Introduction

IgG4 autoimmune diseases (IgG4-AID) were first collectively described in 20151 and include diseases such as myasthenia gravis with antibodies against muscle-specific kinase (MuSK MG), pemphigus vulgaris (PV) or thrombotic thrombocytopenic purpura (TTP)2. IgG4-AID are distinct from other autoantibody-mediated autoimmune diseases, as IgG4 is normally considered as an anti-inflammatory antibody that has structural differences to other IgG subclasses (including functional monovalency) and lacks typical antibody effector mechanisms, such as complement activation3,4,5,6. IgG4 is thought to play a protective role, e.g. in allergy or autoimmunity, by competing with effector antibodies for epitope binding3,7,8,9,10,11. Interestingly, in IgG4-AID the autoantibodies belong predominantly to the IgG4 subclass, and they are directly pathogenic by functional blocking of protein–protein interaction1,12. IgG4 pathogenicity could be demonstrated by passive transfer to experimental animals in (1) MuSK MG (MuSK-IgG4), (2) PV (desmoglein 3-IgG4), (3) pemphigus foliaceus (PF, desmoglein 1 and/or 3-IgG4), (4) chronic inflammatory demyelinating polyneuropathy (CIDP, contactin-1-IgG4), (5) CIDP (neurofascin 155-IgG4), and (6) TTP (ADAMTS13-IgG4)13. Notably, IgG4-AID differ from clinically distinct IgG4-related diseases14 that are therefore not part of our study. IgG4-AID share also further important pathophysiological and therapeutic commonalities15,16,17 including severe disease course, low disease prevalence (equal or less than 5/10,000) and good response to B-cell depletion therapy with rituximab17.

Whether IgG4-AID have distinct genetic risk factors that may predispose for the production of pathogenic IgG4 is unknown. A major contributor to genetic susceptibility to autoimmunity are the highly polymorphic human leukocyte antigen (HLA) genes on chromosome 6p21.3 that encode the major histocompatibility complex (MHC)18,19. HLA-DR, HLA-DQ, and HLA-DP encode the MHC II molecules on antigen-presenting cells and thymic epithelial cells that present self- and foreign antigen peptides to CD4 + T helper cells, which is essential for T-cell activation or the development and maintenance of tolerance20,21.

HLA-DR has been linked to aberrant presentation of self-peptide to autoreactive T helper cells in the thymus22, and genetic polymorphisms in the HLA-DRB1 gene are associated with a range of autoimmune diseases, such as rheumatoid arthritis, diabetes mellitus type I or systemic lupus erythematosus23. There is also evidence that the HLA can influence the production of IgG4: distinct HLA variants were shown to determine the immune response towards autoimmunity or tolerance in animal models24,25 by directly affecting T-cell fate towards conventional (Tconv) or regulatory T cells (Tregs), and production of pro- or anti-inflammatory cytokines, including interleukin-10 (IL-10). IL-10, which is in part produced by Tregs26, induces activation, IgG4 class switch and antibody production in naïve CD40-primed B cells and is therefore a key regulator of IgG4 production26,27,28,29,30,31. Increased IgG4 production was linked to HLA-DRB1*15 in patients with IgG4-related disease32and MuSK MG patients carrying HLA-DRB1*14 expressed elevated levels of IL-10 and MuSK antibodies compared to patients with other HLA variants33. Furthermore, IL-10 was found to be elevated in patients with pemphigus34,35, MuSK-MG36 and thrombotic thrombocytopenic purpura37. This suggests a link between HLA polymorphisms and production of IgG4 via IL-1029,30,31,38. In a previous review, we observed that individual IgG4-AID were frequently reported to be associated with the same recurrent HLA alleles: HLA*DRB1* 04, 11, 14 or 15, and/or HLA-DQB1*0515. GWAS data also suggests that HLA class II gene polymorphisms play a role for susceptibility to several different IgG4-AID15, and specifically the HLA-DRB1 and DQB1 loci were associated with individual diseases39,40. We hypothesized that distinct HLA variants may contribute to a genetic susceptibility resulting in a predominant production of IgG4 subclass antibodies and may therefore be associated with several distinct IgG4-AID. Therefore, we wanted to investigate HLA associations first in individual IgG4-AID to identify disease- specific variants, and then across diseases to identify which HLA variants are shared among different IgG4 associated diseases that may predispose to developing pathogenic IgG4 autoantibodies. To this end, we conducted a systematic review and meta-analysis of case–control studies reporting HLA class II associations in individual IgG4-AID.

