Introduction

Gastric cancer (GC) is one of the most common malignancies and the leading cause of cancer mortality worldwide [1]. It is a heterogeneous disease with diverse molecular phenotypes. The heterogeneity is largely due to genome-wide epigenetic alterations in the malignancy [2], such as aberrant DNA methylation, histone modification, noncoding RNA expression, and RNA editing. Among the alternations, hypermethylation of DNA has been found to gradually increase with progression from normal to premalignant lesions and from premalignant lesions to gastric carcinoma [3].

CpG island methylator phenotype (CIMP) refers to a molecular phenotype which is characterized by genome-wide hypermethylation of promoter CpG island loci and is found in Epstein–Barr virus (EBV)-positive and sporadic microsatellite instability (MSI)-high GCs [4]. Common features of GCs with high levels of CIMP include mutated oncogenes, such as CTNNB1, ERBB2, KRAS, PIK3CA, and ARID1A, high PD-L1/PD-1 expression in the tumor microenvironment, and high CD8 + intratumoral T cells [5].

Sporadic MSI-high GCs attribute their MSI to hypermethylation of the promoter CpG island locus and the associated inactivation of the MLH1 gene [6]. According to previous studies, EBV-positive GCs have the highest prevalence of DNA hypermethylation among all human malignancies, and the vast majority of CpG island loci that are methylated in sporadic MSI-high GCs are also methylated in EBV-positive GCs [4, 7, 8]. Nevertheless, EBV-positive GCs do not always accommodate increased methylation levels of promoter CpG island regions that are increased in MSI-high tumors [4]. Moreover, EBV-positive and MSI-high GCs are considered as mutually exclusive [9, 10]. These findings show that EBV-positive GCs have a unique pathway of carcinogenesis that is distinct from MSI-high GCs [8, 11, 12]. However, the underlying cellular mechanism that allows this ramification is yet to be reported.

Herein, we aimed to identify the putative cause of the mutual exclusivity between EBV-positive and MSI-high GCs by establishing a human GC cell line that is reminiscent of simultaneous EBV infection and MLH1 silenced status. Cellular behavior of EBV-positive GC cell lines with and without MLH1 silencing was compared in vitro and in vivo. This study implies the underlying mechanism of EBV-positive/MSI-high GCs not being observed during natural progression of gastric tumorigenesis.

Methods

Patients

Four hundred and ninety-two advanced GC samples were selected from Seoul National University Hospital as formalin-fixed paraffin-embedded (FFPE) blocks. All samples were obtained from patients who had undergone surgical resection between January 2007 and December 2008. This study was approved by the Institutional Review Board of Seoul National University Hospital, which waived the requirements to obtain informed patient consent (approval no. H-1312-051-542). The clinicopathological features of patients are summarized in Supplementary Table 1.

Cell lines and culture condition

The EBV-positive GC cell lines SNU-719 and NCC24 were purchased from the Korean Cell Line Bank (Seoul, Korea). AGS cell line was cultured to represent microsatellite stable (MSS)/EBV-negative GC cell line. All three cell lines were cultured in RPMI 1640 medium containing 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Cell lines were maintained and passaged for up to 12 months.

MLH1 knockout and western blotting

Single-guide RNAs (sgRNAs) that target MLH1 or control GPF were cloned into LentiCRISPRv2 vector (#52961, Addgene Inco, Cambridge, MA, USA) [13] and transduced by a virus using Virapower packaging mix (Invitrogen, Carlsbad, CA, USA) into SNU-719 and NCC24 cells. Transduced cells were selected in 1 μg/mL puromycin (Sigma-Aldrich, St. Louis, MO, USA) for 7 days. Control and MLH1 knocked-out cells then were seeded as single cells in 96-well plates containing puromycin and expanded. The oligonucleotide sequences used for sgRNA cloning were as follows; AGGTATTCAGTACACAATGC, TGATCCCGGTGCCATTGTCT, and GGGCGAGGAGCTGTTCACCG for MLH1 KO1, MLH1 KO2, and GFP, respectively.

