HLA class II homozygosity confers susceptibility to common variable immunodeficiency (CVID)

Abstract

Most cases of CVID occur sporadically, but familial cases do also occur and 15% of the patients with the disease have first degree relatives with IgA deficiency (IgAD). Our purpose was to study CVID association with HLA class II alleles and to ascertain whether this disease shares a common genetic background with IgAD in our population. Patients with CVID (n = 42), were typed using gene amplification and sequence-specific oligonucleotide probing for HLA-DRB1, DRB3, DQA1 and DQB1 loci and their typing compared with that of 96 IgAD and 334 healthy controls. We observed a positive association between non-Asp residues at position 57 of the HLA-DQβ chain and CVID, although much weaker than in IgAD. Further, we found an association between CVID and homozygosity for genes encoding HLA class II molecules, especially HLA-DQ, not seen in IgAD. The data support the hypothesis that a restricted diversity of HLA class II molecules may contribute to susceptibility to CVID.

INTRODUCTION

CVID is one of the most common primary immunodeficiencies in humans. Affected individuals develop infections and require replacement therapy with immunoglobulin preparations. In the majority of cases CVID is sporadic, but familial forms do occur. An association with IgA deficiency (IgAD), another common primary humoral immunodeficiency, has also been found and about 15% of the patients with CVID have a first degree relative with IgAD [1–3]. The molecular defect of the disease is unknown, but both IgAD and CVID have been associated with alleles of the MHC [4–8], and it has been proposed that both diseases might represent polar ends of a clinical spectrum reflecting a single underlying genetic defect [2–4,7,8]. Association with HLA genes has been reported both in class II [6,9] and class III regions [10]. Recently a susceptibility locus for IgAD/CVID has been located in the MHC between the class III markers D821/D823 and HLA-B8 [3]. We have studied 42 patients with CVID and compared their HLA class II typing with that of 96 IgAD and 334 healthy controls of the same geographical area and the same racial origin. Our purpose was to study CVID association with HLA class II alleles and to ascertain whether this disease shares a common genetic background with IgAD in our population.

SUBJECTS AND METHODS

Study subjects

The study group consisted of 42 CVID and 96 IgAD patients as defined by the World Health Organization Group on primary immunodeficiencies [11], all followed up at the Immunology Unit of La Paz Hospital (Madrid). All patients were unrelated Spanish Caucasians. Similarly, controls (n = 334) were randomly recruited among hospital staff and among biology and medicine students who were invited to give blood samples. All responded to a small questionnaire in order to have their age, sex, name and address taken, and to be sure they were all unrelated healthy Spanish Caucasians from the Madrid region. Informed consent was obtained from all the individuals studied.

DNA isolation and HLA class II and tumour necrosis factor-alpha genotyping

DNA was extracted from fresh peripheral blood leucocytes by a ‘salting out’ procedure with 6 mol/l NaCl following overnight incubation with proteinase K. The hypervariable second exons of the HLA-DRB1, -DRB3, -DQA1 and -DQB1 loci were amplified separately by a polymerase chain reaction (PCR) procedure using the appropriate primers from the 11th International Histocompatibility Workshop [12]. The quality of the PCR product was assessed by agarose gel electrophoresis. From the PCR products, 5 μl (about 20 ng) were slot-blotted onto nylon membranes and cross-linked by exposure to UV light. Hybridization was performed with digoxygenin-labelled allele-specific oligonucleotide probes, from the 11th International Histocompatibility Workshop [12], on amplified DNA to characterize the DRB1, DRB3, DQA1 and DQB1 genes; the results were visualized by exposing the membranes to radiographic films for 5–15 min. Ten generic HLA-DR allele groups, eight DQA1 alleles and 10 DQB1 alleles were identified.

HLA-DR1⁺, -DR2⁺, -DR4⁺, -DR11⁺ and -DR6⁺ DNAs, identified in the HLA-DRB1 generic PCR and hybridization, were then specifically typed for the individual alleles within the HLA-DR1, -DR2, -DR4, -DR11 or -DR6 groups by amplification in HLA-DRB1 group-specific PCR and hybridization with a panel of sequence-specific oligonucleotide probes from the 11th International Histocompatibility Workshop [12].

