Polymorphisms in the coding region of X-ray repair complementing group 4 and aflatoxin B1-related hepatocellular carcinoma


  • Potential conflict of interest: Nothing to report.

  • This study was supported, in part, by the National Natural Science Foundation of China (no. 81160255), the Innovation Program of Shanghai Municipal Education Commission (no. 13YZ035), and the Youth Science Foundation of Guangxi (no. 0832097).*These authors contributed equally to this work.


X-ray repair complementing group 4 (XRCC4) is very important in maintaining overall genome stability and may play an important role in carcinogenesis. We aimed to investigate the role of polymorphisms in the coding region of this gene in hepatocellular carcinoma (HCC) caused by aflatoxin B1 (AFB1). A hospital-based case-control study, including 1,499 HCC cases and 2,045 controls without any liver diseases or tumors, was conducted in a high AFB1 exposure area (the Guangxi region) to assess the relationship between 21 polymorphisms in the coding region of XRCC4 and AFB1-related HCC risk and prognosis. Among these 21 polymorphisms, only rs28383151 modified HCC risk. These individuals with the genotypes of rs28383151 A alleles (rs28383151-GA/AA), compared with the homozygote of rs28383151 G alleles (rs28383151-GG), faced increasing risk of HCC (odds ratio [OR]: 2.17; 95% confidence interval: 1.77-2.67). Significant interactive effects between risk genotypes (OR, >1) and AFB1 exposure status were also observed in the joint-effects analysis. Furthermore, this polymorphism was correlated not only with lower XRCC4-expressing levels, but also with higher AFB1-DNA adducts levels and increasing TP53M and portal vein tumor risk. The rs28383151 polymorphism modified the recurrence-free survival and overall survival of HCC patients, especially under high AFB1 exposure conditions. Additionally, this polymorphism multiplicatively interacted with the glutathione S-transferase M1 polymorphism with respect to HCC risk (ORinteraction = 2.13). Conclusion: Genetic polymorphisms in the coding region of XRCC4 may be risk and prognostic biomarkers of AFB1-related HCC, and rs28383151 is such a potential candidate. (HEPATOLOGY 2013) © 147.

In China, hepatocellular carcinoma (HCC) is the third-most common malignant tumor and accounts for approximately 55% of the world's HCC cases, more than 270,000 each year.1, 2 This tumor occurs more often in eastern and southeastern China, mainly because of high aflatoxin B1 (AFB1) exposure and/or chronic infection of hepatic virus B(HBV) and hepatic C virus (HCV).1, 3 In the high-AFB1-exposure areas, such as Guangxi Zhuang Autonomous Region, this tumor is the most common occurring cancer.3 AFB1 is known as an important I-type chemical carcinogen and can induce various types of DNA damage, such as DNA double-strand break (DSBs), DNA base damage, and oxidative damage.4 Among these forms of DNA damage, DSBs are the most detrimental form, because they may lead to both chromosomal breakage and rearrangement and, ultimately, lead to the tumorigenesis of HCC.3

Although X-ray repair complementing group 4 (XRCC4) is necessary for DNA ligation in the nonhomologous end-joining (NHEJ) pathway, which is responsible for repairing most DSBs.5, 6 Recently, several studies have shown that the polymorphisms in this gene may be associated with dysfunction of XRCC4 and increased tumor risk.7, 8 However, the association between XRCC4 single-nucleotide polymorphisms (SNPs) and HCC has not yet been elucidated. Here, we evaluated whether 21 SNPs in the coding region of this gene modify AFB1-related HCC risk and prognosis. In addition, we also analyzed the effects of XRCC4 polymorphisms on XRCC4-expressing levels, AFB1 DNA adducts levels, portal vein tumor (PVT), and the hot-spot mutation of TP53 gene (TP53M) related to AFB1.


AFB1, aflatoxin B1; AFBO, AFB1-exo-8,9-epoxide; AFP, alpha-fetoprotein; ALB, albumin; CI, confidence interval; DSB, double-strand break; ELISA, enzyme-linked immunosorbent assay; GSTM1, glutathione S-transferase M1; HBV, hepatitis B virus; HBsAg, hepatitis B surface antigen; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HR, hazard ratio; IHC, immunohistochemistry; IRS, immunoreactive score; mRNA, messenger RNA; MRT, median RFS time; NHEJ, the nonhomologous end-joining pathway; OR, odds ratio; OS, overall survival; PCR, polymerase chain reaction; PVT, portal vein tumor; RFS, recurrence-free survival; rs28383151-AA, the homozygotes of rs28383151 A alleles; rs28383151-GA, the heterozygotes of rs28383151 G and A allele; rs28383151-GG, the homozygote of rs28383151 G alleles; rs28383151-GA/AA, the XRCC4 genotypes with rs28383151 A alleles; SNPs, single-nucleotide polymorphisms; TNM, tumor-nodes-metastasis staging system; TP53M, the hot-spot mutation of TP53 gene; WT, wild type; XPC, xeroderma pigmentosum complementation group C; XPD, xeroderma pigmentosum complementation group D; XRCC4, X-ray repair complementing group 4.

Patients and Methods

Human Subjects.

