinct regulatory role during evolutionary processes (Fig. 5B). The domain structure of MLH1 and the positions of these four lysine sites are shown in Fig. 5C.

Deacetylation of MLH1 by HDAC6 blocks the MutL␣-MutS␣ complex formation
To further confirm whether the identified sites could be acetylated in vivo, we mutated each of the four lysine residues to arginine. However, single-site mutation of MLH1 did not show a significant difference in acetylation levels when compared with WT MLH1 (data not shown). Thus, we mutated all four lysine sites of MLH1 either to arginine (4KR) or glutamine (4KQ) to mimic the nonacetylated and acetylated form, respectively. As shown in Fig. 6A, acetylation of the 4KR and 4KQ mutants was abolished, indicating that these four lysine sites are indeed the major acetylation sites in vivo. With a longer exposure of Western blottings for the total anti-acetylated lysine antibody, 4KR and 4KQ showed weak signals (data not shown), suggesting that there are additional minor unidentified acetylation sites in MLH1. To determine the biological role of MLH1 acetylation, we first tested whether acetylation/deacetylation of MLH1 affects MLH1-PMS2 association. As shown in Fig. 6B, no observable difference existed between WT MLH1 and its mutants in terms of their binding to PMS2 (Fig. 6A, lane 4). We next examined whether the acetylation status of MLH1 affects the MutL␣-MutS␣ complex formation. As shown in Fig. 6B, the MLH1-4KR mutant loses its binding affinity to both MSH6 and MSH2, whereas MLH1-4KQ displays the same binding affinity as WT MLH1 (lanes 1 and 2). This result suggests that deacetylation of MLH1 diminishes its binding to the MutS␣ complex. To determine whether binding deficiency of MLH1-4KR is due to the change in acetylation levels of all four lysines or a site-specific effect, we examined the single lysine to arginine mutants and found that none of the single sites affected the binding of both MSH2 and MSH6 (Fig. 6C). We next established HDAC6 Tet-On inducible knockdown U2OS and H292 cell lines to examine the role of HDAC6 in the MutS␣-MutL␣ complex formation. As shown in Fig. 6, D and E, knockdown of HDAC6 increases the binding of MutS␣ to MLH1 in both U2OS and H292 cell lines, indicating that HDAC6 prevents the formation of the MutS␣-MutL␣ complex.

Deacetylation mimetic mutant of MLH1 confers 6-TG resistance
We previously showed that low levels of the MutS␣ complex confer 6-TG resistance (16). However, it is not clear whether the MutL␣ complex is involved in a 6-TG-mediated DNA damage response. To this end, we utilized the human embryonic kidney (HEK) 293T cell line to test whether MutL␣ plays a role in 6-TG sensitivity. HEK293T is considered to be an MMRdeficient cell line because of promoter hypermethylation on MLH1 and extremely low expression of PMS2 (22). We stably transfected vector or MutL␣ (MLH1-PMS2) in 293T cells. As shown in Fig. 7A, compared with the vector control, 293T cells harboring WT MutL␣ (MLH1-PMS2) showed a damage-susceptible phenotype in response to 6-TG, with cleaved PARP-1 protein being used as a surrogate for apoptosis, indicating that ectopically expressed MutL␣ indeed functions as a sensor in 6-TG-induced cell death signaling. We next tested the difference between WT MLH1 and its mutants in response to 6-TG treatment. We stably transfected MLH1(4KR)-PMS2 and MLH1(4KQ)-PMS2 into 293T cells. As shown in Fig. 7B, both WT MLH1 and acetylation mimetic mutant 4KQ, but not deacetylation mimetic mutant 4KR, showed an increase in apoptosis upon 6-TG treatment. As we have shown in Fig. 6B, 4KR cannot form the complex with MutS␣ efficiently, and thus we conclude that the assembly of MutS␣-MutL␣ is necessary for mediating 6-TG-induced apoptosis. The expression levels of exogenous MLH1 and PMS2 are shown in Fig. 7E.

