Haploinsufficiency by minute MutL homolog 1 promoter DNA methylation may represent unique phenotypes of microsatellite instability-gastric carcinogenesis

Promoter DNA methylation of MutL homolog 1 (MLH1) is considered to play a causative role in microsatellite instability (MSI) carcinogenesis in primary gastric cancer, and a high MSI status is associated with treatment sensitivity to human cancers. Nevertheless, clinicopathological analysis is defective for MLH1 methylation status in a quantitative manner. We newly developed quantitative methylation specific PCR using a TaqMan probe and applied it to 138 patients with primary gastric cancer who underwent gastrectomy in addition to basic molecular features such as MSI, Epstein Barr virus, and other DNA methylation status. (1) In primary gastric cancer, median methylation value was 0.055, ranging from 0 to 124.3. First, MLH1 hypermethylation was strongly correlated with MSI-High/MSI-Low status and suppressed immunostaining (P < 0.0001). (2) The MLH1 hypermethylation was associated with advanced age (P = 0.0048), antral location (P = 0.0486), synchronous multiple gastric cancer (P = 0.0001), and differentiated histology (P = 0.028). (3) Log-rank plot analysis identified the most relevant cut-off value (0.23) to reflect gentle phenotypes in MLH1 hypermethylation cases (P = 0.0019), especially in advanced gastric cancer (P = 0.0132), which are designated as haploinsufficiency of MSI (MSI-haplo) phenotype in this study. (4) In synchronous multiple gastric cancer, MLH1 hypermethylation was not necessarily confirmed as field cancerization. (5) MSI-haplo defined by MLH1 methylation status represented distinct prognostic phenotype even after molecular classifications. MLH1 hypermethylation designated as MSI-haplo may represent unique prognostic phenotype during gastric carcinogenesis.


Introduction
Gastric cancer is the fifth most common malignancy and the third leading cause of cancerrelated deaths worldwide [1]. Cancer is a genetic disease, and cancer cell survival is addicted to the specific oncogenes aberrated in tumor cells like HER2 in gastric cancer [2,3]. Therefore, sophisticated molecular understanding is critical for determining therapeutic targets of gastric cancer.
A comprehensive molecular classification of gastric cancer was proposed as Epstein Barr virus (EBV)-associated (8.8%), microsatellite instability (MSI)-associated (21.7%), chromosomal instable (49.8%), and genomic stable (19.0%) gastric cancer [4]. Among these four definite molecular phenotypes, MSI-associated gastric cancer representing defective DNA mismatch repair system is intriguing, because it could be a predictive biomarker for immune checkpoint inhibitors [5][6][7] in addition to classical chemotherapy [8]. It has been reported that the clinicopathological factors in the MSI-associated gastric cancer were elderly, female, intestinal type in Lauren classification, and tumors located in the middle to lower site of the stomach [4]. Moreover, recently, MSI-associated cancers harbored frequent genomic mutations of an SWI/SNF chromatin-remodeling factor such as ARID1A [9,10], and such tumor cells are a good candidate for synthetic lethal target for glutathione metabolism [11].
Nevertheless, genomic mutation of DNA mismatch repair genes has rarely been documented in sporadic gastric carcinoma with MSI-High (MSI-H) [10], and several early papers reported that hypermethylation of the CpG island in the MutL homolog 1 (MLH1) promoter DNA presents with a loss of MLH1 protein expression in almost all cases of MSI-H sporadic gastric carcinomas [10,[12][13][14]. These critical findings suggested that MSI-H is largely caused by MLH1 hypermethylation in sporadic gastric cancer tumor tissues.
Most of the data on MLH1 hypermethylation were evaluated by non-quantitative methylation-specific PCR (MSP). In previous similar reports evaluated by quantitative MSP, the quantified methylation values were not applied to investigate clinicopathological factors. Thus, there have been no reports describing the exact clinicopathological significance of MLH1 methylation status in human cancers in a quantitative assessment. Allowing for the re-emerging importance of the MSI-H phenotype in gastric cancer clinics, it is time that MLH1 methylation status is assessed by a very accurate methylation analysis.
In this paper, we newly developed a DNA methylation quantification system for MLH1 methylation in quantitative MSP, and for the first time clarified the detailed clinicopathological features of MLH1 methylation in primary gastric cancer.

