HLA-A*02 is associated with a reduced risk and HLA-A*01 with an increased risk of developing EBV⁺ Hodgkin lymphoma

It is now well established that approximately 30% to 40% of patients with Hodgkin lymphoma (HL) in the Western world and a far higher proportion in some developing regions carry the Epstein Barr Virus (EBV) in the malignant Hodgkin Reed Sternberg (HRS) cells.^1-3  EBV infection is an early event in the development of HL as the viral genomes are found in a monoclonal form, indicating that infection of tumor cells has occurred before their clonal expansion.⁴ 

Genetic predisposition studies in patients with sporadic and familial HL include a large number of significant associations with the human leukocyte antigen (HLA) region.^5-9  It can be speculated that HLA associations may differ between patients with EBV⁺ and EBV⁻ HL based on presentation of EBV-derived peptides. However, the vast majority of these association studies have not stratified patients with HL according to EBV status. We have recently analyzed DNA samples obtained from more than 200 patients with HL and family-based control participants with a set of microsatellite markers covering the entire HLA region.¹⁰  This genotyping study revealed an association with markers mapping in part of the HL class I region for patients with EBV⁺ HL and not with patients with EBV⁻ HLA. No associations were identified upon stratification of the patients based on histologic subtype. Fine screening of the associated region with single nucleotide polymorphism (SNP) markers using 2 different populations with HL (Dutch and British) demonstrated that HLA-A was the most likely functional candidate gene in this region.¹¹  This indicates that antigenic presentation of EBV-derived peptides in the context of HLA-A may be involved in the pathogenesis of EBV⁺ HL.

A limited set of viral latent proteins is expressed in EBV⁺ HRS cells, which differs from normal EBV-transformed B cells. This so-called latency type II pattern includes EBNA1, LMP1, and LMP2, and 2 small EBV-encoded RNAs (EBER) 1 and 2. The immunodominant EBNA3 is not expressed by HRS cells.¹²  HRS cells in most EBV⁺ patients with HL show expression of HLA class I¹³  and the proteasome components as well as TAP1/TAP2, indicating functional HLA-A peptide presentation. Given the ability of CD8⁺ T cells to target LMP2- and a minority of LMP1-derived peptides in healthy donors,^14-19  an effective cytotoxic T lymphocyte (CTL) response against EBV⁺ HRS cells can be expected. Based on our genetic screening results and the fact that certain HLA-A allele(s) prohibit efficient immune responses, it can be hypothesized that certain HLA-A alleles are predisposed to the development of EBV⁺ HL.

In this study, we genotyped the exons 2 and 3 of the HLA-A gene for 32 SNPs in 70 patients with EBV⁺ HL, 31 patients with EBV⁻ HL, and 59 control participants. The genotyping results were confirmed by immunohistochemistry (IHC) using an HLA-A*02–specific antibody and an HLA-A*02–specific polymerase chain reaction (PCR) on a much larger group (152 patients with EBV⁺ HL and 322 patients with EBV⁻ HL). The genotyping approach is based on comparing the total number of alleles, taking into account the fact that homozygous genotypes have 2 identical alleles and heterozygous genotypes have 2 different alleles. The HLA-A*02–specific IHC and PCR approaches cannot discriminate between HLA-A*02 homozygous and heterozygous samples. Therefore, the genotyping approach has more power to detect significant differences on smaller populations compared with IHC and HLA-A allele–specific PCR approaches.