Acetylation/deacetylation of Lys-33 and Lys-241 in MLH1 may be important for MSH2 binding
Finally, we explored the mechanism by which the four acetylated lysine residues in MLH1 regulate the assembly of the MutS␣-MutL␣ complex. Of the four residues, Lys-361 and Lys-377 are outside the structured area, and Lys-361 is not conserved in the mouse (Fig. 5). We then focused on Lys-33 and Lys-241. As shown in Fig. 8, both Lys-33 and Lys-241 are located in a region enriched with acidic residues. Lys-33 is located near Glu-34 and Glu-37. Lys-241 is located close to Glu-234 and Glu-297. These acidic residues form a continuous negatively charged surface patch that covers the majority of the surface of an elongated groove. If this groove is a potential site for MSH2 binding, the interaction regions in MSH2 are likely positively charged. Without acetylation, Lys-33 and Lys-241 may interfere with MSH2 binding via electrostatic repulsion. On the contrary, the interaction with MSH2 may be enhanced because of lysine neutralization by acetylation.

Discussion
In this study, we have shown that a key DNA mismatch repair protein, MLH1, is a novel substrate of HDAC6. Acetylation/ deacetylation of MLH1 can be regulated by p300 and HDAC6, respectively. Deacetylation of MLH1 leads to disassociation of the MutS␣-MutL␣ complex, a core component of DNA mismatch repair machinery, and induces 6-TG tolerance (Fig. 9).
Much attention has been focused on how germline or somatic mutations, as well as promoter methylation, modulate the MLH1 gene, leading to the accumulation of DNA replication errors (7,23). These errors eventually manifest in a high frequency of microsatellite instability (MSI-H). However, posttranslational modifications of the MLH1 protein, and how these modifications regulate MLH1-mediated DNA damage response, are largely unknown. Our study, for the first time, has revealed that deacetylation of MLH1 by HDAC6 confers 6-TG resistance via disruption of the MutS␣-MutL␣ complex. Because both TSA and nicotinamide treatment significantly increase the acetylation of MLH1 (Fig. 3A and data not shown), and nicotinamide is a sirtuin inhibitor, we speculate that our finding that HDAC6 deacetylates MLH1 may be the tip of the iceberg. Other HDACs and sirtuins could also modulate the status of MLH1 acetylation. Conversely, we suspect that in addition to p300, other HATs may acetylate MLH1. Therefore, multiple HDACs, sirtuins, and HATs could govern MLH1 functionality.
A previous report (24) has shown that the N terminus of MLH1 (1-505 aa) is essential for its binding to the MutS␣ complex, although the C terminus of MLH1 is important for the MLH1-PMS2 heterodimer formation. The results of our study are consistent with this published report. All four acetylated lysines (Lys-33, Lys-241, Lys-361, and Lys-377) are located in the N terminus of MLH1. The deacetylation mimetic mutant of MLH1 exhibits a significantly reduced binding affinity to the MutS␣ complex compared with the WT and acetylation mimetic mutant of MLH1. However, the acetylation mimetic mutant of MLH1 displays a similar binding affinity to the MutS␣ complex compared with the WT MLH1. Previously, we have identify four acetylated lysine sites (Lys-845, Lys-847, Lys-871, and Lys-892) in the C terminus of MSH2, where MSH2 binds to the MutL␣ complex (16). An interesting continuation of this work would be to study whether acetylation/deacetylation of these sites would affect MSH2's binding to MLH1 in the future.
Because MLH1's mutation rate is high in HNPCC, also called Lynch syndrome, we examined whether the four acetylation sites of MLH1 (Lys-33, Lys-241, Lys-361, and Lys-377) are mutated in HNPCC. By searching the online database (https:// preview.ncbi.nlm.nih.gov/clinvar), we found that three out of four sites (Lys-33, Lys-241, Lys-361, and Lys-377) are mutated in HNPCC patients. These mutations are K33E, K241E, K241R, and K361E. Although the phenotypes of these mutations have yet to be reported, we predict that the patients harboring these mutations may exhibit MMR deficiency. Of these sites, Lys-33 is highly conserved among species and located within the ATPase domain. We conducted the ATPase assays using the mutant of MLH1 containing K33R and found that the mutant displays the same ATPase activity as the WT (data not shown), suggesting that HDAC6 does not regulate the ATPase activity of MLH1. Our structural analysis suggests that acetylation of the Lys-33 and Lys-241 sites favors MLH1's binding to MutS␣. Because both Lys-361 and Lys-377 sites are located in the unstructured region, we could not perform molecular surface analysis. We assume that acetylation of both sites may assist MLH1's binding to MutS␣.
Our lab and others have shown that depletion of HDAC6 sensitizes cells to different drugs, including cisplatin, doxorubicin, MNNG, and 6-TG (16,20,21). We revealed that HDAC6 regulates DNA damage response and mismatch repair activities via regulation of MutS␣ homeostasis through sequential deacetylation and ubiquitylation of MSH2 (16). In this study we