The protocol of the study was approved by the ethics committees of the hospitals involved in this study. The design of the Guangxiese HCC study has been previously described.9, 10 Briefly, cases were patients diagnosed with histopathologically confirmed HCC in the Affiliated Hospitals of the two main medical colleges in the Southwestern Guangxi, namely, Guangxi Medical University (Nanning, China) and Youjiang Medical College for Nationalities (Baise, China), during January 2004 and December 2010. Both case and control recruitment are still ongoing. Controls without any clinic hepatitic diseases or tumors were randomly selected from a pool of healthy volunteers who visited the general health check-up center of the same hospitals for their routine scheduled physical exams. To control the effects of confounders, cases were individually matched (1:1 or 1:2) to controls based on sex, ethnicity (Han, Zhuang), age (±5 years), and HBV and HCV infection. After giving written consent, demographic information and clinical pathological data (including cirrhosis, tumor size, PVT, and tumor stage) were collected in the hospitals using a standard interviewer-administered questionnaire and/or medical records.

A total of 1,536 cases and 2,074 controls, representing 95% of eligible cases and 98% of eligible controls, were enrolled and interviewed. At the same time, 4 mL of peripheral blood was obtained for serum analysis and DNA extraction. Surgically removed tumor samples of all cases were collected for analyzing XRCC4 protein expression levels and TP53M. Additionally, for analyzing XRCC4 messenger RNA (mRNA) expression levels, 75 fresh cancerous tissuespecimens were also collected during May 2010-December 2010 according to the following criteria: amount of tumor component (at least 70% of tumor cells) and quality of material (i.e., absence of necrosis or hemorrhage). Thirty-seven cases and 29 controls, respectively, were excluded from the study because of extracted DNA being of low yield or quality and because of lack of information on viral infection status. Thus, 1,499 HCC patients (including 1,156 patients previously studied9) and 2,045 controls (including 1,402 subjects previously studied9) were included for the final analysis. Those hepatitis B surface antigen (HBsAg) positive and anti-HCV positive in their peripheral serum were defined as groups infected with HBV and HCV. Liver cirrhosis was diagnosed by pathological examination, and stages of tumor were confirmed according to the tumor-nodes-metastasis (TNM) staging system.

AFB1 Exposure Data.

In the present study, AFB1 exposure information consisted of exposure levels and years. In Guangxi, because food-consumption types are relatively simple and limited to corn, peanuts, and rice, and AFB1 mainly contaminates these poorly stored foods, especially corn and peanuts, the years of participants having food contaminated by AFB1 were defined as AFB1 exposure years for subjects.11 Because of the abnormal distribution of exposure-years data and significantly different median value of exposure years between HCC cases and controls, AFB1 exposure years were divided into three groups: short (<40 years), medium (40-48 years), and long (>48 years), according to median exposure years among controls (40 years) and cases (48 years). AFB1 exposure levels were ascertained according to serum AFB1 albumin (ALB) adducts levels of peripheral blood. AFB1 ALB adducts levels was tested using the comparative enzyme-linked immunosorbent assay (ELISA) as previously published.12

SNP Selection and Genotyping.

Because missense mutations in the coding region from SNPs lead to an amino acid change in the protein product, and might be associated with the structure and function of corresponding protein,13 we obtained 21 known SNPs in the coding region of XRCC4 using the Ensembl database (Supporting Table 1). For SNP genotypic analyses, laboratory personnel were blinded to case and control status. Genomic DNA was isolated from peripheral blood leukocytes using the standard phenol-chloroform extraction method. Fourteen SNPs (Supporting Table 1) were genotyped by the TaqMan polymerase chain reaction (PCR) assays using the ABI 7900 System (Applied Biosystems, Foster City, CA). For quality control, controls were included in each run, and repeated genotyping and sequencing of a random 5% subset yielded 100% identical genotypes. The other seven SNPs (rs149355996, rs144653114, rs143782027, rs2974446, rs141122119, rs148273490, and rs141304949) were analyzed by a sequencing technique.

XRCC4 Expression Assay.

XRCC4 expression levels were evaluated using both XRCC4 mRNA-expressing levels and protein-expressing levels. Detailed information about XRCC4 expression analysis is described in the Supporting Materials.

DNA Repair Capacity Analysis.

In this study, DNA repair capacity related to AFB1-induced DNA damage was elucidated by using both TP53M and AFB1 DNA adducts levels. Detailed information about DNA repair function analysis is described in the Supporting Materials.

HCC Patient Follow-Up.

Patients were followed and underwent serial monitoring of alpha-fetoprotein (AFP), ultrasonography (US), chest radiograph, and emission computed tomography every 2 months for the first 2 years and semiannually thereafter for detection of any recurrence. Recurrence was diagnosed by imaging techniques, either intra- or extra-hepatically (i.e., lymph nodes and distant metastases). An increase of AFP without radiologic evidence of a new tumor was not diagnosed as recurrence until this became manifest on imaging. The last follow-up day was set on August 31, 2011, and survival status was confirmed by means of clinic records and patient or family contact. In this study, the duration of overall survival (OS) was defined as from the date of curative treatment to the date of death or last known date alive, whereas the duration of recurrence-free survival (RFS) was defined as from the date of curative treatment to the date of tumor recurrence or last known date alive.

