Potentially functional genetic variants in ferroptosis‐related CREB3 and GALNT14 genes predict survival of hepatitis B virus‐related hepatocellular carcinoma

Abstract Background Ferroptosis is a known crucial player in the development of cancers. However, the effect of single nucleotide polymorphisms (SNPs) in ferroptosis‐related genes on survival in hepatitis B virus (HBV)‐related hepatocellular carcinoma (HBV‐HCC) patients remains unknown. Methods We used two‐stage multivariable Cox proportional hazards regression analyses to estimate the associations between 48,774 SNPs in 480 ferroptosis‐related genes and overall survival (OS) of 866 HBV‐HCC patients. Results We identified that two potentially functional SNPs (CREB3 rs10814274 C > T and GALNT14 rs17010547 T > C) were significantly independently associated with the OS of HBV‐HCC patients (CT + TT verse CC, hazards ratio (HR) = 0.77, 95% confidence interval (CI) = 0.67–0.89, p < 0.001 for rs10814274 and TC + CC verse TT, HR = 0.66, 95% CI = 0.53–0.82, p < 0.001 for rs17010547, respectively). Additional joint assessment of protective genotypes of these two SNPs showed that patients with 1–2 protective genotypes had a significantly better OS compared with those carrying 0 protective genotypes (HR = 0.56, 95% CI = 0.45–0.70, p < 0.001). Moreover, the expression quantitative trait loci (eQTL) analysis revealed that the survival‐associated SNP rs10814274 T allele was significantly correlated with reduced CREB3 transcript levels in both normal liver tissues and whole blood cells, while the GALNT14 rs17010547 C allele had a significant correlation with increased GALNT14 transcript levels in whole blood cells. Conclusion These results suggest that genetic variants of CREB3 and GALNT14 may affect the survival of HBV‐HCC patients, likely via transcriptional regulation of respective genes. However, further studies are required to confirm these findings.


| INTRODUCTION
Primary liver cancer is the third most common cancerrelated death worldwide and the second leading cause of cancer deaths in China. 1,2Approximately 90% of liver cancer patients were diagnosed with a histological subtype of hepatocellular carcinoma (HCC). 3Hepatitis B virus (HBV) infection has been found to be a major risk factor for HCC and accounts for 84% of HCC cases in China. 4,5eanwhile, published data showed that 87.7% of HCC patients were seropositive for HBV in southern China. 6mong the treatments for HCC patients, surgery is generally recommended. 7However, the prognosis remains at a poor 5-year survival rate of only 18%. 4 Continuous efforts have been made to improve HCC prognosis, and a series of clinical parameters were identified as predictive prognostic factors.For example, serum alpha-fetoprotein (AFP) levels were found to have a negative impact on HCC prognosis. 7In addition, several classification systems have been derived and applied for the treatment strategymaking and survival prediction for HCC patients, such as the Barcelona Clinic Liver Cancer (BCLC) stage and the TNM system. 8,9However, HCC, as a highly heterogeneous cancer, has distinct prognoses for patients who showed similar clinical characteristics and received the same treatments.Therefore, it is crucial to identify additional prognostic biomarkers that address individual variability and help to improve risk stratification and treatment decision-making for HCC patients.
As one of the most common forms of genetic variation, single nucleotide polymorphisms (SNPs) may affect gene expression and functions and thus have a significant impact on risk and prognosis of cancer. 102][13][14][15] In addition, a recent GWAS has demonstrated some associations between SNPs and prognostic outcomes in HCC patients.For example, Li et al. 16 reported that six SNPs on chromosomes 6p21 and 8p12 were not only associated with risk of chronic hepatitis HBV infection, or HBV-HCC, but also overall survival (OS) of HBV-HCC patients.More recently, Wei et al. 17 identified five SNPs that might serve as a reliable predictive biomarker for survival of HCC patients.However, the vast majority of functional SNPs that did not reach the stringent p-value after multiple test corrections may be ignored.Therefore, in the post-GWAS era, a hypothesis-driven approach using a fewer SNPs of candidate genes in a prognosis-related biological pathway has been applied to avoid the nuisance of multiple tests and contribute to a more effective identification of potentially functional SNPs. 18erroptosis, a unique form of cell programming death, was first coined by Dixon et al. in 2012. 19Unlike other forms of cell death, such as necroptosis and apoptosis, ferroptosis was characterized by iron-dependent lethal accumulation of lipid peroxides. 20Recently, a growing body of evidence has shown that ferroptosis plays an important role in cancer development and progression. 21With regard to HCC, several studies with experiments using HCC cell lines have demonstrated that ferroptosis might serve as the principal mechanism underlying the anticancer effect of sorafenib, suggesting a promising strategy for treating HCC by inducing ferroptosis. 22Meanwhile, recent studies have also demonstrated that ferroptosisrelated genes might serve as prognostic biomarkers for HCC patients.For example, one study using a prognostic model with ferroptosis-related genes showed a great performance in predicting prognosis of HCC patients, showed that patients with 1-2 protective genotypes had a significantly better OS compared with those carrying 0 protective genotypes (HR = 0.56, 95% CI = 0.45-0.70,p < 0.001).Moreover, the expression quantitative trait loci (eQTL) analysis revealed that the survival-associated SNP rs10814274 T allele was significantly correlated with reduced CREB3 transcript levels in both normal liver tissues and whole blood cells, while the GALNT14 rs17010547 C allele had a significant correlation with increased GALNT14 transcript levels in whole blood cells.