Cell lines
The colorectal cancer (CRC) cell lines SW48 and SW480 were used for positive and negative control for the MLH1 methylation, respectively. SW48 and SW480 cells were purchased from the American Type Culture Collection (Manassas, VA, USA). SW48 and SW480 cells were grown in RPMI 1640 medium (GIBCO, Carlsbad, CA) containing 10% fetal bovine serum. The six gastric cancer cell lines (Kato III, KE-97, MKN74, NUGC-4, SH-10-TC, and MKN7) were recently used similarly and described [15].

Patients and tissue samples
This study investigated 138 patients who underwent surgical resection for primary gastric cancer at the Kitasato university hospital, Japan in 2005 as described [15]. We extracted DNA from 138 tumor tissues. These tissue samples were collected from all patients who provided written informed consent for the use of their pathological specimens. The present study was approved by the Ethics Committee of Kitasato University (Number B18-058). Patient characteristics are shown in S1 Table. Patients with neoadjuvant chemotherapy were not included in this study.

DNA extraction and sodium bisulfite conversion
Tissue sections from primary tumors were stained with hematoxylin and eosin, and dissected under a microscope. Genomic DNA was extracted from formalin-fixed paraffin embedded (FFPE) tissues using a QIAamp DNA FFPE Tissue Kit (QIAGEN Sciences, Hilden, Germany). Bisulfite treatment was done according to the manufacturer's instructions of an EZ DNA Methylation-GoldTM Kit (Zymo Research, Orange, CA) as described [15].

Direct sequence and clones' sequence for methylation positive control SW48 cells and negative control SW480 cells
We originally designed bisulfite sequence primers in this study as shown in Fig 1A. The primer sequences used in this manuscript are also included in S2 Table. The PCR products were separated on 1.5-2.0% agarose gel, then visualized them by ethidium bromide staining prior to direct sequence. Distilled water was used as negative control. Cloned sequence was done for 10 clones each in SW48 and SW480 cells, respectively as previously described [16].
qMSP qMSP was carried out using iQ Supermix (Bio-Rad Laboratories, Hercules, CA) in triplicate on the C1000 TouchTM Thermal Cycler CFX96 Real Time System (Bio-Rad) as described [15]. Bisulfite-treated DNA was amplified by the following PCR conditions: 95˚C for 10 min, followed by 40 cycles at 95˚C for 15 sec, annealing temperature (60˚C) for 1 min.
The sequence primers and probes for quantitative assessment were designed with early references regarding conventional methylation-specific PCR [17,18] (Fig 1A) according to a primer design software (MethPrimer in the National Center for Biotechnology Information-NCBI home page). Our unique combination of primers and probe for qMSP is the most similar with that belonging to Shibata [19], which did not investigate clinicopathological information.
Serial dilutions of bisulfite modified DNA from the human colon carcinoma cell line SW48 was used to construct the calibration curve on each plate as a methylation positive control, and the human colon carcinoma cell line SW480 was used as a negative control. The methylation value was defined as a TaqMeth value by the ratio of the amplified signal value of methylated MLH1 to the value for β-actin, which was then multiplied by 100.

PCR amplification of EBV DNA and in situ hybridization (ISH)
For EBV infection, EBV nuclear antigen 1 (EBNA1) DNA was used in this study. Primers and probes of EBNA1 were synthesized as published sequences [20]. Real-time PCR for both EBNA1 and β-actin were performed using iQ Supermix (Bio-Rad Laboratories, Hercules, CA) in triplicate on the C1000 Touch TM Thermal Cycler CFX96 Real Time System (Bio-Rad Laboratories, Hercules, CA). The thermocycling conditions were as follows: 50˚C for 2 min, 95˚C for 10 min, 95˚C for 15 sec and 60˚C for 1 min for 40 cycles. The relative amounts of EBNA1 DNA compared to those of β-actin was calculated as follows: Ct (EBNA1)-Ct (β-actin) for each sample. Relative copy number was determined as 2-ΔΔCt, where ΔΔCt = ΔCt (tumor)-ΔCt (corresponding normal) [21]. An increase in gene copy number was considered when 2-ΔΔCt was 1.0 or more.
EBV encoded small RNA (EBER)-ISH were performed using the Ventana BenchMark XT automatic immunostainer following the manufacturer's recommendations (Ventana Medical Systems, Tucson, AZ, USA) in the recent 2 cases with EBV positive as a routine test, and also assessed DNA copy numbers. Protein removal and nucleic acid exposures used ISH protease 1. The EBER probe of Ventana DNP labeled probes was then applied and hybridized. To detect signals, antibodies (CS1, CS2, CS3, CS4) were employed. The reaction was done with ISH iVIEW BLUE-Research Kit.