HDAC6 deacetylates MLH1
have, for the first time, shown that HDAC6 also regulates the DNA damage response by disrupting the assembly of the MutS␣-MutL␣ complex via deacetylation of MLH1. Interestingly, both acetylation sites Lys-33 and Lys-377 of MLH1 can also be ubiquitinated (25). Expanding the data presented in this study may include testing whether HDAC6 ubiquitinates MLH1. Overall, our past and present studies have demonstrated that HDAC6 modulates the DNA damage response by regulating at least two critically important mismatch repair proteins, MSH2 and MLH1. For all the panels, the acetylated MLH1 bands were quantified and the fold-changes were shown below the those bands. IB, immunoblot.

Cell culture and transfection
All cell lines were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum or Tet-free serum for doxycyclineinducible cell lines, penicillin (100 units/ml), and streptomycin (100 units/ml), except H292-inducible cells that were grown in RPMI 1640 medium with 10% Tet-free serum. Cells were incubated at 37°C with 5% CO 2 . The plasmids were transiently or stably transfected into cells using Lipofectamine 2000 (Invitrogen).

Plasmids, antibodies, and chemicals
The full-length GST-HDAC6 was constructed by inserting the cDNA into the pGEX-4T1 vector. Briefly, HDAC6 full-length cDNA was amplified by PCR with HA-HDAC6-F (16) as a template and primers GST-HDAC6-F and GST-HDAC6-R. The full-length GST-HDAC6 was constructed by inserting the cDNA between SalI and NotI sites into the pGEX-4T1 vector. The F-HDAC6 constructs (full-length, 1-503, 488 -839, and 835-1215) were constructed by inserting cDNAs into pCMV-3Tag-1a between HindIII and SalI sites. The FLAG-MLH1 and its deletion mutants were constructed by inserting cDNA into p3ϫFLAG-CMV10 vector (Sigma) between EcoRI and BamHI sites. The single site mutants (K33R, K241R, K361R, and K337R) of MLH1 were generated using this plasmid as the template. Myc-MLH1 was constructed by inserting cDNA into pCMV-3Tag-2a (Agilent Technologies) between EcoRI and A, lysine 33 is acetylated in MLH1. The peptide was detected with a m/z of 625.3074, which represents an error of 6.1 ppm. The tandem mass spectrum matched the following sequence, PANAIKEMIENCLDAK, indicating that the first lysine was acetylated. B, lysine 241 is acetylated in MLH1. The peptide was detected with a m/z of 1040.0105, which represents an error of 4.2 ppm. The tandem mass spectrum matched the following sequence, TLAFKMNGYISNANYSVK, indicating that the first lysine was acetylated. C, lysine 361 is acetylated in MLH1. The peptide was detected with an m/z of 1237.5928, which represents an error of 1.4 ppm. The tandem mass spectrum matched the following sequence, MYFTQTLLPGLAGPSGEMVKSTTSLTSSSTSGSSDK, indicating that the first lysine was acetylated. D, lysine 377 is acetylated in MLH1. The peptide was detected with an m/z of 858.7390, which represents an error of 4.3 ppm. The tandem mass spectrum matched the following sequence, STTSLTSSSTSGSSDKVYAHQMVR, indicating that the first lysine was acetylated.

