Colorectal cancer (CRC) is the third most common cancer in worldwide contributing to 10% of the total new cancer cases and 9% of cancer deaths in 2012 (Ferlay et al., 2015). The high incidence and mortality demonstrate the global CRC burden. The disease burden is similar in Malaysia where CRC is the second most common cancer diagnosed among males and third common cancer in females. According to the National Cancer Registry (NCR) Report in the year 2011, 2246 CRC cases were reported in 2007, contributing to 12.3% of total cancer cases (Zainal Ariffin and Nor Saleha, 2011). Of these, 1185 cases were males and 1101 cases females.

To date, various treatment strategies have been used to treat CRC depending on the stage at which cancer was discovered. In the early stages of CRC (Stages I and II), surgery is the best option (Holen and Chung, 2008). However, for advanced stages such as stage III and stage IV, treatment often consist of a combination of therapies such as surgery, chemotherapy and/or radiotherapy to minimize the risk of recurrence (Hayat, 2009). Managing CRC remains a clinical obstacle, as more than 15% of cases are resistant to 5-Fluorouracil (5-FU)-based chemotherapeutic regimens which is the mainstay of treatment, and tumor recurrence rates can be as high as 50–60% (Shakibaei et al., 2014). Furthermore, most of recurrence cases occur in the first 3 years after a surgery and the majority started to recur in the second year (Aghili et al., 2010). Various factors have been proposed to play a significant role in CRC recurrence and one of the factors is chemotherapy resistance (Kanwar et al., 2012).

Epigenetic alterations have been suggested as one of the underlying mechanisms of chemotherapy resistance (Zeller and Brown, 2010). One major epigenetic mechanism involved in the progression of cancer is DNA methylation which refer to the enzymatic addition of a methyl group to the 5′ position of cytosine ring by DNA methyltransferase (DNMT) to produce 5-methylcytosine (Bestor, 2000). Altered DNA methylation patterns can influence the expression of the genes by silencing or activating the genes. In general, there are two types of aberrant DNA methylation that could contribute to the development and progression of cancer which are global DNA hypomethylation and CpG island hypermethylation (Wajed et al., 2001).

The term DNA hypomethylation refers to global decrease in the level of methylation in the genome of tumor cells in comparison to normal tissue (Lao and Grady, 2011). On the contrary, hypermethylation is an increase of DNA methylation level at CpG island as compared to normal tissue. Methylation of CpG islands within the promoter of CRC results in silencing of tumor suppressor genes and promote tumor formation (Herman and Baylin, 2003). Methylation of CpG islands at particular genes may also contribute to the acquisition of drug resistance. This was demonstrated by Arnold et al. (2003) whereby methylation of CpG island in the MLH1 gene is associated with resistance to 5-FU and development of chemoresistance could increase the incidence of recurrence CRC. However, this process is reversible and could be achieved by using demethylating agent such as 5-aza-2′-deoxycytidine (5-azadC). Therefore, identification of DNA methylation patterns in cancer may be helpful in predicting prognosis and response to chemotherapeutic agents. In addition, to the best of our knowledge there have been very limited studies which investigate global DNA methylation profile in recurrent CRC. Moreover, at the time this research was conducted there is no study which investigate the global methylation changes in in vitro CRC model upon treatment with combination of 5-FU with 5-azadC. Thus, the aim of the study was to investigate the methylation phenotype in CRC on the global scale using the high density Infinium DNA Methylation 450K DNA assay in order to identify predictive markers for chemotherapy response. The second aim was to investigate the magnitude of 5-azadC in enhancing 5-FU chemosensitivity and to determine the global methylation changes in cells treated with 5-FU, 5-azadC or combination of both agents.

Many studies have shown that DNA methylation is an important underlying mechanism for drug resistance which subsequently posed the increased risk of recurrence. Despite a growing consensus that methylation is a modulator of cancer, the understanding of DNA methylation patterns of recurrent CRC remains limited. To the best of our knowledge, there is no publication addressing the genome-wide DNA methylation analyses of recurrent CRC using the Illumina Infinium HumanMethylation450 Beadchips. In our study, we identified 4787 significantly differentially methylated genes (P < 0.05) on methylation profiling of 43 non-recurrent and five recurrent samples; 3112 genes were hyper- while 1675 genes were hypomethylated. Our results showed that recurrent CRC exhibit increased methylation levels compared to non-recurrent group. Similar patterns have been observed in chordomas where higher methylation level were observed in recurrent chordomas compared to non-recurrent (Alholle et al., 2015).