Conclusion:
These results suggest that genetic variants of CREB3 and GALNT14 may affect the survival of HBV-HCC patients, likely via transcriptional regulation of respective genes.However, further studies are required to confirm these findings.

K E Y W O R D S
ferroptosis, hepatocellular carcinoma, overall survival, single nucleotide polymorphism and they found differences in immune infiltration levels between high-and low-risk groups. 23Another study showed that four ferroptosis-related genes (i.e., FANCD2, CS, CISD1, and SLC1A5) were positively associated with the progression of HBV-HCC, of which higher expression levels of SLC1A5 were associated with tumor progression, immunosuppression, and poorer prognosis in patients with HBV-HCC. 24However, the roles of SNPs in ferroptosis-related genes in HBV-HCC remain unknown.In the present study, we tested the hypothesis that potentially functional genetic variants in the ferroptosis-related genes are associated with survival of HBV-HCC patients in a two-stage analysis.

| Studied populations
A total of 866 histopathologically confirmed HCC patients from Guangxi Medical University Cancer Hospital were recruited between July 2007 and December 2017.More details about the population have been described elsewhere. 25,26In brief, all patients were HBV seropositive and had undergone hepatectomy.Demographic and clinical variables, including age, sex, smoking status, drinking status, AFP level, cirrhosis, embolus, and BCLC stage, were collected.Among all patients, those with BCLC stage B or C had undergone operation according to the Chinese guidelines for treatment of HCC. 27All patients were followed up every 3 months within the first 2 years after hepatectomy and every 6 months in the next year through telephone calls.Survival time was from the date of surgery to the date of death or last follow-up in March 2020.Each patient signed a written informed consent.The present study was approved by the Institutional Review Board of Guangxi Medical University Cancer Hospital (Approval Number: LW2023121).

SNP Selecting
The whole genomic DNA of each HCC patient was extracted by a blood DNA extraction kit (Concert, Xiamen, China).Genotyping was performed using Illumina Infinium Global Screening Assay (Shanghai, China), and the quality control of raw data was described in detail elsewhere. 25We selected ferroptosis-related genes from the FerrDB database (http:// www.zhoun an.org/ ferrdb/ curre nt/ ). 28After removal of duplicated genes and genes in the sex chromosome, a total of 480 genes were selected as the candidate genes.All SNPs in candidate genes and within their ±2 kb flanking regions were extracted using Plink (version 1.09) (http:// pngu.mgh.harva rd.edu/ purce ll/ plink/ ) 29 and then screened by the following criteria: (1) with a calling rate less than 95%; (2) with minor allele frequency <0.05; 3) with Hardy-Weinberg equilibrium p < 1 × 10 −6 .

