Cancer-specific promoter DNA methylation of Cysteine dioxygenase type 1 (CDO1) gene as an important prognostic biomarker of gastric cancer

Background There have been few available prognostic biomarkers in gastric cancer. We rigorously assessed the clinical relevance of promoter DNA methylation of Cysteine dioxygenase type 1 (CDO1) gene, a cancer-specific aberration, in human gastric cancer. Methods Quantitative CDO1 methylation value (TaqMeth V) was initially calculated in 138 gastric cancer patients operated in 2005, and its clinical significance was elucidated. As a subsequent expanded set, 154 gastric cancer patients with pathological stage (pStage) II / III with no postoperative therapy were validated between 2000 and 2010. Results (1) Median TaqMeth V of CDO1 gene methylation of gastric cancer was 25.6, ranging from 0 to 120.9. As pStage progressed, CDO1 TaqMeth V became higher (p < 0.0001). (2) The optimal cut-off value was determined to be 32.6; gastric cancer patients with high CDO1 gene methylation showed a significantly worse prognosis than those with low CDO1 gene methylation (p < 0.0001). (3) A multivariate cox proportional hazards model identified high CDO1 gene methylation (p = 0.033) as an independent prognostic factor. (4) The results were recapitulated in the expanded set in pStage III, where high CDO1 gene methylation group had a significantly worse prognosis than low CDO1 gene methylation group (p = 0.0065). Hematogenous metastasis was unique in pStage III with high CDO1 gene methylation (p = 0.0075). (5) Anchorage independent growth was reduced in several gastric cancer cell lines due to forced expression of the CDO1 gene, suggesting that abnormal CDO1 gene expression may represent distant metastatic ability. Conclusions Promoter DNA hypermethylation of CDO1 gene was rigorously validated as an important prognostic biomarker in primary gastric cancer with specific stage.


Methods
Quantitative CDO1 methylation value (TaqMeth V) was initially calculated in 138 gastric cancer patients operated in 2005, and its clinical significance was elucidated. As a subsequent expanded set, 154 gastric cancer patients with pathological stage (pStage) II / III with no postoperative therapy were validated between 2000 and 2010.

Results
(1) Median TaqMeth V of CDO1 gene methylation of gastric cancer was 25.6, ranging from 0 to 120.9. As pStage progressed, CDO1 TaqMeth V became higher (p < 0.0001). (2) The optimal cut-off value was determined to be 32.6; gastric cancer patients with high CDO1 gene methylation showed a significantly worse prognosis than those with low CDO1 gene methylation (p < 0.0001). (3) A multivariate cox proportional hazards model identified high CDO1 gene methylation (p = 0.033) as an independent prognostic factor. (4) The results were recapitulated in the expanded set in pStage III, where high CDO1 gene methylation group had a significantly worse prognosis than low CDO1 gene methylation group (p = 0.0065). Hematogenous metastasis was unique in pStage III with high CDO1 gene methylation (p = 0.0075). (5) Anchorage independent growth was reduced in several gastric cancer a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Introduction
Gastric cancer is the fifth most common malignancy and the third leading cause of cancerrelated death worldwide [1]. Advanced gastric cancer, defined as depth of muscularis propria or beyond, still exhibited poor prognosis by curative surgery even in combination with effective adjuvant chemotherapy [2,3], and require prognostic factors reflecting their biology to enrich high-risk patients for recurrences. Although several prognostic biomarkers have been reported by immunohistochemistry [4,5] or by mRNA quantification [6,7], they have weak points; the former included issues of intra-tumoral heterogeneity and cut-off line between positive and negative cases, while the latter is instable and not appropriate for routine examination. Hence, stable and quantitative methods have been anticipated to develop like DNA markers [8].
Epigenetic gene silencing of the tumor suppresser genes (TSGs) through promoter DNA hypermethylation is a unique feature in human cancers, whereas such cancer specific methylation is rather a rare event [8,9]. We had developed pharmacologic reversal of epigenetic silencing and uncovered a myriad of transcriptionally repressed genes in human cancers [10][11][12][13], and have finally identified outstanding candidate TSG with frequent promoter DNA methylation, a cysteine dioxygenase type 1 (CDO1) gene in human cancers including gastric cancer [14].
The CDO1 protein is a non-heme structured, iron-containing metalloenzyme involved in conversion of the cysteine to cysteine sulfinic acid (CSA), while it can promote apoptosis by increasing reactive oxygen species (ROS) through suppression of glutathione generation in breast cancer cells [15], thus suggesting that CDO1 is a TSG in the context of pseudo-inflammatory reaction during carcinogenesis. In addition to breast cancer [15,16], promoter DNA hypermethylation of CDO1 gene has been reported to be highly specific to cancer cells [8], and exhibited prognostic relevance in specific cancers such as esophageal [17,18], lung [19], colorectal [20], gallbladder [21], and kidney cancer [22].
Nevertheless, we have not found any reports on the clinical significance of CDO1 gene methylation in primary gastric cancer. In the present study, CDO1 gene promoter DNA methylation was for the first time examined and clarified for detailed clinicopathological factors in primary gastric cancer, and proved great clinical value in gastric cancer clinics. chemotherapy were collected from 1673 gastric cancer patients between 2000 and 2010 for validation. The median follow up term of the expanded one was 100.5 months, ranging from 2 to 148 months. pStage was used according to the Japanese Classification of Gastric Cancer staging system, 14th edition [23]. This study was approved by the Kitasato University Ethics Committee (number B17-251).