Most of the hypermethylated genes were commonly observed in the CpG islands which are associated with promoter regions. Our results are supported by a review from Jin et al. (2011) where most studies showed hypermethylation occurring at CpG-rich promoter regions. By contrast, hypomethylation often occurs in the open sea area of the genome. McCabe et al. (2009) reported that hypomethylation were observed in the intergenic region as well as the gene body (Zheng, 2015) in which these two regions are located in the open sea area (Ching et al., 2015). Pathway enrichment analysis of hyper- and hypomethylated genes revealed that most of the hypermethylated genes of recurrent CRCs were involved in the MAPK signaling pathway which specifically targets the apoptosis process. Cells which are able to escape from the apoptosis process are associated with chemotherapy resistance thus will increase the risk of recurrent cancer (Kasibhatla and Tseng, 2003). In a few studies, TGFβ, CASP3, DAXX, and p38 were found to be hypermethylated and these genes are known to regulate the apoptosis process (Wagner and Nebreda, 2009; Lee et al., 2012; McIlwain et al., 2013; Pan et al., 2013). This is supported by Perry et al. (2006) where most of the hypermethylation in cancer occurred in the genes that regulate the apoptosis process and other cellular functions. On the contrary, the hypomethylated genes of recurrent CRCs are involved in the pathways in cancer particularly in regulating the cell proliferation process and PI3K-AKT signaling pathway. Among the genes involved in the proliferation process and signaling pathway are PPFP, RXR, EGFR, PI3K, CCNA1, and E2F (Balasubramanian et al., 2004; Yamazaki et al., 2005; Rolland, 2006; Yuan and Cantley, 2008; Webster et al., 2009; Reddi et al., 2010). Hypomethylation could increase the expression of these genes (Ehrlich, 2009) leading to the activation of PI3K-AKT signaling pathway and proliferation process. The activation of this pathway may increase the cell survival, hence resulting in the increased risk of recurrence.

Three hypermethylated genes (CCNE1, CCNDBP1, and PON3) and two hypomethylated genes (DDX43 and CHL1) were selected for validation. These genes are involved in cell cycle, cell proliferation and cell growth. CCNE1 belongs to the highly conserved cyclin family and is an important component in the cell cycle regulation. CCNE1 forms a complex with CDK2 and is involved in the transition of G1 to S phase as well as in the DNA replication (Hwang and Clurman, 2005). Increased expression of this gene can cause chromosomal instability and lead to the formation of cancer such as breast cancer, leukemia, non-small lung cancer, and others (Moroy and Geisen, 2004). However, in this study we found that CCNE1 is hypermethylated and its expression was downregulated in recurrent CRCs compared to non-recurrent. This finding is in concordance with the study by Kleivi et al. (2007) where CCNE1 expression was lower in the carcinomatosis group compared to primary cancer. The decreased expression of this gene from primary CRC is associated with poor prognosis, advanced stage and metastasis (Li et al., 2001). Apart from that, a study from Nakayama et al. (2010) showed that increased expression of CCNE1 in primary ovarian cancer was associated with chemotherapy resistant and increased risk of recurrence. Therefore, we believed that hypomethylation of this gene in non-recurrent CRCs is associated with chemotherapy resistant which subsequently contribute to recurrent CRCs. However, further investigation is warranted to confirm this hypothesis.

Cyclin D type binding protein 1 (CCNDBP1), also known as GC1P, is a tumor suppressor gene that is downregulated in some cancers such as breast and prostate (Mostoslavsky et al., 2006; Michishita et al., 2008). In colon cancer, the expression of this gene is lower in tumor tissues compared to normal (Ma et al., 2007). Our study showed that CCNDBP1 is hypermethylated in recurrent CRCs compared to non-recurrent, which could possibly explain the lower expression of this gene. This is the first report of hypermethylation of CCNDBP1 in recurrent CRCs. Paraoxonase 3 (PON3) belongs to the paraoxonase families (PON1, PON2, PON3) that helps in preventing oxidative stress and anti-inflammatory. This gene is also involved in other diseases including cancer (Devarajan et al., 2014). PON3 gene has high expression in cancer tissues of the lung, liver and colon (Schweikert et al., 2012). A study by Schweikert et al. (2012) showed that the expression level of PON3 is different in various stages of cancer. Stage I exhibit decreased expression of PON3 while stage II and III display increased expression of this gene (Schweikert et al., 2012). In our study, we identified hypermethylation of PON3. This is in line with the study by Alholle et al. (2015), where PON3 was hypermethylated in recurrent chordomas compared to the primary tumor. DDX43, also known as helicase antigen (HAGE), is responsible for cell proliferation and has a high expression level in cancer compared to normal (Abdel-Fatah et al., 2014). Our microarray profiling shows that DDX43 is hypomethylated in recurrent CRCs. Hypomethylation of DDX43 induces the expression of the gene and is strongly correlated with advanced disease and poor prognosis in leukemia (Lin et al., 2014). DDX43 expression is increased in 50% of acute myeloid leukemia cases (Adams et al., 2002) and high expression will activate the Ras signaling pathway thus promoting cell proliferation (Ambrosini et al., 2014). The dysregulation of this process may cause the progression of cancer such as metastasis (Linley et al., 2012).