We found that patients with IgG4-AID had significantly increased frequencies of the HLA-DQB1*05 and HLA-DRB1*14 alleles and the HLA-DRB1*14-DQB1*05 haplotype, and a significant negative association with HLA-DRB1*13. Notably, HLA-DQB1*05 is not positively associated with classical autoimmunity and could be a genetic risk factor for the production of IgG4 subclass autoantibodies.

Methods

The systematic review was based on recommendations by the HuGENet™ HuGE Review Handbook, version 1.0 (released by the EQUATOR network, 201541), and MOOSE guidelines for Meta-Analyses and Systematic Reviews of Observational Studies42.

Study design

The protocol, including the research question, search strategy, inclusion/exclusion criteria, data to be extracted, and the planned statistical analysis and bias assessment, was designed at the start of the study. The research question was developed with guidance from the PICOS (PI(E)CO) method43. The population (P) was defined as the participants in case–control studies and the intervention/exposure (I/E) was defined as the presence of distinct HLA alleles. The comparators (C) were the participants (patients and controls) without the distinct HLA allele and the outcome (O) was the occurrence of one of the six class I IgG4 AID. Regarding the study design, only case–control studies were considered, due to the rare nature of the disease. Only case–control studies with patients with IgG4-AID of class I (MuSK MG, PV, PF, TTP and CIDP with autoantibodies against NF155 or CNTN113) and ethnically, age- and gender-matched controls were included in the study.

Search strategy

Three individual researchers (A.P., G.L. and V.B.) used electronic search of 34 bibliographic databases and archives (supplementary Table S1), including PubMed/MEDLINE, Cochrane CENTRAL and Cochrane CDSR, Web of Science (core collection and all databases), BIOSIS, Scopus, Ovid Global Health, clinical trial registries (ClinicalTrials.gov and WHO ICTRP), and databases of systematic reviews (Epistemonikos, PROSPERO), BioOne, Centre for Reviews and Dissemination, CINAHL, DOAJ, EMBASE, EU Clinical Trials Register, GlaxoSmithKline's Clinical Study Register, Godort, HSRProj, JSTOR, Mendeley, metaRegister of Controlled Trials (Current controlled trials), Research gate, Science Citation Index (ISI), Science direct, TRIP Database, U.S. Government Documents, Worldcat, Biorxiv and Medrxiv as well as using other sources including grey literature (open grey) and hand searching. The search strategy included the search for key words, MeSH terms, including the use of a truncation operator (*, e.g. “antibod*” to identify the terms “antibody” and “antibodies”), and misspelling including the terms “HLA,” “human leucocyte antigen”, “DRB1”, “DQB1”,“MuSK myasthenia gravis”, “pemphigus”, “thrombotic thrombocytopenic purpura”, “ CIDP”, “chronic inflammatory demyelinating polyneuropathy”, “MuSK”, “Desmoglein 3”, “Desmoglein 1”, “blistering skin disease”, “ADAMTS13”, Neurofascin 155”, “NF155”, “Contactin-1”, “CNTN1″ and related terms in titles and abstracts and full text using Boolean search strategies. The search was conducted between May 5, 2020 until June 16, 2020. As only a limited number of studies investigating HLA associations in IgG4-AID exist, no restrictions for the date of publication were made and all studies available at the time of the search were included. Due to the language proficiency of the researchers, studies in German, English, Italian and Spanish were considered, if applicable using wildcards (*) for the search terms.