In order to ascertain that no rare allele was missed, sequence-based typing for the DQB1 gene was performed for all patients found to be HLA-DQ homozygotes. Primer pairs located within exon 2 were used for PCR amplification [13]. The positive PCR products were used as templates for the sequencing reaction. Internal labelled sequencing primers were used, located in conserved regions [13]. Automatic sequencing was performed using an ABI PRISM 310 genetic analyser (Perkin Elmer, Applied Biosystems Division, Foster City, CA). Sequence data were processed automatically and evaluated manually matching them with those reported in the 11th International Histocompatibility Workshop [14].

The tumour necrosis factor (TNFα) microsatellite typing was performed as described by Nedospasov et al. [15], with local modifications. Briefly, TNFα microsatellite was amplified using a single PCR reaction with primers IR2 and IR4. IR2 primer was 5′ labelled with HEX fluorescent dye (Appied Biosystems). The amplified samples were electrophoresed and the size of the microsatellite was measured using an ABI PRISM 310 genetic analyser and Genescan software.

Statistical analysis

Allele frequencies in patients and controls were compared by χ² test or Fisher's exact test when an expected cell value was < 5; P values were corrected (P_c) for the number of alleles determined and were considered significant at a level of < 0.05.

RESULTS

Association with HLA alleles

Table 1 shows the frequency of HLA-DR, DQA1 and DQB1 phenotypes in patients with CVID, IgAD and in healthy controls. All the alleles either previously reported to be associated with CVID and IgAD or with a marked difference between controls and CVID (P < 0.05), are shown. Our results confirm previously reported IgAD associations with alleles of the DR1,DQ5, DR3,DQ2 and DR7,DQ2 haplotypes. On the other hand, no statistically significant association was found between any specific HLA class II allele and CVID. When HLA-DR allele frequencies in patients with both diseases were compared by χ² test, a statistically significant difference was found (P < 10⁻⁵).

Table 1.

Main HLA class II phenotype frequency differences between CVID or IgA deficiency (IgAD) and controls

*P < 0.05 when compared with controls; **P_c < 0.05 when compared with controls.

Amino acids at position 57 of the HLA-DQβ chain

It has been reported that aspartic acid at position 57 of the HLA-DQβ chain is associated with resistance to both CVID and IgAD, whereas a neutral amino acid at the same position is associated with disease susceptibility [6]. Table 2 shows that when allele frequencies in the patients' groups and control group were compared, aspartic acid at position 57 of the HLA-DQβ chain was seen to be reduced in CVID, although modestly (36% versus 49% in controls; P = 0.02). This negative association was much more strong in IgAD (23% versus 49% in controls; P < 10⁻⁷). Sixty percent of IgAD patients were non-Asp homozygous compared with 40% CVID patients and only 24% controls (Fig. 1). In CVID the presence of two aspartic residues at position 57 was not associated with resistance, although the presence of two HLA-DQβ chains with neutral amino acids was associated with susceptibility to the disease (Fig. 1). When analysing the data more closely in an attempt to clarify this issue, we observed that in CVID patients, much more than in IgAD, the increased frequency of two HLA-DQβ chains with neutral amino acids at position 57 was associated with an increased presence of homozygosity for HLA-DR, DQA1 and DQB1. In IgAD the increased frequency of homozygotes matched the expected number, in accordance with the increased frequency of associated genes. Fourteen IgAD individuals were homozygous for DQB1*0201, whereas considering the DQB1*0201 gene frequency of 0.396, the expected number of homozygous patients for this allele was 15.1 (compared with six CVID patients for a gene frequency of 0.298 and an expected number of CVID DQB1*0201 homozygous patients of 3.72). No association was found between DQβ57 non-Asp/non-Asp and CVID in the heterozygotes for DR,DQA1,DQB1, whereas a strong association (P < 10⁷) existed with IgAD (data not shown).

Table 2.

Frequency of Asp57 DQB1 alleles in CVID, IgA deficiency (IgAD) and controls

Fig. 1.

