Abstract
Background/Aim: We previously identified that promoter DNA methylation of cysteine dioxygenase type 1 (CDO1) and homeobox only protein homeobox (HOPX) were both cancer specific, and have a clinical potential as prognostic biomarkers in breast cancer (BC). The present study compared the differential prognostic relevance of methylation status of the CDO1 and HOPX genes in BC. Materials and Methods: Methylation levels (TaqMethVs) were quantified in 7 BC cell lines and 133 BC patients by TaqMan methylation-specific PCR and functional traits were explored for CDO1. Results: TaqMethVs were associated between CDO1 and HOPX (r2=0.072, p=0.002). Multivariate Cox proportional hazards model could identify CDO1 hypermethylation as well as Ki-67 as independent prognostic factors related to disease-specific survival (p=0.016, p<0.001). Overexpression of CDO1 decreased the anchorage-independent growth capacity in BC cell lines. Conclusion: CDO1 is a definite tumor suppressor gene, while its prognostic relevance was more than expected in the context of its functional relevance.
Breast cancer (BC) is the most common cancer in women worldwide. Recent estimates by the International Agency for Research on Cancer (IARC) show that BC is the most commonly diagnosed cancer in women, making up 24.2% or about 1 in 4 of all new cancer cases diagnosed in women worldwide (1). The incidence of BC is increasing in the world due to increase in life expectancy, increased urbanization and adoption of western lifestyles. It is thought that the incidence of BC will continue to increase. Therefore, early detection and individualized treatment, in order to improve clinical and survival outcomes, remains the cornerstone of BC control. Thus, novel approaches for the diagnosis and prognosis of BC are still needed.
In BC, genomic events are fewer than in other cancers (2). We previously developed pharmacological reversal of epigenetic silencing (3) and uncovered a myriad of transcriptionally repressed genes, such as PGP9.5 (4), NMDAR2B (5), DFNA5 (6), and HOPX (7) in human gastrointestinal cancers. Until now, we have originally reported cysteine dioxygenase type 1 (CDO1) and homeobox only protein (HOP) homeobox (HOPX) as novel methylation genes in primary BC (8, 9). Through clinicopathological analysis including prognostic information, we have already reported a positive relationship of both genes to prognosis of BC (9, 10).
BC treatment is unique, because personalized therapy is robustly progressing after the subtype classification was proposed at the 2011 St. Gallen consensus meeting was based on gene expression data sets (11, 12). The expressions of hormone receptors, HER2 and Ki-67 are important determinants in the context of the selection of therapy at present. On the other hand, reports on aberrant promoter DNA methylation and their relations to BC prognosis are reported in about 700 papers until 2018, which have covered more than 200 genes, while only 31 genes described by the 48 papers were investigated by quantitative methylation specific PCR (Q-MSP). Then, among the 48 papers, only three genes, CDO1 (10, 13), PITX2 (14, 15), and HOPX (9) were proven to be independent prognostic factors by multivariate prognostic analysis. There have been no reports describing all three genes simultaniously.
From our laboratory, promoter DNA methylation for CDO1 and HOPX was assessed for highly relevant prognostic factors in primary BC, and their prognostic relevance were seen, especially for CDO1 hypermethylation in triple-negative type (10) and HOPX hypermethylation in HER2-negative type (9). In Q-MSP, prognostic value is assessed by the most optimized cut-off values instead of presence or absence of methylation. Using such optimized cut-off values, we analyzed the two original tumor suppressor genes in a comparative manner to confirm better prognostic marker utility in primary BC. Furthermore, functional analysis of the CDO1, a stronger prognostic factor determined by this study, was performed to deeply understand the molecular insights of the prognostic contribution of the CDO1 hypermethylation in primary BC.
Materials and Methods
Human primary BC tissues and methylation data. A previous study by our group obtained methylation values of either CDO1 or HOPX in 172 primary BC patients with no prior chemotherapy who underwent surgical resection of the primary tumors at Kitasato University Hospital between January 1, 1995 and December 31, 1999 (9,10), among whom data of the 133 primary BC patients were available commonly for both genes.
