MSH6 Mutations Arise in Glioblastomas during Temozolomide Therapy and Mediate Temozolomide Resistance

Purpose: Over the past few years, the alkylating agent temozolomide has become the standard-of-care therapy for patients with glioblastoma, the most common brain tumor. Recently, large-scale cancer genome sequencing efforts have identified a hypermutation phenotype and inactivating MSH6 mismatch repair gene mutations in recurrent, post-temozolomide glioblastomas, particularly those growing more rapidly during temozolomide treatment. This study aimed to clarify the timing and role of MSH6 mutations in mediating glioblastoma temozolomide resistance.

Experimental Design:MSH6 sequence and microsatellite instability (MSI) status were determined in matched prechemotherapy and postchemotherapy glioblastomas identified by The Cancer Genome Atlas (TCGA) as having posttreatment MSH6 mutations. Temozolomide-resistant lines were derived in vitro through selective growth under temozolomide, and the MSH6 gene was sequenced in resistant clones. The role of MSH6 inactivation in mediating resistance was explored using lentiviral short hairpin RNA knockdown and MSH6 reconstitution.

Results:MSH6 mutations were confirmed in posttreatment TCGA glioblastomas but absent in matched pretreatment tumors. The posttreatment hypermutation phenotype displayed a signature bias toward CpC transitions and was not associated with MSI. In vitro modeling through exposure of an MSH6 wild-type glioblastoma line to temozolomide resulted in resistant clones; one clone showed an MSH6 mutation, Thr¹²¹⁹Ile, that had been independently noted in two treated TCGA glioblastomas. Knockdown of MSH6 in the glioblastoma line U251 increased resistance to temozolomide cytotoxicity and reconstitution restored cytotoxicity in MSH6-null glioma cells.

Conclusions:MSH6 mutations are selected in glioblastomas during temozolomide therapy both in vitro and in vivo and are causally associated with temozolomide resistance.

The standard of care for newly diagnosed glioblastomas involves surgery and external beam radiation therapy (XRT) in conjunction with the alkylating agent temozolomide (1). The benefit from temozolomide is most noted in patients whose tumors have transcriptional silencing of the O⁶-methylguanine methyltransferase (MGMT) gene mediated by promoter methylation (2), which occurs in approximately half of the tumors (3). Nonetheless, the prognosis for patients with glioblastoma remains bleak. Virtually all patients recur after initial therapy, and average survival remains around 12 months (4).

Our knowledge of the genetic changes underlying glioblastoma, although considerable, is largely confined to pretreatment cases (5). Given that all glioblastomas recur and that the recurrent lesions invariably lead to patient death, there is a pressing need to understand the molecular changes that occur during treatment and that characterize the therapeutically resistant recurrences (6). In this regard, we first identified inactivating somatic mutations in the mismatch repair gene MSH6 in two recurrent glioblastomas treated with temozolomide (7). A survey of the genome in these two tumors revealed large numbers of somatic mutations with mutational signatures consistent with those resulting from defects in DNA mismatch repair. Of note, studies in normal and neoplastic cells had shown that inactivation of MSH6 results in resistance to cytotoxicity mediated by alkylating agents (8–12). We therefore proposed that MSH6 inactivation may be one mechanism underlying temozolomide resistance in glioblastomas. In a follow-up study, we examined a larger series of pretreatment and posttreatment glioblastomas for MSH6 mutations and MSH6 expression (13). MSH6 alterations (mutations and/or absent expression) were not found in any pretreatment glioblastomas and not in any posttreatment glioblastomas given only XRT, but were detected in approximately half of the recurrent glioblastomas treated with temozolomide and XRT. Furthermore, temporal measurements of three-dimensional reconstructed magnetic resonance images showed that MSH6-negative glioblastomas showed more rapid radiologic progression while under temozolomide treatment compared with MSH6-positive tumors. These data supported a role for MSH6 inactivation in the emergence of temozolomide resistance in glioblastoma patients.