Frequencies (%) of homo- and heterozygosity for aspartic acid or non-aspartic acid at position 57 of the HLA-DQβ chain in individual with CVID or IgA deficiency (IgAD) compared with healthy controls.

HLA-DR, DQA1 and DQB1 homozygosity

In order to investigate whether HLA class II homozygosity conferred susceptibility to CVID independently of the presence of amino acids at position DQβ57, we analysed the association of DR,DQA1,DQB1 homozygosity with CVID in all patients and in DQβ57 Asp/Asp CVID. As shown in Table 3, there was an over-representation of DR,DQA1,DQB1 homozygotes in both groups. Conversely, in IgAD neither all patients nor those DQβ57 Asp/Asp showed this association.

Table 3.

Frequency of homozygosity at HLA-DR, DQA1 and DQB1 in all individuals and those carrying Asp/Asp at position 57 in the HLA-DQβ chains

*Fisher's exact test.

HLA-DR subtypes

Serologically defined HLA-DR specificities include several DRB1 subtypes that can be studied by DNA-based methods. When these subtypes were investigated in homozygous CVID patients carrying DR1, DR2, DR3, DR4 and DR5, different subtypes were found in four out of seven. The polymorphic gene DRB3 present in DR3⁺, DR5⁺ and DR6⁺ haplotypes was also typed in CVID patients and one DR3 homozygous was found carrying two different DRB3 alleles (data not shown).

TNFa microsatellite

All 10 patients homozygous for DQA1 and DQB1 were typed for TNFα microsatellite that results in 13 alleles. Only two of these patients were homozygous for this polymorphism located in the MHC class III region (Table 4).

Table 4.

HLA class II genotypes and tumour necrosis factor-alpha (TNF-α) microsatellites of the 10 HLA-DQ homozygous CVID patients

DISCUSSION

Previous studies in white populations have reported a shared positive association with certain MHC haplotypes in CVID and IgAD [3,6,8]. Some reports have shown CVID and IgAD susceptibility associated with haplotypes carrying either HLA-DR1, -DR3 or -DR7 and HLA-DQB1 alleles encoding a non-charged (non-Asp) amino acid at residue 57 of the DQβ chain [6,9]. Others have located a susceptibility locus for IgAD/CVID in MHC class III genes [8,10], and more precisely between the class III markers D821/D823 and HLA-B8 [3]. However, in the present study CVID patients showed no association with specific HLA class II alleles and only a weak association with the presence of a neutral amino acid (non-Asp) at position 57 of the HLA-DQβ chain, unlike what was observed in IgAD. When HLA class II allele frequencies in CVID and IgAD were compared in our patients, a statistically significant difference was observed.

Although both patients and controls were selected for Spanish Caucasian origin, slight differences in the ethnic origin might be a source of error. Nevertheless, it is our belief that our results do not support the hypothesis of a common genetic background in our CVID and IgAD patients. Lines of evidence supporting this hypothesis, in addition to their association with the same HLA class II alleles, include the occurrence of the two disorders in members of the same family [2,3,16] and the existing overlap between the two diseases [17]. However, these two factors were found only in a minority of our CVID patients: only two CVID patients were known to have relatives with IgAD, and in only four out of the 42 CVID patients was a progression from IgAD to CVID documented. Similar findings have been reported in other white populations, with > 80% of CVID patients having no first degree relative with IgAD [1,2], and IgAD and CVID phenotypes being persistent in the vast majority of patients [2,18]. Therefore we favour the hypothesis that only a subset of CVID is genetically related to IgAD. These are likely to be CVID familial cases, that usually have IgAD relatives [2,3,8,16].

CVID is known to be a very heterogeneous disease [1,10], and the possibility of a distinct underlying genetic defect in non-familial as opposed to familial cases would not be surprising. Although the great majority of patients analysed in most genetic studies were non-familial cases [1], there are some hints of a difference in association between familial and non-familial cases. Volanakis et al. found an association between CVID and IgAD and MHC haplotypes encoding DR3 or DR7, but when data were stratified by type of family (multiplex, containing both IgAD and CVID members, and simplex families) statistically significant association was only found in immunoglobulin-deficient patients of multiplex families [8]; also, Olerup et al. found a somewhat weaker association with HLA class II genes in CVID patients than in IgAD (with DR1, DR3 and DR7) [6].

