Association between HLA alleles and Epstein-Barr virus Zta-IgA serological status in healthy males from southern China.

BACKGROUND: Nasopharyngeal carcinoma (NPC), an Epstein-Barr virus (EBV) associated cancer, exhibits an extremely high incidence in southern Chinese. Given that human leukocyte antigen (HLA) plays critical roles in antigen presentation and relates to NPC susceptibility, it is speculated that certain HLA variants may affect EBV reactivation, which is a key pathogenic factor of NPC. Therefore, we attempted to identify HLA alleles associated with the indicator of EBV reactivation, Zta-IgA, in healthy males from NPC endemic area.
METHODS: HLA alleles of 1078 healthy males in southern China from the 21-RCCP study were imputed using genome-wide single nucleotide polymorphism data. EBV Zta-IgA in blood samples were measured using an enzyme-linked immunosorbent assay. Multiple logistic regression analysis was used to evaluate the effect of HLA allele on Zta-IgA serological status and its potential joint association with smoking. The binding affinity for Zta-peptide was predicted using NetMHCIIpan 4.0.
RESULTS: HLA-DRB1*09:01 was found to be associated with a higher risk of Zta-IgA seropositivity (odds ratio = 1.80, 95% confidence interval = 1.32-2.45; p = 1.82 × 10-4 ). Compared with non-smokers without HLA-DRB1*09:01, the effect size increased to 2.19- and 3.70-fold for the light and heavy smokers carrying HLA-DRB1*09:01, respectively. Furthermore, HLA-DRB1*09:01 showed a stronger binding affinity to Zta peptide than other HLA-DRB1 alleles.
CONCLUSIONS: Our study highlighted the pivotal role of genetic HLA variants in EBV reactivation and the etiology of NPC. Smokers with HLA-DRB1*09:01 have a significantly higher risk of being Zta-IgA seropositive, which indicates the necessity of smoking cessation in certain high-risk populations and also provide clues for further research on the etiology of NPC.

INTRODUCTION

Nasopharyngeal carcinoma (NPC) exhibits an extremely high ethnic and geographic disparity, with southern Chinese, especially in Guangdong, Guangxi and Hong Kong, showing an almost 50‐fold higher incidence than that in northern China or Caucasians. Males have two to three times the risk of NPC compared to females. , , The individual genetic susceptibility may convey higher risk of NPC through its main effect, , and interactions with EBV infection and certain environmental risk factors. Much effort has been made to investigate the genetic susceptibility of NPC. Previous studies, including our genome‐wide association study (GWAS) in Guangdong, , have found consistently strong association signals in human leukocyte antigen (HLA) region located on chromosome 6. In addition, it is reported that both HLA class I and class II alleles (eg. HLA‐A2, ‐A11, ‐B13, ‐B46, ‐B58, ‐C03, ‐C04, ‐C07, ‐DRB1*03, ‐DRB1*09 and ‐DRB1*12) were associated with NPC in a case–control study design. , , , , , Considering these genes are mainly responsible for antigen processing and presentation, it is speculated that some high‐risk HLA alleles may have an impact on host immune surveillance against EBV, resulting in the loss of virus–host homeostasis and an increased level of EBV reactivation. Exploring the association between HLA alleles and potential precancerous indicators of NPC in disease‐free subjects from NPC endemic areas may help uncover actual causal factors.

After primary infection, which usually occurred in early human life, EBV stays dormant for most of the time, , , whereas it can be reactivated by certain factors such as tobacco smoking. Antibodies against lytic EBV proteins including capsid antigen (VCA), BZLF1 transcription activator protein (Zta) and early antigen (EA) have shown great potential to stratify NPC cases and controls (area under the curve = 0.90–0.93). , Individuals with elevated levels of immunoglobulin (Ig)A antibodies against lytic EBV proteins have a 20‐ to 30‐fold higher risk of NPC. Longitudinal cohort studies have revealed that the elevated IgA level precedes the occurrence of NPC by 1–15 years. , Our former study conducted in northern and southern China revealed that the seropositive rate of EBV Zta‐IgA in males from NPC endemic areas (21‐RCCP and North populations in Guangdong) was 2.6 times of that in males from non‐endemic areas (Shanxi), after adjusting for age and education. Thus, EBV reactivation probably plays a pivotal role in the initiation and early stage of NPC. The most important protein triggering EBV latent‐lytic switch is Zta encoded by BZLF1, , which is transcribed during immediate‐early stage of the lytic cycle. Zta directly interacts with histone acetylating complexes, and initiates a cascade of early‐lytic gene transcription of EBV. It also plays a critical role in lytic EBV DNA replication. Therefore, it is noteworthy to identify the HLA alleles associated with Zta‐IgA, as an indicator for EBV reactivation, in NPC endemic populations.