Two other studies have now reported MSH6 mutation in glioblastomas following alkylating agent chemotherapy. The Cancer Genome Atlas (TCGA) reported an analysis of 91 glioblastomas with matched peripheral blood, of which 19 cases were recurrent glioblastomas that had received alkylating agent chemotherapy (14). In keeping with our findings, the TCGA reported nonsynonymous MSH6 mutations in five of the recurrent glioblastomas (26% of recurrent tumors), all with a hypermutation phenotype consistent with mismatch repair defects. Two other tumors had a hypermutation phenotype, one with a mutation in another mismatch repair gene. In addition, Maxwell et al. showed 7 nonsynonymous and therefore putative MSH6 mutations (2 of them truncating) out of 27 post-temozolomide samples (26%); of note, this series included some malignant gliomas other than glioblastoma, such as oligodendroglial tumors that in our experience to date do not undergo frequent MSH6 alterations (15). Unfortunately, only two cases with sequence variations had available matched pretreatment samples and both of these had only a common MSH6 polymorphism rather than nonsynonymous mutations. Thus, although the TCGA and Maxwell et al. reports confirm that MSH6 alterations are common in those recurrent glioblastomas exposed to alkylating agents, neither had access to matched pretreatment samples to assess the timing of these mutations.

To further pursue this hypothesis, we undertook additional studies to clarify the timing of MSH6 inactivation in the TCGA clinical cases, as well as to model such inactivation in vitro and to directly evaluate the role of MSH6 in temozolomide resistance in vitro. The results of these studies support the hypothesis that MSH6 inactivation occurs during alkylating agent chemotherapy and that, at least in the common setting of MGMT inactivation, MSH6 inactivation is directly related to therapeutic resistance.

TCGA tissue samples and DNA stocks. Unstained slides from formalin-fixed, paraffin-embedded, anonymous, matched prechemotherapy and postchemotherapy glioblastomas were obtained from the M.D. Anderson Cancer Center, Houston, TX. The posttreatment samples were the same ones used by TCGA for genomic and epigenomic profiling and were identified as displaying the hypermutation phenotype and somatic MSH6 mutations (14). Tumor tissue from unstained formalin-fixed, paraffin-embedded glass slides was deparaffinized in xylene followed by immersion in graded alcohols until rehydration, and genomic DNA was extracted using the Gentra PureGene Kit (Qiagen). DNA quantitation was done using a Nanodrop ND-1000 UV-Vis spectrophotometer (Nanodrop Technologies).

Mutation-targeted PCR and sequencing. Targeted PCR or PCR primers were designed to amplify MSH6 mutations initially identified by the TCGA consortium and posted in the publicly accessible database⁶

(14). PCR was done as described previously (13). Targeted sequencing was done using the standard Sanger method in both forward and reverse directions. Each individual sequencing reaction was repeated for confirmation. Analysis of DNA tracings was carried out using Mutation Surveyor version 3.2 (Softgenetics).

Microsatellite instability testing. PCR was conducted in 20 μL volumes using 1× Platinum Taq PCR buffer, 200 mmol/L deoxynucleotide triphosphates, 2.0 mmol/L MgCl₂, 0.4 mmol/L primers, 1.0 unit of Platinum Taq polymerase (all from Invitrogen) with 40 ng of tumor DNA as template. Primer sets comprised the five reference panel markers recommended by the National Cancer Institute, with 5′ phosphoramidite fluorescent labeling of forward primers as follows: BAT-25 (NED), BAT-26 (6-FAM), D5S346 (VIC), D17S250 (6-FAM), and D2S123 (VIC). The primer sequences for D2S123 were 5′-AACATTGCTGGAAGTTCTGG-3′ (forward) and 5′-GTGTCTTGACTTTCCACCTATGGGACTG-3′ (reverse). Primer sequences for the remaining loci were identical to those previously described (16) except that a 5′-GTGTCTT sequence was added to each reverse primer to facilitate nontemplate adenylation of the 3′ end of the forward strand.

PCR was done in a Mastercycler PCR machine (Eppendorf) with an initial denaturing step at 94°C for 5 min; followed by 38 cycles of denaturing at 94°C for 30 s, annealing at either 50°C or 55°C for 30 s, and extension at 72°C for 30 s; and a final elongation step at 72°C for 10 min. PCR products were pooled and fractionated by size using an Applied Biosystems 3130 DNA Analyzer with GeneMapper software. Microsatellite loci at which tumor DNA showed a novel allele profile not present in the corresponding normal DNA were classified as having microsatellite instability (MSI).