The Dutch study population consisted of 70 EBV⁺ and 31 EBV⁻ HL patients, diagnosed between 1987 and 2000, and 59 family-based healthy controls, recruited from the white population of the Northern Netherlands and has been described previously.¹⁰  All these samples were genotyped and 30 EBV⁺ and 29 EBV⁻ HL patients for whom frozen tissue sections were available were also used for HLA-A*02 status determination by IHC. The frequency of HLA-A*02 positivity in 6107 blood bank donors from the Northern and Eastern part of the Netherlands was used as a control. The HLA-A*02 frequency in the United Kingdom population is about 50%,²⁰  and is comparable to the Dutch population. HLA-A*02 status was also determined in a United Kingdom study population recruited from Scotland and the Northern region of England. The United Kingdom series consisted of 415 patients, of whom 122 were EBV-associated, 219 were male, 243 were in the age range of 15 to 34 years, and 98 were aged 50 years or older. There were 80 patients with mixed cellularity HL (MCHL) and 307 patients with nodular sclerosis HL (NSHL). A total of 297 of these patients were derived from the population-based Scotland and Newcastle Epidemiological study of Hodgkin's Disease (SNEHD) study¹ ; the distribution of cases by age group, sex, EBV status, and histologic subtype was similar in the SNEHD patients and the overall United Kingdom patient group. The Dutch and United Kingdom populations used are both relatively stable with no or less ethnic variation. All subjects (patients and control participants) had given written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by both local and multicenter research ethics committees in the United Kingdom and the medical ethics board of the University Medical Center Groningen. HLA-A*02 status was determined by IHC or HLA-A*02–specific PCR on 152 patients with EBV⁺ HL and 322 patients with EBV⁻ HL.

Exons 2 and 3, encoding the peptide binding groove of the HLA-A gene, were sequenced using 2 different methods. A total of 70 samples (39 patients with EBV⁺ HL and 31 control participants) were sequenced by amplification of the HLA-A gene using primer sets described previously.²¹  PCR reactions were carried out in a final volume of 50 μL, containing approximately 125 ng DNA and 5 U Taq DNA polymerase (Amersham Pharmacia Biotech, Uppsala, Sweden), 10 mM Tris-HCl (pH 9.0), 50 mM KCl, 2.5 mM MgCl_2, 0.2 mM dNTPs (Roche Diagnostics, Mannheim, Germany), and 0.25 μM of each primer (Sigma, Malden, the Netherlands). Thermal cycling was performed on a PTC-225 Thermal cycler (MJ Research, Waltham, MA). Step-down PCR started with incubation for 5 minutes at 95°C followed by 20 cycles of 45 seconds at 94°C, 45 seconds at 68°C, and 4 minutes at 72°C, 5 of the same cycles at an annealing temp of 67°C, and 5 cycles at 65°C. The final cycle ended with 20 minutes at 72°C. The quality and the length of the PCR products were checked on 1% agarose gels and products were then purified with an agarose gel extraction kit (Qiagen, Venlo, the Netherlands). PCR products were sequenced using forward and reverse primers for exon 2 (forward, 5′-CAGACGCCGAGGATGGCC-3′; reverse, 5′-CTCGGACCCGGAGACTGT-3′) and exon 3 (forward, 5′-ACAGTCTCCGGGTCCGAG-3′; reverse, 5′-GGATTCCTCTCCCTCAGGAC-3′) following the standard protocol recommended for the MegaBace 1000 capillary sequencer (Amersham Pharmacia Biotech). Sequences were analyzed using Seqman (DNA Star, Madison, WI).

The remaining samples (31 patients with EBV⁺ HL, 31 patients with EBV⁻ HL, and 28 control participants) were amplified with an HLA-A PCR/Sequencing kit according to the manufacturer's instructions (Atria Genetics, San Francisco, CA). Exons 2 and 3 were sequenced following the instructions recommended for the ABI Prism 3130 genetic analyzer (Applied Biosystems, Warrington, United Kingdom). Sequences were analyzed using the SBT engine program version 1.22.1.1.425 and the IMGT/HLA database version 2.12 (Genome Diagnostics BV, Utrecht, the Netherlands).