We conducted the present study focusing on the associations between HLA alleles and Zta‐IgA in our male population with a higher level of EBV reactivation in Guangdong (21‐RCCP). Four‐digit HLA alleles were imputed using high‐dense single nucleotide polymorphism (SNP) array data from our former GWAS study. , We investigated the potential associations between all available HLA alleles and serum Zta‐IgA. Because smoking could induce EBV reactivation, we also examined the possible synergic effect of HLA alleles and smoking on Zta‐IgA.

MATERIALS AND METHODS

Study population and data collection

In total, 1498 healthy males were recruited from physical examination centers in 21 municipalities of Guangdong Province, during 2005–2007 (21‐RCCP). Details about the design, methods and subjects have been reported previously. , , All participants were Chinese Cantonese, lived in Guangdong province at the time of the study, and had no history of malignancy or immunological diseases. Each subject was interviewed face‐to‐face by well‐trained interviewers to obtain demographic characteristics (age, education level), family history of cancer, lifestyle behaviors (smoking and alcohol consumption). The structured questionnaire was mainly designed based on the questionnaires issued by the University of Arizona Cancer Center (https://cancercenter.arizona.edu/researchers/shared-resources/behavioral-measurement-and-interventions). For cigarette smoking, participants were divided into ever smokers and never smokers. Ever smokers were defined as having smoked at least 100 cigarettes during their lifetime. Detailed information on smoking was collected, including the amount of cigarettes consumption per day and the duration of exposure. Pack‐year was used as the cumulative measurement in the analysis. Participants consuming < 20 and ≥ 20 pack‐years were assigned as light and heavy smokers, respectively. The study was reviewed and approved by the Human Ethics Approval Committee at Sun Yat‐sen University Cancer Center. Informed consent was obtained from all participants.

Serologic test for EBV Zta‐IgA

As stated previously, , a 5–10‐ml blood sample was collected from each subject. Zta‐IgA serostatus was measured with a commercially available enzyme‐linked immunosorbent assay (ELISA) kit (Zhongshan Bio‐technology Co., Ltd, Zhongshan city, China). Zta was produced with purified recombinant peptides specified by BZLF1 (36 + 38‐kDa fine doublet). All samples were detected by the same technicians in the same laboratory of the Sun Yat‐Sen University Cancer Center to limit potential biases. The seropositive/seronegative status of Zta‐IgA was determined based on the ELISA optical density value in accordance with the manufacturer's instructions. To evaluate the reproducibility for the assay, we randomly selected 46 samples to conduct a test–retest assay for Zta‐IgA and calculated intraclass correlation coefficients (ICCs) and the agreement percentage. The ICCs of Zta‐IgA were 0.70 and the agreement percentage of Zta‐IgA was 82.6%.