MSI procedures
MSI analysis was done by evaluating two mononucleotides repeat markers (BAT25 and BAT26) and three dinucleotide repeat markers (D2S123, D5S346, and D17S250) as recommended by the National Cancer Institute workshop for MSI [22]. Each marker was PCRamplified in a separate 10 μL reaction containing 5 μL Type-it Microsatellite PCR kit (QIA-GEN Sciences, Hilden, Germany), 10 pmol each of labeled forward and unlabeled reverse primers, and 10 ng of template DNA. D2S123 and BAT26 were labeled with 6-FAM, BAT25 with VIC, D5S346 with NED, and D17S250 with PET. PCR was done in a Thermal Cycler GeneAtlas S (ASTEC, Fukuoka, Japan).
The reaction conditions were as follows: initial denaturation at 95˚C for 5 min, 28 cycles of (95˚C for 30 sec, 57˚C for 90 sec, 72˚C for 30 sec), followed by a final extension at 60˚C for 30 min. Approximately 1 to 2 ng of each PCR product were mixed with a 1:40 dilution of Hi-Di Formamide-GeneScan LIZ500 Size Standard and visualized on a Genetic Analyzer 3500xL (Applied Biosystems, Foster City, CA, USA). Allelic size alterations were detected using Gene-Mapper Software Version 4.1 (Applied Biosystems, Foster City, CA, USA).
Samples were considered positive for MSI when alternate-sized bands were present in the tumor DNA but absent in the respective control mucosal DNA. Tumors were classified as MSI-H if �2 markers had allelic shifts, low level MSI (MSI-L) if only one of the five markers had allelic shift, and microsatellite-stable (MS-stable) when no marker showed allelic shift [22].

Statistical analysis
Continuous variables were evaluated by ANOVA, Student's t test; categorical variables by Fisher's exact test or the Chi-square test, as appropriate. Clinicopathological characteristics and follow up data were evaluated in terms of overall survival (OS). The follow-up time was calculated from the date of surgery to death or end-point. OS was estimated by the Kaplan-Meier method, and compared using the log-rank test. Variables suggesting potential prognostic factors on univariate analyses were subjected to a multivariate analysis using a Cox proportional-hazards model. A P-value <0.05 was considered to indicate statistical significance. All statistical analyses were conducted using the SAS software package (JMP Pro11, SAS Institute, Cary, NC).

MLH1 promoter DNA hypermethylation in control cell lines
We initially used CRC cell lines, SW48 and SW480 cells, as positive and negative controls for methylation assessment as previously shown in conventional MSP [17]. As expected, SW48 cells with MLH1 hypermethylation were accompanied by its silenced expression, while SW480 cells with no MLH1 hypermethylation exhibited its definite expression ( Fig 1B). RT-PCR proved to be clearly expressed for MLH1 in all six gastric cancer cell lines ( Fig 1B, upper  panel), indicating that no gastric cancer cell lines harbors MLH1 hypermethylation.
The degree of MLH1 promoter hypermethylation by cloned sequences has not been determined anywhere even in the control cells. Using bisulfite treatment of DNA followed by direct sequence (S1 Fig) and cloned sequence analysis, we clarified the methylation level of the MLH1 promoter sequence in SW48 cells (methylation positive control) and SW480 cells (methylation negative control) (Fig 1C). Consequently, we found methylated CpG sites in 99.3% and 0.6% of the promoter DNA in SW48 and SW480 cells, respectively (Fig 1C). These findings suggested that SW48 cells and SW480 cells could be designated as positive and negative controls in our quantitative assessment, respectively.
We subsequently quantified methylated MLH1 after bisulfite DNA treatment using qMSP with primers and probe originally designed for this current study (Fig 1A). The efficacy of PCR amplification of methylated MLH1 was excellent and found to be comparable with that of β-actin ( Fig 1D). MLH1 hypermethylation could be clearly detected for DNA from SW48 cells at concentrations of 1-fold (200 ng/well), 10-fold (20 ng/well), 100-fold (2 ng/well) dilutions, but could not be detected for those of 1000-fold (0.2 ng/well), and 10000-fold (0.02 ng/well) dilutions, whereas no methylation at all could be detected in negative control SW480 DNA templates (200 ng, 20 ng, 2 ng, 0.2 ng, 0.02 ng/well) ( Fig 1D). We used fluorescence of 350 RFU (relative fluorescence unit) as the threshold line in this assay, because the negative control is always below this threshold. These findings suggested that even dense MLH1 methylation such as that found in SW48 cells can barely be detected using qMSP when the DNA is diluted beyond 1000-fold.
Using this optimal condition, MLH1 methylation was never seen in six gastric cancer cell lines, which were abundantly expressed for MLH1 mRNA (Fig 1B, lower panel). Hence, promoter DNA methylation status is completely consistent with mRNA expression status of the MLH1 in the six gastric cancer cells as well as the two CRC cell lines.

