Epistatic Effects Occurring Among Susceptibility and Protective MHC Genes in Iga Deficiency

BRIEF COMMUNICATION Epistatic effects occurring among susceptibility and protective MHC genes in IgA deficiency

A Martı´nez1, L Gual1, M Ferna´ndez-Arquero1, A Nogales2, A Ferreira3, MC Garcia-Rodriguez3, G Fontan3 and EG de la Concha1 1Department of Immunology, Hospital Clinico San Carlos, Madrid, Spain; 2Department of Pediatrics, Hospital 12 de Octubre, Madrid, Spain; 3Immunology Unit, La Paz Hospital, Madrid, Spain

Immunoglobulin A deficiency (IgAD), the most prevalent primary immunodeficiency in Caucasian populations, shows strong evidence of polygenic inheritance with several associated genes being located in the major histocompatibility complex (MHC). Our aims were to determine which previously described MHC associations were primary and not secondary to a decrease or an increase in other MHC haplotype frequencies, to study the genetic interactions between all disease-associated MHC haplotypes and, finally, to ascertain the relative importance of protection vs susceptibility. A relative predispositional effect (RPE) study showed that in addition to the primary positive association of IgAD with HLA-DRB1*0102, DR3/TNFa2b3, and DR7 carrying haplotypes, DRB1*1501 was a marker of a primary protective factor in the Spanish population. Our data also indicate that the combined presence in an individual of two MHC susceptibility haplotypes notably increases the predisposition to the disease and that DRB1*1501 positive haplotypes eliminate the susceptibility conferred by any other MHC haplotype. Genes and Immunity (2003) 4, 316–320. doi:10.1038/sj.gene.6363955