Genotyping and imputation of HLA alleles

Of all the 1498 samples, genetic data of 1080 samples were available and retrieved. HLA allele genotyping was performed on an Infinium Global Screening Array and a Human610‐Quad BeadChip (Illumina, San Diego, CA, USA). Following the quality control and imputation procedures described previously, , samples were removed with genotyping call rates < 95%, duplicates or relatives, and population outliers detected by principal component analysis. Two samples were removed in the process. SNPs were removed with call rates < 95%, minor allele frequency (MAF) < 0.01 and deviation from Hardy–Weinberg equilibrium in controls (p < 10−7). Quality control filtering were conducted using PLINK, version 1.09. Imputation were performed within the extracted entire HLA region on chromosome 6 (29–34 Mb; NCBI Build 37) using SNP2HLA. We imputed four‐digit classical HLA alleles in three HLA class I genes (HLA‐A, HLA‐B and HLA‐C) and five HLA class II genes (HLA‐DRB1, HLA‐DQA1, HLA‐DQB1, HLA‐DPA1 and HLA‐DPB1) based on the HAN Chinese reference panel. Post‐imputation quality control was performed to removed variants with poor imputation quality (INFO score < 0.5, MAF < 0.005, call rate < 95%). We compared the frequencies of imputed alleles with the previously reported data in Southern Chinese population , and with that in the online HLA Frequency Database (http://allelefrequencies.net). The alleles with abnormal frequencies (discrepancy greater than 0.15) were removed. In the end, 1078 samples with 25 four‐digit alleles were included in the subsequent analysis.

Statistical analysis

Comparisons of demographic characteristics, family history, smoking status and alcohol consumption between Zta‐IgA seropositive and seronegative groups were performed using a chi‐squared test for categorical variables and a t test for continuous variables, respectively. Under additive assumption, the effect size (odd ratio [OR]) and p value of HLA dosage allele (0, 1 and 2) was calculated via logistic regression models adjusting for age and smoking status using PLINK, version 1.09. Because 25 four‐digit HLA alleles were included in our study, the significance threshold was set as 2.00 × 10−3 (0.05/25 = 2.00 × 10−3) by applying Bonferroni correction. Logistic regression analyses were used to examine the effect of HLA alleles (Present versus Absent) and smoking (never, light and heavy group) on Zta‐IgA seropositivity using SPSS, version 20.0 (IBM Corp., Armonk, NY, USA). p < 0.05 (two‐sided) was considered tatistically significant.

In silico binding affinity predictions using NetMHCIIpan 4.0

We employed NetMHCIIpan 4.0 to predict and to compare the binding affinity against protein Zta [Human gammaherpesvirus 4] (YP_401673.1) among HLA alleles. Based on the NN‐align, a neural network‐based approach combined with peptide sequence representation, the method is trained on an extensive dataset of over 500.000 measurements of Binding Affinity (BA) and Eluted Ligand mass spectrometry (EL), covering common HLA class II molecules. Using NetMHCIIpan 4.0, a sliding window of 1‐mer overlapping 15‐mer peptides based on the full‐length amino acid sequences of the Zta antigen (245 amino acid) was screened to identify peptides with strong and weak binding affinities. Binding affinities for each peptide are provided as the score of EL (Score_EL). Compared with a set of random natural peptides, the percentile rank value of the score of EL (%Rank_EL) was used to reflect the strength of binding affinities. The threshold for strong binding peptides (%Rank_EL) is 2% and the threshold for weak binding peptides (%Rank_EL) is 10%. Peptides with %Rank_EL > 10% were filtered. The resulting sequence motifs were visualized using WebLogo.

RESULTS

Basic characteristic of the study population

In total, 1078 healthy males were included in the analyses with a mean age of 49.79 years. The positive rate of Zta‐IgA was 15.4% (n = 166). Almost 70% (n = 750) of these individuals had ever smoked. Other characteristics of the participants are presented in Table 1.

TABLE 1

Characteristics of 1078 healthy males in an NPC endemic area

Variables	All, n (%)	EBV Zta‐IgA seropositive, n (%)	EBV Zta‐IgA seronegative, n (%)	p value a
Total	1,078	166 (15.4)	912 (84.6)
Age (mean ± SD)	49.79 ± 10.91	52.79 ± 10.57	49.24 ± 10.89	< 0.001
Education level				0.092
Primary school or less	137 (12.8)	27 (16.5)	110 (12.1)
High school	608 (56.6)	97 (59.1)	511 (56.2)
University or more	329 (30.6)	40 (24.4)	289 (31.8)
Smoking status				0.017
never	328 (30.4)	37 (22.3)	291 (31.9)
ever	750 (69.6)	129 (77.7)	621 (68.1)
Alcohol consumption				0.436
never	464 (43.0)	77 (46.4)	387 (42.4)
<1 drink per week	395 (36.6)	61 (36.7)	334 (36.6)
≥1 drink per week	219 (20.3)	28 (16.9)	191 (20.9)
Family history of NPC				0.448
no	992 (95.8)	153 (94.4)	839 (96.1)
yes	43 (4.2)	9 (5.6)	34 (3.9)
Family history of cancer				0.656
no	759 (73.3)	116 (71.6)	643 (73.7)
yes	276 (26.7)	46 (28.4)	230 (26.3)

A t test and chi‐squared test were used to compare the mean of continuous factors and proportions of categorical factors between the EBV Zta‐IgA seronegative group and the EBV Zta‐IgA seropositive group.