| Differential gene expression analysis
By using data from the TNMplot database (https:// tnmpl ot.com/ ), 36 we analyzed the differential gene expression levels between HCC tissues and paired adjacent normal tissues.To verify the expression levels of genes, we also analyzed the RNA sequencing data using newly collected tumor tissues and paired adjacent normal tissues from additional 100 HCC patients who had undergone hepatectomy in Guangxi Medical University Cancer Hospital.

| Statistical analysis
All HCC patients were randomly assigned to discovery and replication groups in a 1:1 ratio.Then, with adjustment for demographic and clinical variables, multivariable Cox proportional hazards regression analysis was performed to evaluate the associations between SNPs and overall survival of HCC patients in an additive genetic model.The cut-off p-value in survival analyses was set at 0.05.Instead of using the stringent false discovery rate (FDR) because many SNPs were in LD as the results of the imputation, a false-positive report probability (FPRP) analysis was performed with a prior probability of 0.1 to detect a Hazard Ratio (HR) of 1.5 for an association with alleles of each SNP.Only SNPs with FPRP values less than 0.2 in both discovery and replication datasets were selected for further analysis.A stepwise multivariable Cox regression model with adjustment for the above-mentioned variables was used to identify independent SNPs.The stratification analysis by demographic and clinical variables of HBV-HCC patients was also conducted to assess possible interactions between those variables and selected genotypes.The Kaplan-Meier curve was used to depict the associations between genotypes of SNPs and survival of HCC patients, as well as the combination of favorable genotypes.All statistical analyses were performed by R software (3.1.3and 4.0.3versions).The R packages used included "survminer", "survival", "gap", and "GenABLE". 37The Manhattan plot was drawn using the Haploview software. 38The regional plot was plotted using LocusZoom. 39 3 |RESULTS

| Associations between SNPs in ferroptosis-related gene and survival of HBV-HCC patients
As shown in the flowchart of the study design (Figure 1), we randomly assigned the 866 patients into groups of discovery and replication.With detailed characteristics shown in Table S1, a total of 480 ferroptosis-related genes were selected as the candidate genes.After quality control, 48,774 SNPs were available for further analysis in the discovery dataset.In the single-locus analysis, 1186 SNPs were found to be associated with overall survival of HBV-HCC patients (p < 0.05, FPRP<0.2), of which 10 SNPs remained significant after further assessment in the replication dataset (Table 1).The results of selected SNPs were visualized by Manhattan plots (Figure S1) and regional association plots (Figure S2).The baseline characteristics of 866 HBV-HCC patients were described in Table S2.

| The eQTL analysis and functional prediction
To identify those SNPs that may affect gene expression, we performed the expression quantitative trait loci (eQTL) analyses of 10 identified SNPs for their collations with mRNA expression levels of the corresponding genes by using data from the GTEx project.As shown in Figure 2, the CREB3 rs10814274 T allele was correlated with decreased mRNA expression levels in both normal liver tissues and whole blood cells (NES = −0.15,p < 0.001; NES = −0.10,p < 0.001, respectively), compared with the C allele.Compared with the GALNT14 rs17010547 T allele, the C allele was correlated with higher mRNA expression levels of gene in whole blood cells (NES = 0.098, p = 0.01) but not in normal liver tissues (NES = −0.058,p = 0.58).Meanwhile, the rs6543592 G allele was correlated with elevated mRNA expression levels of GALNT14 in whole blood cells (NES = 0.1, p = 0.0023), compared with the A allele, but this trend was not significant in normal liver tissues (NES = 0.0052, p = 0.96).However, no significant correlation between the other seven SNPs and The overall procedures of the present study.
mRNA expression levels of their corresponding genes was found (Figure S3).
We then used online bioinformatics tools to further explore the functions of these three SNPs.According to the HaploReg, these three SNPs (CREB3 rs10814274, GALNT14 rs17010547, and GALNT14 rs6543592) may disrupt the motif of several proteins, affecting the mRNA expression (Table S3).Moreover, CREB3 rs10814274 was predicted to be located on the transcription factor binding site from SNPinfo (Table S3).This observation is supported by the experimental data from the ENCODE project, which suggest that CREB3 rs10814274 is located in the region with enriched H3K4Me1 as well as the binding site of Rbpjl factor (Figure S4).On the other hand, GALNT14 rs17010547 is likely to have an impact on the DNA enhancer (Table S3).Subsequently, the LD analysis showed that both rs17010547 and rs6543592 in GALNT14 were in high LD (r 2 > 0.8; Figure S5).Having considered functional prediction and LD results, we selected CREB3 rs10814274 and GALNT14 rs17010547 for further analysis.