MLH1 promotor DNA methylation and its correlation with clinicopathological factors in primary gastric cancer
On hundred and thirty-eight primary tumor specimens of patients with gastric cancer who underwent surgical resection were assessed by qMSP to evaluate the clinical relevance of the MLH1 methylation status. The median TaqMeth value was 0.055, ranging from 0 to 124.3 in the 138 tumor tissues (T) (Fig 2A).
The correlation of the MLH1 TaqMeth value to clinicopathological factors was initially evaluated by ANOVA in primary gastric cancer. The negative prognostic factors (e.g., staging factors) showed no significant association with MLH1 TaqMeth value by ANOVA (Fig 2B). Conversely, there was significant correlation of a high MLH1 TaqMeth value to synchronous gastric cancer (P = 0.0001, Fig 2C), advanced age (P = 0.0048, Fig 2D), histological type (P = 0.028, Fig 2E), and antral location of the tumor (P = 0.049) ( Fig 2F).
As the MLH1 TaqMeth value was the most strongly associated with synchronous gastric cancer, we then investigated synchronous other tumor tissues together with the corresponding non-cancerous gastric mucosa tissues (Fig 2G). Surprisingly, field cancerization of the double cancer representing MLH1 hypermethylation of the corresponding non-cancerous tissues was unexpectedly limited to only a few cases (6/17 cases, 35%). Threshold was not set to discriminate the corresponding normal mucosa from tumor tissues.

Multivariate prognostic analysis in primary gastric cancer
We further investigated the correlation of the MLH1 TaqMeth value to prognosis (OS) in primary gastric cancer. To determine the optimal cut-off values for predicting poor prognosis, we assessed each p-value and relative risk by the log rank plot analysis as previously described PLOS ONE [15]. The most optimal cut-off value was defined as 0.23 with the highest relative risk in primary gastric cancer (Fig 3A). Intriguingly, receiver operating characteristics curve to predict deaths also recapitulated the same optimal cut-off value of 0.23 (Fig 3B, area under curve = 0.67). These findings suggested that this cut-off value of MLH1 methylation has the most optimally prognostic relevance in primary gastric cancer.
Univariate prognostic analysis showed that age (P = 0.0064), morphological types (P < 0.0001), pathological stage (P < 0.0001), postoperative complication (P = 0.0034), tumor resectability (P < 0.0001), and the MLH1 TaqMeth value were significantly associated with prognosis in primary gastric cancer (S3 Table). The identified univariate prognostic factors were subjected to a multivariate Cox hazards model. Multivariate analysis only identified pathological stage as an independent prognostic factor in gastric cancer, and the MLH1 TaqMeth value was not finally remnant (S3 Table).

Promoter DNA methylation of MLH1 critically affects its protein expression in tumor tissues
MLH1 protein expression was immune-stained in the 10 gastric cancer tumor tissues with the highest and the lowest values of the methylation using anti-MLH1 polyclonal antibody ( Fig  4A). A strong expression of MLH1 protein was observed in tumor tissues with MLH1 hypomethylation, whereas a weak expression was dominant in those with MLH1 hypermethylation. The differential expression of the MLH1 protein categorized into 3 groups (0, 1+, 2+) showed significantly different MLH1 TaqMeth values (P < 0.0001) (Fig 4B), suggesting that MLH1 protein expression is significantly correlated with the MLH1 hypermethylation.