Genomic DNA extraction and bisulfite treatment
Genomic DNA was extracted from cell lines using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). Formalin-fixed paraffin-embedded tissue was cut into six slices of 10 μm thick before genomic DNA extraction using the QIAamp DNA FFPE Tissue Kit (Qiagen). Genomic DNA (2 μg) was bisulfite converted using the EZ DNA Methylation-Gold Kit (Zimo Research, Irvine, CA, USA).

Total RNA extraction and RT-PCR
Total RNA was extracted using an RNeasy Mini Kit (Qiagen). First strand cDNA was synthesized from RNA (2 μg) using SuperScript III reverse transcriptase (Invitrogen) and Oligo (dT) primers (Invitrogen). RT-PCR was carried out using Platinum Taq DNA Polymerase (Invitrogen). Primers sequences for CDO1 and β-actin were previously described [21].

Quantitative methylation-specific PCR (Q-MSP)
For Q-MSP of CDO1 gene, we performed real-time PCR using iQ Supermix (Bio-Rad, Hercules, CA) and CFX96 real-time systems and TaqMeth V was defined as previously described [21]. All reactions were performed in triplicate.

5-Aza-dC and TSA treatment
Cells were seeded in a 10 cm dishes, and were then treated every 24 h for 4 days with either 1 or 5 μM 5-Aza-dC (5-aza-20-deoxycytidine) dissolved in 50% acetic acid or were mock treated with PBS including the same amount of acetic acid. Trichostatin A (TSA; 300 nM; Sigma Aldrich, Inc, St Louis, MO, USA) was added to the medium for the final 24 h. On day 5, the cells were harvested and mRNA was extracted.

Plasmid construction for transfection into cell lines
Full-length CDO1 cDNA was inserted into the pcDNA3.1 myc-His C expression vector (Invitrogen) as previously described [20]. Cells were transfected using Lipofectamine 2000 (Invitrogen) in Opti-MEM (Invitrogen).

Cell proliferation assay
Cell proliferation was assayed using the CytoSelect water-soluble tetrazolium salt (WST-1) Cell Proliferation Assay Reagent (Cell Biolabs, San Diego, CA, USA). On day 1, the cells were cultured in a 96-well plate at a density of 1×10 4 cells per plate. On day 2, the cells were transiently transfected with CDO1. On day 3, cell proliferation was evaluated by measuring the optical density (OD) at 450 nm.

Anchorage-independent colony formation assay
The anchorage-independent colony formation assay was performed. In a six-well plate, 0.72% agarose (Bacto Agar; Becton, Dickinson and Company, Franklin Lakes, NJ) was placed on the bottom. Top agar was made with agarose mixed with 1 × 10 5 cells transfected with CDO1. After 3 weeks of culture, colonies with more than 100 cells were counted in 10 fields of view. The experiment was conducted twice.

Statistical analyses
All statistical analyses were performed using JMP 11 software (SAS Institute Inc., Cary, NC, USA). Continuous variables were evaluated by ANOVA, Student's t test; categorical variables were evaluated by Fisher's exact test or the Chi-square test, as appropriate. Overall survival (OS) was measured from the date of death or censored at the date of the last follow-up evaluation. Survival was estimated using the Kaplan-Meier method and compared by the log-rank test. Differences between results of comparative tests were considered significant if the twosided P value was less than 0.05.