HDAC6 deacetylates MLH1
XhoI sites. FLAG-MLH1 and PMS2 were tandemly cloned to LentiORF pLEX-MCS vector (ThermoFisher Scientific Open Biosystems) with a T2A sequence between two cDNAs to generate FLAG-MLH1/4KR/4KQ-T2A-PMS2 constructs (please see supporting Experimental procedures and Table S1 for detailed information). The bacterially expressed His-MLH1 was constructed by inserting cDNA sequences of full-length MLH1 into the pET-28a vector between EcoRI and XhoI. HA-PMS2 plasmid was constructed by inserting PMS2 cDNA between the BamHI and XhoI sites in pcDNA3.1-C-HA vector. Briefly, PMS2 cDNA was amplified by PCR using pBluescript PMS2 (Addgene Plasmid no. 16457) as a template and primers PMS2F-BamHI, and PMS2R-SalI. HA-PMS2 was obtained by inserting PMS2 cDNA between BamHI and XhoI sites in pcDNA3.1-HA (with a C-terminal HA tag). Please note that SalI and XhoI generate the same cohesive end after cutting. The inducible HDAC6 shRNAs were purchased from Dharmacon Inc., and the targeted sequences are V3THS_330047, TTCGCTTCGAAGTGACACT, and V2THS_71188, TTCT-GTTGAGCATAGCGGG. Please see Table S1 for primers used in generating the above constructs. The anti-HDAC6 (H-300), anti-MLH1 (C-20, N-20, and B-12), anti-PMS2 (C-20), anti-HA (Y-11), and anti-Myc (8E10) antibodies were purchased from Santa Cruz Biotechnology. The anti-AcK antibody (catalog no. 06-933) was purchased from Upstate. The anti-MLH1 antibody was also purchased from Abcam. The anti-HDAC6 antibodies were purchased from Sigma or Santa Cruz Biotechnology, Inc. The anti-MSH2 antibody was purchased from Calbiochem, and the anti-MSH6 antibody was from BD Biosciences. The anti-PARP1 antibody was purchased from Cell Signaling Technology. The anti-FLAG-M2 antibody and agarose beads, anti-␤actin antibody, 6-TG, doxycycline, etoposide, tubastatin A, MTT, and protease inhibitor mixture were purchased from Sigma. Nickel-nitrilotriacetic acid resin was purchased from Clontech.

6-TG treatment
6-TG was dissolved in the medium directly from a stock of 1.5 mM. Cells were washed with PBS and incubated in fresh medium with 6-TG at 37°C for different time periods, as indicated in the figures. For longer treatments, 6-TG was replenished every 2 days.

Immunoprecipitation and immunoblotting
For immunoprecipitations, cells were lysed in LS buffer (PBS, pH 7.5, 10% glycerol, 0.1% Nonidet P-40, protease inhibitor mixture), BC100 buffer (20 mM Tris, pH 7.9, 100 mM NaCl, 10% glycerol, 0.2 mM EDTA, 0.2% Triton X-100, and protease inhibitor mixture), or cell lysis buffer (10 mM Tris, pH 8.0, 85 mM KCl, 5 mM EDTA, 0.5% Nonidet P-40, 0.25% Triton X-100, and fresh protease inhibitor mixture). Lysates were incubated for 2 h with protein A-or protein G-agarose beads for pre-clearing. Primary antibodies were added to the supernatants for 12 h at 4°C, followed by the addition of protein A-or G-agarose beads for an extra 2 h. Immune complexes were collected, washed three times in lysis buffer, and resolved on SDS-PAGE. For immunoblotting, samples were transferred to nitrocellulose membranes and then probed with the indicated antibodies.

Establishment of stable clones
For construction of the stable cell lines, F-MLH1-PMS2, F-MLH1-4KR-PMS2, F-MLH1-4KQ-PMS2, and empty vector were used to transfect 293T cells. One day after transfection, cells were split. After 24 h, 2 g/ml puromycin was added into the medium to select positive cells. Ten days later, stable cell lines were subcloned into 60-mm dishes, and 1 g/ml puromycin was added to the medium to maintain the stable clones in the subsequent culture. The well-isolated single clones were transferred into 24-well plates. The ectopic protein expression was verified by Western blotting analyses using anti-FLAG antibodies. WT MLH1 and its mutant clones with comparable protein expression were isolated for further experiments.

