Open EZH2 Single Nucleotide Variants (SNVs): Diagnostic and Prognostic Role in 10 Solid Tumor Types

: The enhancer of zeste homolog 2 (EZH2) gene encodes a histone methyltransferase that is a catalytic subunit of the Polycomb repressive complex 2 (PRC2) group of proteins that act to repress gene expression. The EZH2 locus is rarely mutated in solid tumors and there is no comprehensive study of EZH2 single nucleotide variants (SNVs) associated with cancer susceptibility, prognosis and response to therapy. Here, for the ﬁrst time, we review the functional roles of EZH2 DNA variants and propose a putative etiological role in 10 various solid tumors including: esophageal, hepatocellular, oral, urothelial, colorectal, lung and gastric cancers. In particular, we found that the C allele of the EZH2 variant rs3757441 is associated with increased EZH2 RNA expression and poorer prognosis (advanced stage) in at least two malignancies such as colorectal and hepatocellular carcinoma. This suggests that the C allele may be a functional risk variant in multiple malignant tumors. We therefore propose that the rs3757441 single nucleotide variant (SNV) be genotyped and real-time PCR assays be performed in large cohort studies in order to conﬁrm this preliminary ﬁnding that could be useful for clinical practice.

The EZH2 gene located on chromosome 7q36 encompasses 20 exons, 19 introns, and encodes 2 functional isoforms (EZH2α and EZH2β) [22]. Within the gene, 41 single nucleotide variants (SNVs) have been identified most of which are in non-coding regions [2]. Emerging evidence indicates that EZH2 SNVs influence tumor initiation, progression and therapeutic response. Therefore, EZH2 DNA variants could be used in diagnostics as screening tools (genetic markers) for various cancers, and the gene could be used as a target for novel, personalized drug therapy to enhance cancer care management [23]. In this review, we aim to discuss the functional and putative etiological roles of EZH2 DNA variants in solid tumors.

EZH2 SNVs Associated with Cancer Susceptibility
Most clinical studies on EZH2 SNVs suggest an association with risk of developing a particular type of cancer (Table 1), but very few have investigated EZH2 protein levels and/or activity. The most studied EZH2 SNVs to potentially predict cancer susceptibility are rs6950683 and rs3757441; carrying the risk T alleles of either rs6950683 or rs3757441 reduces the risk (Odd Ratio: 0.7) of developing oral squamous cell carcinoma susceptibility (OSCC) [24] and particularly small-cell lung cancer (SCLC), when compared to non-small-cell lung cancer patients [25]. Patients carrying the protective C allele at rs3757441 and rs6950683 also showed lower risk of developing hepatocellular carcinoma (HCC) and urothelial carcinomas (UCC), respectively [26,27]. At odds with these studies, the rs3757441 C allele was significantly associated with higher risk of developing ESCC and with higher advanced pathological staging in these tumors [28].
Individuals carrying at the G allele at rs2302427 showed a lower risk of UCC [27]. However, the same genotype at rs2302427 has been associated with higher risks of HCC [29]. Several other EZH2 SNVs have been associated with the risk of developing a specific cancer type in single studies [30,31], ( Table 1).
All of these studies indicate that EZH2 SNVs could be useful biomarkers to predict individual susceptibility to specific cancer types. However, since the effect of each SNV seems to be population and cancer and/or tissue specific, large cohort studies are needed, and more evidence on the functional roles of these SNVs.

