PR3-ANCA-associated vasculitis is associated with a specific motif in the peptide-binding cleft of HLA-DP molecules

Abstract

Objectives

This study aimed to characterize the association between HLA alleles and ANCA-associated vasculitis (AAV) in a genetically homogeneous population, and to analyse the contribution of specific HLA molecule amino acid sequences to the risk of AAV.

Methods

We included 187 Danish patients with AAV and 1070 healthy controls. All were HLA typed at two-field resolution. The association of HLA alleles to PR3- or MPO-AAV was analysed. The contribution of the dominant molecular motifs of the HLA-DPB1 molecule to the risk of AAV was investigated by association studies that included specific amino acid sequences of the hypervariable regions in exon 2.

Results

Ninety-four percent of patients with PR3-AAV were carriers of HLA-DPB1*04:01 while all patients with PR3-AAV were carriers of an HLA-DPB1*04 allele, and 85% were homozygous. This was significantly more than in the control group (P < 0.0001). The association was even stronger when HLA-DPB1*04:02 and -DPB1*23:01 were included. HLA-DPB1*04:01, -DPB1*04:02 and -DPB1*23:01 share amino acids in positions 8–9, 69, 76 and 84–87 within the hypervariable regions, but only positions 69 and 84–87 contributed significantly to the disease risk. HLA-DRB1*15 was associated with an increased risk of developing PR3-AAV, while HLA-DRB1*04, -DRB1*07 and -DQB1*03 were associated with a reduced risk of kidney involvement in PR3-AAV. MPO-AAV was only weakly associated with HLA class I alleles.

Conclusion

PR3-AAV is strongly associated with the HLA-DPB1 alleles HLA-DPB1*04:01, -DPB1*04:02 and -DPB1*23:01, which share amino acid sequences crucial for the peptide-binding groove.

Introduction

ANCA-associated vasculitis (AAV) is a group of rare diseases characterized by small vessel vasculitis and antibodies directed against PR3 or MPO. AAV is a systemic and relapsing disease, which may affect almost all organs, but typically involves the upper and lower respiratory tract, the skin and the kidneys. The aetiology and pathogenesis of AAV remain largely unknown, but there are clinical, histopathological, epidemiological and predisposing genetic differences between MPO-AAV and PR3-AAV. The incidence of MPO-AAV and PR3-AAV varies worldwide and in different ethnic groups. In Europe, PR3-AAV is in generally more common in the northern parts, while MPO-AAV is more common in the southern parts, and the incidence of MPO-AAV is much more frequent in Asia than in Europe compared with PR3-AAV. Patients with PR3-AAV are generally younger, more often have multi-organ involvement and experience more frequent relapses than MPO-AAV patients. The difference between the two entities is thought to be due to both environmental and genetic factors (reviewed in [1]).

Susceptibility and protection against autoimmune diseases, such as type 1 diabetes, multiple sclerosis and Goodpasture’s disease, is associated with specific HLA alleles [2]. HLA molecules present autoantigens to T cells, and the setting by which the presentation occurs determines whether pro-inflammatory or suppressive T cells develop, resulting in either autoimmunity or tolerance and protection from disease. Specific pockets within the peptide binding groove of the HLA molecule are particular important in determining peptide specificity [3], and the sequence of amino acids (aa) forming these pockets is therefore of particular interest. The HLA nomenclature reflects the serological antigen carried by an allotype rather than the actual ability to activate T cells. Thus, two different HLA allotypes may have functional similarities if critical aa sequences are shared. In a genome-wide association study (GWAS) of AAV patients, Merkel et al. showed that a high-risk single-nucleotide polymorphism was related to the glutamine-glutamine-methionine-proline (GGMP) motif in aa positions 84–87 of the HLA-DPB1 chain [4].

