Associations between HLA Class I alleles and the prevalence of nasopharyngeal carcinoma (NPC) among Tunisians

The prevalence of Epstein-Barr virus (EBV)-associated anaplastic nasopharyngeal carcinoma (NPC) might be partially dependent upon genetic background affecting predominantly certain ethnic groups¹. Among them, some regions of Southern Asia suffer about a 100-fold higher prevalence compared with populations not at risk and this high prevalence is maintained upon immigration to Western Countries such as in the U.S. Chinese population Northern African populations (Algerians, Moroccans and Tunisians) referred to as Maghrebians also suffer increased prevalence of NPC which persists upon migration to low risk areas [1, 2] Although heterogeneity of viral strains, dietary, environmental and socio-economical variables have been implicated as predisposing factors, it is likely that genetic traits play a significant role at least in Asians [3, 4].

Genetic predisposition of NPC is paralleled by a strong associations with some HLA class I alleles which have been extensively studied in the Asian population. Of them, HLA-A2, -B14, -B46 and -B58 predispose to the disease while HLA-A11 exercises a protective effect in Asians [5‐12] Moreover, a meta-analysis confirmed a consistent association between HLA-A2, A11, B14, B46 and NPC [13]. With few exceptions [10, 11]., most studies employed serologic HLA typing methods that cannot discriminate alleles within large serologic families [14]. Such discrimination, however, is critical since different ethnic groups display significant differences in frequency of HLA alleles within a serologic family with different capacity to bind and present epitopes that may be relevant to the immune defense against EBV-coded or other NPC-associated antigens. For instance, the HLA-A2 serologic family includes about 100 alleles [15] whose distribution is different among ethnic groups with HLA-A*0201 (~90%) and HLA-A*0205 (~5%) predominant among Caucasian and HLA-A*0203 (~20%) -A*0206 (~15%) and -A*0207 (~40%) in Asians [16‐19] Interestingly, sequence based typing (SBT) in Taiwanese identified a positive association between HLA-A*0207 (but not other HLA-A2 subtypes) and NPC [10]. This association was particularly strong when the observation was extended to the HLA-A*0207/B*4601 haplotype. Since, chromosome 6p, in particular 6p21-23, is characterized by unusually high gene density harboring gene clusters coding for several protein families besides HLA molecules [20], it remains unclear whether NPC/HLA associations reflect distinct antigen presentation potentials of various alleles or rather represent a marker for a susceptibility/protection locus in strong linkage disequilibrium with the HLA region [21].

Maghrebian populations have been less intensely studied at high-resolution SBT. Serological analyses reported positive associations between NPC and HLA: these included increased frequency of HLA-B5 in Algerians (38.2% vs 24.4%) [22], HLA-B13 in Tunisians (15.5% vs 4%) [23] and HLA-B18 in Moroccans, of which only the last was significant after correction for number of specificities tested (relative risk = 4.14) [24]. In 1983, Heirat P et al. [22] reported in a cohort of patients with NPC from Algeria a lower frequency of HLA-Aw33 (3.9% vs 16.8%), -B14 (1.3% vs 16%) and -DR4 (13.2% vs 29.1%) in NPC compared with matched controls. After correction for the number of specificities tested, these differences were not statistically significant and were not pursued further. Subsequently, HLA-A*3301 and HLA-B*1402 were found to belong to the same ancestral haplotype [25]. Further studies done in other Maghrebians did not reproduce these findings. Independently, Mokni-Baizig N et al. [23] noted that the HLA-A23 allele was absent in NPC patients compared to an 18% frequency in non tumor-bearing Tunisians. This is interesting because HLA-A23 is part of an African extended haplotype that includes a -B14 allele (HLA-A*2301/HLA-B*1403) [26]. Moreover, Dardari R et al. [24] reported a significantly lower frequency of HLA-A9 serologic family alleles in NPC patients compared with controls. Since the HLA-A9 serologic family includes HLA-A23, this observation suggests that the reduced frequency of HLA-A9 reflects a lower frequency of the HLA-A*2301/HLA-B*1403 haplotype. Thus, previous studies in Maghrebians suggest that HLA-B14 and HLA-A alleles linked to the HLA-B14 serological family may be under-represented in patients with NPC compared with normal controls. Conversely, HLA-B5, -B13 and -B18 could be associated with increased risk of NPC.