Keywords: IgA deficiency; MHC; haplotypes; RPEs

Immunoglobulin A deficiency (IgAD) is the most this can create misleading deviations in the frequencies prevalent primary immunodeficiency in the Spanish of other haplotypes. To distinguish between these and other Caucasian populations.1 Genetic studies have secondary deviations in the frequencies and a truly unraveled a clear major histocompatibility complex genuine (positive or negative) association, a stepwise (MHC) genetic component.1 Although the genes respon- procedure called relative predispositional effects (RPEs) sible for the increased susceptibility to IgAD are not has been developed.7 This method proceeds in several known, they have been seen associated with HLA-DR1-, rounds of case–control comparisons, haplotypes show- -DR3- and -DR7-positive MHC haplotypes and some ing the strongest associations being sequentially authors have suggested that HLA-DR2 was associated removed from the analysis. However, when this ap- with protection.2 proach is used in a classically designed case–control A recent report from our group3 indicated that there study, the problem of low number of samples in the last are distinct MHC susceptibility genes, located in differ- rounds arises. ent regions on different MHC haplotypes. On HLA-DR1- Moreover, in case–control studies, association between and -DR7-positive haplotypes, the susceptibility locus a disease and a genetic marker can arise as an artifact mapped to the class II region, whereas on haplotypes of population structure, and genetic linkage cannot be carrying HLA-DR3, it mapped to the telomeric end of the shown. Therefore, we thought a family-based study was class III region in strong association with the TNFa2b3 warranted. Some recent studies on IgAD genetics had microsatellite alleles. already been performed in families, but they The study of resistance genes in polygenic diseases had focused either on a single family8 or on a single lags well behind the analysis of susceptibility factors, but haplotype.9 There has been no study analyzing together a few examples are known. For instance, the DRB1*1501/ all susceptibility and resistance markers reported within DQB1*0602 haplotype confers protection against type I the MHC on a family basis. diabetes mellitus,4 and the extended MHC haplotype We have studied 183 IgAD patients as defined by the HLA-B7/DRB1*1501 has been shown to carry on its class World Health Organization Group on Primary Immuno- III region a gene that diminishes susceptibility to multi- deficiencies,10 all ascertained at the Immunology Unit of ple sclerosis5 and protects against achalasia.6 Never- La Paz Hospital (Madrid) and 342 healthy controls. All theless, it has to be noted that when one or several patients and controls were unrelated Spanish Caucasians haplotypes in a region appear associated with a disease, and both groups were essentially the same studied in a previous report.3 Additionally, we studied parents and siblings of 100 of these patients. All parental MHC Correspondence: Dr EG de la Concha, Immunology Department, Hospital haplotypes could be established in these families Universitario San Carlos, 28040 Madrid, Spain. (n ¼ 400). Those present among IgAD children or affected E-mail: [email protected] parents (n ¼ 212) were defined as ‘IgAD’ haplotypes. The MHC associated genes in lgA deficiency A Martı´nez et al 317 remaining parental haplotypes (n ¼ 188) were defined individual factor. The strongest positive association was as ‘control’ haplotypes, the so-called AFBAC (affected found between IgAD- and HLA-DRB1*0102-positive family-based controls) method.11 The study was ap- genotypes, whereas the other two did not fall within proved by the hospital ethics committee. the DRB1*0102 odds ratio (OR) 95% confidence interval Table 1,12–14 first round, shows the allelic distribution (Table 2). As DR3/TNFa2b3/B8 is a very well-conserved of DRB1 group alleles in IgAD and control MHC extended haplotype, all individuals carrying it are haplotypes. Previous results had shown the association expected to have essentially identical DNA, at least over of IgAD with the HLA-DRB1*0102 subtype of the HLA- the HLA-DR/DQ-HLA-B genomic region, and are, DR1 group,15, 3 and with the DR3/TNFa2b3/B8 haplo- therefore, expected to carry the same susceptibility type,3, 8, 9, 16, 17 and therefore these have been explicitly gene/s. Hence, the different risk conferred by included in the table. In agreement with what was DRB1*0102- and DR3/TNFa2b3-positive genotypes is previously described, DRB1*0102, DR7 and the DR3/ another datum pointing to different susceptibility genes TNFa2b3 haplotype were seen more frequently in the being carried by each genotype (see Table 2). For HLA- IgAD haplotype group than in controls. DR7-positive heterozygous genotypes, relative risk is Contrarily, three HLA-DR groups (DR2, DR5, and even lower (RR ¼ 3.2), but as there are several extended DR8) were found to be at higher frequency in the control haplotypes carrying this allele, there is always the haplotypes. In order to ascertain whether any of the possibility that not all of them carry the susceptibility positive or negative associations were secondary to the gene/s. deviations in the frequencies of other alleles, we We then studied whether, in the population carrying a performed a step-by-step analysis. MHC susceptibility factor, the combined presence, in the A hierarchy of associated haplotypes, marked by their same individual, of a second MHC susceptibility factor HLA-DR alleles was obtained with an haplotypic RPEs.7 (either the same or a different susceptibility factor), With this method, the most associated haplotype (the increased the risk for IgAD. As can be seen in Table 2, 48 haplotype with the highest w2 value and Po0.05) was out of 147 (32.7%) IgAD patients carrying genotypes with sequentially removed from both IgAD and control an MHC susceptibility factor carried a second MHC haplotypes, and then a new round of w2 and P-values susceptibility factor as compared to 15 out of 137 (10.9%) was recalculated. The process ends when no remaining controls (P ¼ 0.00001; OR ¼ 3.94). haplotype is differently distributed in IgAD compared to We then calculated the relative risks conferred by each control haplotypes. combination of the two susceptibility factors. When a This study (Table 1 first through third round) showed combination was present in less than 10 controls, its that positive associations with all three haplotypes were genotypic frequency was estimated under the assump- not artifactually caused by another hypothetical protection of Hardy–Weinberg equilibrium with the formulas tive haplotype. When all three susceptibility haplotypes p2 and 2pq for homozygous and heterozygous genotypes, were removed, a significant decrease in frequency was respectively. The estimated RR values for each one of still observed for HLA-DR2 (Table 1, fourth round). This the three possible pairs were as follows: DRB1*0102/ showed as well that this was a true protective associa- DR3-TNFa2b3 ¼ 10.82, DRB1*0102/DR7 ¼ 17.85, and tion, not secondary to the presence of the positively DR3-TNFa2b3/DR7 ¼ 19.48. The RRs for the homozy- associated haplotypes. As DR2 had already been gous individuals were 40.57, 25.97, and 10.13 for reported to be a protective allele, no Bonferroni correc- DRB1*0102, DR3-TNFa2b3 and DR7, respectively. tion was performed. In contrast, the lower frequencies of Therefore, the RR was greatly enhanced when two DR5 and DR8 lost their significance, and they were doses of the susceptibility factors were present either in considered secondary to the other associations and not homo- or heterozygosity. Although for each individual dissected further in this study. genotype, 95% confidence interval was very large and no When the HLA-DR2-positive samples were specifi- statistically significant results could be obtained, the data cally typed for the individual subgroup alleles (HLA- were rather impressive and added to that of their sum DRB1*1501, *1502, *1601, and *1602 alleles), a statistically shown in the previous paragraph (P ¼ 0.00001; significant difference was only found for HLA- OR ¼ 3.94) point to a dose-related effect in most DRB1*1501, the most frequently observed DR2 subtype instances. It has been recently reported,17 studying in our population (0/88 in IgAD vs 10/154 in controls, individuals positive for the HLA-DR3/B8 haplotype, P ¼ 0.01). that increased frequencies of IgAD are only found in It is known that gene interactions modulate the homozygotes for this conserved, extended haplotype susceptibility to polygenic diseases. In Crohn’s disease, and have suggested a recessive inheritance with a low different mutations in the NOD2 gene notably increase penetrance. Although we have observed that all three disease susceptibility when combined in an individual.18, susceptibility MHC haplotypes (those carrying either 19 In type I diabetes mellitus, the protective effect of the DRB1*0102, DR7, or DR3-TNFa2b3) are sufficient for HLA-DQB1*0602 allele appears to be dominant over conferring susceptibility to IgAD, we also see that they susceptibility conferred by HLA-DR3/DR4.20 However, are greatly influenced by genes on the other MHC in IgAD no attempt has been made to investigate the haplotype, and the strong dosage effect of the HLA- interactions between susceptibility and protective genet- DR3/B8 haplotype has to be acknowledged. ic factors. The small ORs of DR3-TNFa2b3 and DR7 in hetero- After having identified, in a family-based study, the zygosity, and the notably increased risk when two doses most relevant positive and negative MHC susceptibility of these haplotype markers, irrespective of them being in factors, we investigated in a case–control study their homo- or heterozygosity, exist in the same individual interactions in different genotypic contexts. We first would suggest that DR3-TNFa2b3 and DR7 might be studied the strength of the susceptibility caused by each functionally equivalent. However, as our previous data