The association between AAV and HLA is mostly studied by GWAS in ethnically heterogeneous groups using single-nucleotide polymorphisms with variable resolution. A large European GWAS demonstrated an association between PR3-AAV and HLA-DP [5]. This association was confirmed in two GWAS in Americans with European ancestry [4, 6] and further localized to HLA-DPB1 [4]. GWAS allow for identification of associations to both coding and non-coding polymorphisms in larger genetic regions, which may not be identified using sequence-based HLA typing, whereas complete sequencing ensures that all biologically relevant immuno-genetic information relating to the antigen-presenting region of HLA may be analysed. Hilhorst et al. [7] confirmed the association between PR3-AAV and HLA-DPB1 and further specified it to HLA-DPB1*04:01 in two European AAV cohorts. They also found that PR3-AAV patients homozygous for HLA-DPB1*04:01 had a higher risk of relapse but did not confirm previously findings by Chang et al. reporting that HLA-DPB1*04:02 was associated with greater all-cause mortality in Chinese AAV patients [8]. The association between HLA-DR and AAV has been studied in a Dutch population by Stassen et al. [9] and in a North American population by Cao et al. [10]. Among North Americans, especially among African Americans, PR3-AAV was associated to HLA-DRB1*15:01, but no such association was found in the Dutch cohort. MPO-AAV has been found to be weakly associated with HLA-DQ, but not HLA-DP [4, 5].

To characterize the possible association of HLA alleles to AAV we performed HLA typing of HLA-A, -B, -C, -DRB1, -DQB1 and DPB1 loci at genomic sequence resolution in a genetic homogeneous Caucasian population. The genotype was associated with the disease burden at onset as well as disease outcomes (kidney involvement, multi-organ disease or relapse), and related to ANCA type (PR3- or MPO-ANCA positive). We further performed sequence analysis of the hypervariable regions in exon 2 of HLA-DPB1 [11] to evaluate if specific sequences of the HLA-DP types were associated with disease risk in AAV.

Methods

Design

The study was a retrospective study. It was approved by the local Committee on Health Research Ethics and the Danish Data Registry and informed content was obtained from all patients before they were included in the study. We invited prevalent and eligible AAV patients followed at the Outpatient Clinics of Departments of Nephrology and Rheumatology, Aarhus University Hospital and the Department of Nephrology at Aalborg University Hospital in Denmark to participate in the study. To be included patients must be older than 18 years, have a diagnosis of granulomatosis with polyangiitis or microscopic polyangiitis according to the Chapell-Hill classification, and be either PR3- or MPO-ANCA positive as well as anti-GBM negative, if there were signs of kidney involvement.

Clinical and laboratory data were collected from patients’ medical records from the time of diagnosis and up to 3 years after. These included: age, gender, ethnicity, ANCA subtype, treatment with antihypertensive and/or antidiabetic medication, smoking, plasma (p)-CRP, blood sedimentation rate, p-creatinine and albuminuria (urine albumin/creatinine ratio, urine-albumin in a single sample or the albumin excretion of 24-h collection). Organ involvement at time of diagnosis was noted and characterized as kidney, upper and lower airways, CNS, the peripheral nervous system, the gastrointestinal system and ‘other organs’. The type of induction treatment at diagnosis was also recorded. The following information was recorded at 1, 2 and 3 years after diagnosis: end stage renal disease (chronic dialysis or kidney transplantation), p-creatinine and albuminuria, major or minor relapses, as well as any immunosuppressive medication the time. All laboratory analyses were performed by local clinical biochemical laboratories. Kidney involvement was defined by signs albuminuria that could not be explained by another condition, or a kidney biopsy consistent with AAV. Remission was defined as no clinical or biochemical signs of active disease, and relapse as recurrence or new appearance of any signs of disease activity. Follow-up data at either 1, 2 or 3 years were missing for 16 (8.6%) of the patients, primarily because of inadequate hospital record keeping.

Healthy blood donors from the Department of Clinical Immunology, Aarhus University Hospital served as controls for HLA distribution.

HLA typing

Patients and controls were HLA typed simultaneously by an external laboratory (Histogenetics, Ossining, NY, USA) and typing included HLA-A, -B, -C, -DRB1, -DQB1 and -DPB1 loci at a minimum of two-field resolution to allow assignment of the antigen-presenting region of the HLA molecules.