Statistical Analysis.

All statistical analyses were done using SPSS version 18 (SPSS, Inc., Chicago, IL). The two-sided chi-square test was used to evaluate differences in frequency distributions of demographic characteristics, AFB1 exposure information, and XRCC4 genotypes between cases and controls. Based on individually matched design, we did conditional logistic regression (with multivariate factors, including known causes of HCC among the Guangxi population) to estimate odds ratios (ORs) for risk of HCC and their 95% confidence intervals (CIs). The test for screening the main effects of 21 SNPs was based on the additive model, treating genotype as an ordinal variable (wild type [WT] coded as 0, heterozygote as 1, and homozygotes variant as 2). The correction for multiple testing in the screen stage was done using the correlation matrix-based method, which takes into account the linkage disequilibrium between SNPs.14 The effective number of independent SNPs (Meff) was determined using the spectral decomposition,14 and the threshold for significance was calculated as αcorrect = 0.05/Meff. Based on this method, we obtained a Meff estimate of 20.99, and we therefore set the significance threshold to αcorrect = 2.38 × 10−3. In addition, to correct for testing the associations also in the AFB1 exposure years and levels, genotype frequencies in these groups were further adjusted for multiple comparisons using Bonferroni's method, and the significance threshold was lowered to αcorrect = 1.19 × 10−3correct = 2.38 × 10−3/2).

Spearman's r test was used to analyze the correlation between XRCC4 genotypes and XRCC4 protein expression. Kaplan-Meier's survival analysis with the log-rank test was used to evaluate the relationship between this polymorphism and HCC prognosis. Risk factors for HCC prognosis were first selected using Cox's multivariate regression model (including age, sex, race, HBV and HCV infection, AFB1 exposure information, tumor size, tumor TNM stage, and all possible multiplicatively interactive variables) with step-wise forward selection based on the likelihood ratio test. Hazard ratios (HRs) and 95% CIs for risk factors were next calculated in the same multivariate Cox's regression model (simultaneously including all risk variables and multiplicatively interactive variables).


Clinical and Demographic Characteristics of Cases and Controls.

Our final analysis included 1,499 HCC cases and 2,045 controls (Supporting Table 3). There were no significant differences between cases and controls in terms of distribution of age, sex, race, and HBV and HCV status as a result of individual matching (P > 0.05). These results suggest that HCC patient data were comparable to control data.

AFB1 Exposure Increased HCC Risk.

The AFB1 exposure information for the entire study population is shown in Supporting Table 4. We observed that the AFB1 exposure years were associated with an increased risk for HCC (OR = 2.53 for medium-exposure years; OR = 5.85 for long-exposure years). We also found that HCC cases (28.38 fmol/mg) had higher serum levels of AFB1 ALB adducts than controls (11.57 fmol/mg). For statistical analysis, values were logarithmically transformed and then were divided into three stratus: low (<2.18 ln fmol/mg), medium (2.18-2.98 ln fmol/mg), and high (>2.98 ln fmol/mg), according to the mean logit value of serum AFB1 ALB adducts among controls and cases (Supporting Table 4). Regression analysis showed that HCC risk gradually increased with an increasing number of exposure levels (adjusted OR = 2.10-6.52; P < 0.01).

XRCC4 rs28383151 Polymorphism Increased HCC Risk.

XRCC4 genotypes from all peripheral blood DNA samples are listed in Table 1. For these SNPs, the genotype distribution of controls was consistent with those expected from Hardy-Weinberg's equilibrium (Supporting Table 5; P > 0.05). Among 21 SNPs, only rs28383151 (no. 3) was observed to modify HCC risk. Logistic regression analyses exhibited that the adjusted OR for HCC for those individuals carrying the heterozygotes of rs28383151 G and A allele (rs28383151-GA) versus those with the homozygotes of the rs28383151 G alleles (rs28383151-GG) was 1.94 (95% CI: 1.54-2.43); the corresponding OR for those featuring the homozygotes of rs28383151 A alleles (rs28383151-AA) was 3.10 (95% CI: 2.10-4.58). Because our previous studies have shown that several polymorphisms in other genes (including xeroderma pigmentosum complementation group C [XPC] codon 939, xeroderma pigmentosum complementation group D [XPD] codon 751, and glutathione S-transferase M1 [GSTM1] polymorphisms) modify HCC risk in the Guangxi population,9-11 we investigated their possible modifying effects on XRCC4 risk value and observed a similar risk value for the XRCC4 rs28383151 polymorphism after adjusting for these three SNPs (Supporting Table 6). Additionally, we also found some evidence of multiplicative interaction between XRCC4 and GSTM1 (ORinteraction = 2.13 [95% CI: 1.87-2.42]; Pinteraction = 1.56 × 10−30; data not shown).

Table 1. Polymorphisms of XRCC4 and the Risk of HCC
  GenotypeControls (n = 2,045)HCCs (n = 1,499)HCC Risk*
No.SNPxx/xy/yynxx/nxy/nyynxx/nxy/nyyOR (95%CI/Ptrend)_xy§OR (95%CI/Ptrend)_yyOR (95%CI/Ptrend)_xy/yy
  • *

    HCC risk value was calculated using OR conditional on matched set adjusted for AFB1 exposure years and AFB1 exposure levels (corresponding OR and 95% CI values can be found in Supporting Table 7).