Establishment of HDAC6 inducible knockdown stable clones
The TRIPZ-inducible lentiviral vectors containing shH-DAC6-047 and -088 were transfected into U2OS and H292 cells. One day after transfection, 1 g/ml puromycin was added into the medium to select positive cells. Two weeks later, stable cell lines were subcloned to 60-mm dishes. Then cells were cultured for another 4 weeks with puromycin for further selection. Positive cells were confirmed by the manufacturer's protocol. For induction of HDAC6 knockdown, vehicle or doxycycline (0.5 g/ml) was added to each positive clone for 4 days. Anti-HDAC6 Western blot analysis was used to verify HDAC6 knockdown.

Colony formation assay
FLAG-MLH1/4KR/4KQ-PMS2 or empty vector was stably transfected; 293T cells were seeded in triplicates (500 cells per 60-mm dish); and the cells were incubated overnight at 37°C to allow for adherence to the dishes. Cells were then treated with a vehicle control or 6-TG at 0.5 M for 10 days. Afterward, cells were directly fixed and stained with crystal violet (0.05% w/v, 1% formaldehyde, 1% methanol in PBS) for 20 min. Colonies on each plate were scanned and counted using OpenCFU software.

MTT assay
2,000 cells were seeded in 96-well plates and incubated at 37°C overnight. 6-TG was added into the medium for 2 days at the indicated concentrations. Afterward, 15 l of MTT was added into medium for 4 h before being replaced and dissolved with DMSO. The absorbance was quantified by spectrophotometer at 570 nm wavelength.

Identification of acetylation sites of MLH1 by LC-MS/MS
The gel band containing acetylated MLH1 was excised and treated with tris(2-carboxyethyl)phosphine and iodoacetamide. Trypsin in-gel digestion was carried out at 37°C overnight. The extracted peptides were analyzed by LCMS/MS. A nanoflow liquid chromatograph (U3000, Dionex, Sunnyvale, CA) coupled to an electrospray ion trap mass spectrometer (LTQ-Orbitrap, ThermoFisher Scientific, San Jose, CA) was used for tandem MS peptide-sequencing experiments. The sample was first loaded onto a pre-column (5 mm ϫ 300 m inner diameter packed with C18 reversed-phase resin, 5 m, 100 Å) and washed for 8 min with aqueous 2% acetonitrile and 0.04% TFA. The trapped peptides were eluted onto the analytical column, (C18, 75 m inner diameter ϫ 15 cm, Pepmap 100, Dionex, Sunnyvale, CA). The 120-min gradient was programmed as follows: 95% solvent A (2% acetonitrile ϩ 0.1% formic acid) for 8 min and solvent B (90% acetonitrile ϩ 0.1% formic acid) from 5 to 50% for 35 min, then solvent B from 50 to 90% for 2 min, and holding at 90% for 5 min, followed by solvent B from 90 to 5% for 1 min and re-equilibration for 10 min. The flow rate on the analytical column was 300 nl/min. Tandem mass spectra were collected in a data-dependent manner following each survey scan. The MS scans were performed in Orbitrap to obtain accurate peptide mass measurement, and the MS/MS scans were performed in a linear ion trap using 60 s exclusion for previously sampled peptide peaks. Sequest and Mascot searches were performed against the Swiss-Prot human database downloaded on February 10th, 2009. Two trypsinmissed cleavages were allowed, and the precursor mass tolerance was 1.08 Da. MS/MS mass tolerance was 0.8 Da. Both MASCOT and SEQUEST search results were summarized in Scaffold 2.0.

GST pulldown
GST fusion proteins were purified as described previously (16). For in vitro binding assays, His-tagged MLH1 protein was incubated with 10 l of GSH-agarose beads or beads with 1 g of the GST protein or GST-HDAC6 protein in BC100 at 4°C overnight with gentle rotation. The GST beads were washed four times with BC100 buffer and eluted with 40 l of BC100 plus 20 mM reduced GSH for 2 h with gentle rotation. Half of

HDAC6 deacetylates MLH1
the elution was resolved on an SDS-polyacrylamide gel and detected by Western blotting.