Quantification of promoter DNA methylation of CDO1 gene in primary gastric cancer
Q-MSP for CDO1 gene was initially performed in 138 primary gastric cancer. Median Taq-Meth V of CDO1 gene was 25.6, ranging from 0 to 120.9 ( Fig 1A). The CDO1 TaqMeth V tended to become higher as pStage progressed. There was a significant difference between pStage IV and pStage I / II / III (p < 0.0001 / p = 0.01 / p = 0.03, respectively)( Fig 1B). Age

Prognostic analysis according to CDO1 gene TaqMeth V in primary gastric cancer
For prognostic analysis, the optimal cut-off value for OS was determined by ROC curve predicting death event. The most optimized TaqMeth V was determined to be 32.59 (AUC of 0.70, p < 0.0001, sensitivity 76.6%, specificity 56.8%) (Fig 2A). Gastric cancer patients were divided into two groups of H group (n = 47): high CDO1 TaqMeth V group (TaqMeth V � 32.6) and L group (n = 91): low CDO1 TaqMeth V group (TaqMeth V <32.6). The H group had a significantly poorer prognosis (5-year OS 49.5%) than the L group (5-year OS 82.0%) (p < 0.0001) ( Fig 2B). Age, procedures of gastrectomy, lymph node dissection, radical resection, tumor location, morphological type, pStage, and CDO1 TaqMeth V were significant (p < 0.05) prognostic factors in a univariate analysis. These univariate prognostic factors were applied to Cox proportional hazards model. As a result, pStage and CDO1 TaqMeth V (HR 2.28, CI 1.07-4.95, p = 0.033) were finally remnant independent prognostic factors in multivariate analysis ( Table 1).
As the H group included higher pStage than the L group did (p = 0.0006, Table 2). prognosis was then compared in individual pStage. There was no significant difference between the H group and the L group in pStage I / IV (p = 0.2846/p = 0.62, respectively) (left and right panels of Fig 2C), while the H group (n = 22) exhibited significantly poor prognosis than the L Group (n = 32) in pStage II / III (p = 0.0077) (Fig 2D, left panel). In 2005, pathological stage II / III patients were recommended for postoperative adjuvant chemotherapy as a phase III clinical trial (ACTS-GC), the prognosis may have been modified by the adjuvant chemotherapy because the ACTS-GC proved prognostic efficacy [27]. We then restricted prognostic analysis to the 39 patients with pStage II / III gastric cancer without postoperative adjuvant chemotherapy, and still proved that the H group (n = 16) had a significantly poorer prognosis (5-year OS 37.7%) than the L group (n = 23, 5-year OS 81.7%) (p = 0.0011) (Fig 2D, right panel).
In order to verify the results of prognostic significance in pStage II / III gastric cancer without adjuvant chemotherapy, prognostic analysis was performed on 154 patients in pStage II / III advanced gastric cancer without adjuvant chemotherapy who were collected between 2000 to 2010 as an expanded set. The result again proved similar prognosis with the learning set. CDO1 gene hypermethylation could predict poorer prognosis of pStage II / III gastric cancer without adjuvant chemotherapy than CDO1 gene hypomethylation (5-year OS 53.8% / 76.3% in the H / L group, respectively) (p = 0.0067) (Fig 2E, left panel). Statistical difference of OS was not recognized in pStage II (Fig 2E, middle panel), but in pStage III (Fig 2E, right panel); the 5-year OS of the H group was 40.1%, while 71.3% in the L group, and the prognosis was significantly poorer in the H group as compared with that in the L group (p = 0.0065).
The recurrence pattern of pStage II/III gastric cancer patients with no adjuvant chemotherapy in the expanded set was then clarified to explain the cause of poor prognosis by CDO1 gene hypermethylation. There were 8 recurrent cases (14.0%) in pStage II and 43 recurrent cases (44.3%) in pStage III, with significant more relapses in pStage III cases than in pStage II cases (p < 0.0001). There were 15 cases (9.6%) of initial recurrences at lymph nodes, 25 cases (15.9%) of initial recurrences at distant organ, 22 cases (14.0%) of initial recurrence at peritoneum, and 3 cases of initial recurrences at local location (1.9%). First of all, there were significantly more recurrences in the H group than in the L group among the pStage III gastric cancer patients (p = 0.0052), and the most outstanding features were characterized by more distant organ metastasis in the H group than in the L group in pStage III (p = 0.0075) ( Table 3).

