A functional variant in the RAD51 3' UTR is associated with survival of hepatocellular carcinoma patients.

The RAD51 gene plays an important role in DNA repair by homologous recombination, and is involved in the development and progression of multiple cancers. Single nucleotide polymorphisms in RAD51 have been previously described to impact the prognosis of patients with cancers, however, it is still unclear whether this is also true for hepatocellular carcinoma (HCC). This study therefore aimed to identify genetic variants in RAD51 and determine the effect on the survival of patients with HCC. In this study, we performed genotyping assays for RAD51 polymorphisms in a cohort of 368 patients with HCC who had underwent hepatectomy. Using multivariate cox proportional hazards model and Kaplan-Meier analyses with log-rank tests, we compared the survival of patients with HCC according to RAD51 SNP genotypes. We identified one potential functional variant, rs12593359, located in a microRNA (miRNA) binding site in the RAD51 3' untranslated region, to be an independent predictor of overall survival of patients with HCC in the dominant model. Patients carrying GT/TT genotypes had a significantly increased risk of death when compared with those carrying GG genotype (adjusted hazard ratio = 1.34, 95% confidence interval = 1.02-1.76, P = 0.035). Kaplan-Meier curve analysis showed a markedly shorter survival time for patients with HCC carrying GT/TT genotypes of SNP rs12593359 (19.0 months vs. 36.0 months; Plog-rank = 0.012). Notably, this effect was particularly pronounced in several subgroups of patients (e.g., males, Hepatitis B virus-positive patients, patients with a single tumor nodule, patients with alpha-fetoprotein (AFP) < 400 ng/ml, or patients who were cancer embolus-free). Additional expression analysis of quantitative trait loci showed that SNP rs12593359 was significantly associated with RAD51 mRNA expression levels in 483 cell-cultured fibroblasts (P = 1.1×10-4). These findings provide evidence that RAD51 rs12593359 is associated with HCC survival and may serve as a promising predictor of survival in patients with HCC.