In our population the association with HLA class II alleles was confirmed in IgAD, but in CVID only a very weak association with non-Asp at position 57 of the HLA-DQβ chain was observed. This suggests a different genetic association in the majority of our CVID compared with IgAD. Surprisingly, HLA class II homozygosity was associated with the risk of CVID, but not with IgAD. A statistically significant increase of DR,DQA1,DQB1 homozygotes was observed both in non-Asp/non-Asp DQβ57 as in Asp/Asp DQβ57 CVID. Although there is a tight linkage disequilibrium between DR and DQ genes in MHC haplotypes, some DR groups of alleles can be associated with more than one DQA1,DQB1 pair of alleles. That is the case in our population for DR2 (associated with DQA1*0102 or *0103, and DQB1*0502, *0602 or *0601), DR4 (in linkage disequilibrium with DQB1*0301 or 0302) and DR7 (associated either with DQB1*0202 or *0303). In all CVID patients, when homozygosity was found for HLA-DR, both haplotypes encoded a unique HLA-DQ molecule.

In HLA-DR serologically defined specificities (DR1-DR10), several subtypes can be differentiated by DNA-based methods. These subtypes share the non-polymorphic α-chain and all the β-chain, including the first and second hypervariable regions, except for one or a few amino acids in the third hypervariable region. Analysis of these subtypes revealed differences in four out of 10 patients encoding the same HLA-DR in both haplotypes. Similarly, one out of the remaining six encoded different DRB3 chains in both haplotypes. Therefore all 10 patients that have been classified as homozygotes for DR,DQA1,DQB1 have identical DQA1 and DQB1 alleles in both haplotypes, but half of them have slight differences in the DRβ chains.

Similarly we investigated whether CVID patients homozygous for DQA1 and DQB1 were also homozygous for a marker within the class III region of the MHC. Only two patients proved homozygous for the TNFα microsatellite.

Therefore, the presence of identical DQA1/DQB1 alleles on both haplotypes appeared associated with an increased susceptibility to CVID, independently of which particular alleles they carried, and of TNFα microsatellites.

HLA-DR and DQ molecules determine the ability to respond immunologically to foreign antigens. They are involved in antigen presentation to T cell receptors and T cell–B cell interactions. Polymorphism at their peptide-binding site is critical for protection against infection, and in all likelihood they have been subject to pathogen-driven selective pressures for diversification [19]. A higher than expected ratio of heterozygosity to homozygosity at MHC loci has been seen in several different populations, most notably in inbred populations, where a reduction to about one third of the statistically expected level has been observed [20,21]. As both DQ α- and β-chains are polymorphic and the DQ α-chain encoded by one chromosome can combine either with the DQ β-chain encoded by this or the second chromosome, four different DQ molecules may be formed in heterozygotes for DQA1 and DQB1, compared with only one in the homozygotes. A larger number of distinct class II molecules possessed by a given person expands significantly the potential of that person's antigen-presenting cells and therefore increases the likelihood that that person will generate an immune response to the foreign antigen [22]. Moreover, in some instances homozygosity for HLA has been reported to be associated both with an increased difficulty in clearing infections [23] and with susceptibility to autoimmune diseases [24].

In conclusion, in our CVID patients we found only a weak association with the presence of a neutral amino acid (non-aspartic acid) at position 57 of the HLA-DQβ chain, whereas homozygosity for HLA class II molecules, especially HLA-DQ, was seen to be much more closely associated. It is tempting to speculate that homozygosity for HLA class II MHC genes, and the ensuing lower number of distinct class II molecules present at the cell surface of antigen-presenting cells, will hamper the likelihood of generating an immune response to foreign antigens and might be involved in the B cell differentiation defect occurring in CVID. Nevertheless, CVID appears a complex disease, where this susceptibility factor is neither sufficient nor necessary and does not exclude other genetic susceptibility factors, some of them already published, that may be present inside or outside the MHC.