The depletion of MLH1 protein expression was validated by western blotting. To extract proteins, cells were lysed with RIPA buffer supplemented with a protease inhibitor (Roche, Mannheim, Germany) and a phosphatase inhibitor cocktail (Roche). Cell debris was removed by centrifugation. The protein concentration was measured and diluted, so that the final concentration of the cell lysates was 30 μg per 15 μL. Cell lysates were separated on 8% sodium dodecyl sulfate (SDS)-polyacrylamide gel, electrophoretically transferred to nitrocellulose membranes (Millipore, Bedford, MA, USA), and blocked with 5% not-fat dry milk in phosphate-buffered saline for 1 h. After blocking, the membranes were incubated with primary antibodies anti-MLH1 (1:1000, CST-3515, Cell Signaling Technology, Danvers, MA, USA) and anti-β-tubulin (1:1000, ab21057, Abcam, Cambridge, MA, USA). Horseradish peroxidase-conjugated anti-mouse IgG (1:2000, SA001-500, GenDEPOT, Barker, TX, USA) was used as secondary antibody. Proteins were detected by chemiluminescent reagent ECL solution (Biosesang, Korea). Protein detection was were repeated thrice.

Proliferation and tumorsphere formation assays

GC cells were seeded in 24-well plates and incubated for up to 72 h. Each well was stained with 0.2% crystal violet solution in 20% methanol for 20 min followed by washing and air drying. Crystal violet stain was dissolved in 10% sodium dodecyl sulfate (SDS) for 20 min at room temperate and its absorbance was measured at 570 nm for intensity to detect cell proliferation. At least four wells were duplicated for each condition. Proliferation assays were performed approximately five times before the tumorsphere assay.

For tumorsphere formation, 200 GC cells were seeded into each 96-well ultra-low attachment plates (Corning, Lowell, CA, USA) in 200 μL of RPMI 1640 medium containing 10 mM HEPES, 10 ng/mL human recombinant basic fibroblast growth factor, and 20 ng/mL recombinant epidermal growth factor. Each cell line was seeded in at least 10 wells. After 7 days of culture, the number of tumorspheres was measured using computerized image from microscopy.

Cell invasion and migration assays

A 24-well Insert System using an 8 μm polyethylene terephthalate membrane (Corning) was obtained and incubated with PBS-diluted Matrigel for 24 h (Corning) for the invasion assay. After removing PBS, cells suspended in serum-free RPMI 1640 (5 × 105 cells/insert) were added to the top of each insert. The lower chamber contained identical medium with 10% FBS. Incubation was performed for to 72 h and non-invaded cells were removed from the top of each insert with cotton swab before crystal violet staining. To quantify the invaded cells, each insert was dissolved with 10% SDS, and absorbance at 570 nm was measured with an ELISA reader. For the migration assay, an identical protocol was applied but without the coating of the insert with Matrigel. Furthermore, the number of cells per each insert was reduced (105 cells/insert). For both assays, each condition was evaluated with at least four inserts.

MSI tests

The MSI status of cell lines was assessed with the genomic DNA of cells using capillary electrophoresis and parallel sequencing. Assessment with capillary electrophoresis was conducted as previously described [14], with modification of the evaluation process. Two markers were tested for screening (BAT25 and BAT26). When instability was detected in both markers, all five markers (BAT25, BAT26, D2S123, D5S345, and D17250) were tested for confirmation of MSI-high status. Targeted parallel sequencing with 21 markers identified cell lines as MSI-high when more than 30% of the markers showed instability.

Xenograft tumor model, immunohistochemical staining, and EBV in situ hybridization

GC cells were injected subcutaneously into BALB/c nude mice. NCC24 cell lines were injected in both flanks of four mice and SNU-719 cell lines were injected in five mice. The tumor volume was measured periodically until tumor extraction. Tumor masses derived from xenografts were removed surgically and fixed with buffered formalin. Immunohistochemical staining with anti-MLH1 antibody and EBV in situ hybridization with silver staining were performed on glass slides as described previously [15]. Animal procedures were approved by the Biomedical Research Institute of Seoul National University Hospital (approval no.3520160085).