Immunofluorescence staining
H1299 cells were plated in a 12-well plate containing a circular glass coverslip in 1 ml of RPMI media. 24 h post-plating, etoposide was added to a final concentration of 10 M. At the indicated time, glass coverslips were removed from the wells and fixed in 4% paraformaldehyde in a 24-well plate at room temperature with shaking for 10 min and stored in PBS at 4°C until all time points were collected. Then, coverslips were washed with PBS, then washed with 0.3% Triton X-100 in PBS for 10 min two times, then washed with 2% BSA 0.3% Triton X-100 in PBS for 15 min. Then, primary antibodies, with a 1:800 dilution of rabbit anti-HDAC6 (Sigma) and mouse anti-MLH1 (Abcam G168-15, ab14206), were added to a mixture of 150 l of 2% BSA and 0.1% Triton X-100 in PBS, which was added to the coverslips. The plate was sealed with paraffin wrap and incubated on the shaker in a cold room (4°C) for 48 h. After 48 h, the plate was removed from the cold room and placed on the room temperature shaker for 20 min. Primary antibody was removed, and coverslips were washed with PBS for 15 min three times. Then for the secondary antibodies, a 1:800 dilution of AlexaFluor 488 -tagged anti-rabbit secondary and AlexaFluor 594 -tagged anti-mouse secondary (ThermoFisher Scientific) was added to a solution of 2% BSA and 0.1% Triton X-100 in PBS. 150 l of this mix was added to each coverslip well, and the plate was incubated at room temperature in the dark for 4 h. Orientation of the plate was reversed halfway through the incubation. Then, coverslips were washed with 0.3% Triton X-100 in PBS for 15 min three times and mounted on glass slides with a drop of Vectashield antifade mounting medium with DAPI. Coverslips were sealed to the slide with clear nail polish. Images were taken using a Zeiss LSM 780 confocal microscope paired with Carl Zeiss Zen 2012 SP1 (black edition) (64-bit) software. Acquisition settings were the same for all images. Dimensions were 1024 ϫ 1024, three channels, 8-bit; objective was EC Planneofluar ϫ40/1.30 oil Ph3 M27; pixel dwell was 1.58 microsiemens; average was line 8.

Histone deacetylation assay
The ac-Myc-MLH1 and the F-HDAC6 full-length proteins were purified from 293T cells. Briefly, Myc-MLH1 was transfected into 293T cell for 2 days and treated with TSA overnight prior to harvest. Purified ac-Myc-MLH1 on the Myc beads was incubated with either F-HDAC6 or buffer control in 50 l of ice-cold HD buffer (20 mM Tris, pH 8.0, 150 mM NaCl, and 10%  Upon 6-TG treatment, 6-thioguanine ( s G) and S 6 -methylthioguanine (S 6 mG) can be inserted into the DNA, so that s G:T and S 6 mG:T mispairs can be formed. These mispairs would be recognized by the MSH2-MSH6 heterodimer (MutS␣), and MutS␣ would then recruit the MLH1-PMS2 heterodimer (MutL␣) and additional components to initiate futile mismatch repair, leading to apoptosis. However, HDAC6 can deacetylate MLH1 and prevent MutL␣ from being recruited to MutS␣, leading to 6-TG tolerance.

HDAC6 deacetylates MLH1
glycerol) at 37°C for 2 h. The reaction was terminated by adding the SDS-PAGE loading buffer.

Nuclear/cytoplasmic fraction preparation
The adherent cells in 60-mm plates were washed with cold PBS, and cells were scraped in PBS. Then cells were suspended in 0.4 ml of lysis buffer (10 mM Hepes, pH 7.9, 10 mM KCl, 0.1 mM EDTA, and 0.1 mM EGTA with protease inhibitors) and left on ice for 15 min. Then, 12.5 l of 10% Nonidet P-40 was added, and the mixture was vortexed for 10 s. The mixture was spun down for 1 min at 14,000 rpm at 4°C. The supernatant was kept, which is the cytoplasmic fraction. Then 40 l of extraction buffer (20 mM Hepes, pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA with protease inhibitors) was added. The sample was left on ice for 30 min, and it was vortexed for 3-5 s every 5 min. Then it was spun for 5 min at full speed (14,000 rpm) at 4°C. The resultant supernatant was the nuclear fraction.