HLA-DPB1 motifs

The HLA-DPB1 aa polymorphisms include six hypervariable regions in exon 2 [11] (Table 1). We specified the aa at positions 8–9, 35–36, 55–56, 69, 76 and 84–87 for each DPB1 allele in every patient and control using the IPD-IMGT/HLA homepage (https://www.ebi.ac.uk/ipd/imgt/hla).

Statistics

Differences between groups were evaluated using unpaired t-test or by Wilcoxon–Mann–Whitney test for continuous variables, and by Fisher’s exact test for categorical variables. Rates to first relapse were analysed by Kaplan-Meier survival curve and evaluated by the log-rank (Mantel-Cox) test.

The distribution of HLA types or aa in the DPB1 hypervariable regions was compared between PR3- or MPO-ANCA positive cases and controls using logistic regression analysis. Predictors were inserted in an additive approach. A priori, HLA-DPB1*04 was hypothesized to associate with disease and analyses were performed unadjusted. Results for the distribution of HLA-A, B, C, DR and DQ are reported unadjusted and Bonferroni adjusted. Linkage disequilibrium was assessed, and conditional analysis carried out when appropriate. Associations between HLA types and other outcomes among patients with either PR3- or MPO-ANCA positive AAV were assessed in a similar manner. Analyses were adjusted for sex.

Results

One-hundred and eighty-seven patients with symptoms of vasculitis and PR3- (n = 140) or MPO-positive (n = 47) ANCA were included (Table 2). Fifty-eight percent were male, and all were Caucasian except for one MPO-AAV Asian patient. Patients with MPO-AAV were on average 10 years older than patients with PR3-AAV. There was no difference in age between male and female patients (data not shown). PR3-ANCA-positive patients with kidney involvement were older than patients without kidney involvement. We did not find such difference among MPO-AAV patients (Table 2). Interestingly, male patients with PR3-AAV more frequently had kidney disease at onset than female PR3-AAV patients (61 vs 36% for male and female patients, respectively, P = 0.007). This difference was not observed among MPO-AAV patients (72 vs 68% for male and female patients, respectively, P = 1.0). More patients with MPO-AAV had hypertension at the time of diagnosis than PR3-AAV patients (32 vs 15% for MPO-AAV and PR3-AAV patients, respectively, P = 0.03). There was no difference in the frequency of diabetes or smoking between PR3-ANCA- and MPO-ANCA-positive patients. There was no difference in the proportion of patients with single- and multi-organ disease between MPO-AAV and PR3-AAV patients, or in the number of PR3-ANCA- and MPO-ANCA-positive patients experiencing a relapse during follow-up (Table 2). Only four patients died during the follow-up period.

HLA-DPB1 association

In total 188 patients (140 with PR3-AAV and 48 with MPO-AAV) and 1070 healthy controls were HLA typed. The single, Asian patient was excluded from this part of the study to ensure a homogeneous genetic background.

As expected, HLA-DPB1*04:01 was the most common allele among PR3-AAV patients (Table 3). Ninety-four percent of PR3-AAV patients were carriers of HLA-DPB1*04:01 (132/140), but even more striking was that all patients were carriers of at least one HLA-DPB1*04 allele, while 85% (119/140) carried two HLA-DPB1*04 alleles. The odds ratio (OR) for developing PR3-AAV when carrying the HLA-DPB1*04:01 allele was 6.16 (95% CI 2.98, 12.74, P = 2.4 × 10⁻⁸). Sixty percent (84/140) were homozygous for HLA-DPB1*04:01. This differed significantly from the distribution in the control group, where 16.4% were non-carriers and only 36.9% were homozygous for HLA-DPB1*04 (P < 0.0001), and the OR for developing PR3-AAV when homozygous for HLA-DPB1*04 was 9.68 (95% CI 5.99, 15.65, P = 2.7 × 10⁻²⁷).