Flow cytometry

Each GC cell was incubated with TrypLE Express (Gibco, Carlsbad, CA, USA) for 10 min for detachment, with minimal loss of surface protein expression [16] . After washing with cold incubation buffer (comprised of PBS, 0.5% bovine serum albumin, and 0.05% sodium azide), cells were incubated on ice with antibodies anti-CD44 (1:50, ab81424, Abcam) or anti-CD133 (1:10, 130-098-826, Miltenyi Biotec, Germany) for 30 min. FACSCanto II System (BD Biosciences, San José, CA, USA) was used to detect CD44 and CD133 expression via allophycocyanin and phycoerythrin absorbance, respectively.

Statistical analysis

Experimental data were analyzed using R software (3.5.3). To analyze data from The Cancer Genome Atlas (TCGA), R package TCGAbiolinks was used [17]. Student’s t test was used to determine the significance of the results. P value of < 0.05 was considered as statistically significant.

Results

Exclusivity between CIMP-high phenotypes and MLH1 promoter methylation status in GC

EBV-positive and MSI-high GCs are mutually exclusive molecular phenotypes that show significant difference in MLH1 mRNA expression levels (Fig. 1a, b). According to methylation data from TCGA, MLH1 is one of the rare genes with CpG sites that are hypermethylated in MSI-high GC than in EBV-positive GC (Fig. 1c). A closer look at the methylation status of the MLH1 gene revealed that CpG sites in the MLH1 promoter region to be hypermethylated for MSI-high GC, whereas other molecular phenotypes, including EBV-positive GCs, are unmethylated in the same region (Fig. 1d).

Fig. 1
figure 1

Exclusivity between molecular phenotypes and DNA methylation status of gastric cancer (GC) according to molecular phenotype. a Proportion of each molecular phenotype shows exclusivity between MSI-high (MSI +) and EBV-positive (EBV +) GCs in SNUH GC cohort and TCGA-STAD. b Expression of MLH1 mRNA in EBV-positive and MSI-high GCs according to TCGA-STAD. c Volcano plot of genes CpG sites that are differentially methylated in EBV-positive and MSI-high GCs. Hypermethylated CpG sites in MSI-high GC associated with MLH1 are marked in orange. d Methylation status of CpG sites adjacent to MLH1 promoter. Only MSI-high tumors are hypermethylated in this region

Full size image

Down-regulation of MLH1 does not affect proliferation, invasion, migration, and short-term MSI status in EBV-positive GC cells

To determine the effect of MLH1 deficiency in the EBV-positive GC cells SNU-719 and NCC24, we selected cells that showed decreased expression of MLH1 in western blot after MLH1 knockout by CRISPR-Cas9 (Fig. 2a). In both cell lines, MLH1-knockout cell lines (KO1 and KO2) showed identical cell proliferation as the control cell lines (P = 0.836 and P = 0.491, respectively, for SNU-719, and P = 0.395 and P = 0.078, respectively, for NCC24) (Fig. 2b, c). Furthermore, PCR-based assessment of MSI within the first 3 months of knockout showed that MSI was not induced in GC cell lines (Fig. 2d). Two EBV-positive GC cell lines did not show any alteration in their ability to invade (P = 0.423 and P = 0.206, respectively, for SNU-719, and 0.262 and P = 0.247, respectively, for NCC24) (Fig. 2e) and migrate (P = 0.836 and P = 0.603, respectively, for SNU-719, and P = 0.707 and P = 0.253, respectively for NCC24) (Fig. 2f) after MLH1 knockout.

Fig. 2
figure 2

Initial evaluation of cell lines after MLH1 knockout. a Western blot assay of cell lines. b Tumor growth of SNU-719 cell lines. c Tumor growth of NCC24 cell lines. d Capillary electrophoresis for screening MSI status in cell lines. e Insert membrane photos (× 200) and relative absorbance of crystal violet-stained cells from migration assy. f Insert membrane photos (× 200) and relative absorbance of crystal violet-stained cells from invasion assay

Full size image

Prolonged culture of MLH1-negative, EBV-positive GC cells gradually induce instability in microsatellites

According to studies by Germano et al. [18], mouse tumor cells showed instability in MSI test after more than 100 days of Mlh1 knockout. Therefore, we extended our culture of GC cell lines up to 12 months while screening for MSI via capillary electrophoresis and parallel sequencing. MLH1-negative NCC24 was transformed to MSI-high tumors after 12 months, while MLH1-negative SNU-719 remained as microsatellite stable (Fig. 3a, b). However, parallel sequencing revealed that both MLH1-negative cell lines gradually gained instability (Table 1).

