Identification of Prognostic miRNAs Targeting EZH2 in Hepatocellular Carcinoma Using The Cancer Genome Atlas Database


 Background: Enhancer of zeste homolog 2 (EZH2) gene have a prognostic role in hepatocellular carcinoma (HCC). This study aimed to identify the prognostic microRNAs (miRNAs) targeting EZH2 in HCC. Methods and Results: We downloaded the gene and miRNA RNA-seq data from The Cancer Genome Atlas (TCGA) database. Differences in EZH2 expression between tumor and control samples and those between tumors with different clinical variables were analyzed using the Mann-Whitney U test. Association of EZH2 expression with prognosis in HCC patients was detected using Cox regression analysis. We also identified miRNAs targeting EZH2 with negative correlations, compared the miRNA expression profiles between tumor and control tissues, and identified pathways and protein-protein interaction pairs related to EZH2. The miRNA-EZH2-pathway network was constructed accordingly. EZH2 was significantly upregulated in HCC tumors compared with control samples (p<0.0001) and in tumors with advanced T classifications (3/4 vs. 1/2, p=0.0039) and stages (III/IV vs. I/II, p=0.0028). The Cox regression analysis showed that TCGA HCC patients who had high EZH2 expression levels showed a short survival time (HR=1.677, 95% CI 1.316-2.137; p<0.0001). Among miRNAs targeting EZH2, seven miRNAs, including hsa-let-7c-5p, were negatively correlated with EZH2 expression and were significantly downregulated in HCC tumor samples compared with controls (p<0.0001). The miRNA-EZH2-pathway network included seven downregulated miRNAs and four pathways, including hsa00310: Lysine degradation. Hsa-let-7c-5p was associated with prognosis in HCC (HR=0.849 95% CI 0.739-0.975; p=0.021). Conclusions: EZH2-hsa-let-7c-5p has a significant association with HCC prognosis and the mechanism worth investigating.


Introduction
Hepatocellular carcinoma (HCC) is the most frequent type of primary liver cancer (comprising 75%-85% of cases) and has a high global incidence and mortality rate globally. HCC-related death is estimated to approximately 662,000 annually [1] and is approximately 781,000 in 2018 [2]. HCC is the second or third most common cause of cancer-related death worldwide [2]. The main risk factors for HCC are infections of hepatitis B virus (HBV), hepatitis C virus (HCV), heavy alcohol intake, smoking, and obesity [2,3].
However, the clinical prognosis of advanced HCC remains poor and had high incidence rates of tumor recurrence and metastasis. Accordingly, identi cation of diagnostic and prognostic factors for HCC remains necessary.
The aim of this study was to identify prognostic miRNAs interplaying with EZH2 in HCC. Association of EZH2 expression with prognosis in HCC patients was investigated. MiRNAs targeting EZH2 were screened from online databases and prognostic miRNAs negatively correlated with EZH2 in HCC patients were then identi ed using integrated bioinformatics analysis. This study provided a novel and important molecular mechanism of HCC development and prognosis. Analysis of EZH2 expression Differences in EZH2 expression levels between tumor samples and controls, as well as between samples had different pathologic stages (I/II vs. III/IV), pathologic T classi cations (1/2 vs. 3/4), ages (< 65 yrs vs.

Data materials
≥ 65 yrs), genders (male vs. female), races (Asian, White, or Black), without and with prior malignancy were analyzed using the non-parametric Mann-Whitney U test or the Kruskal-Wallis H test.

Cox regression analysis for EZH2 expression
The association of EZH2 expression with prognosis in TCGA HCC patients was analyzed using the Cox regression analysis. Also, associations of EZH2 expression pro le and clinical factors (including pathologic stage, pathologic T classi cation, age, gender, race, and prior malignancy) with prognosis in TCGA HCC patients were analyzed using univariate and multivariate Cox regression analysis. We also performed the Cox regression survival analysis to investigate the difference in survival percent between patients with high and low EZH2 expression levels, which were divided using the median expression value.

Expression of miRNAs in HCC tumor samples
Differences in miRNA expression levels between tumor and control samples were analyzed using the nonparametric Mann-Whitney U test. We also analyzed the correlations of miRNAs with HCC prognosis using Cox regression analysis.

Statistical analysis
All the statistical analyses were performed in the SPSS 22.0 software (IBM, Chicago, USA). Differences in patients' age, male ratio, and race were analyzed using the t-test, Chi-square test, and Wilcoxon rank sum test, respectively. Differences in miRNA and EZH2 expression levels between groups were analyzed using the non-parametric Mann-Whitney U test, and those across more than three groups were analyzed using the non-parametric Kruskal-Wallis H test, respectively. We used the Cox regression analysis to identify the association of miRNAs, EZH2, and clinical variables with prognosis in HCC patients. Spearman correlation analysis was conducted to identify miRNAs negatively correlated with EZH2 expression. Hazard ratio (HR) and 95% con dent interval (CI) were calculated following Cox regression and KM survival analyses. For all analyses, the signi cant criterion was set at p = 0.05.

Demographics of TCGA HCC patients
The demographics of TCGA HCC patients (n = 371) and controls (n = 50) are shown in 92.18%), the ratio was high than controls (80.00%, p < 0.0001; Table 1). Among tumor samples, most were at early T, N, and M classi cations, without prior treatment (n = 369) and malignancy (n = 336). The median overall survival time was 598.5 (6.00-3675.0) days.  Fig. 2A) and from Asian patients compared with White/Black patients (p = 0.0059; Fig. 2C). There was no difference in EZH2 expression level between patients with and without prior malignancy (p = 0.1931; Fig. 2C) and between female and male patients (p = 0.4084; Fig. 2D). These results might indicate that EZH2 expression is related to age and race in HCC patients.