Functional assessment of CDO1 gene transfection on gastric cancer cells
Expression of CDO1 gene was never observed in 6 gastric cancer cell lines, as compared with positive expression control of HepG2 cells at mRNA level (Fig 3A), where DNA hypermethylation was confirmed in all the 6 cell lines and DNA hypomethylation was seen in HepG2 cells (Fig 3B). Reactivation by demethylation treatments using 5-Aza-dC and Trichostatin A was confirmed in all 6 the cell lines (Fig 3C), suggesting that expression of CDO1 gene must be suppressed by epigenetic manners such as promoter DNA methylation. Immunostaining for CDO1 protein confirmed its localization in the cytoplasm of non-cancerous gastric mucosa gland cells (data not shown) or cancer cells harboring low value of CDO1 gene methylation (L group) (Fig 3D, upper panels). On the other hand, representative specimens of H group were weakly positive for CDO1 protein expression (Fig 3D, lower panels).
The CDO1 full length vector was transfected to all 6 gastric cancer cell lines and the expression was confirmed by RT-PCR and Western blotting (Fig 3E). Intense expression of CDO1 protein was confirmed in MKN7, SH-10-TC, NUGC-4 and Kato III, but not in MKN74 and KE-97 ( Fig 3E). After CDO1 gene transfection, cell proliferation was suppressed in NUGC-4 by simple cell count (Fig 3F, p = 0.026) and WST-1 cell viability assays (Fig 3G, p = 0.03).
Among gastric cancer cell lines tested in this study, only Kato III exhibited anchorage-independent colony formation. For demonstrating the tumor suppressive activity of CDO1 gene, only Kato III cells were considered to be appropriate. The CDO1-transfected cells showed suppressed capacity of anchorage independent growth compared to MOC-transfected cells (p = 0.0245) in Kato III cells (Fig 3H).