  • xx/xy/yy represents WT homozygote/heterozygote/variant-type homozygote.

  • nxx/nxy/nyy = number of subjects with xx genotype/number of subjects with xy genotype/number of subjects with yy genotype.

  • §

    OR (95%CI/Ptrend)_xy represents HCC risk value of xy genotype, compared to xx genotype.

  • OR (95%CI/Ptrend)_yy represents HCC risk value of yy genotype, compared to xx genotype.

  • OR (95%CI/Ptrend)_xy/yy represents HCC risk value of combination of xy genotype and yy genotype, compared to xx genotype.

    Abbreviation: ND, not determined.

01rs142575170GG/GA/AA1,767/270/81,282/210/71.01 (0.79-1.30/0.91)1.17 (0.34-4.08/0.81)1.01 (0.80-1.27/0.92)
02rs28383138CC/CG/GG1,758/231/561,262/183/540.97 (0.75-1.24/0.78)1.51 (0.96-2.40/0.08)1.06 (0.85-1.33/0.61)
03rs28383151GG/GA/AA1,758/225/621,056/303/1401.94 (1.54-2.43/10−8)3.10 (2.10-4.58/10−8)2.17 (1.77-2.67/10−14)
04rs61762970TT/TC/CC1,865/172/81,369/126/41.19 (0.89-1.60/0.24)0.95 (0.21-4.34/0.94)1.19 (0.89-1.58/0.25)
05rs149355996AA/AT/TT1,959/84/21,459/39/10.80 (0.51-1.25/0.34)1.70 (0.13-22.84/0.69)0.82 (0.52-1.28/0.38)
06rs144653114GG/GA/AA1,292/630/123950/466/831.03 (0.86-1.23/0.75)0.91 (0.64-1.29/0.59)1.01 (0.86-1.19/0.91)
07rs79561451TT/TC/CC1,883/152/101,376/117/61.15 (0.85-1.56/0.37)0.86 (0.26-2.81/0.80)1.13 (0.84-1.52/0.41)
08rs28360135TT/TC/CC1,669/309/671,219/234/460.97 (0.78-1.21/0.77)0.99 (0.63-1.55/0.97)0.97 (0.79-1.19/0.78
09rs56334522TT/TG/GG1,971/72/21,431/67/11.14 (0.77-1.70/0.52)0.76 (0.01-8.51/0.83)1.13 (0.76-1.68/0.54)
10rs28360136GG/GC/CC1,362/616/67994/454/511.14 (0.95-1.35/0.16)1.23 (0.79-1.93/0.37)1.14 (0.97-1.36/0.12)
11rs140143447CC/CT/TT1,988/57/01,462/37/00.92 (0.56-1.51/0.73)ND0.92 (0.56-1.51/0.73)
12rs201604424AA/AG/GG1,614/372/591,178/276/451.05 (0.85-1.29/0.67)0.88 (0.54-1.43/0.61)1.02 (0.84-1.25/0.82)
13rs143782027GG/GA/AA1,562/428/551,151/303/450.92 (0.76-1.13/0.44)1.36 (0.84-2.19/0.21)0.97 (0.80-1.17/0.75)
14rs151332331GG/GA/AA1,685/290/701,231/219/491.02 (0.81-1.28/0.87)1.19 (0.76-1.85/0.45)1.05 (0.85-1.30/0.65)
15rs140579916CC/CA/AA1,827/218/01,337/162/00.88 (0.68-1.14/0.34)ND0.88 (0.68-1.14/0.34)
16rs2974446AA/AC/CC2,029/16/01,484/15/01.14 (0.50-2.62/0.76)ND1.14 (0.50-2.62/0.76)
17rs141122119AA/AG/GG1,949/94/21,438/60/10.89 (0.59-1.32/0.55)0.73 (0.03-16.00/0.84)0.88 (0.59-1.31/0.54)
18rs138837678AA/AC/CC2,000/43/21,461/37/11.37 (0.80-2.37/0.25)0.25 (0.02-3.11/0.28)1.27 (0.74-2.16/0.39)
19rs61749611AA/AC/CC1,832/207/61,342/153/41.13 (0.87-1.48/0.37)1.24 (0.90-5.34/0.77)1.14 (0.87-1.48/0.35)
20rs148273490GG/GC/CC1,904/140/11,392/106/11.22 (0.90-1.67/0.21)0.78 (0.05-12.51/0.86)1.22 (0.89-1.66/0.22)
21rs141304949TT/TC/CC1,847/180/181,353/134/120.98 (0.74-1.29/0.86)0.83 (0.34-2.05/0.69)0.96 (0.74-1.26/0.78)

To assess possible interactive effects of matching factors and rs28383151 polymorphism on HCC risk, we performed a series of bivariate stratified analyses by matching factors, such as HBV and HCV infection, age, race, and sex, on this polymorphism and did not find that these factors modulated the effect of this polymorphism on HCC risk (Pinteraction > 0.05; Supporting Table 8). This implied that these matching factors should be effectually manipulated and should not modify the association between rs28383151 polymorphism and HCC related to AFB1 exposure.