Due to strong linkage disequilibrium the contributions from the six hypervariable regions in HLA-DPB1 at the positions 8–9, 35–36, 55–56, 69, 76 and 84–87 to the risk of getting AAV were assessed by conditional analysis by stepwise logistic regression. For all motifs, the allele(s) associated with DPB1*04 were coded as additive. Only the presence or absence of a lysine at position 69 and the presence or absence of a glycine-glycine-proline-methionine at position 84–87 contributed independently to the model. The models were compared using pseudo R², which revealed a better fit for the combined effect of the motifs in position 69 and 84–87 than for HLA-DPB1*04 (pseudo R²: 0.17 and 0.15, respectively). When all three predictors, i.e. HLA-DPB1*04 and the two motifs at position 69 and 84–87, were inserted in the model only the two motifs were independent predictors of PR3-AAV. A lysine at position 69 and glycine-glycine-proline-methionine at positions 84–87 are all included in the HLA-DPB1*04:01, -04:02 and -23:01 alleles. Consequently, HLA-DPB1*23:01 alone was also a strong predictor of PR3-AAV (OR 3.79, 95% CI 1.12, 12.77, P = 0.03). Among the 140 PR3-AAV patients, 131 were homozygous for the glycine-glycine-proline-methionine motif (Table 4). Since homozygosity at position 69 and position 84–87 appeared strongly associated with PR3-AAV, we combined the two motifs in a separate analysis (Table 5), showing that the OR for developing PR3-AAV associated with homozygosity was 12.1 (95% CI 7.2, 20.4, P = 8.6 × 10⁻³⁰) when compared with the group of heterozygous and negative individuals.

More controls than PR3-AAV patients carried the HLA-DPB1*02:01 allele (19.3% of controls vs 4.3% among patients, Table 3), but we could not find an independent effect of this allele (OR 1.49, 95% CI 0.61, 3.68, P = 0.38). We also compared the effect of HLA-DPB1*02:01 with the effect of all other non-HLA-DPB1*04 alleles and did not find any difference (data not shown).

The distribution of HLA-DPB1*04 alleles in patients with MPO-AAV was not significantly different from the control group (P = 0.56, data not shown).

HLA-A, -B, -C, -DR and -DQ

HLA-DRB1*15 [OR 1.84, 95% CI 1.37, 2.48, P < 0.001, Bonferroni adjusted P-value (aP) <0.01] and HLA-B*07 (OR 1.69, 95% CI 1.25, 2.27, P = 0.001, aP = 0.038) are overrepresented among PR3-AAV patients. HLA-A*03 is associated with PR3-AAV unadjusted, but the association disappears when adjusted for multiple comparisons (OR 1.48, 95% CI 1.09, 2.00, P = 0.011, aP = 0.418). HLA-DRB1*07 (OR 0.50, 95% CI 0.29, 0.88, P = 0.015, aP = 0.57) and HLA-DRB1*13 (OR 0.63, 95% CI 0.40, 0.99, P = 0.045, aP = 1) seem to be less frequent among PR3-AAV patients, but the association disappears when adjusted for multiple comparisons. Conditional analysis for linkage disequilibrium, including all predictors above in addition to HLA-DPB1*04, showed that only HLA-DPB1*04 (OR 7.74, 95% CI 4.89, 12.24, P = 2.1 × 10⁻¹⁸) and HLA-DRB1*15 (OR 1.46, 95% CI 1.08, 1.98, P = 0.015) had independent effects.

Only weak, possible positive associations to HLA-B*07 (OR 1.68, 95% CI 1.03, 2.75, P = 0.038, aP = 1) and HLA-B*40 (OR 1.85, 95% CI 1.04, 3.27, P = 0.036, aP = 1) were identified among MPO-AAV patients, while HLA-DRB1*03 was less frequent in MPO-AAV patients (OR 0.40, 95% CI 0.17, 0.93, P = 0.032, aP = 1).

HLA associations to clinical presentation

We did not find any difference in HLA alleles between PR3-AAV patients that relapsed during the follow-up period and non-relapsing patients (data not shown). Notably, no significant difference was found in the number of patients homozygous for HLA-DPB1*04:01 between the two groups; 64% (37/58) of the PR3-AAV patients that relapsed and 54% (38/71) of the patients that did not were homozygous for HLA-DPB1*04:01 (P = 0.28). Moreover, we did not find any association of homozygosity of HLA-DPB1*04:01 to time-to-first-relapse or to major vs minor relapses (data not shown). Additionally, we did not find any difference between relapsing and non-relapsing patients carrying two of the HLA-DPB1*04:01, -DPB1*04:02 or -DPB1*23:01 alleles; 88% (51/58) of the relapsing patients carried two of the three alleles, compared with 87% (62/71) of the non-relapsing patients (P = 1.0).