MLH1 hypermethylation and MSI-associated gastric cancer in the context of molecular features
We then compared the MLH1 TaqMeth values with MSI status. In the 136 gastric cancer samples, MSI-H was confirmed in 12 (8.8%), MSI-L in 5 (3.7%), and MSS in 119 (87.5%) (Fig 5A).
Recent molecular classification of gastric cancer is categorized into EBV-associated, MSIassociated, and other gastric cancers [4]. In this study, we initially detected EBV-associated gastric cancer by simple PCR assessment (Fig 6A, left panel) in gastric cancer cases which had been confirmed positive by ISH in the recent routine tests (representative ISH shown in Fig  6B). We also applied quantitative PCR to 138 cases (right panel of Fig 6A) as well as the 2 recent cases (Fig 6B) (designated as simple EBV positive, sEBV in this study). EBV positive gastric cancers were mutually exclusive from MSI-H/L as previously shown [4], and almost not redundant with MLH1 methylation status (2 cases redundant in MSI-haplo) (Fig 6A, right  panel).
This simple molecular classification revealed that the best prognosis was again shown in MSI-haplo (n = 59) in contrast to sEBV (n = 7), and other gastric cancers (n = 72) in Fig 6C  (P = 0.0167), and MSI-haplo showed 83.9% OS in gastric cancer.
We recently demonstrated that Cysteine dioxygenase type 1 (CDO1) hypermethylation was cancer-specific in primary gastric cancer and its hypermethylation was correlated with poor prognosis in the same patient cohort [15]. As the CpG islands methylator phenotype (CIMP) including MLH1 hypermethylation has been repeatedly proposed to be enriched in the specific patients [23], we finally compared MLH1 methylation and CDO1 methylation in gastric cancer. MLH1 hypermethylation was significantly associated with CDO1 hypermethylation (R = 0.433, P < 0.0001) (S3 Fig), and MSI status was also significantly correlated with CDO1 hypermethylation like MLH1 hypermethylation (Fig 4C), suggesting that MSI carcinogenesis may be associated with pan-epigenetic alterations. Combination of MLH1 and CDO1

PLOS ONE
methylation status has important prognostic information in gastric cancer (Fig 4D), suggesting that prognosis is differentially affected by oncological function of the individual genes.