XRCC4 rs28383151 Polymorphism Interacted With AFB1 Exposure on HCC Risk.

To study the correlation between rs28383151 polymorphism and AFB1 exposure years in the risk for HCC, we analyzed the joint effects of AFB1 exposure years and XRCC4 genotypes on HCC risk (Table 2). In this analysis, we used as a reference the lowest risk group: those who had rs28383151-GG and short-term AFB1-exposure years. We observed that increasing the number of exposure years consistently increased HCC risk; moreover, this risk was more pronounced among subjects with the risk genotypes of XRCC4 (OR, >1). We found some evidence of multiplicatively interactive effects of genotypes and exposure years on HCC risk (19.61 > 5.28 × 1.98) according to the previously published formula (OReg > OReg' × ORe'g).15 Additionally, a similar increased-risk trend was also found in the sequential joint-effects analysis of this polymorphism and AFB1 exposure levels for HCC risk (11.26 > 5.76 × 1.35; Table 2).

Table 2. Joint Effects of AFB1 Exposure and XRCC4 rs28383151 Polymorphism on HCC Risk
AFB1 exposureSNPn%n%OR (95%CI) Ptrend
  • *

    OR conditional on matched set adjusted for AFB1 exposure levels (corresponding OR and 95% CI values can be found in Supporting Table 7).

  • Combination of rs28383151 GA genotype and rs28383151 AA genotype.

  • OR conditional on matched set adjusted for AFB1 exposure years (corresponding OR and 95% CI values can be found in Supporting Table 7).

 GA1215.9583.91.74 (1.21-2.51)*3.00 × 10−3
 AA271.3231.53.07 (1.66-5.69)*3.57 × 10−4
 GA/AA1487.2815.41.98 (1.43-2.75)*3.59 × 10−5
MediumGG46022.532921.92.57 (2.08-3.18)*3.75 × 10−18
 GA753.7976.54.33 (3.04-6.18)*5.98 × 10−16
 AA291.4483.23.65 (2.18-6.11)*9.15 × 10−7
 GA/AA1045.11459.74.12 (3.02-5.61)*2.86 × 10−19
LongGG36918.048532.45.28 (4.29-6.51)*4.15 × 10−55
 GA291.41489.915.75 (10.16-24.42)*6.56 × 10−35
 AA60.3694.638.93 (16.33-92.85)*1.49 × 10−16
 GA/AA351.721714.519.61 (13.17-29.19)*1.24 × 10−48
Likelihood ratio test for interaction, P = 4.74 × 10−6
 GA1165.7372.51.04 (0.7-1.54)0.85
 AA291.4231.52.59 (1.47-4.55)1.00 × 10−3
 GA/AA1457.1604.01.35 (0.97-1.88)0.07
MediumGG54326.628619.11.72 (1.42-2.10)4.73 × 10−8
 GA613.01167.76.27 (4.47-8.79)1.70 × 10−26
 AA110.5281.98.27 (4.06-16.85)5.72 × 10−9
 GA/AA723.51449.66.57 (4.81-8.98)4.37 × 10−32
HighGG27813.648432.35.76 (4.71-7.03)2.56 × 10−66
 GA482.315010.010.26 (7.22-14.59)1.89 × 10−38
 AA221.1895.913.40 (8.25-21.79)1.17 × 10−25
 GA/AA703.423915.911.26 (8.36-15.18)6.19 × 10−57
Likelihood ratio test for interaction, P = 2.12 × 10−4

XRCC4 rs28383151 Polymorphism Was Correlated With XRCC4 Expression.

To investigate the potential effects of rs28383151 polymorphism on XRCC4 expression, we analyzed the association between this polymorphism and XRCC4 protein using immunohistochemistry (IHC) in the cancerous tissues of 1,499 HCC cases. The data showed that the genotypes with rs28383151 A alleles were significantly related to decreased XRCC4 expression in hepatocellular tumor tissues, compared with rs28383151-GG (Fig. 1A; P < 0.01). To further analyze this correlation, subjects were divided into three groups based on XRCC4 expression scores in the tumors, representing low (immunoreactive score [IRS]: 1-3), medium (IRS, 4-6), and high (IRS, >6) expression of XRCC4. Spearman's r test exhibited this polymorphism negatively related to the levels of XRCC4 protein (r = −0.242; Supporting Table 9). Representative photographs exhibit the aforementioned correlation between genotypes and expression levels (Fig. 1B). Moreover, mRNA levels of XRCC4 in cancerous tissues with rs28383151-GA or -AA were significantly lower than those with rs28383151-GG (Fig. 1C; P < 0.01). Together, these results suggest that this polymorphism should modify XRCC4 expression.

Figure 1.