Genes and Immunity MHC associated genes in lgA deficiency A Martı´nez et al 318 Table 1 HLA-DR alleles carried in IgAD and control haplotypes. Relative predispositional effect study

Allele IgAD haplotypes Control haplotypes w2P

First round DR1 (except DRB1*0102) 11 15 1.28 NS DRB1*0102 33 11 9.61 0.0019 DR2 2 17 14.45 0.00014 DR3 (TNFa2b3 negative) 16 19 0.82 NS DR3/TNFa2b3 positive 25 8 7.48 0.0062 DR4 17 22 1.54 NS DR5 8 23 9.98 0.0016 DR6 31 40 3.02 NS DR7 62 19 25.33 2.0 Â 10À6 DR8 3 9 3.89 0.048 DR9 1 2 0.47 NS DR10 3 3 0.02 NS Total 212 188

Second round: DR7 removed DR1 (except DRB1*0102) 11 15 0.25 NS DRB1*0102 33 11 16.04 6.2 Â 10À5 DR2 2 17 10.80 0.0010 DR3 (TNFa2b3 negative) 16 19 0.03 NS DR3/TNFa2b3 positive 25 8 12.20 0.00048 DR4 17 22 0.21 NS DR5 8 23 6.20 0.013 DR6 31 40 0.41 NS DR8 3 9 2.43 NS DR9 1 2 0.23 NS DR10 3 3 0.02 NS Total 150 169

Third round: DRB1*0102 removed DR1 (except DRB1*0102) 11 15 0.00 NS DR2 2 17 8.56 0.0034 DR3 (TNFa2b3 negative) 16 19 0.16 NS DR3/TNFa2b3 positive 25 8 16.92 3.9 Â 10À5 DR4 17 22 0.002 NS DR5 8 23 4.00 0.045 DR6 31 40 0.05 NS DR8 3 9 1.58 NS DR9 1 2 0.11 NS DR10 3 3 0.14 NS Total 117 158

Fourth round: DR3 a2b3 removed DR1 (except DRB1*0102) 11 15 0.23 NS DR2 2 17 6.61 0.010 DR3 (TNFa2b3 negative) 16 19 1.03 NS DR4 17 22 1.32 NS DR5 8 23 2.25 NS DR6 31 40 1.36 NS DR8 3 9 0.91 NS DR9 1 2 0.03 NS DR10 3 3 0.37 NS Total 92 150

Fifth round: DR2 removed DR1 (except DRB1*0102) 11 15 0.05 NS DR3 (TNFa2b3 negative) 16 19 0.49 NS DR4 17 22 0.20 NS DR5 8 23 3.17 NS DR6 31 40 0.47 NS DR8 3 9 1.24 NS DR9 1 2 0.06 NS DR10 3 3 0.24 NS Total 90 133

DNA was isolated by standard methods from peripheral blood leukocytes. DRB1, DQA1, and DQB1 typing was carried out by polymerase chain reaction (PCR) amplification and hybridization with allele- specific oligonucleotides.12 The TNFa and b microsatellites typing was performed with primers described previously.13 The PCR fragments were subsequently run on an ABI Prism 310 automatic sequencer. DR2 subtyping was performed by a Taq-Man assay.14