| Identification of independent SNPs associated with HBV-HCC survival
Next, a stepwise multivariable Cox regression analysis with adjustment for demographic and clinical variables was used to assess the independence of two selected SNPs.As a result, both CREB3 rs10814274 C > T (HR = 0.76, 95% CI = 0.66-0.88 and p < 0.001) and GALNT14 rs17010547 T > C (HR = 0.69, 95% CI = 0.58-0.83and p < 0.001) remained statistically significantly associated with HBV-HCC OS (Table S4).In addition, we performed 1000 times random validations with bootstrapping, and the results returned normally distributed HR values for both SNPs, suggesting rs10814274 and rs17010547 were not chosen by chance (Figure S6).Based on the above-mentioned evidence, both CREB3 rs10814274 and GALNT14 rs17010547 were considered independently potentially functional SNPs and thus were selected for the final analysis.

| Associations of CREB3 rs10814274 and GALNT14 rs6543592 genotypes with HBV-HCC survival
When evaluating the associations of two selected SNPs with HBV-HCC OS, we found that the HRs of both CREB3 rs10814274 C > T and GALNT14 rs17010547 T > C were larger than 1 (p trend <0.001 for both SNPs).Specifically, patients with CT + TT genotypes of CREB3 rs10814274 or TC + CC genotypes of GALNT14 rs17010547 were associated with a better OS (HR = 0.62, 95% CI = 0.51-0.77,p < 0.001;  HR = 0.66, 95% CI = 0.53-0.82,p < 0.001, respectively), compared with their wild genotypes.To assess the collective effect of two selected SNPs on HBV-HCC OS, we combined their protective genotypes (i.e., CREB3 rs10814274 CT + TT and GALNT14 rs17010547 TC + CC) into a genetic score as the number of protective genotypes (NPG).We found that an increasing NPG was associated with a better OS in a dosedependent manner (p trend <0.001).We also found that the risk of death for those patients with one or two NPG was about 44% lower (HR = 0.56) than for those who did not carry these protective genotypes (p < 0.001) (Table 2).These observed associations are visualized in the Kaplan-Meier survival curves (Figure 3).

| Stratified analysis between protective genotypes and HBV-HCC survival
To evaluate whether the combined effect of NPG on HBV-HCC survival was modified by other variables, we performed stratified analysis by age, sex, smoking status, drinking status, cirrhosis, AFP, embolus, and BCLC stage.As shown in Table 3, those patients with one or two NPG had a better OS in all subgroups except for female patients.No interactive effects between the variables and survival-associated genotypes were observed in all subgroups, except for an interaction between NPG and AFP (p interaction = 0.024), which may be confounded by some unknown factors.

| Different mRNA expression analysis
To further explore the roles of CREB3 and GALNT14 in HCC survival, we assessed the mRNA expression levels of these two genes in paired HCC tumors and adjacent normal tissues, as well as the correlation between gene expression levels and survival of HCC patients.As shown in Figure 4, the expression levels of CREB3 were higher in the tumor tissues, and higher expression levels were associated with The expression quantitative trait loci (eQTL) analysis of three SNPs.The eQTL results of CREB3 rs10814274 in normal liver tissues (A) and whole blood cells (B); the eQTL results of GALNT14 rs17010547 in normal liver tissues (C) and whole blood cells (D); the eQTL results of GALNT14 rs6543592 in normal liver tissues (E) and whole blood cells (F). a poorer HCC OS (Figure 4A-C).Interestingly, the expression levels of GALNT14 were lower in tumor tissues, and higher expression levels of GALNT14 were associated with a better progression-free survival (PFS) (Figure 4D-F), but the correlations between GALNT14 expression levels and HCC OS were not significant.