TNM classification was made according to the latest 7th edition of the Union for International Cancer Control (UICC). This study was performed in accordance with the clinical research guideline of the ethics committee of Kitasato University School of Medicine (B15-161).
BC cell lines. The seven BC cell lines, SK-BR3, YMB-1, CRL, MDA-MB231, YMB-1E, MDA-MB453, MCF-7 cells were previously described (10). Colorectal cancer cell line, DLD1 was provided from the Cell Resource Center for Biomedical Research Institute of Development, Aging and Cancer, Tohoku University (Sendai, Japan). The hepatocellular carcinoma cell line, HepG2, was also purchased from RIKEN BioResource Centre (Ibaraki, Japan) as control cell lines.
RNA purification and reverse transcription-polymerase chain reaction (RT-PCR). Total RNA from cell lines were extracted using Rneasy Mini Kit and RNeasy FFPE Kit (QIAGEN, Hilden, Germany). Total RNA was reverse-transcribed with SuperScriptIII reverse transcriptase kit (Invitrogen, Carlsbad, CA, USA). RT-PCR was performed, and the PCR products were separated on 1.5-2.0% agarose gel, then visualized by ethidium bromide (10).
Plasmid and transient transfection. The full-length cDNA sequence of CDO1 was isolated using PCR and sub-cloned into pcDNA™3.1D/V5-His-TOPO vector with confirmation of no mutation of CDO1 (Invitrogen). The vector with self-ligation was used as a mock control. Plasmid vectors were transiently transfected into 7 BC cell lines using Lipofectamine 2000 reagent (Invitrogen).
Anchorage-independent colony formation assay. Anchorage-independent cell growth was analyzed by plating 0.36% top agarose (Bacto™ Agar, Becton Dickison and Company, Franklin Lakes, NJ, USA) containing 1.0×105-5.0×105 cells on a surface of 0.72% bottom agarose in 6-well plates. Cells were fed weekly by overlying fresh soft agar solution. Colonies were visualized with ethidium bromide after 2-3 weeks of incubation. Two independent experiments were performed, and each experiment was done in triplicate.
Proliferation assay. Cell proliferation and viability (1.0×105 cells/well) were measured using the Premix WST-1 Cell Proliferation Assay System (Takara Bio, Tokyo, Japan) in 96-well plates. Dates are expressed as an absorbance at 450nm. Experiments were performed in triplicate.
Invasion assay. The invasive property of cell was measured using CytoSelect™ 96-well Cell Invasion Assay Kit (CELL BIOLABS, San Diego, CA, USA). Cells were seeded at density of 1.0-5.0×105 cells/ml in serum free media in the membrane chamber. Media containing 10% fetal bovine serum was added in the feeder tray and the membrane chamber was placed back into the feeder tray. After incubation for 24 h, the membrane chamber was fixed and stained by 4X Lysis Buffer/CyQuant® GR dye solution. The fluorescence was measured at 480/520nm. Experiments were performed in triplicate.
Statistical analysis. For continuous variables, Student's t-test or analysis of variance (ANOVA) was used for comparison between two groups, or among multiple variables, respectively, while for categorical variables, we used χ2 test or Fisher exact test. Clinicopathologic characteristics and follow-up dates were analysed in terms of disease specific survival (DSS). The follow-up time was calculated from the date of surgery to death or end-point, and patients with other disease deaths were censored. DSS was calculated by Kaplan-Meier methods, and survival differences were assessed in the log-rank test. Variables suggesting potential prognostic factors on univariate analyses (p<0.05) were subjected to multivariate analyses using a Cox proportional-hazards model. All statistical analyses were performed using JMP®11 software (SAS Institute Inc., Cary, NC, USA).
Results
Correlation of promoter DNA methylation status of CDO1 and HOPX with prognosis in 133 primary BC patients. We initially compared quantitative methylation values of both CDO1 and HOPX with prognosis in 133 primary BC patients who were informative for methylation data of both genes (9, 10). Clinicopathological characteristics of the 133 BC patients of this study are shown in Table I. For prognostic analysis, we used 3 kinds of cut-off values (high value, median value, and low value).