Genes and Immunity MHC associated genes in lgA deficiency A Martı´nez et al 319 Table 2 Distribution of susceptibility factors in IgAD patients and in insulin-dependent diabetes mellitus (IDDM). More- controls over, its most frequent conserved, extended haplotype in our population (HLA-B7/DRB1*1501) harbors a resis- Susceptibility factor Patients Controls RR (CI) tance allele for multiple sclerosis and achalasia.5,6 The protective element in IgAD seems to be more related to DRB1*0102 34 19 10 (5.0–21) the one found in diabetes mellitus, as it seems not to DR3/TNFa2b3 21 26 4.6 (2.2–9.5) be confined to the HLA-B7/DRB1*1501 haplotype, but DR7 44 77 3.2 (1.9–5.6) * Two susceptibility factors 48 15 18.2 (8.8–38.2) instead to be present in every HLADRB1 1501-positive No susceptibility factor 36 205 1 haplotype. Total 183 342 In summary, we have studied the positive and negative associations of IgAD with MHC genes and haplotypes. As in many other immunologically mediated diseases, we have found important associations with Table 3 Distribution of HLA-DRB1*1501 in patients and controls MHC. However, as any other genetically complex grouped by their genotype disease, other genes in other chromosomes are also likely to be involved in IgAD susceptibility. Concerning Genotype IgAD patients Controls P the MHC genes, our results show that several genes are primarily involved in IgAD and that there are important 1501+ Total 1501+ Total interactions between them. HLA-DRB1*1501, seen as the only protective association, has a dominant protective DRB1*0102/xa 0 34 2 19 0.12 effect. In addition, there are several genes in different DR3 TNFa2b3/x 0 21 4 26 0.08 MHC haplotypes associated with different degrees DR7/x 0 44 9 77 0.01 x/x 2 36 51 205 0.01 of susceptibility. This susceptibility appears greatly enhanced when two of these haplotypes appear in compound genotypes either in homo- or heterozygosity, ax denotes any allele except DRB1*0102, DR3/TNFa2b3, or DR7. but is completely wiped out when DRB1*1501 is present. The important epistatic effects between all these MHC genes in IgAD suggest that they probably act in the same show that IgAD susceptibility genes on DR3-TNFa2b3 or related biological pathways. and DR7-positive haplotypes are different, we can speculate that these two genes might perform related functions. As we mentioned above, HLA-DR2, and its subtype Acknowledgements DRB1*1501, were the only markers found in statistically significant negative association with IgAD in the RPE This work was supported by Grant 01/348 from the study. DRB1*1501 is a subtype of HLA-DR2, and thus a Fondo de Investigaciones Sanitarias (Spain) much more accurate genetic entity than DR2. Therefore, We are most grateful to Carmen Martı´nez for her we decided to investigate the protective role of expert technical assistance. DRB1*1501, when present together with any of the three independent susceptibility markers, or in their absence (see Table 3). It was rather impressive to observe that in individuals with a susceptibility marker, DRB1*1501 References was not present in any patient as compared to 10.9% in 1 Burrows PD, Cooper MD. IgA deficiency. Adv Immunol 1997; controls (P ¼ 0.002; OR (CI) ¼ 0 (0.00–0.36)). As also 65: 245–276. shown in Table 3, in individuals without any suscept- 2 Olerup O, Smith CI, Hammarstrom L. Different amino acids ibility marker, DRB1*1501 was also present less fre- at position 57 of the HLA-DQ beta chain associated with quently in patients (5.3 vs 19.9% in controls; P ¼ 0.01; OR susceptibility and resistance to IgA deficiency. Nature 1990; (Cl) ¼ 0.18 (0.03–0.80)). 347: 289–290. Therefore, our data indicated that this protection was 3 de la Concha EG, Fernandez-Arquero M, Gual L et al. MHC dominant over susceptibility and acted upon the three susceptibility genes to IgA deficiency are located in different susceptibility haplotypes. As different haplotypes carry regions on different HLA haplotypes. J Immunol 2002; 169: 4637–4643. distinct susceptibility genes, the protective gene marked 4 Noble JA, Valdes AM, Cook M, Klitz W, Thomson G, Erlich by HLA-DRB1 obviously cannot be an allelic variant of HA. The role of HLA class II genes in insulin-dependent all those genes, but must instead (at least in some diabetes mellitus: molecular analysis of 180 Caucasian, multi- instance) be a different gene. We would refer to the plex families. Am J Hum Genet 1996; 59: 1134–1148. genetic interaction between HLA-DRB1*1501 and the 5 Allcock RJ, de la Concha EG, Fernandez-Arquero M et al. susceptibility alleles located in the MHC as epistasis. Susceptibility to multiple sclerosis mediated by HLA-DRB1 is When the alleles of two genes show some degree of influenced by a second gene telomeric of the TNF cluster. Hum epistasis, they probably act in the same or related Immunol 1999; 60: 1266–1273. biological pathways. The ‘dominant’ or epistatic gene 6 de la Concha EG, Fernandez-Arquero M, Conejero L et al. Presence of a protective allele for achalasia on the central (in this case HLA-DRB1*1501) is usually placed down- region of the major histocompatibility complex. Tissue Anti- stream in the pathway, closer to the final phenotypic gens 2000; 56: 149–153. 21 outcome. 7 Payami H, Joe S, Farid NR et al. Relative predispositional HLA-DRB1*1501/DQA1*0102/DQB1*0602, the main effects (RPEs) of marker alleles with disease: HLA-DR alleles DR2 haplotype, has been described as a resistance factor and Graves disease. Am J Hum Genet 1989; 45: 541–546.