Discussion
MLH1 hypermethylation has been repeatedly examined so far by conventional MSP; however, quantitative analysis related to clinicopathological relevance is defective. The classical conventional MSP was judged by visual inspection, in which MLH1 hypermethylation was recognized in 20-30% in gastric cancer and consistent with MSI-H [13,24]. If the cut-off of the MLH1 TaqMeth value was determined as the highest top 20% in our current study, it was corresponding to 3.5, which is close to that (2.3) to identify MSI-H/L phenotype (Fig 5D).
TaqMeth values of 0-5 in qMSP could be corresponding to visual detection by the inspection study [16], so our current data representing 2.3 of MLH1 TaqMeth value may be consistent with the numerous previous literatures regarding MLH1 methylation frequencies by the conventional MSP [13,24].
Conversely, our current assessment clarified the best optimal cut-off value of 38.55 to identify MSI-H (Fig 5E), and such methylation level is consistent with silenced expression of MLH1 protein (Fig 4A). These findings, for the first time, proposed that MLH1 methylation degree can affect MSI phenotype (MSI-H/MSI-L) in a stepwise manner. Our findings can also scientifically support the early report that MSI-H can develop from MSI-L or absence of MSI due to accumulation of DNA methylation during progression of early-stage gastric cancer [25].
From prognostic point of view, conversely, the best optimal cut-off value of the TaqMeth value of MLH1 was 0.23 (designated as MSI-haplo-the top 45.6% in this study, Fig 5F), proposing a potential entity of MSI-haplo phenotype that shows good prognosis in our current study (Fig 3). This prognostic outcome is consistent with early reports exhibiting good prognosis in gastric cancer with conventional MSI-H or high MLH1 methylation [26]. We are considering that haplo-MSI phenotype could be microsatellite-instable despite MSS defined by the current definition. The definition of MSI-H/L is based on the specific relevant markers (Bethesda panel), and they could not be covered with minor abnormalities of microsatellites caused by haplo-MSI that is corresponding to the lowest hypermethylation of MLH1 (> 0.23) (Fig 5F). Prognostic stratification defined by the most optimal cut-off value (0.23, but not 2.3) might propose bona-fide entity of MSI phenotypes including haplo-MSI in gastric cancer. Hence, MSI-haplo phenotype determined by MLH1 methylation in qMSP is still considered to sustain unique phenotypes of the conventional MSI from a prognostic point of view. We examined the methylation of MLH1 in each MSI phenotype (MSI-H, MSI-L, and MSS) using a direct sequencing. It was considered that the sensitivity to detect methylation of direct sequencing is limited to MLH1 hypermethylation cases, and MSI-haplo can only be detected by real-time PCR (S4 Fig). Our data also showed that MSI-haplo was significantly associated with differentiated histology and antral location in a multivariate analysis (Table 1), and such phenotypes have been considered to be unique for conventional MSI phenotypes [4,[27][28][29][30]. Hence, MSI-haplo may be the newly emerging entity representing MSI carcinogenesis.
Among the clinicopathological features in our current study, MLH1 hypermethylation was the most strongly associated with synchronous gastric cancer. This finding proposed MLH1 involvement in field cancerization of gastric carcinogenesis described elsewhere [31]. In our data, however, MLH1 was not necessarily hypermethylated in synchronous other gastric cancer, and/or the corresponding non-cancerous mucosal tissues (Fig 2F). Our current findings suggest that MLH1 hypermethylation may be included among other epigenetic alterations among pan-epigenetic genes during gastric carcinogenesis.
We herein hypothesized field cancerization by multiple concurrent epigenetic alterations. Gastrointestinal cancer with the CIMP was proposed, which have included MLH1 hypermethylation [32]. In fact, MLH1 hypermethylation occurs predominantly in the setting of other gene hypermethylation such as HPP1 reported from other group [19] and CDO1 in the current study.
Conversely, CDO1 hypermethylation was proven to be closely associated with MLH1 hypermethylation (R = 0.433, P < 0.0001), and CDO1 hypermethylation was also recognized specifically in MSH-H/L phenotypes (P = 0.0005, Fig 4C). These findings suggest that MLH1 mediated MSI carcinogenesis is a part of pan-epigenetic carcinogenesis. We cannot explain field cancerization by MLH1 hypermethylation alone.
MSI-H tumors have been reported to have marked infiltration of CD8 T cells [33,34], suggesting that cancer immunity is highly activated. Tumor cells attacked by cancer immunity may be exposed to high regenerative oxygen species [35,36], and the redox system including NF-E2-related factor 2 (NRF2) activation is required for cancer cell survival [37]. Tumor growth requires NRF2 without CDO1 expression [38], so CDO1 hypermethylation with its gene silencing was recognized in MSI-H tumors.
Conversely, CDO1 hypermethylation is a prognostic marker of poor prognosis [15], while MLH1 hypermethylation is a marker of good prognosis. In the current study, we clarified that CDO1 hypermethylation with MLH1 hypomethylation showed the worst prognosis in gastric cancer and vice versa (Fig 4D). The differential roles of the two discrete epigenetic genes in prognosis suggest that both genes do not necessarily share the common functional role in gastric carcinogenesis, although epigenetic mechanism is shared.
Smyth et al reported that surgery alone for resectable gastroesophageal cancer had the best prognosis of MSI-H or Mismatch repair (MMR) deficiency (including MLH1 methylation) groups compared to MSI-L or MSS groups [39]. On the other hand, the prognosis of MSI-H or MMR deficiency group with chemotherapy plus surgery was poor. Therefore, MSI-H or MMR deficiency groups may be resistance to chemotherapy. An et al also found that methylated MLH1 tended to have a better prognosis than unmethylated (MST: 68.4 vs 28.4 months, P = 0.49), and that CIMP-high due to multiple methylated genes including MLH1 was CIMPlow. They reported that the prognosis was significantly better than that of CIMP-negative (P = 0.04) [40].
We also investigated EBV-associated gastric cancer in this study. EBV-associated gastric cancer was recognized in 7 of 138 (5.1%). This frequency is less than that of The Cancer Genome Atlas data (8.8%) [4], but recapitulates the recent large-scale report of EBV-associated gastric cancer (4.6%) [41], where EBV-associated gastric cancer was mutually exclusive of MSI-associated gastric cancer. Even in the current study, EBV-associated gastric cancer was exclusive of MSI-H/MSH-L. However, EBV-associated gastric cancer showed MSI-haplo in 2 cases (case 48 and case 103 of Fig 1A, right panel) among the 7 positive cases. EBV-associated gastric cancer almost inevitably harbored PI3 kinase mutation and p16 gene silencing [4,41], which may be due to epigenetic carcinogenesis and therefore not completely exclusive of MLH1 methylation in the previous literature [42].

Conclusions
In this study we, for the first time, present critical data of clinicopathological relevance of MLH1 hypermethylation assessed in qMSP in primary gastric cancer. We found that minute hypermethylation of MLH1 may result in haploinsufficiency of MLH1 expression, which represents unique phenotypes of MSI-gastric carcinogenesis. The frequency of MSI-haplo is much higher than that reported previously by the conventional MSI. As it is clinically distinct PLOS ONE from otherwise gastric cancer, keeping MSI-phenotype, this discovery may be beneficial for the development of therapeutic strategy for gastric cancer according to molecular classification.