The EBV Zta‐IgA seropositive individuals were significantly older than the seronegative individuals (52.79 ± 10.57 versus 49.24 ± 10.89, p < 0.001). A significantly higher proportion of ever‐smokers were found in the Zta‐IgA seropositive group, with a proportion of 77.7% compared to 68.1% in the Zta‐IgA seronegative group (p = 0.017). No significant differences were observed for other characteristics, including education levels, alcohol consumption, family history of cancer and family history of NPC between the seropositive and seronegative groups (Table 1).

HLA‐DRB1*09:01 is related to EBV Zta‐IgA serological status in healthy males from southern China

After genotyping and imputation, 25 four‐digit HLA alleles of eight genes were ultimately included in the subsequent analysis. In the process of quality control, some common HLA‐A, HLA‐B and HLA‐C alleles in the Southern Chinese population, such as HLA‐A*02:07, HLA‐A*11:01, HLA‐B*46:01 and HLA‐C*01:02, were filtered as a result of poor imputation quality or abnormal frequencies. Thus, in our study population, the frequencies of the most common alleles for each HLA gene were: for HLA‐A: HLA‐A*26:01 (0.019); for HLA‐B: HLA‐B*13:01 (0.079) and HLA‐B*58:01 (0.079); for HLA‐C: HLA‐C*07:02 (0.228); for HLA‐DPA1: HLA‐DPA1*02:01 (0.077); for HLA‐DPB1: HLA‐DPB1*21:01 (0.053); for HLA‐DQA1: HLA‐DQA1*06:01 (0.134); for HLA‐DQB1: HLA‐DQB1*05:02 (0.180); for HLA‐DRB1: HLA‐DRB1*09:01 (0.131).

When compared between Zta‐IgA seropositive and negative males, HLA‐DRB1*09:01 showed much higher frequencies in the seropositive group compared to the seronegative group (0.202 versus 0.125; p = 1.58 × 10−4) and HLA‐DPB1*21:01 showed much lower frequencies in the seropositive group compared to the seronegative group (0.021 versus 0.054; p = 0.011).

After adjusting for age and smoking, we found that only HLA‐DRB1*09:01 was significantly associated with a higher risk of Zta‐IgA seropositivity (OR = 1.80, 95% confidence interval [CI] = 1.32–2.45; p = 1.82 × 10−4) when applying Bonferroni correction. Other 24 HLA alleles were not found to be associated with Zta‐IgA serological status (Table 2).