Genes and Immunity MHC associated genes in lgA deficiency A Martı´nez et al 320 8 Schroeder Jr HW, Zhu ZB, March RE et al. Susceptibility locus sequence-specific priming (TaqMan assay). Tissue Antigens for IgA deficiency and common variable immunodeficiency in 1999; 54: 508–516. the HLA-DR3, -B8, -A1 haplotypes. Mol Med 1998; 4: 72–86. 15 Fiore M, Pera C, Delfino L, Scotese I, Ferrara GB, Pignata C. 9 Vorechovsky I, Cullen M, Carrington M, Hammarstrom L, DNA typing of DQ and DR alleles in IgA-deficient subjects. Webster AD. Fine mapping of IGAD1 in IgA deficiency and Eur J Immunogenet 1995; 22: 403–411. common variable immunodeficiency: identification and char- 16 Volanakis JE, Zhu ZB, Schaffer FM et al. Major histo- acterization of haplotypes shared by affected members of 101 compatibility complex class III genes and susceptibility to multiple-case families. J Immunol 2000; 164: 4408–4416. immunoglobulin A deficiency and common variable immu- 10 WHO. Primary immunodeficiency diseases. Report of a WHO nodeficiency. J Clin Invest 1992; 89: 1914–1922. scientific group. Clin Exp Immunol 1997; 109: 1–28. 17 Alper CA, Marcus-Bagley D, Awdeh Z et al. Prospective 11 Thomson G. Mapping disease genes: family-based association analysis suggests susceptibility genes for deficiencies of IgA studies. Am J Hum Genet 1995; 57: 487–498. and several other immunoglobulins on the [HLA-B8, SC01, 12 Kimura A, Sasazuki T. Eleventh international histocompat- DR3] conserved extended haplotype. Tissue Antigens 2000; 56: ibility workshop reference protocol for the HLA DNA-typing 207–216. technique. In: Tsuji K, Aizawa M, Sasazuki T (eds). HLA 1991. 18 Ogura Y, Bonen DK, Inohara N et al. A frameshift mutation Proceedings of the Eleventh International Histocompatibility in NOD2 associated with susceptibility to Crohn’s disease. Workshop and Conference. Oxford University Press: Oxford, Nature 2001; 411: 603–606. 1991, pp 397–419. 19 Hugot JP, Chamaillard M, Zouali H et al. Association of NOD2 13 Nedospasov SA, Udalova IA, Kuprash DV, Turetskaya RL. leucine-rich repeat variants with susceptibility to Crohn’s DNA sequence polymorphism at the human tumor necrosis disease. Nature 2001; 411: 599–603. factor (TNF) locus. Numerous TNF/lymphotoxin alleles 20 Redondo MJ, Eisenbarth GS. Genetic control of autoimmunity tagged by two closely linked microsatellites in the upstream in type I diabetes and associated disorders. Diabetologia 2002; region of the lymphotoxin (TNF-beta) gene. J Immunol 1991; 45: 605–622. 147: 1053–1059. 21 Marsh L, Neiman AM, Herskowitz I. Signal transduction 14 Tremmel M, Opelz G, Mytilineos J. High-resolution typing during pheromone response in yeast. Annu Rev Cell Biol 1991; for HLA-DRB1*15 amd -DRB1*16 by fluorescence-marked 7: 699–728.

Genes and Immunity