Based on HLA sequence data from the IMGT/HLA database²²  of the international ImMunoGeneTics project, 11 SNPs in exon 2 and 21 SNPs in exon 3 showing different alleles between the 3 most common HLA-A types (*01, *02, and *03) were selected (Table 1). Most of these SNPs (26 of 32) result in nonsynonymous amino acid residues. According to Bjorkman and Parham²³  and Reche and Reinherz²⁴ , 24 SNPs have potential peptide and/or T-cell receptor (TCR) contact positions.^23,24  Genotypes of the selected 32 SNPs discriminating between the HLA-A*01, *02, and *03 types were determined for all samples of the Dutch study population (70 patients with EBV⁺ HL, 31 patients with EBV⁻ HL, and 59 family-based healthy control participants). Haplotypes were generated based on these 32 SNPs using the 2SNP program.²⁵  Haplotype association analysis was performed by comparison of the frequencies of the different haplotypes between patients with EBV⁺ HL, patients with EBV⁻ HL, and control participants using the Fisher exact test to assess the significance of the difference in distribution. Differences in frequencies per haplotype and genotype were calculated by χ² or Fisher exact tests. HLA-A alleles of the most frequent haplotypes on a level of 2 digits were determined using the SCORE program (version 2.13)²⁶ . This program can distinguish between all different HLA-A alleles and indicates if a given sequence is specific for only 1 HLA-Aallele or not. Using the 32 selected SNPs, we can reliably discriminate between HLA-A*01, *02, and *03 alleles.

HLA-A*02 status was determined by IHC or HLA-A*02–specific PCR on 152 patients with EBV⁺ HL and 322 patients with EBV⁻ HL. IHC using the clone BB7.2 (AbB Serotec, Oxford, United Kingdom) was performed on frozen tissue sections of 30 patients with EBV⁺ HL and 29 patients with EBV⁻ HL from the North Netherlands population, which identifies HLA-A*02 and the much less abundant HLA-A*28.²⁷  HLA-A*02 PCR is specific for this specific HLA-A type and does not yield PCR products for other HLA-A alleles. This PCR was performed on DNA samples from 122 patients with EBV⁺ HL and 293 patients with EBV⁻ HL from the United Kingdom. Primers, 5′-labeled with [(2,5-Dioxo-1-pyrrolidinyl)oxo]carbonyl]-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (FAM), were specific for exons 2 and 3 of the HLA-A*02 gene and resulted in an amplicon of 813 bp.²⁷  Primers that were 5′-labeled with 4,7,2′,4′,5′,7′-hexachloro-6-carboxyfluorescein (HEX), specific to the β₂ microglobulin gene, and amplifying a 330-bp fragment were also included in each assay. DNA from the JY cell line, which is homozygous for HLA-A*0201, was used as a positive control in each assay; a negative control consisting of water was included after 2 samples. Reactions were performed in a 50-μL volume containing 1 [tims] HotStarTaq Mastermix (Qiagen, Crawley, United Kingdom) and 1.0 μM primers. Thermal cycling was performed using an Applied Biosystems GeneAmp PCR System 9700 as follows: initial denaturation at 95°C for 15 minutes; 5 cycles of 95°C for 1 minute, 65°C for 1 minute, and 72°C for 2 minutes; 25 cycles of 95°C for 1 minute, 55°C for 1 minute, and 72°C for 2 minutes; followed by a final polymerization step at 72°C for 7 minutes. PCR products were analyzed on a 3100 Genetic Analyzer with the GS 1000 ROX marker using GeneScan Analysis software v3.7 (Applied Biosystems). The frequency of HLA-A*02 positivity was compared between patients with EBV⁺ HL, patients with EBV⁻ HL, and with the 6107 control participants using χ² tests to assess significant differences.