TABLE 2

Association between HLA alleles and EBV Zta‐IgA serological status

HLA alleles	Allele frequencies			OR (95% CI) a	p value
	Total (n = 1078)	Zta‐IgA seropositive (n = 166)	Zta‐IgA seronegative (n = 912)
HLA‐A
26:01	0.019	0.030	0.018	1.85 (0.88–3.88)	0.104
HLA‐B
13:01	0.079	0.066	0.076	0.88 (0.54–1.43)	0.610
58:01	0.079	0.054	0.081	0.68 (0.41–1.13)	0.134
07:05	0.022	0.012	0.026	0.48 (0.17–1.34)	0.158
13:02	0.021	0.009	0.025	0.38 (0.11–1.23)	0.106
HLA‐C
07:02	0.228	0.259	0.226	1.16 (0.88–1.54)	0.290
08:01	0.115	0.090	0.113	0.81 (0.54–1.22)	0.309
03:02	0.068	0.054	0.069	0.79 (0.48–1.31)	0.365
15:02	0.035	0.027	0.033	0.75 (0.36–1.53)	0.422
04:03	0.022	0.015	0.026	0.60 (0.23–1.55)	0.292
HLA‐DPA1
02:01	0.077	0.051	0.081	0.67 (0.40–1.11)	0.116
HLA‐DPB1
21:01	0.053	0.021	0.054	0.38 (0.17–0.82)	0.014
HLA‐DQA1
06:01	0.134	0.142	0.130	1.11 (0.78–1.56)	0.568
01:04	0.086	0.066	0.088	0.74 (0.47–1.18)	0.205
05:01	0.064	0.057	0.067	0.84 (0.51–1.39)	0.492
HLA‐DQB1
05:02	0.180	0.178	0.173	1.00 (0.74–1.36)	0.986
06:01	0.121	0.090	0.130	0.66 (0.44–1.00)	0.048
02:01	0.065	0.060	0.068	0.88 (0.54–1.43)	0.599
03:02	0.059	0.066	0.058	1.13 (0.70–1.81)	0.620
05:03	0.054	0.036	0.056	0.62 (0.34–1.14)	0.122
05:01	0.033	0.027	0.036	0.76 (0.37–1.53)	0.437
06:09	0.012	0.012	0.013	0.96 (0.33–2.84)	0.947
HLA‐DRB1
09:01	0.131	0.202	0.125	1.80 (1.32–2.45)	1.82 × 10 −4
12:02	0.130	0.130	0.127	1.03 (0.72–1.48)	0.869
13:02	0.019	0.018	0.020	1.00 (0.41–2.42)	0.996

Odds ratios (ORs) and 95% confidence intervals (CIs) of HLA dosage alleles were calculated using logistic regression adjusting for two covariates (age, smoking status) under additive assumption.

Bold indicates significant p values after Bonferroni correction for multiple testing (0.05/25 = 2.00 × 10−3).

HLA‐DRB1*09:01 enhanced the effect of smoking on Zta‐IgA seropositivity

To investigate the potential joint association of smoking and HLA‐DRB1*09:01 with Zta‐IgA serological status, ORs were calculated for heavy and light‐smokers (≥ 20 and < 20 pack‐years) with or without HLA‐DRB1*09:01 compared to the never‐smokers without HLA‐DRB1*09:01, which was set as the reference group. For those without HLA‐DRB1*09:01, heavy smokers (≥ 20 pack‐years) had a moderate risk of Zta‐IgA seropositivity (OR = 1.72, 95% CI = 1.02–2.90; p = 0.041), whereas the OR for light smokers (< 20 packy‐years) was not significant (OR = 1.19, 95% CI = 0.67–2.09; p = 0.554) compared to the never‐smokers without HLA‐DRB1*09:01.

However, when smoking was combined with potential risk allele HLA‐DRB1*09:01, the effect size increased almost two‐fold, especially for heavy smokers. The heavy smokers carrying HLA‐DRB1*09:01 had a 3.70‐fold increased risk of Zta‐IgA seropositivity (95% CI = 2.01–6.80; p < 0.001) compared to never‐smokers without HLA‐DRB1*09:01, and light smokers carrying HLA‐DRB1*09:01 also had a 2.19‐fold increased risk of Zta‐IgA seropositivity (95% CI = 1.15–4.15; p = 0.017) (Figure 1).

FIGURE 1

The effect of smoking and HLA‐DRB1*09:01 on EBV Zta‐IgA serological status. We divided smoking categories by cumulative amount (never, < 20, ≥ 20 pack‐years) for the analysis. Never smokers without HLA‐DRB1*09:01 were defined as the reference group. Odds ratios (ORs) for the Zta‐IgA seropositive group (Zta‐IgA+) versus Zta‐IgA seronegative group (Zta‐IgA‐) among different carrier status of HLA‐DRB1*09:01 and smoking status were calculated using logistic regression models. Test was adjusted for age. The test was two‐sided and p < 0.05 was considered statistically significanr. Squares indicate study‐specific odds ratios; horizontal lines indicate study‐specific confidence intervals (CIs); the solid vertical line indicates an odds ratio of 1.0

Zta protein showed a stronger binding affinity for HLA‐DRB1*09:01 in silico predictions