EZH2 associates with prognosis in HCC patients
We analyzed the association of EZH2 with prognosis in TCGA HCC patients using the stepwise Cox regression analysis. Univariate Cox regression analysis showed that pathologic T, pathologic stage, and EZH2 expression were associated with overall survival in HCC patients (p < 0.0001;  Table 2). Cox regression survival analysis showed that HCC patients with a high level of EZH2 had a lower survival percent compared with patients who had a low level of EZH2 (Fig. 3). We also observed the association of EZH2 with HCC prognosis in online databases: GEPIA (logrank p = 5.6e-05; Fig. 4A), UALCAN (logrank p < 0.0001; Fig. 4B), and Oncolnc (logrank p = 5.87e-05; Fig. 4C). The results indicated that a high level of EZH2 was correlated with a poor prognosis in HCC patients.

Screening of miRNAs associated with HCC prognosis
At last, we screened out HCC prognosis-associated miRNAs among the seven downregulated miRNAs using Cox regression analysis. Univariate Cox regression analysis identi ed that only hsa-let-7c-5p was associated with prognosis in HCC patients (HR = 0.849, 95% CI 0.739-0.975; p = 0.021; Table 5). These results showed that hsa-let-7c-5p-EZH2 might be an important miRNA-mRNA axis in HCC. HR, hazard ratio. CI, 95% con dent interval.

Discussion
Our present study con rmed that the EZH2 gene had a higher expression level in tumors compared with controls and in advanced tumors compared with early-stage tumors. Hsa-let-7c-5p was upregulated in HCC tumors compared with controls and its expression was associated with a good prognosis in HCC patients. These results showed that the hsa-let-7c-5p-EZH2 axis might have a crucial role in the progression of HCC.
EZH2 regulates downstream genes and signalings in a PRC2-independent manner [13]. EZH2 functions as both an oncogenic gene or as a tumor suppressor gene by activating its downstream target genes and signalings through a PRC2-independent way [13]. For instance, EZH2 directly binds to the promoter of the large tumor suppressor 2 (LATS2) gene and induces its H3K27me3 in gallbladder cancer cells [26]. EZH2 also decreases forkhead box C1 (FOXC1) expression by promoting H3K27me3 in breast cancers [27]. LATS2 is a member of the large tumor suppressor family [28,29] and the FOXC1 transcription factor also as an oncogenic gene by regulating cell proliferation, senescence, angiogenesis, and metastasis [30,31]. This evidence reveals the important role of EZH2 in human cancers.
Recent evidence shows that EZH2 regulates immune responses in human cancers, including HCC [4,13,32,33]. Elevated EZH2 level was positively correlated with immunosuppression in HCC and was negatively correlated with the contents of Class I major histocompatibility complex molecules [4]. However, loss or knockdown of EZH2 in regulatory T cells and natural killer cells enhance their recruitment and anti-tumor immunity [34,32]. Therefore, many efforts have been made to inhibit EZH2 methyltransferase activity, break PRC2' structure, suppress EZH2 expression, or develop EZH2 inhibitors with low toxicity, high e ciency, and high selectivity in cancer treatment.
Song et al [46] con rmed the downregulation of hsa-let-7c-5p in HCC tumor tissues. They showed that a high level of hsa-let-7c-5p was correlated with a long overall survival period. Li et al.
[47] con rmed that hsa-let-7c-5p and EZH2 were downregulated and upregulated in breast cancer tissues compared with controls, respectively. They also showed that the inhibition of hsa-let-7c-5p increased EZH2 expression in MDA-MB-231 breast cancer cells. However, there is poor information on the regulation of hsa-let-7c-5p on EZH2 and the axis in HCC. Therefore, identi cation of the hsa-let-7c-5p-EZH2 axis might provide a novel and reference to HCC mechanism. Also, the strategy focusing on inhibiting EZH2 might provide a reference for the treatment of HCC.

Conclusions
Our study showed that the EZH2 gene was upregulated in HCC tumor samples and its level was correlated with prognosis in HCC patients negatively. Also, EZH2 was negatively targeted by hsa-let-7c-5p, which had a lower level in HCC tumors compared with control and a positive correlation with overall survival in HCC patients. However, important roles of the hsa-let-7c-5p-EZH2 axis in HCC development and prognosis should be validated by experiments. The expression pro le of EZH2 in hepatocellular carcinoma tumors. A, difference in EZH2 level between tumor and normal control tissues. B-C, differences in EZH2 levels between tumor tissues with different pathologic T classi cations and stages, respectively. Differences were analyzed using the non-parametric Mann-Whitney U test.

Figure 2
Expression pro le of EZH2 in hepatocellular carcinoma tumors from patients with different clinical variables. The difference in EZH2 expression level tumors samples from patients aged over and younger 65 years (A), with and without prior malignancy (B), Asian, White, and Black (C), and male and female (D).
Differences were analyzed using non-parametric Mann-Whitney U test (in gure A, B and D) and the Kruskal-Wallis H test (in gure C).

Figure 3
Cox regression survival analysis for EZH2 expression in hepatocellular carcinoma. The TCGA cohort (n=371) was used for this analysis. HR, Hazard ratio. CI, 95% con dent interval.

Figure 6
The miRNA-EZH2-pathway network in hepatocellular carcinoma. Upregulated EZH2 is shown by red node and downregulated miRNAs are indicated by green diamonds. Pathways and EZH2-associated genes are presented by empty squares and circles, respectively.