Genotypes of the selected 32 SNPs discriminating between the HLA-A*01, *02, and *03 types were determined for all sequenced samples, and haplotypes were reconstructed. The 3 most common (> 5%) haplotypes explained approximately 70% of the observed haplotypic variation over the 32 SNP loci (Table 2). The patients with EBV⁺ HL showed a significantly different haplotype frequency distribution compared with control participants and patients with EBV⁻ HL (P < .001). One haplotype was overrepresented in the patients with EBV⁺ HL (37.1%) versus control participants (15.3%; P < .001) and patients with EBV⁻ HL (19.4%; P = .012). The HLA-A type corresponding to this haplotype was HLA-A*01. Another haplotype was underrepresented in the patients with EBV⁺ HL (15.0%) versus control participants (42.4%; P < .001) and patients with EBV⁻ HL (30.6%; P = .01). The HLA-A type corresponding to this haplotype was HLA-A*02. The third haplotype was equally present in patients with EBV⁺ HL (12.1%), control participants (9.3%), and patients with EBV⁻ HL (8.1%), and corresponded to HLA-A type HLA-A*03. Other less common haplotypes showed no different frequencies between patients with EBV⁺ HL, patients with EBV⁻ HL, and control participants.

Patients with EBV⁺ HL were significantly more often homozygous for HLA-A*01 compared with control participants (18.6% versus 1.7%; P < .001; Table 3) and more often heterozygous for HLA-A*01 (31.4% versus 18.6%), although this latter difference was not significant (P = .097). In contrast, patients with EBV⁺ HL were significantly less often homozygous for HLA-A*02 compared with control participants (1.4% versus 20.3%; P < .001) and less often heterozygous for HLA-A*02 (21.4% versus 35.6%), although this latter difference was also not significant (P = .074; Table 3).

Determination of the HLA-A*02 status by IHC or HLA-A*02–specific PCR showed that 35.5% of the patients with EBV⁺ HL (n = 152) and 50.9% of the patients with EBV⁻ HL (n = 322) were HLA-A*02⁺ (P = .001). The percentage of HLA-A*02⁺ samples in the 6107 control samples was 53.0%, which is significantly different from the percentage in patients with EBV⁺ HL (P < .001; Table 4).

A total of 30 patients with EBV⁺ HL and 29 patients with EBV⁻ HL were analyzed with both IHC and SNP genotyping. All samples positive for HLA-A*02 by IHC were either homozygous or heterozygous for HLA-A*02 by genotyping, and all HLA-A*02⁻ IHC samples carried no HLA-A*02 haplotype. This indicates that results obtained using the 2 methods are consistent.

In a previous study, a genetic association was found with the HLA class I region, including the HLA-A gene, and EBV⁺ HL.^10,26  In this study, the HLA-A gene was analyzed for association with EBV⁺ HL. The results confirmed an association; HLA-A*02 was underrepresented in patients with EBV⁺ HL (15%), and HLA-A*01 was overrepresented in patients with EBV⁺ HL (37.1%). HLA-A*03 and other HLA-A alleles occurred at similar frequencies in patients with EBV⁺ HL, patients with EBV⁻ HL, and control participants.

HLA-A*02 can present various immunogenic EBV-derived peptides of the latency type II antigens,^15-18,29,30  and can mediate a cellular immune response.³¹  A possible protective effect of HLA-A*02 was proposed many years ago by Poppema et al³²  and Bryden et al.²⁷  Poppema et al used IHC and studied a relatively small group of patients with EBV⁺ HL. They found no difference in the number of HLA-A*02⁺ samples between patients with EBV⁺ HL (10 of 19) and patients with EBV⁻ HL (25 of 53). Bryden et al used flow cytometry and HLA-A*02–specific PCR and showed that patients with EBV⁺ HL were less likely to be HLA-A*02⁺ (14 [40%] of 35) compared with patients with EBV⁻ HL (52 [54%] of 97), but this difference was not significant. Nevertheless, on the basis of our genotyping data, the HLA-A*02 allele frequency is significantly different between patients with EBV⁺ HL (15.0%), control participants (42.4%), and patients with EBV⁻ HL (30.6%). In addition to genotyping, we analyzed HLA-A*02 positivity by use of IHC and HLA-A*02–specific PCR on a much larger group (122 patients with EBV⁺ HL and 293 patients with EBV⁻ HL). The percentage of HLA-A*02⁺ samples was significantly decreased in patients with EBV⁺ HL (35.5%) compared with patients with EBV⁻ HL (50.9%), which confirmed our genotyping results. This suggests that the sample groups studied by Poppema et al and Bryden et al were probably too small to reveal a significant difference. Moreover, this study demonstrates that the genotyping approach has more power to detect significant differences on small populations (70 patients with EBV⁺ HL, 31 patients with EBV⁻ HL, and 59 control participants) compared with IHC and HLA-A*02–specific PCR approaches. In particular, the genotyping data showed a significantly decreased number of individuals homozygous for HLA-A*02 in the EBV⁺ HL group compared with the control group, while IHC and HLA-A*02–specific PCR approaches cannot discriminate between HLA-A*02 homozygous and heterozygous samples.