Given the function of HLA‐DR in EBV infection, we predicted and compared the binding affinity for Zta protein among three successfully imputed HLA‐DR alleles (HLA‐DRB1*09:01, HLA‐DRB1*12:02 and HLA‐DRB1*13:02). Table 3 shows the peptides predicted to have strong binding affinities for HLA‐DR alleles, as well as how these 9‐mer peptides cores were selected. Among the total 231 peptides of the Zta protein, there were twelve peptides of Zta protein exhibiting potential strong binding affinity (%Rank_EL ≤ 2%) for HLA‐DRB1*09:01 in the prediction analysis. By contrast, only three peptides for HLA‐DRB1*12:02 and one peptide for HLA‐DRB1*13:02 showed potentially strong binding affinity. The complete results of the prediction analysis showing all Zta‐peptides with any binding affinity (%Rank_EL ≤ 10%) among the total 231 peptides for HLA‐DRB1 alleles is provided in the Supporting information (Table S1). Furthermore, WebLogo plots (see Supporting information, Figure S1) showed that the sequence motif for 9‐mer core peptides that bind preferentially to HLA‐DRB1*09:01, which was particularly rich in hydrophobic and polar AAs, presents a significantly different binding motif pattern compared to those of HLA‐DRB1*12:02 or HLA‐DRB1*13:02.

TABLE 3

The EBV Zta‐peptides with predicted strong binding affinities for the HLA‐DRB1 alleles

HLA allele	Pos a	Zta‐peptide	Zta‐peptide Core	Score _EL b	%rank _EL c	Bind Level d
HLA‐DRB1*09:01
	197	LQHYREVAAAKSSEN	YREVAAAKS	0.78	0.29	Strong
	196	LLQHYREVAAAKSSE	YREVAAAKS	0.73	0.47	Strong
	198	QHYREVAAAKSSEND	YREVAAAKS	0.70	0.60	Strong
	83	ENAYQAYAAPQLFPV	YQAYAAPQL	0.67	0.77	Strong
	82	PENAYQAYAAPQLFP	YQAYAAPQL	0.67	0.78	Strong
	60	QLTAYHVSTAPTGSW	YHVSTAPTG	0.67	0.80	Strong
	61	LTAYHVSTAPTGSWF	YHVSTAPTG	0.66	0.84	Strong
	59	GQLTAYHVSTAPTGS	YHVSTAPTG	0.66	0.87	Strong
	195	QLLQHYREVAAAKSS	YREVAAAKS	0.61	1.19	Strong
	62	TAYHVSTAPTGSWFS	YHVSTAPTG	0.55	1.67	Strong
	81	APENAYQAYAAPQLF	YQAYAAPQL	0.54	1.80	Strong
	84	NAYQAYAAPQLFPVS	YQAYAAPQL	0.53	1.86	Strong
HLA‐DRB1*12:02
	174	ELEIKRYKNRVASRK	IKRYKNRVA	0.36	1.40	Strong
	190	RAKFKQLLQHYREVA	FKQLLQHYR	0.35	1.48	Strong
	189	CRAKFKQLLQHYREV	FKQLLQHYR	0.32	1.98	Strong
HLA‐DRB1*13:02
	174	ELEIKRYKNRVASRK	IKRYKNRVA	0.49	1.17	Strong

Pos, peptide position (starting from 0).

Score_EL, raw eluted ligand likelihood prediction score.

%Rank EL, rank of the predicted eluted ligand likelihood score compared to a set of random natural peptides.

Binding level: Strong = strong binding. The threshold for strong binding peptides (%Rank_EL) is 2%. Peptides with %Rank_EL > 2% are not shown.

Bold indicates the 9‐mer core of EBV Zta‐peptide showing strong binding affinity for the corresponding HLA‐DRB1 allele.