Tissue culture. The human glioblastoma cell lines A172 and U251 were originally obtained from the American Tissue Culture Collection. The primary human glioblastoma cell culture Gli60 was established from recurrent tumor xT3162 post-temozolomide and radiotherapy; the tumor had the somatic mutation p.Val⁸⁰⁹X in MSH6, which results in premature termination, and had lost the remaining copy of chromosome 2, leading to null expression of the protein (7, 13). Temozolomide was purchased from Sequoia Research Products Limited (Pangbourne), reconstituted to a stock concentration of 100 mmol/L, and stored at −80°C. O⁶-benzylguanine (O⁶-BG) was obtained from Sigma-Aldrich, reconstituted to stock concentration of 320 mmol/L with DMSO, and stored at −80°C. A172, U251, and Gli60 cells were maintained in DMEM supplemented with 10% FCS, 1% l-glutamine (Invitrogen), and grown at 37°C humidified atmosphere containing 5% CO₂. Cells were confirmed to be free of Mycoplasma using the Lonza MycoAlert detection kit.

Generation of temozolomide-resistant glioblastoma subclones. The glioblastoma cell line A172 was treated with temozolomide at 100 μmol/L or DMSO solvent control at a final concentration of 0.1% for 3 wk. To generate temozolomide-resistant glioma sublines, A172 cells were cultured in six-well plates and allowed to adhere overnight at a 37°C incubator. Control groups were treated with 0.1% DMSO alone. Cell treatment was repeated every 24 h for 5 consecutive days and then exposure to the fresh temozolomide was done every 3 d to a total of 3 wk. Each single clone was grown to derive stable resistant cell lines for subsequent study. MSH6 and MGMT expression was measured by Western blot using the cell lysates treated with temozolomide.

Reconstitution of MSH6 expression in Gli60 using lentiviral approach. The vector backbone for lentiviral reconstitution of Gli60 and the viral packaging procedure have been described previously (17, 18). Briefly, cDNA for MSH6 was cloned into the construct under the control of the cytomegalovirus promoter, and cDNA for the fluorescence marker GFP was under the control of internal ribosomal entry sequence. We used empty constructs expressing only GFP as an infection control. Gli60 cells were infected with lentiviral particles in the presence of protamine sulfate (Sigma) at a multiplicity of infection of 10:1. Successful infection was confirmed by the appearance of green fluorescing cells and Western blot evaluation for MSH6 protein.

Generation of lentiviral constructs for shRNA-mediated MSH6 knockdown. Bacterial plasmid constructs containing shRNA sequence candidates against MSH6 mRNA were obtained from the Massachusetts General Hospital shRNA core (Dr. Toshi Shioda, Massachusetts General Hospital). Subsequent steps in the production of lentiviral particles and infection of glioblastoma cells were done according to protocols listed in the RNA Consortium/Broad Institute of MIT and Harvard (TRC/BROAD) website.⁷

Lentiviral supernatants were removed from U251 cells after overnight incubation and the cells were allowed to grow in DMEM with 10% FCS for 48 h before selection with increasing doses of puromycin from 0.5 to 2.0 μg/mL (Sigma-Aldrich). MSH6 and MGMT protein levels in U251 cells stably infected with the five candidate shRNA constructs were assessed by Western blot analysis.

Cell viability assay. To compare cell viability between parental glioblastoma cell lines and temozolomide-treated, resistant, or MSH6 knockdown cell lines, we used the Promega CellTiter96 Aqueous cell proliferation assay cytotoxicity assay measuring absorbance at 490 nm (Promega). Briefly, 1,000 cells were seeded per well in 96-well tissue culture plates before the drug treatment. Cells were treated daily with fresh temozolomide (100 μmol/L) for 7 d. U251 and U251 Sh1 cells were additionally treated with temozolomide in the presence of O⁶-BG (40 μmol/L). Gli60 cells reconstituted with wild-type MSH6 and with empty control vector were incubated with various concentrations of temozolomide for 5 d and cell growth was evaluated by the MTS assay.