For CDO1, almost all cut-off values showed a statistical significance of prognosis between hypermethylation group and hypomethylation group by log-rank plot analysis, so we herein used a median value of 58 as definitive prognostic cut-off value as in the previous studies (10), where BC patients with CDO1 hypermethylation (n=62) showed significantly worse prognosis than those with CDO1 hypomethylation (n=71) (p=0.009) (Figure 1A). Additionally we used high cut-off value of 212.9 and low cut-off value of 42.5, leading to change of patient numbers distributed between hypermethylation/hypomethylation (n=13/120, 80/53, respectively). The high and low cut-off values were determined by plotted p-value and relative risk according to the log-rank plot analysis for TaqMethVs (10).
On the other hand, for HOPX, two kinds of very unique cut-off values showed outstanding relevance of prognosis between hypermethylation group and hypomethylation group by log-rank plot analysis, so we also herein used high value of 24 and low value of 4.3 as definitive prognostic cut-off value as in the previous press (9). BC patients with HOPX hypermethylation (n=45) showed worse prognosis than those with HOPX hypomethylation (n=88) (p=0.121) based on the cut-off value of 24, while HOPX hypermethylation (n=25) showed worse prognosis than HOPX hypomethylation (n=108) (p=0.229) based on the cut-off value of 4.2 (Figure 1B). Furthermore, as analyzed by the median cut-off value in CDO1, we used median value as a cut-off value of 16.9 in HOPX, leading to change of patient numbers distributed between hypermethylation/hypomethylation (n=60/73) (9). Prognostic relevance of HOPX methylation status was eliminated in primary BC, putatively due to smaller numbers tested in the present study as compared to the early reported one (9).
A Cox proportional hazards model was employed to conduct a multivariate prognostic analysis (Table II). In univariate analysis, high pT (p=0.002), pN factor (p<0.001), positive hormone receptor (p=0.006), Ki-67 (p<0.001), subtype (p=0.002), post-operative adjuvant therapy (p=0.003), anthracycline chemotherapy (p=0.006), hormone therapy (p=0.002) and CDO1 methylation (cut-off value was 58, p=0.009) showed significant prognostic factors. In multivariate analysis, Ki-67 positivity (p<0.001) and high TaqMeth V of CDO1 (p=0.016) were independent prognostic factors related to DSS in primary BC.
The association of methylation values between CDO1 and HOPX in 133 primary BC patients. We then investigated the association of methylation values between CDO1 and HOPX in 133 primary BC tissues, because previous studies pointed out CpG island methylator phenotypes (CIMP) (16). The cases were arranged in descending order for the TaqMeth V of CDO1 (upper panel of Figure 2A) and those of HOPX were shown in the cases which are same with CDO1 (lower panel of Figure 2A). There was a significant association of the TaqMeth Vs between CDO1 and HOPX (p=0.002, r2=0.072), however the association was unexpectedly weak (Figure 2B). If the promoter DNA methylation status was delineated by the median values, the methylation status of both CDO1 and HOPX was highly associated (p<0.001) (Figure 2C). In conclusion, methylation status of both genes was sure to be associated with each other, but the association was not so strongly paralleled. Therefore, prognostic relevance of both genes should be separately debated.
The promoter DNA methylation of CDO1 predicts poor prognosis in primary BC. We further analyzed prognosis in 133 BC patients by differential combination of the TaqMeth Vs of CDO1 and HOPX. The BC patients were classified into 3 patterns based on the differential cut-off values (low, median, and high). The 133 BC patients were classified into 4 groups (high/high, low/high, high/low, and low/low methylation status for CDO1/HOPX methylation status, respectively) by each cut-off value. Group I was defined as high/high methylation for CDO1/HOPX. Group II was defined as low/high methylation for CDO1/HOPX. Group III was defined as high/low methylation for CDO1/HOPX. Group IV was defined as low/low methylation for CDO1/HOPX. According to the Kaplan-Meier curves of the 3 patterns, groups I and III reproducibly showed poorer prognosis than otherwise groups (groups II and IV) (Figure 3). These 2 groups (I and III) were always represented by CDO1 hypermethylation. This finding indicated that the promoter DNA methylation of CDO1 is the prognostic marker irrespective of HOPX methylation status.