DISCUSSION

Because EBV reactivation probably plays pivotal roles in NPC etiology, Zta‐IgA, an indicator of EBV reactivation, may present a risk for NPC in high‐endemic areas. In case–control studies in NPC endemic areas, such as Zhongshan City and Taiwan Province, consistent evidence has been observed indicating that increased Zta‐IgA seropositive rates are associated with NPC. , , Previous studies have provided evidence for genetic variants in the HLA region that may affect host immune surveillance against EBV being associated with the risk of NPC. , Therefore, association analyses exploring potential HLA alleles affecting Zta‐IgA in a high‐risk population of NPC were more conducive for revealing the pathogenesis of NPC. The present study was conducted in males from Guangdong, China, where the incidences of NPC were the highest worldwide. Consistently, the seropositive rate of Zta‐IgA is 15.4% in our study subjects, and much higher than thre rate of 6.3% in NPC non‐endemic areas. We found an HLA class II allele, HLA‐DRB1*09:01, was significantly associated with Zta‐IgA seropositivity in Guangdong healthy males. In addition, the known association between smoking and Zta‐IgA tended to be enhanced by HLA‐DRB1*09:01 because smokers with HLA‐DRB1*09:01 exhibit a 119–270% higher risk of Zta‐IgA seropositivity compared to never‐smokers without HLA‐DRB1*09:01. Our further epitope prediction bioinformatics analysis showed that epitopes in HLA‐DRB1*09:01 had a stronger binding affinity for peptides of Zta.

HLA‐DRB1*09 has been reported as being related to NPC in Asian populations as a risk allele (OR = 1.37). , This is consistent with our results, indicating that HLA‐DRB1*09 alleles may contribute to the risk of NPC via its positive impact on EBV reactivation. As a cell surface receptor binding to T cell receptor, HLA‐DR is involved in presenting peptides from outside the cell to CD4+ T lymphocytes (helper T cells), which in turn stimulate B cells to produce antibodies. Furthermore, HLA‐DR is also considered to be the co‐factor of EBV infection of B lymphocytes, which assists EBV in infecting cells by binding with EBV‐specific glycoprotein gp42. Our epitope prediction analysis proposed that HLA‐DRB1*09:01 might have more potential to bind Zta antigen and thereby elevate the efficiency of cell surface antigen presentation to T cell receptors. We considered that this would sequentially induce a higher level of IgA response. Further laboratory experiments are still required to confirm the HLA‐DRB1*09:01 restricted EBV epitopes.

We did not observe any significantly associations between other NPC‐associated HLA alleles such as HLA‐A*26:01 and Zta‐IgA serological status in our study. This may partly be a result of their low frequency. Nevertheless, two genome‐wide association studies focusing on seroreactivity to EBV consistently found the strongest signals in the HLA class II region, especially in the HLA‐DRB region. , In addition, previous studies focusing on other EBV‐associated diseases such as Hodgkin's lymphoma and multiple sclerosis in Europe and America provided similar results, with anti‐EBV IgG being significantly related to HLA‐DRB1 alleles. , , However, Zta‐IgA, representing EBV reactivation and risk of NPC, had seldom been studied. Determination of the Zta‐IgA related HLA allele may help to identify high‐risk groups and provide clues for the etiology of NPC in Southern China.

In our results, ever‐smokers were more likely to be Zta‐IgA seropositive than never smokers, which was consistent with previous reports. Experimental studies also confirmed that cigarette smoke extract could promote EBV reactivation and replication, possibly as a result of its function in inflammation and the cell‐mediated immune response. , , , Although there were no statistically significant interactions between HLA‐DRB1*09:01 and smoking on Zta‐IgA seropositivity, our study provided evidence for a joint association of HLA‐DRB1*09:01 and cigarette smoking with Zta‐IgA seropositivity. In comparison with never‐smokers without HLA‐DRB1*09:01, the effect of smoking on Zta‐IgA serological status increased when carrying HLA‐DRB1*09:01, especially for the heavy smokers (OR = 3.70). These findings suggest that some HLA genetic variations might aggravate the effect of cigarette smoking on Zta‐IgA seropositivity. Smoking affects both innate and adaptive immune responses by aggravating pathological immune responses or weakening defense immunity. Interactions between HLA alleles and smoking have been observed in EBV‐associated diseases such as multiple sclerosis. , It has been found that smoking triggers HLA‐DR shared epitope‐specific immune reactions in rheumatoid arthritis, indicating that the effect of smoking in immune response may be regulated by HLA alleles. , Further functional research is still warranted aiming to uncover the underlying biological mechanism of the synergistic effect of HLA alleles and smoking on EBV activation.