XRCC4 rs28383151 polymorphism correlated with XRCC4 expression and DNA repair capacity. (A) XRCC4 protein expression was evaluated using the IHC scores of IRS system according to the following formula: IRS = SI × PP, as described elsewhere.36 XRCC4 expression scores are shown as box plots, with horizontal lines representing the median, the bottom and the top of the boxes representing the 25th and 75th percentiles, respectively, and vertical bars representing the range of data. We compared genotypes with rs28383151 A alleles (rs28383151-GA and rs28383151-AA) with rs28383151-GG using the Student t test. (B) Representative images show that different expression levels were observed in cancerous tissues from cases with different XRCC4 genotypes (scale bars: 50 μm). (C) Quantitative real-time PCR analysis of XRCC4 mRNA level in HCC tumor tissues from 75 subjects. Data were analyzed using the Student t test. Data are shown as means ± standard error of three independent experiments. (D) AFB1 DNA adducts formation in AFB1-treated QSG-7701 cells with overexpression of XRCC4 (see Supporting Materials). XRCC4 expression in AFB1-treated QSG-7701 cells was identified by western blotting. Levels of AFB1 DNA adducts were tested using comparative ELISA. Data were analyzed from three independent tests using the Student t test and are shown as means ± standard deviation. ast;P value (8.99 × 10−9) for the mutant type of XRCC4 rs28383151 (XRCC4mt) versus the WT of XRCC4 rs28383151 (XRCC4wt). (E) Percentage of TP53M-positive tumors in three groups with different genotypes of rs28383151 from 1,499 subjects. Data were analyzed using Pearson's chi-square test.

XRCC4 rs28383151 Polymorphism Was Associated With DNA Repair Capacity-Related AFB1-Induced DNA Damage.

To explore whether rs28383151 polymorphism effects DNA repair capacity, we analyzed the effects of a mutant type of rs28383151 polymorphism (XRCC4mt) on AFB1 DNA adducts levels in AFB1-treated QSG-7701 cells with overexpression of XRCC4 (see Supporting Materials). We found that XRCC4mt had higher AFB1 DNA adducts levels (0.68 ± 0.05 nmol/μg DNA), compared to the WT of rs28383151 polymorphism (XRCC4wt, 0.44 ± 0.05 nmol/μg DNA, P < 0.01; Fig. 1D; Supporting Table 10). Because TP53M is the most important molecular signature of AFB1-induced HCC,16 we investigated whether this polymorphism modified this mutation in the 1,499 cancerous tissue subjects. Results showed that risk genotypes of XRCC4 increased the frequency of TP53M (Fig. 1E), and corresponding risk values were 1.60 and 3.92 for rs28383151-GA and rs28383151-AA, respectively (Table 3). Taken together, these findings suggest that the rs28383151 polymorphism might correlate with DNA repair capacity for repair of DNA damage caused by AFB1 exposure.

Table 3. XRCC4 rs28383151 Polymorphism and TP53M Risk
 TP53M (−)TP53M (+)  
rs28383151n%n%OR (95% CI)* Ptrend
  • *

    Adjusted for age, sex, race, HBV and HCV infection, AFB1 exposure years, and AFB1 exposure levels (corresponding OR and 95% CI values can be found in Supporting Table 7).

  • Combination of rs28383151 GA genotype and rs28383151 AA genotype.

GA5915.924421.71.60 (1.16-2.20)4.28 × 10−3
AA133.512711.33.92 (2.17-7.08)5.94 × 10−6
GA/AA7219.437133.02.02 (1.51-2.70)2.24 × 10−6

XRCC4 rs28383151 Polymorphism Was Related to a Poor Prognosis of HCC Patients.

To assess the clinical relevance of rs28383151 polymorphism, we analyzed the survival follow-up information of all HCC patients. Among these subjects, 1,092 without decompensated cirrhosis received the same curative resection treatment, according to Chinese Manage Criteria of HCC,17 and were included for final survival analysis. Association analysis between risk genotypes (namely, genotypes with rs28383151 A alleles; rs28383151-GA/AA) or nonrisk genotype (rs28383151-GG) and the clinic-pathologic characteristics of HCC were first performed separately (Supporting Table 11). We observed a significant distribution difference of genotypes among different AFB1 exposure levels, tumor size, and recurrence status, but not in AFB1 exposure years, gender, minority, HBsAg, cirrhosis, or TNM stage (Supporting Table 11). Survival analysis next showed that HCC cases carrying rs28383151-GA/AA, compared with those with rs28383151-GG, had shorter RFS (median RFS time [MRT] was 16 versus 33 months; Fig. 2A) and higher recurring risk (Table 4), particularly under high AFB1 exposure conditions (Fig. 2A). Additionally, this polymorphism was related to the OS of HCC cases (Fig. 2B), and some evidence of multiplicative interaction was found for rs28383151 polymorphism and AFB1 exposure (Pinteraction < 0.05; Table 4).

Figure 2.

Association between survival and XRCC4 rs28383151 polymorphism in 1,092 HCC cases receiving curative treatment. (A) Genotypes with rs28383151 A alleles (rs28383151-GA/AA) were found to have a shorter MRT than rs28383151-AA (left). RFS of HCC patients (middle) with many AFB1 exposure years or (right) high AFB1 exposure levels was associated with rs28383151 genotypes. (B) Rs28383151 polymorphism modified median OS time (MST). Left: OS plot for all 1,092 HCC cases; middle: OS plot for these cases having long AFB1 exposure years (n = 363); right: and OS plot for these cases having high AFB1 exposure levels (n = 531). Cumulative hazard function was plotted by Kaplan-Meier's methodology, and P value was calculated with two-sided log-rank tests.