According to our genotyping results, the HLA-A*01 allele frequency is significantly higher in the group of patients with EBV⁺ HL, and these patients are more often homozygous for HLA-A*01 compared with control participants (18.6% versus 1.7%). However, we have to be aware that the population used for genotyping is relatively small, though our previous study showing a haplotype association (based on SNPs near the HLA-A gene) with EBV⁺ HL also revealed a significantly different frequency of homozygous genotypes comparing patients with EBV⁺ HL and control participants,¹¹  supporting our current data. Moreover, individuals homozygous for HLA-A*01 are also homozygous for the risk alleles of the 2 SNPs flanking the associated haplotype (rs2530388 and rs6457110) as observed in our previous study. Together, these data confirm the results of this study that patients with EBV⁺ HL are significantly more often homozygous for HLA-A*01 and less often homozygous for HLA-A*02 compared with control participants.

Many years ago, Hafez et al reported a significantly higher frequency of HLA-A*01 in patients with HL (53.8%) compared with healthy control participants (16.2%) in the Egyptian population.³³  The relative risk showed that persons carrying HLA-A*01 were 6 times more susceptible than those lacking it. These data were supported by Falk and Osoba, who also reported an increased incidence of HLA-A*01 in Canadian patients with HL.³⁵  Unfortunately, the EBV status of the patients with HL in these studies is not known. However, Falk and Osoba reported significantly more HLA-A*01 in patients with MCHL (57.6%) compared with those with NSHL (29.3%), which might indicate that EBV is involved, since patients with MCHL are more likely to be EBV⁺ than patients with NSHL.³⁶  These studies support a potential general character for the genetic association with HLA-A as observed in our study, but this needs to be confirmed in other ethnic populations.

HLA-A*01 and HLA-A*02 show different amino acids at most of the positions corresponding to the 32 SNPs that were analyzed. According to the studies of Bjorkman and Parham²³  and Reche and Reinherz²⁴ , 19 of these 32 SNPs are located at potential peptide-binding positions, and 13 SNPs are positioned at potential TCR-binding sites.^23,24  Binding of EBV-derived peptides to the HLA-A molecule is based on the amino acid residues at the binding positions, which are different for HLA-A*01 and HLA-A*02. It is known that HLA-A*02 can present EBV-derived peptides of the EBV latency type II genes. However, reports on HLA-A*01 are rare,³⁶  which might suggest that EBV-specific immune responses by HLA-A*01 are not commonly observed. In addition, to identify and compare potential HLA-A*02– and HLA-A*01–restricted epitopes within latency type II EBV antigens, we analyzed the amino acid sequences with a computer program designed to predict HLA-binding peptides, based on an estimation of the half-time dissociation of the HLA-peptide complex (BIMAS, http://bimas.dcrt.nih.gov/molbio/hla bind/index.html). This analysis revealed that HLA-A*02 binds peptides with high half-time dissociation scores (highest 10 results varied between 441 and 25 714), and HLA-A*01 binds peptides with very low half-time dissociation scores (highest 10 results varied between 0.9 and 2.5). These data support the idea that the HLA-A*01 allele cannot present high immunogenic EBV-derived peptides. Since the HLA system is codominant (both alleles are expressed), we suggest that individuals carrying HLA-A*02 are able to present immunogenic EBV-derived peptides, while individuals carrying only HLA-A*01 are either not able to present immunogenic peptides or only present EBV-derived peptides of low immunogenicity. Consequently, HLA-A*01⁺ individuals can most likely not evoke an efficient CTL response. Based on our genotyping data, we can conclude that HLA-A*01⁺ individuals are more susceptible to EBV⁺ HL, whereas HLA-A*02⁺ individuals have a diminished risk of developing EBV⁺ HL.