Because HLA alleles in the present study were genotyped from GeneChips (Illumina, San Diego, CA, USA) data, we were hardly able to evaluate extremely rare variants. In addition, the imputation was performed on two GeneChips using the HAN reference panel, resulting in some common alleles in the Southern Chinese population, such as HLA‐A*11:01, HLA‐A*02:07, HLA‐B*46:01 and HLA‐C*01:02, having to be excluded after quality control. However, the imputation was conducted based on the most common reference panel of the HAN population and under strict quality control. Thus, the four‐digit HLA alleles imputed were reliable in our study subjects. Moreover, multicenter studies with a large sample size are still warranted to explore the effect of HLA alleles, especially those rare HLA alleles with respect to Zta‐IgA seropositivity in NPC endemic areas.

In summary, we found that an NPC risk allele, HLA‐DRB1*09:01, is an important genetic factor for EBV reactivation and may play a role in NPC development. This risk allele could exaggerate the effect of smoking with respect to EBV reactivation, as represented by an increased seropositive rate of Zta‐IgA. The strong binding affinity to Zta peptide confirmed the specificity of HLA‐DRB1*09:01 in the process of Zta antigen presentation. Our study highlighted the important roles of HLA variations regarding the etiology of NPC in southern China. From a public health perspective, determination of this high‐risk HLA allele may help to identify individuals at risk of NPC. Future approaches for inhibiting EBV reactivation may serve as a method of preventing NPC in individuals with a higher potential to reactivate EBV. The synergy of HLA‐DRB1*09:01 and smoking also indicates the necessity of smoking cessation in certain high‐risk groups of NPC and provides clues for further research on the etiology of NPC.

CONFLICT OF INTEREST

The authors declare that they have no conflicts of interest.

AUTHOR CONTRIBUTIONS

W‐HJ was responsible for study conceptualization. L‐LY, C‐MD, W‐QX, Y‐QH and T‐MW were responsible for the methodology. L‐LY and C‐MD were responsible for software. L‐LY, C‐MD, Y‐JJ, W‐QX, Y‐QH and T‐MW were responsible for study validation. L‐LY and C‐MD were responsible for the formal analysis. Y‐JJ, TZ, Z‐YW, YL, M‐QZ, L‐JC, W‐LZ, D‐HL, R‐WX, L‐TL, X‐TT, Y‐XW and J‐WH were responsible for the investigation. Y‐QH and T‐MW were responsible for data curation. L‐LY, C‐MD and W‐QX were responsible for writing – original draft preparation. W‐HJ, W‐QX, Y‐QH, T‐MW, and D‐WY were responsible for writing – review and editing. L‐LY was responsible for visualization. J‐BZ was responsible for supervision. W‐QX and Y‐QH were responsible for project administration. All authors have read and approved the final version of the manuscript sumbitted for publication.

Table S1. The EBV Zta‐peptides with predicted strong or weak binding affinities for the HLA‐DRB1 alleles

Figure S1. Sequence motifs that characterize 9‐mer core epitopes for EBV Zta‐peptides that are predicted to bind preferentially to the corresponding HLA‐DRB1 allele. (a) HLA‐DRB1*09:01, (b) HLA‐DRB1*12:02 and (c) HLA‐ DRB1*13:02. Binders were defined as those with ≤ 10%Rank_EL for alleles (see Supporting information, Table S1). In silico predictions for 15‐mer peptides were generated across Zta protein using NetMHCIIpan 4.0. Sequences were plotted using WebLogo, as referenced in the main text. The colors of the amino acids (AA) correspond to their chemical properties: polar AA are shown in green (G, S, T, Y, C; note G is non‐polar when it forms a bond with other AA) or purple (Q, N); basic AA in blue (K, R, H); acidic AA in red (D, E); hydrophobic AA in black (A, V, L, I, P, W, F, M). The size of the letter relates to the frequency with which the AA is found in this position of core 9‐mers, the overall peak height on the y‐axis indicates the degree of conservation for specific epitopes at this location (measured in bits).

Click here for additional data file.