Screening and study selection

After deduplication, three researchers screened the obtained records for eligibility independently of each other, based on inclusion/exclusion criteria in two phases (first phase: screening of title/abstract, second phase: screening of the full text) using Rayyan software44. The inclusion criteria for selecting the studies were as follows:

  1. a.

    Studies in humans with a case-control design that

  2. b.

    Reported the association of gene variants of the HLA class II gene locus, including allele, genotype or haplotype frequency.

  3. c.

    Studies in which cases were patients with class I IgG4-AID that were tested positive for the corresponding antibodies ((MuSK myasthenia gravis (antibodies to MuSK) pemphigus vulgaris (antibodies to desmoglein 3), pemphigus foliaceus (antibodies to desmoglein 1 and/or 3), peripheral neuropathies, including CIDP, (antibodies against Contactin 1), peripheral neuropathies, including CIDP, (antibodies against Neurofascin 155), thrombotic thrombocytopenic purpura (antibodies against ADAMTS13)) by standardized laboratory tests, including the following tests: ELISA, cell-based assay (CBA), radioimmuno(precipitation)assay (RIA), direct or indirect immunofluorescence test.

  4. d.

    Studies with a minimum of 1 control per case, that were age and gender matched, and that described the controls as either ethnically matched or with controls from the same geographical region as the patient cohort.

  5. e.

    Studies with controls that are either healthy individuals or patients with a different type of the same disease that were negative for their relevant autoantibodies, as well as any other IgG4 associated autoantibody or with an unrelated disease.

The exclusion criteria were as follows:

  1. f.

    Studies in which the controls had any immunodeficiencies/abnormalities in the HLA locus,

  2. g.

    Studies in which data on an individual patient level was not available,

  3. h.

    Studies including subjects already included in other published studies,

  4. i.

    Studies where a full text version was not available and

  5. j.

    Studies where insufficient data was available to calculate an odds ratio.

During the two screening phases, the three researchers were blinded to each other’s decisions in order to prevent bias. Any discrepancies in the assessment after unblinding were resolved via discussion. The search and selection of studies was documented and visualized with a PRISMA flow chart45.

Data extraction

Data was extracted from tables and running text in the included manuscripts and collected in Excel (Microsoft Office, USA). If data was incomplete, unpublished or unavailable, it was attempted to retrieve the data by contacting the corresponding authors of the study by email. The following information from each included study was extracted: primary author, year of publication, full bibliographic information, demographic information of patients and controls (sex, age), country of study site, type of IgG4-AID and/or type of autoantibody, affected organ, type of control, HLA typing method, sample size, genotype frequency, allele frequency, haplotype frequency, HLA supertype frequency, OR and 95% CI. Combinable data was analysed by meta-analysis for association between HLA alleles (genotype or allele frequency) and class I IgG4-AID individually and collectively.

Statistics

Combinable data (haplotype, genotype and allele frequencies of HLA class II alleles, analyzed separately) was included in the analysis. To study genetic associations with the HLA class II alleles, we used the dominant genetic model of association to analyze genotype frequencies and the allelic model of association to analyze allele frequencies46. Depending on the information provided in the individual studies, haplotype analysis was conducted either with the dominant genetic model of association (based on haplotype frequency, defined as the number of individuals with a specific haplotype out of the number of total individuals (n)), or using the allelic model of association (based on haplotype frequency, defined as the total number of a specific haplotype out of the total number of alleles of all study participants (2n)). Both datasets were analyzed and presented separately.

The combined effect of the included studies (pooled OR) was calculated using the Review Manager ((RevMan) [Computer program] Version 5.4.1 The Cochrane Collaboration, 2020).

Mantel–Haenszel tests for the ORs were performed with a random-effects model for different studies, which was important to address heterogeneity in the studies between diseases, and visualized using forest plots.

The heterogeneity of the included studies was measured using X2, I2 and Tau2. The publication bias was inspected by funnel plots. To overcome bias due to the predominance of pemphigus studies, which comprised > 50% of the studies, the analysis was repeated, excluding studies on pemphigus. Uncorrected p values < 0.05 were reported as statistically significant. To correct for multiple testing, we applied a Bonferroni-correction for k = 40 tests (20 marker positions, analyzed either as alleles or genotypes), which leads to a corrected significance level of p < 0.00125. Results that were only significant at the uncorrected level are shown in italics.