| DISCUSSION
In the present study, we evaluated the associations between 48,774 SNPs of 480 ferroptosis-related genes and HBV-HCC survival in multivariable Cox regression analysis.We demonstrated that both CREB3 rs10814274 C > T and GALNT14 rs17010547 T > C were associated with a better OS of HBV-HCC patients.The underlying mechanism for the observed effects of both SNPs on HCC survival is likely through the regulation of their corresponding gene expression.
In ferroptosis, many genes participate in cancerassociated signaling pathways, leading to cell death, which may affect tumor progression and response to treatment.However, few have studied the roles of ferroptosisrelated genes in tumor progression and clinical outcomes.
For example, one study reported that genetic variants of ferroptosis-related APOE, BCL3, and ALOX5AP were associated with the risk of thyroid cancer. 40Another study showed that SNPs in IDH1 105CCT were associated with ferroptosis in inflammatory intrahepatic cholangiocarcinoma (ICC). 41To the best of our knowledge, the present study is the first to report associations between genetic variants in ferroptosis-related genes and survival of HBV-HCC patients.
CREB3 belongs to the cyclic AMP-response elementbinding protein (CREB) family.While CREB3 functions as a transcription factor that regulates numerous genes in the central nervous system and responds to the Golgi stress, 42,43 the whole CREB family is considered often overexpressed and functions as a tumor mediator in multiple cancers. 446][47][48] Similarly, Shen et al. reported that the mRNA expression levels of CREB3 were higher in HCC tissues than in adjacent normal tissues, and higher expression of CREB3 was associated with a worse 10year overall survival in patients with HCC. 49As for HBV  T A B L E 2 Associations between two selected SNPs and survival of HBV-HCC patients.
infection, recent studies have demonstrated that CREB plays a crucial role in HBV replication.For example, one study demonstrated that a CREB motif in the HBV pre-S2 region contributed to basal S promoter activity, and they also found that CREB/PKA signal transduction pathways in hepatocytes may be utilized by HBV to enhance HBsAg expression during homeostasis and hepatic inflammation. 50Another study showed that both replication and gene expression of HBV require functional CREB and HBV-CRE, and CRE decoy oligonucleotides and the overexpression of CREB mutants could effectively block the HBV life cycle. 51Notably, it has been reported that ferroptosis was negatively regulated by CREB in lung adenocarcinoma.To be more specific, CREB could positively regulate the transcription level of glutathione peroxidase 4 (GPX4) through binding to its promoter region, which is marked by the upregulation of lipid reactive oxygen species (ROS).Knockdown of CREB led to ferroptotic-like cell death, which also supports the oncogenesis role of CREB. 52Therefore, we speculate that changing expression levels of CREB3 may alter the inhibiting effect of ferroptosis in HCC cells and subsequently affect the prognosis of HCC.In the present study, we found that the CREB3 rs10814274 T allele was associated with a better survival of HBV-HCC patients and a significant association of decreased CREB3 mRNA expression levels in both liver tissues and whole blood cells.In addition, the CREB3 mRNA expression levels were much higher in HCC tissues than in adjacent normal tissues, and higher expression levels of CREB3 mRNA were associated with a worse prognosis in HCC patients, which also supports the oncogenesis role of CREB3.However, these findings need to be substantiated in further molecular biology experiments and clinical studies.GALNT14, located on chromosome 2, is one of the polypeptide N-acetylgalactosaminyltransferase (GALNT) family members.The GALNT family could catalyze protein O-glycosylation, and the abnormal expression of GALNT could lead to distinct results through alteration of O-glycosylation in several cancers. 53,54Although the mechanisms of GALNT14 in HCC are largely unclear, it was reported that GALNT14 promoted cell proliferation and migration, and silencing GALNT14 enhanced cell sensitivity to anticancer drugs, implying an important functional role of GALNT14 in HCC. 55Concerning ferroptosis, Li et al. 56 reported that downregulation of GALNT14 could induce ferroptosis by inhibiting the mTOR/EGFR pathway, which then suppresses the protein levels of SLC7A11 and GPX4.This anti-ferroptosis role is consistent with the growthpromoting role of GALNT14 in HCC, but the role of GALNT14 in HCC has not yet been fully elucidated.