Discussion
Promoter DNA of the CDO1 gene is frequently hypermethylated in human cancers including gastric cancer [14], which showed the highest AUC (0.95) to differentiate tumor tissues from the corresponding non-cancerous tissues [8]. Hypermethylation in tumor tissues beyond 60% was designated as highly relevant methylation gene (HRMG), and CDO1 gene is the most common HRMG among human cancers [8]. Using the best optimized cut-off value of Taq-Meth V to discriminate tumor from non-cancerous tissues, CDO1 gene hypermethylation is found in 72~91% in various cancers [14,18,21,28]. These frequencies were determined, based  Fig 2D) Kaplan-Meier survival curves for OS comparing pStage II / III gastric cancer patients without postoperative adjuvant chemotherapy with CDO1 TaqMeth V below 32.6 and those with CDO1 TaqMeth V equal to or over 32.6 (p = 0.0011). In 154 patients as an expanded set of pStage II / III gastric cancer without postoperative adjuvant chemotherapy. (Left panel of Fig 2E) Kaplan-Meier survival curves for OS comparing gastric cancer patients with CDO1 TaqMeth V below 32.6 and those with CDO1 TaqMeth V equal to or over 32.6 (p = 0.0067). (Middle panel of Fig 2E) In expansion set of pStage II. Kaplan-Meier survival curves for OS comparing gastric cancer patients with CDO1 TaqMeth V below 32.6 and those with CDO1 TaqMeth V equal to or over 32.6 (p = 0.3735). (Right panel of Fig 2E) In expansion set of pStage III. Kaplan-Meier survival curves for OS comparing gastric cancer patients with CDO1 TaqMeth V below 32.6 and those with CDO1 TaqMeth V equal to or over 32.6 (p = 0.0065).
https://doi.org/10.1371/journal.pone.0214872.g002 on comparison of tumor tissues to the corresponding non-cancerous tissues, and affected by the methylation level of non-cancerous tissues. For example, corresponding non-cancerous tissues were relatively highly methylated for CDO1 gene in gallbladder cancer, and threshold cut-off value became high and the frequencies were underestimated [21]. Actual methylation frequencies are therefore considered higher than the report (~90% in almost human cancer). Anyway, CDO1 gene has an outstanding feature with regard to cancer-specific methylation in human cancer. The current study is the first report describing clinicopathological relevance of CDO1 gene promoter DNA methylation status in primary gastric cancer. Recent literatures of the significant association between CDO1 methylation and poor prognosis have been reported in breast [15,16], esophageal [17,18], renal cells carcinoma [22], HPV associated malignancies [29], prostate cancer [30], gallbladder cancer [21], and colorectal cancer [20]. Nevertheless, there has never been reported with regard to prognostic relevance in primary gastric cancer.
In the present study, CDO1 TaqMeth V was rigorously validated as prognostic factor of primary gastric cancer. The most importantly, it could still show the prognostic relevance in pStage II/III gastric cancer patients without postoperative adjuvant chemotherapy. We have to clearly recognize the difference of patients with adjuvant chemotherapy and those without it from a prognostic point of view, because recent adjuvant chemotherapy is really effective to pStage II/III advanced gastric cancer [2,3]. In order to know the biological role of CDO1 gene during natural clinical course, we had better not include gastric cancer with adjuvant chemotherapy. In our current study, CDO1 gene promoter methylation definitely accumulates as disease progressed, and it was significantly associated with the initial recurrences at distant organs in pStage III gastric cancer. CDO1 gene actually suppressed anchorage independent growth in Kato III cells, suggesting that it plays a functionally critical role in distant metastasis of gastric cancer.
The best optimized cut-off TaqMeth V with regard to prognosis was set as 32.6. This optimized cut-off value is always higher than those delineated tumor from non-cancerous tissues in various cancers [14,17,20,21]. Moreover, cancer patients with CDO1 gene hypermethylation showed suppressed expression of CDO1 protein in immunohistochemistry [20,21]. These findings suggested that higher promoter DNA methylation status represents strong suppression of CDO1 protein expression, which may be linked to tumor aggressiveness. Almost cancer cell lines including gastric cancer did not recognize CDO1 gene expression, and it is therefore difficult to conduct an experiment to suppress expression of CDO1 gene by RNA interference using siRNA. Only HepG2, a liver cancer cell line expressed CDO1 gene, and suppression of CDO1 gene by RNA interference resulted in invasive capacity as described by Brait M [14]. Reflected by this functional experiment, CDO1 gene promoter DNA methylation is an excellent prognostic marker, because it is DNA (that is stable in any environment), and it could be quantified by Q-MSP differently from immunohistochemistry. Although there was no significant difference of the initial recurrences at peritoneum according to CDO1 gene TaqMeth V in gastric cancer, it can be used as a cancer detection marker, because it is highly specific to cancer cells [14]. We recently reported the usefulness of DNA diagnosis using the CDO1 gene methylation in DNA cytology test using the peritoneal lavage of gastric cancer [31]. DNA cytology of peritoneal lavage had higher diagnostic ability compared to the conventional cytology test of peritoneal lavage. Currently, prospective study is conducted to validate the clinical utility (UMIN000026191).
Cysteine biology has recently focused on cancer stem cell features [32]. CD44 variant interacts with xCT, a glutamate-cystine transporter, and permits intracellular increase of glutathione that protects stem cells, which is associated with inflammatory processes. Ablation of CD44 induced loss of xCT from the cell surface and suppressed tumor growth. CDO1 protein catalyzes the oxidation of cysteine to CSA [33], reducing intracellular cysteine concentration and subsequent reduction of glutathione. This molecular mechanism may be associated with tumor suppressive function of CDO1 gene [15].
Our DNA was extracted from the formalin-fixed, paraffin embedded (FFPE) tumor tissues, and not from noninvasive biopsy. Unlike fresh frozen samples, FFPE samples were demonstrated to exhibit deterioration in the quality of RNA [34], while verification of DNA methylation analysis using FFPE specimens has been done, and the usefulness has been confirmed [35]. For this reason, methylation analysis was also carried out using FFPE samples in this study.
In conclusion, CDO1 TaqMeth V was rigorously validated to be an important prognostic factor in primary gastric cancer. It was considered that the CDO1 methylation together with its aberrant expression may be causatively involved in the distant metastasis, resulting in poor prognosis of gastric cancer. If it is possible to predict distant metastasis, selection of patients requiring preoperative adjuvant chemotherapy can be effectively made, and can be highly expected for precision medicine.