Introduction
Liver cancer was the fourth leading cause of cancer-related deaths worldwide in 2018, with approximately 782,000 deaths occurring annually [1] . Hepatocellular carcinoma (HCC) accounts for 75-85% of all primary liver cancer cases [2] . Since more than 50% of HCC cases are diagnosed in the late stages of the disease, HCC prognosis has only marginally improved over the time, with a 5-year survival rate of just 18.1%, making HCC the second most lethal tumor after pancreatic cancer (which has an 8.5% 5-year survival rate) [3] . Therefore, it is imperative to improve our ability to predict which patients are likely to have a poor survival outcome. Clinically, the known characteristics of HCC such as tumor size, serum alpha-fetoprotein (AFP) levels, and Barcelona Clinic Liver Cancer (BCLC) stage are commonly used to predict the prognosis of patients with HCC [4,5] . However, these histopathological features of primary tumors do not provide su cient information for assessing tumor malignancy. The response of individuals to treatments such as hepatectomy is heterogeneous, suggesting that genetic factors may also play an important role in the prognosis of this cancer and investigating such molecular mechanisms could reveal promising prognostic or predictive factors for HCC [6][7][8] .
Homologous recombination (HR) plays an important role in tumorigenesis and the proteins involved in HR protect chromosomes against endogenous damage affecting both DNA strands, such as doublestrand breaks [9] . They maintain genomic stability by supporting DNA replication and repairing DNA damage, which helps cells to avoid gene mutations [10] . RAD51 is one of the most important human HR genes. It is located on chromosome 15 at position q15.1 and encodes a product of 339 amino acids [11] . RAD51 performs an irreplaceable function by forming nucleoprotein laments in single-stranded DNA, which mediates homologous pairing, and strand exchange reactions between single-and doublestranded DNA during the repair process [12] .
Several studies have provided evidence demonstrating that abnormal RAD51 expression is associated with poorer outcomes in several tumor types [13][14][15][16] . For example, RAD51 overexpression was shown to correlate with malignant phenotypes of colorectal cancer and may predict poor prognosis for these patients [17] . Elevated expression of RAD51 has been associated with decreased survival of patients with non-small-cell lung cancer, which may be due to enhanced tumor cell survival, anti-apoptotic activity, and chemo-/radioresistance [18] . In vitro studies have demonstrated that the abnormal increase of RAD51 expression favors tumor progression by inhibiting caspase-3-mediated apoptosis, as well as by interfering with other intracellular pathways, including p53, p21 and Bcl-2, which may cause improper repair of damaged DNA or further amplify the damage of sites in a recombinant DNA strand [19] . In addition, a previous study showed that miR-103 was able to dramatically enhance the chemosensitivity of U2OS osteosarcoma cells by speci cally targeting the 3′ untranslated region (3′ UTR) of RAD51 to reduce its expression [20] . Recently, many reports have also demonstrated that RAD51 polymorphisms, located in potential functional sites, could modulate its expression and function, thereby affecting cancer development and progression [21][22][23] .
MicroRNAs (miRNA) are a class of evolutionarily conserved small noncoding RNAs that are known to regulate gene expression by binding to the 3′ UTR of target mRNAs, leading to mRNA cleavage and/or translational repression [24] . MiRNAs are proposed to regulate the expression of more than 30% of the protein-coding genes in the human genome [25] . However, the binding process can be in uenced by SNPs within miRNA binding sites, and disrupting miRNA-mRNA interactions may result in the deregulation of target gene expression, especially with respect to certain oncogenes, tumor suppressor genes, or genes involved in oncogenic pathways [26] . A variety of molecular epidemiological studies have shown that SNPs located in miRNA-binding sites in DNA repair genes, including RAD51, were associated with the risk and prognosis of lung cancer, breast cancer, and head and neck carcinoma [27][28][29] . Nevertheless, the associations between genetic variants in RAD51 and the survival of patients with HCC are still unclear.
In the present study, to test the hypothesis that genetic variants in the predicted miRNA binding sites of the RAD51 gene are associated with the survival of patients with HCC, we rst performed bioinformatic prediction for RAD51 gene SNPs with a minor allele frequency (MAF) of ≥ 0.05 in the Chinese Han population. We selected one SNP, rs12593359, located in the 3′ UTR of the RAD51 gene, for further analysis as the location of this polymorphism has the potential to in uence miRNAs binding activity.
Then, we analyzed data from a cohort of 368 patients with HCC to determine the effects of this SNP on overall survival, and assessed whether the RAD51 rs12593359 polymorphism could be a potential biomarker to predict HCC prognosis.

Study population
A total of 368 patients with newly diagnosed and pathologically con rmed HCC were recruited between July 2007 and December 2013 from Guangxi Medical University cancer hospital. All patients had undergone radical surgery resection and their les contained complete clinical information. Detailed clinical information was collected through medical chart review or by consultation with treating physicians, and included age at diagnosis, sex, smoking status, drinking status, hepatitis B virus infection, tumor size, number of tumors, BCLC staging, AFP level, and cancer embolus status. All patients were followed up to collect data via telephone or hospital visits after surgery until patients died or up to the date of the last follow-up in June 2019. Peripheral blood sample (5 mL) was collected in a vacuum EDTA anticoagulant tube from each study participant. The research protocol was approved by the Ethical Committee Review Board of Guangxi Medical University, and written informed consent was obtained from each of the participants.