Western blot analysis. Antibodies were obtained from the following sources: MSH6 monoclonal antibody from BD Biosciences, MGMT from Lab Vision Co., and β-actin monoclonal antibody from Santa Cruz Biotechnology. Cells were lysed in radioimmunoprecipitation assay protein extraction buffer (Sigma-Aldrich) together with protease and phosphatase inhibitors at 1:100 dilutions (Sigma-Aldrich) and then centrifuged 10 min at 4°C to harvest the supernatant. The concentration of the extracted protein was determined using the Bio-Rad protein assay kit (Bio-Rad Laboratories). Twenty micrograms per lane of the extracted protein were loaded onto 4% to 12% Tris-glycine gels (Invitrogen) for electrophoresis, and electro-transferred to Pure Nitrocellulose Membrane (Bio-Rad Laboratories). The membrane was blocked with 5% nonfat dry milk in Tris-buffered Saline-Tween buffer [10 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 0.05% Tween 20] for 1 h at room temperature, followed by incubation with different antibodies in the blocking buffer overnight at 4°C. After washing with TBS-T buffer, the membrane was incubated with horseradish peroxidase–conjugated anti-mouse IgG antibody (Promega Co.) at a dilution of 1:2,000 in blocking buffer for 2 h at room temperature. The membrane was developed using the enhanced chemiluminescence system (Perkin-Elmer, Inc.) and exposed to Biomax XAR film (Kodak).

Analysis methods. GraphPad Prism 3.0 was used to determine statistically significant differences between cytotoxicity curves. Best-fit curves for the growth of A172/A172TR3 lines, U251 line, and U251-Sh1 were calculated using the program-derived “exponential growth” nonlinear regression equation with the starting points of all curves held constant at the average absorbance value for all day 1 points. Second-order polynomial nonlinear regression was used for the analysis of survival fractions at the end of the 5-d incubation with temozolomide in Gli60 cells reconstituted with MSH6 and control vector. To determine significance, only the mean Y value was considered for each replicate point and rate constants of the calculated best-fit curves were compared using the t test. Due to multiple testing within each cytotoxicity experiment, statistical significance was defined as P < 0.01.

Genetic analysis of matched TCGA glioblastoma samples: MSH6 mutations, hypermutation phenotype, and MSI status. We previously reported somatic MSH6 mutations in 3 of 11 recurrent glioblastomas treated with alkylating agents and XRT, but in no pretreatment tumors or in posttreatment tumors treated with XRT only, suggesting that MSH6 mutations arise specifically after alkylating agent chemotherapy (13). Given that the TCGA has now found somatic MSH6 mutations in 5 of 19 recurrent glioblastomas, we sought to determine the timing of these genetic changes. We obtained matched pretreatment and posttreatment unstained formalin-fixed, paraffin-embedded tumor tissue sections from four of the five cases identified by TCGA as having somatic MSH6 mutations (TCGA-02-0043, -02-0083, -02-0099, and -02-0114). To show that sequencing from formalin-fixed, paraffin-embedded tissues was of adequate sensitivity and specificity to detect mutations, we sequenced the posttreatment samples and confirmed all five MSH6 mutations reported by TCGA in the four cases (the TCGA-02-0114 posttreatment specimen had two distinct MSH6 somatic mutations). Notably, however, sequencing of the matched pretreatment glioblastomas (including two prechemotherapy specimens from TCGA-02-0114) showed no mutations (Table 1; Fig. 1). The absence of germline and of somatic pretreatment MSH6 mutations strongly suggests that MSH6 mutations do not contribute to the development of glioblastoma. Rather, the presence of MSH6 mutations in the posttreatment samples is consistent with de novo alterations in the tumor cell genome in association with treatment.

The TCGA report noted a hypermutation phenotype in all four of these MSH6-mutant glioblastomas (14). However, the published data did not clarify the sequence context of the mutations beyond classifying them as CpG or non-CpG. We therefore undertook a detailed analysis of the somatic mutation data of these four cases, and confirmed our prior findings of a hypermutation phenotype with a preponderance of C:G>T:A transitions at CpC dinucleotides that is striking when compared with recurrent, post-temozolomide glioblastomas without MSH6 mutations (Table 2; ref. 7). For example, TCGA-02-0083 contains 94 C>T somatic mutations, of which 60 are within the context of CpC dinucleotides. Interestingly, this recurrent tumor also harbors somatic mutations in two other mismatch repair genes, MSH2 and MLH1, which could account for the larger number of such mutations compared with the other hypermutant cases. The other three posttreatment TCGA cases in this series (TCGA-02-0043, -0099, and -0114) also contained markedly higher numbers of somatic mutations and a preponderance of mutations in the context of CpC dinucleotides. On the other hand, these cases did not show high MSI; we did not detect MSI-high (>3/5 unstable loci) in any cases and detected only minor shifts in <2 loci in two cases (TCGA-02-0043 and -02-0083, data not shown). These findings are in keeping with a recent evaluation of the role of MSI as a surrogate marker for MSH6 inactivation in recurrent malignant gliomas; Maxwell et al. (15) showed no correlation between MSH6 mutations and MSI, as assessed by a panel of five mononucleotide loci. That study, however, did not determine whether the recurrent tumors with MSH6 mutations had a hypermutation phenotype. In our study, the overwhelming number of somatic mutations in these tumors, in conjunction with MSH6 mutations, are wholly consistent with MSH6 inactivation causing the hypermutation phenotype, and the absence of MSI is consistent with known MSH6 function (7, 13, 19).