EZH2 SNVs Associated with Prognosis and Response to Therapy
Several of these EZH2 SNVs have been associated with cancer prognosis ( Table 2). The C alleles of rs6950683 and rs3757441 may predict a higher risk of developing lymph-node metastasis in HCC [26]. In two independent studies, we found that the C/C genotype at rs3757441 was significantly associated with shorter progression free survival (PFS) and overall survival (OS) (p < 0.01 and p < 0.05, respectively) in metastatic colorectal cancer (mCRC) patients [32,33]. Interestingly, we also found that the C/C genotype (rs3757441) was predictive of poorer response to chemotherapy in mCRC patients. The same genotype was also associated with increased tumor size, poorer differentiation, higher T stage and higher pathological stage in ESCC patients [28].
Our bioinformatic analyses indicate that the rs3757441 C allele creates a binding site for the X-Box binding Protein-1 (XBP1) transcription factor [32]. Since the XBP1 transcription factor is associated with an enhancer [34] we have hypothesized that the C allele induces higher EZH2 expression. This hypothesis is corroborated by our findings in colorectal cancer [32] and blood cells, showing that the CC genotype is associated with significantly higher EZH2 expression [35]. Further molecular studies are needed to definitively test this hypothesis. The rs3757441 C allele is also associated with EZH2 and H3K27me3 up-regulation in primary CRC specimens [35]. Based on these results, it is conceivable that the rs3757441 CC genotype triggers higher EZH2 expression and therefore poorer cancer prognosis, at least in colorectal cancer (CRC) and ESCC cells (Figure 1). It remains to be determined why the same genotype reduces cancer risk in other neoplasms. Notably, this contradictory pattern suggests that the combinations of low penetrance genes make EZH2 either a risk or protective gene, depending on the particular genetic background. Few studies have demonstrated a possible prognostic role for other EZH2 SNVs. For example, the G allele of rs2302427 has been associated with lower tumor stage in UCC patients [29]. The rs887569 TT genotype was correlated with a significantly longer OS in gallbladder cancer patients, [36] (Table 2).
These data indicate that EZH2 SNVs could predict prognosis and response to therapy in specific cancer types. Table 2. EZH2 SNV associated with prognosis and response to therapy. We reported protective/risk alleles wherever possible. When the heterozygous genotype showed higher or lower risk than both homozygous genotypes, we reported full genotypes. Cancer types are abbreviated (UCC: urothelial cell carcinoma, HCC: hepatocellular carcinoma, OSCC: oral squamous cell carcinoma, mCRC: metastatic colorectal cancer, SCLC: small-cell lung cancer, ESCC: esophageal squamous cell carcinoma risk; PFS: progression free survival; OS: overall survival; CI 95%: confidence interval 95%).

Conclusions
This review has identified 10 EZH2 SNVs that could potentially predict cancer risk and/or prognosis in 10 types of solid tumors. However, without conducting large cohort studies it is uncertain as to whether these variants are functional in cancer progression or in linkage disequilibrium with a functional variant.
The rs3757441 and rs6950683 SNVs are the most promising diagnostic and functional SNVs to-date. Notably, the T allele of both SNVs has been associated with increased risk of developing four different solid neoplasms, including liver and lung cancers. However, the rs3757441 C allele may be associated with more aggressive phenotypes and poorer prognosis in ESCC and mCRC. Bioinformatic data suggest that the C variant is associated with increased binding of the XBP1 transcription factor and therefore with higher EZH2 expression. We believe that the tissue-and disease-dependent role of the rs3757441 could be explained by the fact that the XBP1 transcription factor is more active/expressed in certain tissues (colon, esophagus) than in others (liver, lung). Where the XBP1 transcription factor is sufficiently expressed, the molecular model proposed in Figure 1 is valid. Where XBP1 levels are very low, other factors could mediate the function of the rs3757441 SNV. Hence, the functional role of the rs3757441 needs to be investigated in different population and cellular contexts, to fully explain its tissue-and disease-specific function.
Furthermore, EZH2 somatic mutations (which are beyond the scope of this review) seem to play a key role in some melanomas. Melanoma is the most frequently EZH2-mutated solid neoplasm [37]. Melanoma cells harbor the Y641 activating mutation within the catalytic SET domain that was first described in Diffuse large B-cell (DLBC) lymphoma. Numerous studies have identified an association between aberrant EZH2 expression and melanoma progression [38]. Although EZH2 SNVs have not been comprehensively investigated in this setting, further translational studies are warranted.
In conclusion, we believe that the clinical usefulness of EZH2 SNVs should prompt the development of more studies in two main directions:

1)
Systematic study of the role of all EZH2 SNVs (and their linkage) in specific neoplasms. Ideally, these studies should be prospective and conducted on at least two, independent cohorts if (as we did in [32,33]). 2) Functional studies on the role of clinically relevant SNVs, identified in step 1. These studies should be conducted at least in vitro and in human tissues, to confirm the effects of each variant upon EZH2 expression or other function.
EZH2 inhibitors have been shown to be active against a wide range of human neoplasms, and some of these new drugs are being tested in clinical trials [39,40]. This emerging evidence reinforces the need for identifying reliable and non-invasive biomarkers to predict the expression and function of EZH2 in patients. These novel diagnostic tools could help predict individual patient response to therapy and identify novel drug therapy targets. Our group and others are currently working on translating these findings into useful clinical applications.