In most patients with EBV⁺ HL, the HRS cells show expression of HLA class I in the diagnostic lymph node biopsy.^13,37,38  This is confirmed in our HL population, in which 71% of the EBV⁺ patients showed membranous HLA class I expression compared with 14% in patients with EBV⁻ HL. Diepstra et al analyzed the SNPs near the HLA-A gene¹¹  for association with membranous HLA class I expression in HRS cells and observed similar allele frequencies in patients with EBV⁺ HL with and without HLA class I expression. This indicates that the genetic association with EBV is stronger than the genetic association with positive membranous HLA class I staining on the HRS cells (Diepstra et al, manuscript in preparation). Apparently, antigen presentation (of EBV-antigenic peptides) in the context of HLA-A is important, especially in early disease pathogenesis. At a later stage in the development of EBV⁺ HL, HLA class I expression may be lost. Malignant transformation and tumor progression are frequently reported to be associated with loss of HLA class I expression in different tumor types.³⁹ 

Another possibility is that HLA-A*01 alleles are predisposed to increased frequency of nonneoplastic EBV-infected B cells due to impaired immune surveillance. A high level of EBV-infected peripheral blood cells may increase the change of developing EBV⁺ HL. Gallagher et al reported that EBV DNA was more frequently detected in serum from patients with EBV⁺ HL than patients with EBV⁻ HL.⁴⁰  Furthermore, Khan et al showed an increased frequency of EBV-infected circulating memory B cells in pretreatment samples of patients with EBV⁺ HL versus those with EBV⁻ HL. They argued that elevated levels of EBV-infected B cells within the peripheral blood is a risk factor for developing EBV⁺ HL.⁴¹ 

Based on the functional link of HLA-A alleles in the context of presentation of EBV-derived peptides, this gene represents the most likely causal association with EBV⁺ HL. However, the associated HLA class I region also includes 9 pseudogenes and the protein-coding HCG9 gene.¹¹  It cannot be excluded that other disease-influencing variations are present in the associated HLA class I region. In conclusion, individuals carrying HLA-A*02 have a diminished risk of developing EBV⁺ HL, while those individuals carrying the HLA-A*01 haplotype homozygously are more susceptible to developing EBV⁺ HL.

This work was supported by a grant of the Dutch Cancer Society, KWF grant RUG 2000–2315. Work at the LRF Virus Center is supported by a LRF specialist program grant, and sample collection was supported by the Kay Kendall Leukemia Fund.

Contribution: The study was designed by S.P., G.J.t.M., and A.v.d.B. A.D. was involved in retrieval of Dutch tissue samples. R.F.J. was involved in retrieval of United Kingdom samples and data. M.N. was involved in sequencing, data analyzing, and overall coordination. B.H. and N.K. contributed to the HLA-A typing of Dutch patients. I.M.N. did the statistical analyses. M.P. contributed to HLA-A sequencing. C.P.D. and A.G. performed HLA-A*02–specific PCR. L.V. did the HLA-A*02 immunohistochemistry. M.N. wrote the paper with contributions from all other authors.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Anke van den Berg, UMCG PO Box 30001, 9700 RB, Groningen, the Netherlands; e-mail: a.van.den.berg@path.umcg.nl.