Molecular characterization of in vitro derived temozolomide-resistant glioblastoma cell line. We next sought to model the phenomenon of MSH6 inactivation in vitro. To do so, we exposed the human glioblastoma cell line A172 to temozolomide to generate drug-resistant clones and found reduced MSH6 protein level in a highly resistant line, A172TR3. There was no difference in growth rate between parental A172 and A172TR3; however, the resistant clone showed significantly enhanced survival in the presence of 100 μmol/L temozolomide compared with the parental A172 (P < 0.01; Fig. 2A). We further reasoned that, similar to the situation in human glioblastomas, the development of a temozolomide-resistant clone with reduced MSH6 protein could result from in vitro somatic MSH6 mutation. In fact, sequencing of MSH6 in the resistant line identified a novel MSH6 alteration that was not present in the parental A172 cells. This clone, A172TR3, had a c.3656C>T MSH6 somatic mutation, altering threonine to isoleucine at amino acid position 1219 (Fig. 2B). This same mutation has been identified in two TCGA recurrent glioblastomas with the hypermutation phenotype (TCGA-02-0043 and -02-0099) in the malignant melanoma cell line MZ7-mel derived from a postchemotherapy splenic metastasis (Welcome Trust Sanger Institute, COSMIC database, The Cancer Genome Project⁸

), and as a germline mutation in a colorectal cancer and a case of complex nonatypical endometrial hyperplasia (20). Notably, the MZ7-mel cell line also contains a significant number of somatic mutations consistent with that of the hypermutation phenotype. The Thr¹²¹⁹Ile mutation is therefore likely important to MSH6 function and suggests that the in vitro induced mutation of Thr¹²¹⁹Ile in the temozolomide-resistant A172TR3 cells is biologically and functionally significant.

It is also noteworthy that we were able to derive the MSH6 Thr¹²¹⁹Ile mutation after temozolomide exposure because the same mutation has been reported independently in glioblastomas that failed therapy with other alkylating agents. TCGA-02-0043 was treated with the alkylating agent CCNU (lomustine) and TCGA-02-0099 was treated with the PCV combination therapy (procarbazine, lomustine, and vincristine). In this regard, it is interesting to note that temozolomide is a S_N1-type methylating agent that mediates guanine modification, resulting in base pair mismatch during replication, whereas CCNU causes interstrand cross-linking (12, 21, 22). Moreover, whereas the role of the mismatch repair pathway in facilitating temozolomide-mediated cell death is well understood (12), the role of mismatch repair proteins in mediating CCNU cytotoxicity is less clear. These results may suggest a separable convergence in downstream signaling function from DNA damage recognition, as has been hypothesized (19, 23).

MSH6 inactivation may be heterozygous and is expectedly not associated with high MSI. Some of the recurrent TCGA glioblastomas with a hypermutation phenotype contain heterozygous somatic MSH6 mutations rather than biallelic inactivation. In addition, A172TR3 showed a heterozygous Thr¹²¹⁹Ile MSH6 mutation that was clearly associated with temozolomide resistance in vitro. In addition, patients with colorectal cancers and MSH6 missense mutations often have preserved MSH6 immunoreactivity; in fact, colorectal cancer cells and cells from endometrial complex nonatypical hyperplasia with the germline Thr¹²¹⁹Ile missense mutation have been reported to have preserved MSH6 immunoreactivity (20). It would therefore seem that partial MSH6 inactivations, both in vivo and in vitro, may be associated with functional effects.