Close Homolog of L1 (CHL1) is important for brain development and neuron activities (Manderson et al., 2009). It also plays a role in cancer progression (Qin et al., 2008) and contributes to the metastasis process (Siesser and Maness, 2009). In recurrent CRCs, we identified hypomethylation of this gene compared to non-recurrent CRCs. Hypomethylation of CHL1 was also observed in metastatic oral squamous cell carcinoma (OSCC) compared to non-metastatic tumor (Huang et al., 2013). A previous reported the low expression of CHL1 in majority of primary cancers and its expression is increased in the metastatic or invasive tumor (Senchenko et al., 2011).

The potential of demethylating agent such as 5-azadC in sensitizing resistant cancer cells to chemotherapeutic agent in CRC in vitro model has already been studied by several groups (Fang et al., 2004; Mossman et al., 2010; Khamas et al., 2012; Flis et al., 2014). Nevertheless, the global methylation changes following the treatment is not well-characterized and the aforementioned studies pre-selected certain genes for further investigation. The action of 5-azadC is not gene-specific and vary across different cell lines thus it is important to know which genes or pathway will be affected by this agent. Furthermore, the chemotherapeutic agent used by previous studies was Trichostatin A (TSA) (Fang et al., 2004; Mossman et al., 2010) which is not commonly practiced by our oncologist. To be more relevant with our population, we investigated the effect of 5-FU chemosensitization by 5-azadC in hypermethylated and hypomethylated cell lines and profiled the global methylation changes caused by the agents whether singular or in combination. We observed that the reduction in cancer cell viability is significantly greater in SW48 compared to HT29 when treated with both 5-FU and 5-azadC. This is in concordance with other study which showed that there was an increase in inhibition of CRC cells growth after combinatorial treatment with both agents (Flis et al., 2014).

SW48 cells treated with combined drugs showed decrease in methylation level, particularly in genes that involved in MAPK signaling pathway which specifically target p53 signaling pathway. P53 is a tumor suppressor gene that helps in regulates apoptosis process and cell cycles (Nieto et al., 2004). DNA damage induced by chemotherapy have been shown to activate p53 signaling pathway which can trigger the apoptosis process. Our study revealed demethylation of genes involve in p53 signaling pathway in SW48 which then could possibly induce the activation of these genes. As these genes become activated, the presence of 5-FU will cause DNA damage and therefore p53 signaling pathway could be re-activated and lead to apoptosis process. Hypomethylation of genes involved in Fas mediated apoptosis pathway which were FAS, DAXX, ASK1, and MAX was also observed in SW48 treated with combined drugs. DAXX gene will bind to FAS death domain, and overexpression of DAXX will initiates Fas-induced apoptosis through the activation of ASK1, MKK6, p38, and MAX genes (Yang et al., 1997).

This study has several obvious limitations. First, the number of recurrent CRC cases used in this study is small thus we compensated with increasing the sample size of the non-recurrent CRCs and performed genome-wide profiling instead of targeting specific genes. Secondly, we did not manage to obtain a matched primary and recurrent cancer tissues from the same patient. It would be ideal to perform the comparison using the cancer tissues from the same individual.

To the best of our knowledge, our study is the first to address the genome wide methylation changes in recurrent CRCs and non-recurrent CRCs. The increased in global methylation levels were observed in the recurrent cases. Our study revealed several potentially biologically important genes – CCNE1, CCNDBP1, PON3, CHL1, and DDX43 that may represent new potential therapeutic targets for patients with chemoresistant phenotype. Our in vitro findings suggest that 5-azadC-mediated restoration of 5-FU sensitivity is mediated at least in part by MAPK signaling pathway. Further investigation is needed in order to completely understand the involvement of these genes in the cancer recurrence.

This study was carried out in accordance with the recommendations of Universiti Kebangsaan Malaysia Research Ethics Committee (Reference number: UKM 1.5.3.5/244/UMBI-001-2014) with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Universiti Kebangsaan Malaysia Research Ethics Committee.

RB performed the experiment as described in the manuscript, statistical analysis, interpretation of the data and drafted the manuscript. N-SAM performed interpretation of the data and gave critical insight to the manuscript. SO assisted in the analysis of the methylation profiling data. IS is a colorectal surgeons which assisted in specimen retrieval. IR is a pathologist which evaluated the histopathology of the specimens. NM and RJ are mentors, assisted in coordination of the study and gave critical evaluation of manuscript. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.