SNP selection and genotyping
We used the National Institute of Environmental Health Sciences (NIEHS) SNP database (https://snpinfo.niehs.nih.gov/) to identify functional RAD51 SNPs using the following criteria: (1) SNPs located in the 3′ -UTR of RAD51 in potential miRNA target sites; (2) minor allele frequency (MAF) in Chinese Han population > 0.05; and (3) low linkage disequilibrium using an r 2 threshold of < 0.8 for each other. Ultimately, only one SNP (rs12593359, located in the miR-129-3p binding site in the RAD51 3′ UTR) ful lled all of these criteria. Genomic information regarding RAD51 rs12593359 and the linkage disequilibrium plot are shown in Table 1S and Figure 1S.
For genotyping, we extracted genomic DNA by using a standard phenol/chloroform extraction method and stored at -80°C. Then, we performed the genotyping using the Agena MassARRAY system (Agena, San Diego, CA, USA) according to the manufacturer's instructions. The primers for PCR were designed using the Assay Designer software package of the Sequenom system (San Diego, CA, USA). The primers used for RAD51 rs12593359 were: F: 5′-ACGTTGGATGTAGGTTTGGCACAAGACTCC-3′ and R: 5′-ACGTTGGATGCAGTAAAACTCTCAAGCAGG-3′. The genotyping results were analyzed using the MassARRAY Typer software version 4.0.

Statistical analysis
Overall survival (OS) was de ned as the interval between the date of surgery and the date of death, or the date of the last follow-up. We used the Cox proportional hazards regression model to evaluate the effect of genotypes and clinicopathological variables on OS, calculated as hazard ratios (HRs) with their corresponding 95% con dence intervals (CIs). All HRs were adjusted for age at diagnosis, sex, smoking status, drinking status, hepatitis B virus infection, tumor size, number of tumors, BCLC staging, AFP level, and cancer embolus status. We performed a stepwise conditional logistic regression analysis to explore the best model for predicting survival outcome. Kaplan-Meier curves and log-rank tests were also used to assess differences in OS.
We performed expression quantitative trait loci (eQTL) analysis to investigate correlations between SNP rs12593359 and RAD51 mRNA expression levels, which were obtained from the genotype-tissue expression (GTEx) Portal (http://www.gtexportal.org/home/). The differences in RAD51 mRNA expression between HCC and normal tissues and associations between RAD51 expression levels and HCC overall survival were examined using data from the Gene Expression Pro ling Interactive Analysis (GEPIA) database (http://gepia.cancer-pku.cn/index.html), including data from 369 HCC tumor tissues, 50 normal liver tissues in The Cancer Genome Atlas (TCGA) dataset, and 110 normal liver tissues in the GTEx dataset. Finally, to provide additional supporting evidence, we also used the online Kaplan-Meier plotter tool (http://kmplot.com/analysis/) to assess the association between RAD51 expression levels and survival of liver cancer patients.
All statistical analyses were performed using the SPSS software, version 18.0 (SPSS Institute, Chicago, IL, USA) or GraphPad Prism, version 6.0 (GraphPad Software Inc., San Diego, CA, USA). All statistical tests were two-sided and P < 0.05 was considered to be statistically signi cant.

Association between RAD51 rs12593359 and clinical outcome in HCC patients
The effect of RAD51 rs12593359 on HCC prognosis is shown in Table 2. We observed that the dominant genetic model of the SNP rs12593359 had a signi cant correlation with HCC prognosis. Compared to the rs12593359 GG genotype, the GT/TT genotype conferred a signi cant increase in the risk of death in multivariate Cox proportional hazards model adjusted for all variables (adjusted HR = 1.34, 95% CI = 1.02-1.76; P = 0.035). Kaplan-Meier curve analysis showed a markedly shorter survival time in HCC patients with GT/TT genotypes of SNP rs12593359 compared with that of patients carrying a GG genotype (19.0 months vs. 36.0 months; P log-rank = 0.012) (Figure 1a).

Stepwise regression analysis of HCC prognosis
To determine the independent prognostic factors associated with HCC OS, we performed a stepwise multivariate Cox proportional hazards regression analysis using covariates listed in Table 1 (Table 3).