MMR proteins function in multimeric complexes; MSH6 dimerizes with MSH2 to form the eukaryotic equivalent of the bacterial mutSα complex (19). The MSH6:MSH2 dimer functions to detect single nucleotide mismatches, and subsequent corrective actions are undertaken by complexes of other MMR family members. In this regard, it has been reported that compromise of MMR function can occur following the mutation of one allele. This may occur through a dominant-negative effect through “soaking up” of the normal binding partner—MSH2 in the case of MSH6 (24). Alternatively, inactivation of one copy of MSH6, as seen in the TCGA recurrent tumors and by us, could contribute to compromised MMR function through a “gene-dosage” effect, which could be exaggerated in an environment with strong selective pressures, as in the tumor microenvironment in the presence of temozolomide. Notably, analyses of changes in MSH6 protein tertiary structure secondary to the missense mutations in the four recurrent TCGA glioblastoma show significant alterations of protein folding in functional domains—in the case of Thr¹²¹⁹Ile, a putative protein-protein interaction domain.⁹

These data suggest that the elucidation of the differential effects of homozygous versus heterozygous MSH6 inactivation in glioblastoma therapeutic resistance will be an important subject for further inquiry.

As noted, inactivation of MSH6 is typically associated with the MSI-L phenotype and not with high MSI (25, 26). Moreover, for MSH6, there exists functional dichotomy between mismatch repair (or surveillance) function and apoptotic signaling secondary to cytotoxic agents (24). This might further explain the lack of correlation between missense MSH6 mutations and level of MSI in recurrent glioblastomas. There also exists a substantial body of literature on the role of MSH6 in mediating somatic hypermutation in the generation of antibody gene diversity (27, 28). This is not surprising because MSH6 functions solely in the survey of the genome for single base-pair mismatches and does not participate in their subsequent repair (29). These observations are in keeping with our findings and argue that the role of MSH6 in temozolomide response is not dependent on MSI.

Genetic knockdown of MSH6 in U251 glioblastoma cell line and correlation with temozolomide resistance. To investigate the specific functional role of MSH6 in glioblastomas, particularly in response to temozolomide treatment, we knocked down the expression of wild-type MSH6 protein in the human glioblastoma cell line U251 through a lentiviral-mediated shRNA approach. We obtained clones of shRNA against the human MSH6 gene from the TRC/BROAD consortium and generated lentiviral constructs of the five candidate clones (Table 2, Supplemental Data). Infection of U251 cells by lentiviral constructs expressing the five MSH6 shRNA candidates was followed by puromycin selection of successfully infected cells. We then carried out Western blot analyses of the MSH6 protein to identify the shRNA candidates with most efficiently down-regulated MSH6 protein (Fig. 3A). The shRNA construct 1 (U251-Sh1) showed 90% knockdown of MSH6 compared with 50% in construct 5 (U251-Sh5). We proceeded to examine the response of these cells to temozolomide in in vitro cytotoxicity assays. Both U251-Sh1 and U251-Sh5 are significantly more resistant to temozolomide compared with parental controls when exposed to 100 μmol/L temozolomide for 7 days (Fig. 3A; data for U251-Sh5 not shown; P < 0.01). Nonetheless, U251-Sh1 and U251-Sh5 exhibited similar proliferation indices to parental U251 cells, showing that enhanced temozolomide resistance in the MSH6-knockdown cells is not due to alteration in cell growth kinetics. These results confirm a role for MSH6 in mediating temozolomide cytotoxicity.

Restoration of temozolomide cytotoxicity in Gli60 with reconstitution of MSH6 expression. We had previously derived and characterized the primary glioblastoma cell culture Gli60 from a recurrent glioblastoma in a patient who had recurred quickly during treatment with temozolomide and radiotherapy. Gli60 has the somatic MSH6 mutation delG2425, which results in premature protein termination, and loss of the remaining wild-type MSH6 gene, with resulting null expression of MSH6 on Western blotting; Gli60 also exhibits the characteristic hypermutation phenotype of C→T transitions preferentially at CpC dinucleotides (7, 13). Gli60 thus represents a unique resource for the study of post-temozolomide recurrent glioblastoma. To support our hypothesis that somatic inactivating MSH6 mutations result in a survival advantage to these cells, we restored MSH6 expression in Gli60 and examined the subsequent response to temozolomide. Gli60 cells were infected with a lentiviral construct expressing MSH6 under the control of the cytomegalovirus promoter (Gli60-MSH6) or with an empty control vector (Gli60-Con vector). We then incubated the cells with different concentrations of temozolomide and examined the cytotoxicity using the MTS assay. Gli60-MSH6 showed restored temozolomide sensitivity compared with Gli60-CON (P < 0.01; Fig. 4). This confirms that restoration of MSH6 expression in glioma cells from a patient who failed temozolomide treatment conferred temozolomide sensitivity in vitro, and functionally links MSH6 function to temozolomide sensitivity in a glioblastoma that harbors the characteristic hypermutation phenotype.