Table 4. Cox's Proportional Hazard Model Analysis for Multivariate Analysis of Potential Predictor Factors for HCC Cases With Curative Treatment
VariableHR (95% CI/P)*HR (95% CI/P)
  • *

    For diseases-free survival of HCC cases.

  • For OS of HCC cases.

Tumor size, cm  
 >51.49 (1.21-1.83/1.38 × 10−4)1.64 (1.38-1.95/2.53 × 10−8)
TNM stage  
 II2.28 (1.65-3.17/8.16 × 10−7)2.74 (2.10-3.58/1.51 × 10−13)
AFB1 exposure years
 Median2.98 (2.26-3.92/5.63 × 10−15)2.96 (2.32-3.78/2.41 × 10−18)
 Long5.87 (4.50-7.66/9.33 × 10−39)6.31 (4.99-7.97/3.08 × 10−53)
AFB1 exposure levels
 Median1.34 (1.10-1.70/5.47 × 10−3)1.10 (0.94-1.28/0.24)
 High2.78 (1.95-2.89/3.97 × 10−18)1.23 (1.06-1.42/5.06 × 10−3)
XRCC4 rs28383151 genotypes
 GA1.80 (1.52-2.14/2.07 × 10−11)1.02 (0.87-1.20/0.77)
 AA2.69 (2.12-3.42/4.21 × 10−39)1.40 (1.14-1.73/1.60 × 10−3)
Interaction of XRCC4 and AFB1 exposure years
 AA × long exposure years2.53 (1.35-4.73/3.70 × 10−3)1.98 (1.15-3.42/0.01)

XRCC4 rs28383151 Polymorphism Increased the Risk of PVT.

On the basis of a recent report showing that the dysregulation of XRCC4 is related to tumor metastasis,18 we investigated whether the rs28383151 polymorphism influenced the risk of PVT, the most common metastasis type of HCC,19, 20 in 1,092 HCC cases during follow-up after curative treatment (Table5). Results exhibited that rs28383151 A alleles significantly increased PVT risk (OR = 2.26; P = 3.94 × 10−5). In the stratified analysis based on tumor size, this risk role was only observed in these patients having more than 5-cm tumors, but not in those with less than 5-cm tumors (Supporting Table 12).

Table 5. XRCC4 Polymorphisms and PVT Risk
 PVT (−)PVT (+)  
rs28383151n%n%OR (95% CI)* Ptrend
  • *

    Adjusted for age, sex, race, HBV and HCV infection, tumor size, AFB1 exposure years, and AFB1 exposure levels (corresponding OR and 95% CI values can be found in Supporting Table 7).

  • Combination of rs28383151 GA genotype and rs28383151 AA genotype.

GA3713.616620.21.71 (1.12-2.61)1.23 × 10−2
AA51.88910.96.13 (2.40-15.69)1.55 × 10−4
GA/AA4215.425531.12.26 (1.53-3.33)3.94 × 10−5


In Guangxi Zhuang Autonomous Region, where more than 90% of HCCs develop, the major environmental factors are chronic infection with HBV and ingestion of foodstuffs contaminated with AFB1.3 AFB1, an important chemical carcinogen, is produced by fungi of the Aspergillus species. Because these fungi readily grow on such foodstuffs as corn and groundnuts stored in damp-hot conditions, high AFB1 exposure areas are distributed in tropical areas such as Guangxi Region. Once ingested, this carcinogen is mainly metabolized by cytochrome P450 into the genotoxic metabolic, AFB1-exo-8,9-epoxide (AFBO). AFBO can bind to DNA and results in such DNA damage as DSBs and then induces HCC carcinogenesis.4, 21, 22 Our present study also shows that HCC risk is related to different AFB1 exposure status. However, increasing epidemiological evidence has exhibited that although many people are exposed to the same risk factor, only a relatively small proportion of individuals with chronic AFB1 exposure develop HCC.23-26 These results imply that an individual susceptibility related to genetic factors (e.g., DNA repair capacity) might be correlated with the carcinogenesis of HCC caused by chronic AFB1 exposure, whereas NHEJ genes play an important role in the repair of DSBs resulting from exogenous attacks, such as AFB1, and might be important cancer-correlated genetic factors.27-29

XRCC4, located on chromosome 5q14.2, is an important NHEJ gene.7 The encoded protein of this gene consists of 336 amino acid residues (DDBJ/EMBL/Genbank accession no. AAD47298) and interacts directly with Ku70/Ku80 in the NHEJ pathway.30 It is hypothesized that XRCC4 serves as a flexible join between Ku70/Ku80 and its associated protein, Ligase IV.30 XRCC4 is required for precise end-joining of blunt DNA DSBs in mammalian fibroblasts, and the mutant, XRCC4, results in more-deficient NHEJ capacity.31 A gene-targeted mutation study has also shown that differentiating neurons and lymphocytes strictly require XRCC4 end-joining proteins. The targeted inactivation of this gene leads to late embryonic lethality accompanied by defective neurogenesis and defective lymphogenesis.32, 33 These results demonstrate that XRCC4 is essential for the DNA repair capacity of NHEJ.