Strati ed analysis of association between SNP rs12593359 with OS by HCC patient characteristics
We further analyzed the effect of the SNP rs12593359 on OS in HCC patients strati ed by clinical characteristics. As shown in Table 4, the signi cant increase in risk of death conferred by the SNP rs12593359 was observed in males (adjusted HR = 1.40, 95% CI = 1.05-1.87; P = 0.021), but not in females (P = 0.707); in HBV-positive patients (adjusted HR = 1.37, 95% CI = 1.01-1.85; P = 0.042), but not in HBV-negative patients (P = 0.901); in those with a single tumor nodule (adjusted HR = 1.51, 95% CI = 1.11-2.06; P = 0.009), but not in those with multiple tumor nodules (P = 0.940); in those with AFP levels < 400ng/mL (adjusted HR = 1.58, 95% CI = 1.10-2.29; P = 0.014), but not in those with levels ≥ 400ng/mL (P = 0.451); and in patients who were cancer embolus-free (adjusted HR = 1.55, 95% CI = 1.10-2.20; P = 0.014), but not in those with a cancer embolus (P = 0.461). No signi cant association between SNP rs12593359 and HCC OS existed in other subgroups (P > 0.05). Figure 1b-f, log-rank tests indicated that there was a signi cant difference in overall survival between patients with TT/TG genotypes and those with the GG genotype of SNP rs12593359 in the above-mentioned subgroups.

As indicated in
Correlation between rs12593359 genotypes and the mRNA expression level of RAD51 We performed the eQTL analysis to correlate the rs12593359 genotypes and mRNA expression levels of RAD51 using data of the 483 cell-cultured broblasts from the GTEx database. The rs12593359 GG genotype was signi cantly associated with lower expression levels of RAD51 mRNA than the TT genotype, but not TG genotype (P = 1.1×10 -4 , Figure 2).
Association of RAD51 mRNA expression levels with survival of patients with HCC By using the GEPIA database, we evaluated mRNA expressions levels of RAD51 in 371 HCC and 160 normal tissue samples. HCC tissues exhibited a signi cantly higher expression of RAD51 as compared to that of the normal tissues (P < 0.001, Figure 3), and patients with higher RAD51 mRNA expression in tumor tissues showed a poorer overall survival (HR=1.6, P log-rank = 0.0054, Figure 4). Furthermore, we also found that a higher expression level of RAD51 was associated with poorer survival in 364 liver cancer patients from the online tool Kaplan-Meier plotter (P = 0.00097, Figure 2S).