Relationship of MSH6 alterations to MGMT status. Temozolomide acts by adding methyl groups to the O⁶ position of guanine nucleotides. The first line of response to repair this chemotherapeutic event is mediated by MGMT, which removes these methyl groups (30). However, a substantial proportion of glioblastomas have transcriptional silencing of the MGMT gene through promoter hypermethylation, and cases with intact MGMT expression can have MGMT down-regulated by the MGMT inhibitor O⁶-BG (31, 32). In such cases, the mismatch repair pathway becomes a primary mediator of O⁶-methylguanine cytotoxicity, and defects in the mismatch repair pathway can therefore serve as an alternate resistance mechanism for cancer cells (12, 33). As a result, to develop resistance after temozolomide exposure, cancer cells that cannot up-regulate MGMT expression would be expected to inactivate this mismatch repair pathway.

In the present experiments, in keeping with what has been found in primary human glioblastomas, both the TCGA cases and the studied cell lines had inactivation of MGMT. All four of the TCGA cases had methylation of the MGMT promoter (14), which has been correlated with decreased expression of MGMT and improved response to temozolomide in initially treated primary tumors (2, 34). The A172 cell line does not express MGMT, nor does the temozolomide-treated resistant clone A172TR3 (data not shown). The U251 cell line does not express MGMT, and MGMT expression is not induced in the knockdown cells (Fig. 3A). Nonetheless, to confirm that MGMT did not play a role in the U251 knockdown experiments, we also added the irreversible MGMT inhibitor O⁶-BG to both parental U251 and the MSH6-knockdown U251-Sh1 and U251-Sh5 clones, and this did not alter drug sensitivity in both the parental and the two MSH6-knockdown clones (P < 0.01; Fig. 3B). Thus, the difference in sensitivity to temozolomide between the parental and the genetically modified clones was due principally to MSH6. At the same time, O⁶-BG is being considered as an adjuvant in temozolomide therapy due to its ability to deplete cellular MGMT (31, 35). In this regard, pharmacologic inhibition of MGMT by O⁶-BG in vivo or escalation of temozolomide dosing through “dose dense” treatment scheduling could result in accelerated selection pressure to develop alternate escape mechanisms to temozolomide-mediated cytotoxicity (33), one of which would be somatic mutations of a mismatch repair gene such as MSH6. Future adjuvant therapies aimed at overcoming MGMT activity could therefore potentially increase the frequency by which we observe alternate escape pathways such as mismatch repair inactivation (36).

In summary, MSH6 mutations are frequent in recurrent glioblastomas that have been treated with alkylating agents, but have not been found in any prechemotherapy glioblastomas. Combining our data (3 of 11, 27%) with those of the TCGA (5 of 19, 26%) and Maxwell et al. (7 of 27, 26%), MSH6 mutations have now been found in 15 of 57 (26%) glioblastomas after alkylating agent chemotherapy, with similar incidences found in the three series (13–15). We have also shown that in vitro derivation of temozolomide-resistant cells can be associated with MSH6 inactivation and mutation. Moreover, in vitro inactivation of wild-type MSH6 protein in glioblastoma cells can result in increased temozolomide resistance, and in vitro reconstitution of MSH6 expression can restore temozolomide sensitivity in glioblastomas lacking MSH6. These multiple approaches support an integral role for MSH6 inactivation in mediating temozolomide resistance in glioblastomas. It is also likely that defects in other mismatch repair proteins could play similar roles: As mentioned above, one case from the TCGA data (TCGA-02-0083) had somatic mutations in two other mismatch repair genes, MSH2 and MLH1, and in another large-scale genome-wide study of glioblastomas, one recurrent tumor post-temozolomide/XRT had the hypermutation phenotype, but MSH6 mutations were not found in this tumor (37). Thus, mismatch repair defects may be a common resistance pathway for treated glioblastomas that have already inactivated MGMT.

No potential conflicts of interest were disclosed.