Fig. 3
figure 3

MSI test by capillary electrophoresis after 12 months of passage. a No instability was observed in the control NCC24 cell line. b Four out of five markers showed instability (arrows) in MLH1 knocked-out NCC24 cell line

Full size image
Table 1 MSI of GC cell lines after MLH1 knockout evaluated through parallel sequencing
Full size table

Sphere-forming and xenograft tumor assay indicate regulation of stemness by MLH1 in EBV-positive GC cells

Tumorsphere formation assay and xenograft tumor model were developed to determine the effect of stemness in MLH1-negative GC cells. MLH1-negative SNU-719 and NCC24 cells showed a significant decrease in sphere number compared with control cell line (P < 0.001 and P < 0.001, respectively, for SNU-719, and P = 0.003 and P < 0.001, respectively, for NCC24) (Fig. 4a). However, AGS cell line did not show decreased number of tumor spheres when MLH1 expression was downregulated (Supplementary Fig. 1). In the xenograft tumor model, tumors with MLH1 deficiency were significantly decreased in tumor volume compared to those injected with control cell lines (P < 0.001 and P < 0.001 for SNU-719 and NCC24, respectively) (Fig. 4b). We confirmed that MLH1-negativity and EBV-positivity were not affected during the xenograft procedure (Fig. 4c). Cell surface proteins CD44 and CD133 were selected to determine stemness of cell lines, since they are considered as cancer stem cell makers in EBV-positive cancer or gastric cancer [19,20,21]. A decrease of surface stem cell marker CD44 expression, induced by MLH1 knockout, was observed significantly in NCC24 but was not as significant in SNU-719 cells (Fig. 4d). The surface protein expression of CD133 was not detected in both cell lines, which was also confirmed by immunohistochemistry of the xenograft tumor (Supplementary Fig. 2). Therefore, regardless of the MSI status, reduction in stemness was observed in MLH1-negative EBV-positive GC cell lines both in vitro and in vivo.

Fig. 4
figure 4

Tumor stemness, MLH1 expression, and maintenance of EBV-positivity in GC cell lines. a Tertiary tumorspheres (left) and its count in GC cell lines (right). b In xenograft tumor models, control cell lines were injected to the right flank and MLH1 knocked-out cell lines were injected to the left flank (left). Tumor size periodically measured before tumor extraction (right). c Tumor mass obtained from xenograft were immunostained with MLH1 (upper) or stained with silver for EBV in situ (lower). d Flow cytometry for CD44 expression in GC cell lines

Full size image

Discussion

GC is characterized by its heterogeneity and complexity of molecular phenotypes, and therefore, elucidating the properties in different molecular subtypes could lead to the development of novel therapeutic interventions. MLH1 deficiency, MSI-high status, or DNA mismatch repair (MMR) deficiency has been reported to be associated with increased [22, 23] or decreased features related to tumor stemness in human malignancies. [24] Unlike other cancers, DNA methylation is higher in GC with EBV-positive than in GC with MSI-high phenotype [25], and therefore, silencing MLH1 in a hypermethylated DNA environment is an uncharted territory. This study reports, for the first time, the underlying mechanism of exclusivity between EBV-positive GC and MSI-high GC in relation to the stemness of GC cells.

Herein, we generated MSI-high status in the EBV-positive GC cell line NCC24 after 12 months of culture following MLH1 knockout via CRISPR-Cas9. Although the EBV-positive GC cell line SNU-719 did not reach MSI-high status from the identical procedure, parallel sequencing revealed that MLH1-knockout cell lines have been promoting a gradual increase in instability, which might eventually lead to MSI-high status with passage longer than 12 months. Silencing of MLH1 in EBV-positive GC cells was accompanied by reduction in stemness but not in tumor proliferation. Interestingly, these events seemed to be independent from gaining of instability, since no prolonged passage was required to observe these outcomes. Furthermore, these features remained unchanged even after prolonged passage (data not shown). Unlike EBV-positive GC cells, EBV-negative/microsatellite stable GC cells did not show decrease in cellular features associated with stemness when MLH1 was downregulated. These findings indicate that reduction of stemness from MLH1-knockout is likely limited to blocking EBV infection-specific signaling pathways involved in tumor progression.