Molecular typing provides enhanced discrimination for population studies and large data bases have accumulated in African populations including Maghrebians [25‐29] These studies, however, did not include Tunisians (TU). Therefore, the aim of this study was to 1) establish a high resolution SBT data base in normal Tunisians matched for gender, age and geographical residence to a similar cohort of NPC patients; 2) compare HLA allele and haplotype frequencies between normal healthy Tunisians and the previously characterized Northern Moroccan Berber population (ME) [28] to provide a broader frame for the interpretation of the results from the Tunisian study; 3) test the validity of the reported decreased prevalence of HLA-B14 and increased prevalence of HLA-B5, -B13 and -B18 in patients with NPC based on SBT-derived, high resolution information; 4) Find novel alleles and haplotypes associated with NPC in Tunisian that may serve as putative genetic markers for further susceptibility studies.

NPC is among common cancers the one whose prevalence is most likely affected by a genetic predisposition that can be uncovered through population studies. This seems to particularly apply to the Southern Asian population repeatedly shown to be affected at high incidence. Although viral and environmental factors play an important role in the ethio-pathogenesis of NPC, genetic factors may also predispose as documented by several family and/or case-control studies [45, 46]. The short arm of the human chromosome 6, is not the only genomic area that has been implicated in NPC predisposition as a large family study located a predisposition area in chromosome 3 [3]. However, many case-control studies pointed at HLA class I as a genomic region strongly associated with the prevalence of NPC in Southern Chinese [5‐12, 47‐49] Such associations may bear two possible explanations; first the antigen presenting function of HLA molecules may be responsible for an altered effectiveness of the immune response against tumor associated antigens expressed by NPC in individuals bearing different HLA alleles [10]. This logical explanation is, however, challenged by the lack of candidate epitopic determinants associated to particular HLA class I alleles that clearly affect the natural history of NPC. In addition, most HLA class I associations reported in Asian populations indirectly support this concept since they are primarily associated with increased risk of developing NPC. This suggests that lack of antigen presenting capacity may result in reduced protection against the growth of NPC. For instance, the HLA-A*0207 allele association reported by Hildesheim et al. [10] in Taiwanese Chinese was explained by suggesting a reduced antigen presentation potential of this allele compared with other HLA-A*02 alleles present in the same population such as HLA-A*0201 which are not associated with higher prevalence of NPC. Similarly, the higher prevalence of HLA-A*0207 in Asians compared to Caucasians [10] may also explain the higher prevalence of the disease in Asian populations. However, associations of HLA phenotypes with increase prevalence of NPC are difficult to attribute to a single allele considering the high frequency of heterozygosity and the co-dominant expression of the three HLA class I loci aimed at broadening the antigen presenting repertoire of each individual.

An alternative explanation for the association between HLA and the prevalence of NPC attributes a broader significance of HLA class I genes as markers for other predisposition factors genetically controlled within a locus in close linkage disequilibrium [21]. For instance, even in Hildesheim's report, the extended haplotype HLA-A*0207/HLA-B*4601 was characterized by a higher odds ratio to develop NPC than the HLA-A*0207 allele alone suggesting a more complex genetic explanation for such association than the displacement by the HLA-A*0207 allele of other HLA class I alleles belonging to the HLA-A*02 family. In addition, our and others' experiences do not suggest strong differences in epitope binding and antigen presentation between HLA-A*0201 (the prototypic Caucasian HLA-A*02 allele) and HLA-A*0207. These alleles differ only by one amino acid substitution from a Tyrosine to a Cysteine in position 99 within the binding groove of the HLA molecule. This substitution does not seem to significantly affect binding and presentation of canonical HLA-A*0201-associated epitopes [50‐52] as well as EBV-associated epitopes [53]. Finally, lack of antigen presentation capacity by a specific HLA class I allele should be compensated by other alleles expressed by the same individual; therefore, favouring homozygosity of the unfavourable allele in affected individuals. However, most reports do not suggest higher homozygosity values in high prevalence populations suggesting lack of balancing selection. Thus, dominant genetic traits directly related to oncogenesis may play a more important role than HLA class I alleles which are co-dominantly expressed.

Maghrebs have been less exhaustively studied. However, serological analyses observed positive associations between NPC and HLA class I alleles. These include increased frequency of HLA-B5 in Algerians (38.2% vs 24.4%) [22], HLA-B13 in Tunisians (15.5% vs 4) [23] and HLA-B18 in Moroccans (relative risk = 4.14) [24]. In addition, negative associations have been reported such as HLA-Aw33, -B14 in Algerians with NPC. However, after correction for the number of specificities tested, these differences were not statistically significant and were not pursued further. HLA-A*3301 and HLA-B*1402 have subsequently been recognized to belong to the same ancestral haplotype [25] suggesting that this haplotype may harbor more relevance to NPC than previously recognized. This impression was supported also by the identification of a negative association between HLA-A23 (known to belong to the extended HLA-A*2301/HLA-B*1403 haplotype [26]) and the prevalence of NPC [23]. This observation converges toward the relevance of HLA-B14 related haplotypes which may bear protective effects against NPC in North Africans. Moreover, Dardari R et al. [24] reported a significantly lower frequency of the HLA-A9 serologic family alleles in NPC patients compared with controls. Since the latter study did not provide molecular resolution, we can only speculate that the reduced frequency of HLA-A9 reflected a lower frequency of the HLA-A*2301/HLA-B*1403 haplotype since HLA-A23 is part of the HLA-A9 serological family. Thus, previous studies in Maghrebians suggested that HLA-B14 and HLA-A alleles linked to the HLA-B14 serological family may be under-represented in patients with NPC compared with normal controls. Conversely, HLA-B5, -B13 and -B18 could be associated with increased risk of NPC.