More than 40 polymorphisms have been reported in the XRCC4 gene (SNP500Cancer database), some of which are correlated with malignant tumors, such as oral, gastric, and bladder cancers.7 In this study, we only analyzed 21 known SNPs in the coding region of this gene because these polymorphisms localize at conserved sites of this gene. They change the coded amino acids and may be associated with a decreased DNA repair capacity and an increased cancer risk.7 Among 21 SNPs, we only found that the rs28383151 polymorphism increased HCC risk; furthermore, this polymorphism would interact with AFB1 exposure status in the process of HCC induced by AFB1. Given that our several reported polymorphisms (including XPC codon 939, XPD codon 751, and GSTM1 polymorphisms) were also correlated with AFB1-related HCC risk, we analyzed the relative contribution of the XRCC4 rs28383151 polymorphism after adjusting for these three polymorphisms and observed a similar risk value. These data suggest that the rs28383151 polymorphism should be an important genetic susceptible factor of cancer risk. In addition, we found some evidence of XRCC4-GSTM1 interactive effects on HCC risk, possibly because this gene-gene interaction results in a more-obvious accumulation of DNA damage and, consequently, correlates with a higher risk for HCC.

To explore the possible pathogenesis that the rs28383151 polymorphism increases AFB1-related HCC risk, we analyzed the effects of this polymorphism on XRCC4 expression and DNA repair function. We found that rs28383151 A alleles were significantly associated with down-regulated levels of XRCC4 expression, including protein and mRNA expression. Regarding the association between rs28383151 polymorphism and DNA repair capacity, we elucidated this association using the levels of AFB1 DNA adduct and the frequency of TP53M. This was done primarily because AFB1 DNA adducts are a major type of DNA damage induced by AFB1 exposure and corresponding levels are related not only to AFB1 exposure, but also to DNA repair capacity,2, 4 whereas TP53M is the characteristic genetic change correlated with AFB1 exposure, and higher frequency of this mutation predicts higher AFB1 exposure and higher HCC risk.2, 16 This suggests that AFB1 DNA adducts and TP53M could be regarded as biomarkers of DNA repair function related to AFB1 exposure. Our results showed that the rs28383151 polymorphism increased the level of AFB1 DNA adducts and frequency of TP53M. Together, these findings suggest that this polymorphism may decrease the DNA repair capacity of the NHEJ pathway through modulating XRCC4 expression levels and function. The DNA damage induced by AFB1 cannot be repaired effectively, with higher adduct levels leading to induction of mutations, such as in the p53 gene, and higher risk of hepatocellular carcinogenesis. Therefore, the rs28383151 polymorphism may play an important role in the carcinogenesis of Guangxiese HCC caused by AFB1.

Another interesting finding of this study is that the rs28383151 polymorphism was associated with poor HCC prognosis, possibly because it correlates with the fact that this polymorphism increased the risk of PVT. Supporting our results, recent studies have shown that dysfunction of XRCC4 relates to tumor metastasis.18

When we were investigating the association between genetic polymorphisms and AFB1-related HCC, it was important to establish effective methods for collecting sufficiently large samples for gene-environment interaction analysis, avoiding effects of confounders, and evaluating AFB1 exposure information.15 In this study, the effects of possible confounders, including age, sex, race, and HBV and HCV infection status, were controlled with an individually matched design. In the stratified analysis, no significant interactive effects were found, suggesting that these factors should be effectually manipulated and not modify the correlation between the rs28383151 polymorphism and HCC related to AFB1 exposure.

This study had several limitations. Because of the hospital-based study, potential selection bias might have occurred. Because the liver disease itself may affect the metabolism of AFB1 and modify the levels of AFB1 DNA adducts, the increased risk with AFB1 exposure status noted in this study was probably underestimated. In spite of the relatively large sample sizes of our studies, the power to elucidate gene-environmental interactions was limited because of the very small magnitudes of the overall associations and the relatively low frequency of risk genotypes. Although the status of TP53M was investigated in cases of HCC, other mutations of the TP53 gene were not evaluated. Additionally, despite the analysis of 21 SNPs in the coding region of XRCC4, we did not analyze the polymorphisms of other genes involved in the NHEJ pathway possibly being able to modify the risk of AFB1 for HCC.5, 34 Therefore, more genes deserve further elucidation based on a large sample and the combination of genes and AFB1 exposure.

In summary, this study is, to the best of our knowledge, the first report that describes XRCC4 polymorphisms and their associations with AFB1-related HCC risk and prognosis. Our study showed that the rs28383151 polymorphism might modulate HCC risk and prognosis related to AFB1 exposure. Particularly, the association was stronger for gene-environmental interactions than for a single gene or environmental factor. Our findings might have prevention implications through identifying an at-risk population as well as add significant predictive value to the traditional predictors of cancer prognosis (e.g., stage and surgery) once these findings are replicated by other studies based on a larger scale or prospective studies.


The authors thank Drs. Qiu-Xiang Liang, Yun Yi, and Yuan-Feng Zhou for their sample collection and management and Dr. Hua Huang for his molecular biochemical technique. The authors also thank all members of the Department of Medical Test and Infective Control, Affiliated Hospital of Youjiang Medical College for Nationalities, for their help.