Discussion
In the present study, we evaluated the prognostic value of the RAD51 SNP rs12593359, in a cohort of 368 patients with HCC in a population of Southern China. The most important nding was that SNP rs12593359, located in a miRNA binding site in the RAD51 3′ UTR, was signi cantly associated with HCC OS, and this effect was particularly pronounced in several subgroups of HCC patients (males, HBV positive patients, patients with a single tumor nodule, patients with AFP < 400 ng/ml, or patients with no cancer embolus). Furthermore, we observed that the genotypes of SNP rs12593359 was signi cantly associated with RAD51 mRNA expression levels in 483 cell-cultured broblasts.
RAD51 encodes a homolog of the E. coli RecA protein that catalyzes DNA repair via homologous recombination, an evolutionarily conserved mechanism for the repair of DNA damage and the generation of genetic diversity [30] . RAD51 expression is tightly controlled in normal cells, as it promotes strand exchange between unimpaired and impaired homologous DNA fragments, insuring a highly reproducible quality of DNA repair. Even slight changes in RAD51 gene expression may lead to DNA instability [31] . Indeed, overexpression of RAD51 is detected in multiple human tumor cells and correlates with poor prognosis [32][33][34][35] . In this study, we evaluated the prognostic value of RAD51 in patients with HCC using the GEPIA database. Consistent with studies in several other tumors, we found that RAD51 mRNA levels were higher in HCC tissues as compared to that of normal tissues, and the higher expression levels were associated with poorer survival. These ndings suggest that RAD51 may serve as an independent prognostic marker of survival in HCC patients, and raises the question of why certain individuals experience elevated expression of RAD51 and eventually develop HCC with a poor prognosis.
Accumulating evidence has suggested that miRNAs actively take part in the regulation of the DNA damage/repair network by affecting the expression of DNA repair genes such as RAD51, ultimately regulating their biological functions [36,37] . Polymorphisms in the predicted miRNA target sites of these genes have been shown to be strongly involved in regulating the expression of the mRNAs, either by providing mutated binding sites which alter miRNA-mRNA interactions, or by forming hairpin loop structures which stabilize and thus slow down mRNA degradation [38,39] . Our functional prediction analysis indicated that the rs12593359 SNP in the 3-UTR region of RAD51 was located within a predicted miRNA binding site for hsa-miR-129-3p (http://snpinfo.niehs.nih.gov/snpfunc.htm). The hsa-miR-129-3p may act as an important tumor suppressor that could inhibit tumor development and progression in several cancers [40][41][42] . Cui et al. reported that hsa-miR-129-3p was downregulated in HCC tissues, and patients with low hsa-miR-129-3p levels had a poorer prognosis than those with higher levels of this miRNA. However, re-expression of hsa-miR-129-3p was able to reverse epithelial-mesenchymal transition, and inhibit invasion and metastasis of HCC cells through targeting of the Aurora-A kinase [43] .
In our study, we found that the RAD51 SNP rs12593359 is an independent predictor of HCC OS, and that patients with HCC carrying GT or TT genotypes had a signi cantly increased risk of death when compared with those carrying GG genotypes. In subsequent genotype-mRNA expression correlation analysis, the TT genotype was signi cantly associated with increased mRNA levels of RAD51 in cellcultured broblasts. Chen et al. demonstrated that rs12593359 may be a putative variant mediating posttranscriptional regulation of the RAD51 gene, and the rs12593359 GT or TT genotype was found to yield signi cantly higher RAD51 mRNA levels compared with the GG genotype in lymphoblastoid cell lines [44] . Based on these results, together with our own ndings, we speculate that the functional SNP rs12593359 in the RAD51 3′ UTR may disrupt the binding ability of has-miR-129-3p, thus inhibiting its regulation of target gene expression. At the same time, such a change might lead to increased expression of RAD51, thus modifying the overall survival of HCC patients. However, the mechanisms underlying this phenomenon need to be studied further. Such ndings would help us to better assess the prognosis of patients with HCC, and develop a timely and individualized treatment plan, or even provide new ideas for targeted drugs.
It is important to acknowledge the limitations of the present study. Firstly, our sample was not large enough to detect minimal associations, which may have limited the statistical power of our analyses.
Secondly, whether these ndings (identi ed in a population of Southern China) can be applied at a more general level needs to be further validated in other district and ethnic populations. Thirdly, we did not determine the expression of RAD51 in our own HCC cohort, and assess its relationship with the rs12593359 genotypes and the survival of our patients. Finally, no direct functional investigations were conducted to provide mechanistic insights into this SNP, which will be an important topic of investigation in future studies.
In conclusion, our results suggest that the RAD51 3′-UTR polymorphism rs12593359 may be functional and may affect the survival of patients with HCC by altering RAD51 expression. Further studies using a larger population and functional exploration of the underlying molecular mechanisms are warranted to con rm these ndings.

Declarations
Ethics approval and consent to participate

Availability of data and materials
The data used to support the ndings of this study are available from the corresponding author upon request.    Table 1.
Bold values are statistically signi cant. Figure 1 Kaplan-Meier overall survival analysis for HCC by RAD51 rs12593359 genotypes (in dominant model) (a)

Figures
All HCC patients (b) Males with HCC (c) HBV-positive patients (d) HCC patients with a single tumor nodule (e) HCC patients with AFP < 400 ng/ml (f) HCC patients with cancer embolus-free.

Figure 2
The correlation between RAD51 rs12593359 genotypes with its mRNA expression levels in cell cultured broblasts from the genotype-tissue expression database.

Figure 3
Higher expression levels of RAD51 were found in HCC tissues than in 160 normal liver tissues (50 adjacent normal tissues in The Cancer Genome Atlas dataset and 110 normal liver tissues in genotypetissue expression data.) Page 19/20

Figure 4
Association of RAD51 mRNA expression with the survival of HCC from the GEPIA database.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Table1S.docx