Hypermethylation of CDO1 predicts worse prognosis even in primary BC with Ki-67 positive, an alternate independent prognostic factor. As mentioned above, positive Ki-67 was a significant prognostic factor independent of CDO1 hypermethylation (median cut-off of 58) in the multivariate prognostic analysis (Table II). We thus examined the survival curve separately in primary BC patients with K-67 negative (n=104) and Ki-67 positive (n=29) according to the median TaqMeth V of CDO1 (Figure 4A). Hypermethylation group of CDO1 showed poorer prognosis than hypomethylation group in both Ki-67 positive cases and negative cases (p=0.159, and p=0.024, respectively) (Figure 4A). We further classified the 133 BC patients into 4 groups by the most important prognostic factors of CDO1 methylation and Ki-67 by differential cut-off values (low, median, and high) of CDO1. Group I was defined as Yes/Yes for CDO1 hypermethylation/Ki-67 positivity. Group II was defined as Yes/No for CDO1 hypermethylation/Ki-67 positivity. Group III was defined as No/Yes for CDO1 hypermethylation/Ki-67 positivity. Group IV was defined as No/No for CDO1 hypermethylation/Ki-67 positivity. Group I was showed the poorest prognosis, while group IV exhibited the best prognosis (Figure 4B). Group I was showed extremely poor prognosis in all patterns according to the low, median, and high cut-off values (p<0.0001 in all cases) (Figure 4B). The promoter DNA methylation of CDO1 could therefore be designated as potent indicator of prognosis in combination with Ki-67 positivity in primary BC.
Tumor suppressive functions of CDO1in human BC cell lines. From the clinical data, hypermethylation of CDO1 was strongly associated with poor prognosis in primary BC. Since such outstanding clinical relevance must be represented by the functional involvement of the CDO1 in aggressive BC phenotype, we examined BC cell lines to assess the expression and function of the CDO1.
CDO1 expression was never observed in 7 BC cell lines by RT-PCR (left panel of Figure 5A, a positive control was HepG2, a hepatocellular cancer cell line, and a negative control was DLD1, a colorectal cancer cell line). First, transient transfection of the plasmid vector with the full-length CDO1 into the 7 BC cell lines such as MCF-7, SK-BR3, YMB-1, CRL, YMB-1E, MDA-MB231 and MDA-MB453 confirmed CDO1 expression at mRNA level by RT-PCR (right panel of Figure 5A).
Overexpression of CDO1 into YMB-1, YMB-1E and CRL cells showed a significantly decreased colony formation compared to the counterparts (mock) in anchorage-independent colony formation assay (left panel of Figure 5B). The reduction rate of the colonies was shown, compared to the mock cells as 100% (right panel of Figure 5B). The reduction of the colonies was confirmed in almost BC cell lines except MDA-MB231 cells. MDA-MB453 cells were not formed a colony in nature.
In addition, overexpression of CDO1 also suppressed the log-phase growth in proliferation assay, especially in YMB-1E cells (p=0.021) (left panel of Figure 5C). On the other hand, overexpression of CDO1 did not show significantly suppressed activity in matrigel invasion assay (right panel of Figure 5C). These results indicated that CDO1 has a definite tumor suppressive activity, however the strength of the tumor suppressive function was weaker than expected from the prognostic relevance in primary BC.
Discussion
There have been numerous reports describing epigenetic prognostic factors of BC (17, 18), however a final validation has never been performed, and definitive prognostic factors are still highly demanded in addition to hormone receptor, HER2, and Ki-67. In the present study, we compared the 2 potential epigenetic prognostic markers of CDO1 hypermethylation and HOPX hypermethylation using the same BC samples, and the final focus was given on CDO1 hypermethylation. Surely, methylation status of both genes are closely associated to each other, however complete matching was not confirmed, so their prognostic relevance should be separately debated. In any combination of the cut-off values, HOPX methylation could not stratify BC patient prognosis instead of CDO1 methylation.