Low resolution (two digits level) analysis of allele frequency in our study quite remarkably and specifically confirmed previous positive associations described between NPC and HLA class I alleles in Berber populations. Both HLA-B*18 and -B*51 (the latter representing a split of the serologically-defined HLA-B5 family) were confirmed by our study (Table 4). In addition, a novel association was observed between HLA-B-57 (belonging to the HLA-B17 serological family) and the prevalence of NPC. Only the positive association observed in Tunisians between HLA-B13 and NPC prevalence was not reproduced by this study [23]. Negative associations were also identified by this study of which only HLA-B*14 (OR = 0.09, Fisher's test p-2 value = 0.0001) strongly reproduced previous findings while no association was noted with HLA-A*23 nor HLA-A*33. Interestingly, however, another strong negative association was noted between HLA-Cw*0802 and NPC (OR = 0.18, Fisher's test p-2 value = 0.003). This is of particular significance because both HLA-B*1402 and HLA-Cw*0802 belong to the same ancestral haplotype which was present not only in Tunisians but also Moroccan Berbers but practically absent in Tunisian patients with NPC (OR = 0.12, Fisher's test p-2 value = 0.001). This strong negative association suggests that this haplotype may bear a locus in strong linkage disequilibrium with these HLA alleles with dominant protective effects against the onset of NPC in North Africans such as a tumor suppressor gene with stronger stability in this phenotype. Since protective alleles have also been described in Asian populations (such as HLA-A11) [13], it could be postulated that some HLA alleles may bear immunological functions facilitating the immune surveillance against NPC. However, there is no known functional similarity between alleles described in Chinese and North Africans to support this hypothesis suggesting that other factors may bear on this predisposition. Indeed, in Asian populations, micro-satellite analyses within the short arm of Chromosome 6 identified associations between putative loci and NPC prevalence that are stronger than those with HLA class I alleles [12]. Hu C-C et al. [54] using high-resolution micro satellite mapping recently described a carcinoma-susceptibility locus located in a segment of Chromosome 6p21.3 containing the class I loci. Similarly, Nh SP et al. [55] concluded that an HLA-related recessive mutation is responsible for the prevalence of NPC in Chinese. Finally, Tian W et al. [56] identified in the Southern Chinese Han population a positive association between a predisposition to develop NPC and exon 5 of the MICA-STR encoded within the HLA class I region. Moreover, a recent meta-analysis of chromosomal aberrations associated with NPC [57] identified a small region in Chromosome 6p21.2-p23 that is frequently amplified in NPC and contains various candidate genes that may be responsible for NPC [58]. Thus, our findings suggest that there are haplotypes within or close to HLA class I region, particularly the HLA-B and -C loci that may bear genetic characteristics favoring or protecting against the onset of NPC in Tunisian populations at risk. This information may help identify individuals belonging to different ethnic groups whose genetic risk to develop NPC could be marked by HLA class I haplotypes. These prototypic phenotypes could be scanned with novel high throughput technologies covering genomic areas in high linkage disequilibrium with such haplotypes. Since the analysis of the Tunisian control population did not identify differences between Tunisians and Moroccans, considering that a common Berber substratum is shared by North African countries [59‐63], it is possible that our findings could be relevant to the other Northern African populations justifying the extension of future studies to Moroccans and Algerians [24].

Some of the alleles discussed by this study have been associated with other pathologies. HLA-B*1403 is the predominant allele in patients affected by ankylosing spondylitis from West Africa, where HLA-B*27 is uncommon [64]. HLA-B*51 and HLA-B*15 alleles have been associated with Behcet disease [65], B*44, B*51, B*57, B*15 with HPV infection and protection/susceptibility to in cervical carcinoma [66]. The present study, may offer novel insight on the molecular genetics of this region that may influence disease, although obvious relationships between the information presented and it causality is still missing.