No other associations between HLA alleles and relapse were found (data not shown).

The risk of kidney involvement at the time of diagnosis was reduced in PR3-AAV patients expressing HLA-DRB1*04 (OR 0.40, 95% CI 0.22, 0.73, P = 0.003, aP = 0.11), -DRB1*07 (OR 0.23, 95% CI 0.06, 0.90, P = 0.035, aP = 1) or -DQB1*03 (OR 0.57, 95% CI 0.34, 0.96, P = 0.034, aP = 1) (data not shown), even though the association was lost when adjusted for multiple comparisons. In contrast, no HLA class II alleles were associated with kidney involvement at the time of diagnosis among patients with MPO-AAV. We did not find any associations between HLA class I molecules and kidney involvement at the time of diagnosis, except for HLA-B*44, which appears to be associated with a reduced the risk of kidney involvement in AAV patients independent of ANCA type (OR 0.37, 95% CI 0.18, 0.78, P = 0.009, aP = 0.34) (data not shown).

We did not find any associations between HLA, PR3- or MPO-AAV and single or multi-organ involvement, nor did we find any associations with relapses among MPO-AAV patients (data not shown).

Discussion

We confirmed a strong association between HLA-DPB1*04:01 and risk of PR3-AAV and, in addition, identified an exclusive association between PR3-AAV and HLA-DPB1*04 in this Danish Caucasian patient population. All PR3-AAV patients in our cohort carry the HLA-DPB1*04 allele and 85% of PR3-AAV patients were homozygous for this allele. Moreover, we showed that the association to HLA-DPB1 is even stronger when HLA-DPB1*04:02 and -DPB1*23:01 are included. HLA-DPB1*04:02 and HLA-DPB1*23:01 share the same aa as HLA-DPB1*04:01 in the hypervariable regions located at positions 8–9, 69, 76 and 84–87. These positions are critical to the shape of the peptide-binding cleft and consequently for the peptide binding itself and for the presentation of antigenic peptides to T cells. The aa sequence of HLA-DPB1 at positions 84–87 seems of particular interest since the GGMP motif at this position is shared by all alleles associated with PR3-AAV, and is consistent with findings by Merkel et al. demonstrating that a high-risk single-nucleotide polymorphism was related to the GGMP motif in aa position 84–87 of the HLA-DPB1 chain [4]. The GGMP motif forms an essential element of the peptide-anchoring pocket 1 (Fig. 1). Interestingly, HLA-DPB1*02:01 and -DPB1*02:02 also carry this motif but are not associated with PR3-AAV. This points to an additional role of the basic lysine residue at position 69 present in all the PR3-AAV risk alleles, which is substituted with the acidic glutamic acid residue in HLA-DPB1*02:01 and -DPB1*02:02. The functional importance of these aa substitutions for the presentation of PR3 protein fragments remains to be elucidated, but the association probably reflects the differential presentation of PR3 fragments in HLA-DP molecules depending on the aa composition in the hypervariable regions.