In a study on murine breast, colorectal, and pancreatic cell lines [18], inactivation of Mlh1 had resulted in MMR deficiency status. MMR-deficient cells showed comparable growth with their proficient counterparts in vitro and on transplantation in immunocompromised mice. However, in immunocompetent syngeneic mice, MMR-deficient cells grew much poorly. These cells also showed increased mutational burden and dynamic generation of neoantigens, which eventually lead to increased recruitment of tumor-infiltrating lymphocytes and subsequent decrease in tumor growth. Enhanced generation of neoantigens and increased density of tumor-infiltrating lymphocytes led to MMR-deficient cells, which greatly increased responsiveness to immunotherapy. Recent findings have also suggested that immunotherapy, such as PD-1 blockage, induces clonal replacement of T cells by immune cells from outside the tumor tissue, such as peripheral blood, rather than expanding immune cells from the adjacent tumor microenvironment [26]. Taken together, these findings suggest that checkpoint blockage activity is largely affected by the intrinsic ability of the tumor cells to recruit external immune cells, and responsiveness to immunotherapy might not be predetermined by pre-existing tumor-infiltrating immune cells.

Herein, MLH1 silencing has led to MSI-high transformation in GC cells. Although EBV-positive metastatic GCs responded to PD-1 targeted therapy [27], whether an increase in mutational load and neoantigens due to MMR deficiency status could still be beneficial in boosting responsiveness to immunotherapeutic reagents in EBV-positive GC remains uncertain. However, the long interval before induction of MSI-high status in both murine and human cell lines warrants attention before clinical applications. Potential method of accelerating this process includes controlling oxygen concentration or hypoxia-associated proteins within the tumor cells [28]. Moreover, we downregulated MLH1 in established GC cells instead of nonmalignant gastric epithelial cells. Since MMR deficiency and EBV infection are both strong drivers of early carcinogenesis, this study might not have uncovered the whole story of mutual exclusivity.

Our study had a few limitations. Although EBV and MSI are both important drivers of multistep carcinogenesis, our study was limited to MSI-high status developed in EBV-positive GC cell lines and did not include results from EBV infected MSI-high GC cell lines. Further studies should investigate whether EBV infected MSI-high GCs also show reduced stemness. The potential reduction of cancer stem cells caused by downregulation of MLH1 expression was not fully evaluated with cancer stem cell-associated mRNA signature but with limited flow cytometry and immunohistochemical markers. In EBV-positive GC cells, EBV latent membrane protein 2A (LMP2A) increased stemness mediated by the NF‐κB pathway [29]. Under stressed conditions, MLH1-deficient cancer cells failed to activate NF‐κB compared to the wild-type MLH1 cancer cells [30]. Therefore, NF‐κB pathway is a potential link between EBV-positive GC and reduced stemness in MLH1-deficient status. Finally, our study could not answer the question of how MSI-high status induced in EBV-positive GC cells could affect stemness apart from the depletion of MLH1. Since re-expression of MLH1 will not correct MSI-high status, we believe that reduced stemness will persist due to oncogenic addiction regardless of MLH1 re-expression once MSI-high status has been induced [31].

In conclusion, MSI-high status and EBV-positivity, which are known to be mutually exclusive, were established as a result of prolonged culture of EBV-positive GC cells after MLH1 knockout. These cells showed decreased stemness compared to EBV-positive GC cells, both in vitro and in vivo. Our current findings suggest that a double positive GC phenotype is not observed in clinical conditions due to the novel introduction of DNA MMR gene deficiency having disadvantage on tumor stemness when EBV-positivity is already present. Future studies should investigate the signaling pathway specific to EBV-positive GCs that shows reduction of stemness from MLH1 deficiency.