Ki-67 is a well-known prognostic factor in BC, and in our study, it was again proved to be the most potent independent prognostic factor as well as CDO1 hypermethylation in BC. This result is consistent with world-wide consensus of prognosis of BC (19), and recapitulated our previous finding in terms of Ki-67 (20). In our early report, CDO1 hypermethylation was a potent prognostic indicator in triple-negative BC, while in this study, CDO1 hypermethylation was for the first time proven to be a potent prognostic indicator in BC even with positive Ki-67. Triple-negative BC included more patients with Ki-67-positive than otherwise patients (20), so CDO1 hypermethylation may represent excellent prognostic factor even in aggressive BC.
Promoter DNA of CDO1 has been frequently found methylated in various human cancers, and prognostic relevance of its hypermethylation has been extensively reported in breast cancer (10), esophageal SCC (21), esophageal adenocarcinoma (22), gallbladder cancer (23), colorectal cancer (24) from our laboratory, and breast cancer (13, 25), prostate cancer (26), renal cancer (27), and lung cancer (28) from other research groups. These indicated that CDO1 hypermethylation is a common landmark explaining dismal prognosis of human cancer.
We wanted to investigate how much CDO1 is involved in metastatic phenotypes of BC, and CDO1 was transfected in seven kinds of BC cell lines. It actually suppressed anchorage-independent growth in several BC cell lines, suggesting a definite involvement in cancer metastatic ability. On the other hand, it did not affect the invasive capacity of BC cells. These functionally modest findings were not as expected, based on the prognostic relevance of CDO1 methylation in BC, since they were considered ideal prognostic biomarkers with perfect prognostic profiles correlated with CDO1 methylation status (10). The higher CDO1 methylation the tumors harbored, the poorer prognosis the patients exhibited, so almost all cut-off values exhibited statistically significant difference between CDO1 hypermethylation and CDO1 hypomethylation (Figure 3), and risk of death increased as methylation values of CDO1 became higher (10). Allowing for such discrepancy between robust prognostic relevance and modest functional involvement, CDO1 hypermethylation may represent the time from cancer initiation, which could be designated as cancer clock, rather than its functional involvement.
Clinical utility of the CDO1 hypermethylation can be approved as a prognostic marker of BC, just after it could be validated in specific clinical (homogenous) conditions. CDO1 hypermethylation was actually demonstrated to be a significant prognostic factor in BC with lymph node metastases (13), with anthracycline treatment (25), with triple-negative (10), or with Ki-67-positive. Current therapeutic strategies against BC are based on subtype, so prognostic relevance of CDO1 methylation must be validated according to the subtype of BC such as luminal type, HER2 type, and triple-negative type. For luminal type BC, prognosis is excellent in terms of the current clinical therapy, however its incidence is the largest among BC and BC patients who have recurrent disease and finally died largely included luminal type. So, prognostic factors are still highly demanded. In the present study, BC patients even with negative Ki-67 showed marginal difference according to CDO1 methylation status. Future validation is still anticipated regarding prognostic relevance for luminal type BC.
In conclusion, CDO1 hypermethylation is a definite tumor suppressor gene, while its prognostic relevance was more than expected in the context of its functional relevance. The prognostic relevance of CDO1 hypermethylation in the context of other cancers may indicate time since cancer initiation.
Acknowledgements
Not applicable.
Footnotes
Authors' Contributions
Conception and design: YT, K. Yamashita, MW; Acquisition of data: YT, MW, K. Yokota, HH, TK, MK, NM, HN, HK; Analysis and interpretation of data: YT, YK, K. Yamashita; Drafting the article: YT, K. Yamashita; Critically revising the article for important intellectual content: YT, K. Yamashita, MW; Final approval of the version to be published: YT, K. Yamashita, MW; All authors read and approved the final manuscript.
Conflict of Interests
The Authors declare that they don't have any conflicts of interest.
- Received March 25, 2019.
- Revision received April 15, 2019.
- Accepted April 17, 2019.
- Copyright© 2019, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved