Mismatch repair and treatment resistance in ovarian cancer

Ovarian cancer is the leading cause of death from gynecological cancer in the Western world [1]. The treatment of ovarian adenocarcinoma has improved over the last 20 years owing to the combined treatment of cytoreductive surgery and chemotherapy [2]. Although the response of the primary tumor to taxane and platinum-based chemotherapy is high, about 20% of patients never achieve a clinical response and the majority of the patients will relapse and eventually die of drug-resistant disease [3].

If it would be possible to predict primary platinum resistance, patients might be spared an ineffective but toxic platinum-containing therapy and might benefit from an early therapy with different drugs. Recently, several molecular profiling studies, including our study, have revealed gene sets that can predict response to platinum-based chemotherapy in ovarian cancer [4‐6]. We discovered a nine-gene set which predicts response with a sensitivity of 89% and a specificity of 59% [5]. One of these nine genes was proliferating cell nuclear antigen (PCNA). PCNA is a DNA sliding clamp that interacts with several proteins involved in cell cycle control, DNA methylation, DNA replication and DNA repair including mismatch repair [7]. In this study, we have focused on DNA mismatch repair and its role in platinum-based chemotherapy resistance in ovarian cancer.

DNA mismatch repair (MMR) is divided into three steps: initiation, excision and resynthesis (Figure 1). Several proteins are involved in the initiation of MMR including the three MutS-homologs, MSH2, MSH3 and MSH6. The MutS homologs form a heterodimer that recognizes DNA damage; the MSH2 and MSH6 dimer (the hMutSα complex) recognizes base-base mismatches and single base loops whereas the MSH2 and MSH3 dimer (hMutSβ complex) recognizes insertion/deletion loops of more then one base. After the recognition of the DNA damage the binding of a heterodimer of the MutS-homologs MLH1 and PMS2 (the hMutLα complex) leads to the further initiation of MMR. Other known and still unknown proteins involved in the last two steps of MMR, the excision of the damaged strand and the resynthesis, are recruited subsequently. Proteins known to be involved are exonuclease ExoI, proliferating cell nuclear antigen (PCNA), DNA polymerase δ and perhaps ε and in addition based on its association with DNA polymerase δ and PCNA, DNA ligase I [8, 9].

Inactivation of MMR leads to the occurrence of unrepaired deletions in mono- and dinucleotide repeats resulting in variable lengths of these repeats. This is called microsatellite instability (MSI) and MSI is therefore used as a marker for MMR deficiency. MSI can be caused by genetic or epigenetic inactivation of several genes involved in MMR. Mouse knockout models have demonstrated that MSH2^-/-, MSH3^-/-, MLH1^-/- and PMS2^-/- leads to a high frequency of MSI while MSH6^-/- and PMS1^-/- cause a low frequency (reviewed by Wei et al. [10]). However, in hereditary nonpolyposis colon cancer (HNPCC) families (which are known to have a high frequency of MSI) germline mutations in MSH2 and MLH1 are responsible for the MSI, while MSH6 and PMS2 are less frequently involved [9]. The lesser role for MSH6, PMS2, PMS1 and especially MSH3 inactivation in MSI seen in HNPCC patients could be due to functional redundancy of these genes.

Interestingly, a number of in vitro studies have suggested a relationship between MMR deficiency and platinum-drug resistance. Several resistant sublines of ovarian and melanoma cell lines generated by cisplatin selection, appeared to be MMR deficient [11‐15]. In addition, a colon (HCT116) and an endometrioid cell line (HEC59), deficient for MLH1 and MSH2 respectively, were 2.1 and 1.8 fold resistant to cisplatin compared to cell lines complemented with chromosome 3, containing MLH1, or chromosome 2, containing MSH2 [15‐17]. These in vitro studies suggest that inactivation of proteins involved in the initiation of MMR might cause cisplatin resistance. It is thought that the DNA damage caused by platin-drugs is recognized by MMR. The cell will then undergo several unsuccessful repair cycles, finally resulting in the induction of apoptosis. When MMR is inactive the DNA damage caused by platin-drugs will not be picked up and will therefore not result in apoptosis rendering the cells resistant to platin-drugs.

Several studies have determined the frequency of MMR inactivation in ovarian cancer using MSI as a marker [18‐37]. However, there was a wide range observed (0%–39%) and so far only a few studies have linked MMR inactivation to platinum-based chemotherapy resistance. Thus, there is still no general agreement about the frequency of MMR inactivation and its possible involvement in the platinum-based chemotherapy resistance seen in ovarian cancer patients.

The aim of this study is to determine the frequency of mismatch repair (MMR) inactivation in ovarian cancer and whether it is associated with platinum-based chemotherapy resistance. To this purpose we analyzed seventy-five ovarian carcinomas and eight ovarian cancer cell lines. In the discussion, the results were compared to that of twenty similar studies in the literature including in total 1315 ovarian cancer patients.

In this study we aimed to address two questions; 1) what is the frequency of MMR inactivation in ovarian cancer, and 2) is it associated with platinum-based chemotherapy response.

First we analyzed eight ovarian cancer cell lines, i.e. SKOV6, HOC7, SKOV3, 2774, OVCAR3, KB3.1, CAOV3 and A2780. Microsatellite instability (MSI), which is a marker for MMR inactivation, was detected in three out of eight cell lines i.e. SKOV3, 2774 and A2780. This results in a frequency of MMR inactivation in ovarian cancer cell lines of 38%. The MSI in SKOV3 can be explained by the loss of MLH1 mRNA expression which, however, was not caused by promoter methylation. This is in agreement with the loss of MLH1 protein expression seen in SKOV3 described in a study of the 60 NCI cancer cell lines [44]. In concordance with our findings, 2774 was also described to be MSI [45]. One of the MSI positive A2780 sublines showed a strong methylation of the MLH1 promoter without MLH1 mRNA expression, while the other subline showed a low level of methylation and relative high mRNA expression. Strathdee et al. described that one MLH1 allele was methylated in A2780 [12] which is comparable with the methylation status we saw in A2780, moreover one of our A2780 sublines showed complete methylation. On the other hand, another study did not detect MSI in A2780 [11]. Interestingly, Aquilina and colleagues suggested there is a subpopulation of A2780 cells, estimated to be around one per 10⁶ cells [46], which are MLH1 deficient and heterozygous for the p53phe172 mutation [46, 47]. Since these cells have a growth advantage, prolonged culturing of the A2780 cell line can result in selection of this subpopulation. Thus over time, separately cultured A2780 can have varying percentages of cells belonging to this subpopulation which may explain the discrepancies in MMR status seen in the A2780 cell lines analyzed by us.

Next we studied the association between MMR inactivation and cisplatin resistance in these cell lines. MMR inactivation seen in SKOV3 and 2774 might result in the relative resistance to cisplatin compared to the other cell lines. On the other hand, A2780 which has clearly an inactive MMR, was most sensitive to cisplatin. Overall, there seems to be no association between the response to cisplatin and MMR status in these eight cell lines. This is similar to a study in the 60 NCI cell lines which also showed no association between response to cisplatin and MMR status based on the MLH1 and/or MSH2 protein expression [44].

Since ovarian cancer is a heterogeneous disease characterized by various histological types which may have different MSI frequencies, the number of specimens analyzed is very important in characterizing a feature that may be uncommon such as MSI. We therefore, made a summary of 20 studies totaling 1315 ovarian carcinomas, to compare the findings of these studies with our results (Table 2). The MSI frequencies determined in these studies ranged from 0% to 39%. Overall, MSI was detected in 165 of the 1315 ovarian carcinomas tested, suggesting an overall incidence of 13% [18‐37].

Multiple differences between these studies could have caused the wide range in the MSI frequency (0–39%). One of these is the number and variety of microsatellite markers analyzed to determine the MSI. The NCI recommended five markers comprising the National Cancer Institute Consensus Panel (NCI-CP) for the detection of MSI, i.e. markers for the mononucleotide repeats BAT25 and BAT26 and the dinucleotide repeats D2S123, D5S346 and D17S250 [48]. Table 2 shows per study the number of MS markers used and specifies how many of these are part of the NCI-CP. Interestingly, the studies that used all NCI-CP markers to determine the MS status also showed a wide range in MSI frequency (8–39%) which is similar to the overall range (0–39%). Therefore, the various MS markers used cannot be the sole cause for the wide range. Moreover, Gras et al. suggest that the reliability of the mononucleotide markers BAT25 and BAT26 is so high that most MSI can be predicted by evaluating these two markers exclusively [27], confirming the less stringent role for the various markers used for the analysis.

Another difference between the studies is the distribution of the various histological types of the ovarian carcinoma tissues analyzed (Table 2). This difference in the distribution could be a cause for the wide range in the MSI frequency especially since it has been suggested that certain histological types have a higher frequency of MSI. To determine whether there is a relation between histology and MSI within these studies, we looked at the frequency of MSI per histological type for the 628 patients with known histology (Table 2). The summary of these studies suggests that the frequency of MSI is higher in the mucinous and endometrioid adenocarcinoma compared to clear cell and serous adenocarcinoma (the overall frequencies of MSI were 22%, 16%, 9% and 8%, respectively) (Table 2). We hypothesize that mucinous and endometrioid histology might be prone to a higher MSI frequency since sporadic endometrial carcinoma, which is closely related to endometrioid ovarian cancer, has a MSI frequency of 20–30% [49‐51] and MSI is almost universal present in the colorectal tumors of the hereditary nonpolyposis colon cancer (HNPCC) syndrome which all have a mucinous histological type. Therefore, the different histology's of the ovarian carcinomas included in the several studies seems to be a plausible cause for the wide range in MSI frequency reported in these studies.

Next we addressed the second part of the aim of this study, is MMR inactivation associated with resistance to platinum-based chemotherapy in ovarian cancer. Forty-six of the 75 ovarian carcinomas we analyzed had been treated with platinum-based chemotherapy, eleven did not respond and 34 did. For one patient the response was not known. Methylation of the MLH1 promoter was detected in two of the eleven non-responders (18%) and four of the 34 responders (12%) and this was not significantly different (p = 0.664). Since we did not detect any MSI, the resistance seen in the eleven patients could not be associated with MSI and MMR inactivation.

The relation between MMR deficiency and platinum-drug resistance has been investigated in only a few in vivo studies. Similarly to our result, no MSI was detected by Mesquita et al. [18] who studied 34 ovarian carcinomas of which seven did not respond to cisplatin/paclitaxel therapy. So the resistance seen in these seven nonresponding patients was also not associated with MMR inactivation. In contrast, Samimi et al. [52] found an inverse relation between MLH1 protein expression and the response to platinum-based chemotherapy in 54 ovarian carcinomas. Again, the number of ovarian carcinomas included in these studies is small and no further conclusion can be drawn from these results.

Since platinum-drug resistance is thought to be multifactorial the involvement of other resistance mechanisms could have overruled the possible contribution of MMR status. However, platinum treatment does seem to select for MMR deficient cells since in vitro enrichment for MLH1 deficient colon cancer HCT116 cells in a mixed cell population was seen after cisplatin treatment [53]. In addition, several in vivo studies found an increase in the percentage of MSI and MLH1 methylation after platinum-based chemotherapy as well as a decrease in the percentage of cells positive for MLH1 and MSH2 [14, 19, 25, 52]. Moreover, an increase in MLH1 methylation after platinum-based chemotherapy was associated with poor survival in ovarian cancer patients [19]. These results as well as the in vitro studies mentioned in the introduction, suggest that MMR inactivation causes a low level resistance to platinum-based chemotherapy which does not play a significant role in intrinsic resistance. However, due to selection during chemotherapy MMR inactivation might play a greater role in the acquired resistance. We therefore propose that the role of MMR inactivation in acquired resistance in ovarian cancer should be further investigated.

No MMR inactivation was detected in 75 ovarian carcinoma specimens and no association was seen between MMR inactivation and resistance in the ovarian cancer cell lines as well as the ovarian carcinomas. We hypothesize that MMR inactivation is not clearly associated with intrinsic resistance in ovarian cancer. However, it might play a role in acquired resistance due to selection of MMR deficient cells during platinum-based chemotherapy, but this needs further investigation.