Hilhorst et al. [7] demonstrated a similar association between HLA-DPB1*04:01 and PR3-AAV when examining two cohorts from Maastricht and Lübeck, respectively. In the Maastricht cohort, only ∼35% of PR3-AAV patients were homozygous for HLA-DPB1*04:01 while 23% did not carry this HLA-subtype, while in the Lübeck cohort ∼55% were homozygous and 13% were non-carriers. The HLA-DPB1 alleles in the non-carriers were not described. The differences between the findings in these and our cohorts probably reflect different genetic pools, including differences in HLA class II molecules other than HLA-DPB1, which was not examined by Hilhorst et al. Hilhorst et al. [7] also identified an association between homozygosity of HLA-DP*04:01 and the risk of relapse in white PR3-AAV patients of European ancestry. In a Chinese cohort of patients with renal AAV, HLA-DRB1*04:05 was associated with a poor response to treatment while HLA-DPB1*04:02 was associated with mortality, but most patients in this cohort were MPO-ANCA positive [8]. All PR3-AAV patients in our cohort were carriers of an HLA-DPB1*04 allele, and 95% (all but three) of the patients who relapsed in the follow-up period carried the HLA-DPB1*04:01. In any case, we did not find any difference in the number of patients carrying HLA-DPB1*04:01 between relapsing and non-relapsing PR3-AAV patients, and in contrast to Hilhorst et al. [7] we did not find that patients homozygous for HLA-DPB1*04:01 had a higher risk of relapse.

We show that PR3-AAV is not only associated to HLA molecules with similarities in the aa forming the peptide-binding groove, probably through the presentation of an as yet unknown autoantigen, but is also independently associated with HLA-DRB1*15. This association is weak and may reflect functional epistatic interactions between the different HLA–peptide–T cell interactions [12, 13]. The association between PR3-AAV and HLA-DRB1*15 was also demonstrated by Cao et al. [10], but not identified in two GWAS [4, 5] or in a study by Stassen et al. [9]. This probably reflects the study of different cohorts, especially in terms of ethnicity since the HLA repertoire differs among ethnic groups.

Interestingly, there is a tendency for particular HLA alleles (HLA-DRB1*04, -DRB1*07 and -DQB1*03) to be associated with reduced risk of kidney involvement in PR3-AAV. This may reflect HLA-associated differences in the risk of a ‘second hit’ during disease progression as it evolves from a localized process to systemic disease. Patients with localized GPA are more often ANCA negative than patients with systemic disease. For example, in a study of patients with GPA with orbital masses ∼40% were PR3-ANCA negative (no patients were MPO-ANCA positive), and extra-ENT involvement was less frequent at the time of manifestations of orbital masses than at the diagnosis of GPA [14], indicating that acquiring PR3-ANCA positivity increases the potential to become a systemic disease. Another explanation could be that in the progression of disease, neo-antigens are generated and differential presentation of these antigens by different HLA molecules could influence the further course of the disease. In any case, we did not find any difference in HLA associations between patients with single- or multi-organ involvement, and the association needs conformation in other and bigger cohorts.

MPO-AAV was found to be weakly associated with HLA-DQ in two GWAS [4, 5]. We did not find such an association.

The study is limited by its retrospective design, and thus patients that died early in the disease course were not included. Only four patients died during the follow-up period of 3 years (data not shown). Furthermore, systematic scoring of disease activity, e.g. by BVAS, was not available, which limited the possibility to assess the association between HLA genotype and disease severity. Also, the number of patients is limited, especially patients with MPO-AAV, which notably weakens the analysis of HLA associations and clinical presentation.

Unfortunately, we do not have a validation cohort allowing us to further validate the findings. The strong association to HLA-DPB1*04:01 has been found in other cohorts, but it will be interesting if the additional HLA associations found in this study, especially the exclusive association with HLA-DPB1*04, can be validated in other Scandinavian cohorts or in cohorts of patients of other ethnicity and similar disease patterns.

In conclusion, our findings confirm a strong association of HLA-DPB1*04:01 with PR3-AAV and show that the association is even stronger when HLA-DPB1*04:02 and -DPB1*23:01 are included. These HLA molecules most likely have functionally important similarities in the aa shaping the peptide-binding groove leading to the presentation of the same autoantigen(s). We also found that other HLA class II molecules are weakly associated with disease course in PR3-AAV, but not MPO-AAV, supporting other compelling evidence that PR3- and MPO-AAV are two different disease entities and should be recognized as such in terms of epidemiology, genetics, pathogenesis, course of disease and possibly treatment as well [1, 15].

Funding: This work was supported by AP Moeller Foundation, Helen and Ejnar Bjoernows Foundation, The Danish Kidney Society, The Danish Society of Nephrology and the Danish Medical Association.

Disclosure statement: The authors have declared no conflicts of interest.

References