In the present study, we found that lower expression of GALNT14 was associated with a worse PFS, implying that GALNT14 may play a tumor suppressor role in HCC, but the molecular mechanisms of GALNT14 in predicting survival of HCC patients need to be further investigated.Moreover, we found that the GALNT14 rs17010547 C allele was associated with a protective effect on HBV-HCC survival and a significant association of increased GALNT14 mRNA expression levels in whole blood cells, but the underlying mechanisms also need to be further investigated.c p inter : p-value for interaction analysis between NPG and covariables.
One previous GWAS study has demonstrated the association between the GALNT14 rs9679162 genotypes and the response to chemotherapy in advanced HCC patients. 57In later studies, rs9679162 was also found to be an effective predictor in HCC patients who received transarterial chemoembolization (TACE) or sorafenib treatment. 58,59However, by using data from 160 patients with advanced HCC, Chu et al. 60 did not find any significant association between rs9679162 genotypes and survival of HCC patients.Consistent with their findings, we did not find any significant associations between rs9679162 genotypes and survival of HBV-HCC patients in the present study (Table S5), and rs9679162 was not in LD with rs17010547.One possible explanation is the relatively small sample size in our study, which may limit the power to detect a weak association between rs9679162 and HBV-HCC survival, particularly in HBV-HCC patients who had undergone hepatectomy.Therefore, additional experimental investigations are required to determine how GALNT14 rs17010547 T > C may influence HCC.
Several inherent limitations should be mentioned in the present study.Firstly, all patients were recruited in the same hospital, which might cause selection bias.Secondly, the sample size of the present study was relatively small, and further studies with a larger population are needed to verify our findings.Lastly, the molecular mechanisms underlying the survival-associated SNPs and the observed association call for direct biological experiments for functional validation.
In summary, the present study identified statistically significant associations between two potentially functional genetic variants (CREB3 rs10814274 and GALNT14 rs17010547) in ferroptosis-related genes and survival of HBV-HCC patients.The protective genotypes of these two SNPs contributed to a better OS in a doseresponse manner in the combined analysis.Such protective effects on OS are likely through SNP-associated F I G U R E 4 Differential mRNA expression analysis and overall survival analysis of the two genes.The expression level of CREB3 mRNA from TNMplot (A) and our data (B).The KM survival curves for CREB3 mRNA expression levels with OS (C) of HCC patients.The expression level of GALNT14 mRNA from TNMplot (D) and our data (E).The KM survival curves for GALNT14 mRNA expression levels with PFS of HCC patients (F).OS, overall survival; PFS, progression-free survival.
expression regulation of CREB3 and GALNT14.These results provide some valuable information for risk stratification and treatment strategy-making for HBV-HCC patients.

T A B L E 1
Associations between SNPs of ferroptosis-related genes and survival of HBV-HCC patients in discovery, replication, and combined dataset.

b
NPG: number of protective genotypes (rs10814274 CT and TT; rs17010547 TC and CC).
Stratified analysis of the protective genotypes of selected SNPs in HBV-HCC patients.
T A B L E 3Abbreviations: CI, confidence interval; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HR, hazards ratio; SNP, single nucleotide polymorphisms.a NPG: number of protective genotypes (rs10814274 CT and TT; rs17010547 TC and CC).b Adjusted for age, sex, smoking status, drinking status, cirrhosis, AFP, embolus, and BCLC stage.