Consent for publication

All authors declare their consent for publication.

Results

Number and characteristics of included studies

After search and screening, 52 full-text articles with a total of 64 datasets (Tables 1,2) were included in the qualitative synthesis and 51 full-text articles with 62 datasets in the quantitative synthesis (Fig. 1). The following number of studies was identified: 36 on pemphigus, seven on TTP, five on MuSK MG, three on CIDP. Allele, genotype or haplotype frequencies were extracted and analyzed separately.

Table 1 Characteristics of included studies.
Full size table
Table 2 HLA class II associations identified in IgG4-AID.
Full size table
Figure 1
figure 1

PRISMA flow chart of study identification and eligibility screening. For specific inclusion and exclusion criteria see methods. Figure modified from Moher et al.47.

Full size image

Due to lack of data on HLA-DP, only polymorphisms in the HLA-DR and HLA-DQ genes were extracted. The following studies and datasets were included in the qualitative synthesis but excluded from the meta-analysis as they did not fit all selection criteria: the study by Joly et al., 2020, and one dataset from the Delgado study (1997).

Genetic associations with individual IgG4-AID

We wanted to study genetic HLA associations of the individual diseases. Data of 15 HLA-DRB1 alleles (DRB1*01–16) and five HLA-DQB1 alleles (DQB1*02–06) could be extracted from studies on pemphigus, MuSK MG and TTP (summarized in Tables 1,2). Due to a lack of data, no separate analysis for CIDP was undertaken.

Pemphigus

In a substantial proportion of the studies, there was little to no distinction between pemphigus vulgaris and pemphigus foliaceus. In this study, we therefore analyzed all subtypes of pemphigus collectively (supplementary Figs. S46–S66). The pooled ORs and 95% CIs indicated that four HLA variants were associated with a significantly increased frequency in pemphigus patients: HLA-DRB1*04 (genotype: OR 4.86; 95% CI 3.61–6.54; p < 0.00001; allele: OR 4.18; 95% CI 3.14–5.56; p < 0.00001), HLA-DRB1*14 (genotype: OR 4.81; 95% CI 2.88–8.05; p < 0.00001; allele: OR 6.14; 95% CI 4.98–7.58), HLA-DQB1*03 (genotype: OR 2.77; 95% CI 1.56–4.92; p = 0.0005; allele: OR 1.99; 95% CI 1.39–2.83; p = 0.0002) and HLA-DQB1*05 (genotype: OR 4.3; 95% CI 2.53–7.28; p < 0.00001; allele: OR 3.04; 95% CI 2.10–4.41; p < 0.00001). Eight variants were significantly decreased in pemphigus patients, suggesting a protective role: HLA-DRB1*03 (genotype: OR 0.34; 95% CI 0.25–0.47; p < 0.00001; allele: OR 0.35; 95% CI 0.17–0.70; p = 0.003), HLA-DRB1*07 (genotype: OR 0.38; 95% CI 0.25–0.58; p < 0.00001; allele: OR 0.45; CI 95% 0.32–0.61; p < 0.00001), HLA-DRB1*09 (genotype: OR 0.57; 95% CI 0.43–0.77; p = 0.0002; allele: OR 0.62; 95% CI 0.47–0.81; p = 0.0005), HLA-DRB1*11 (genotype: OR 0.42; 95% CI 0.27–0.65; p < 0.0001; allele: OR 0.47; 95% CI 0.31–0.72; p = 0.0005), HLA-DRB1*13 (genotype: OR 0.51; 95% CI 0.31–0.82; p = 0.006; allele: OR 0.44; 95% CI 0.32–0.6; p < 0.00001), HLA-DRB1 *15 (genotype: OR 0.47; 95% CI 0.37–0.59; p < 0.00001; allele: OR 0.37; 95% CI 0.3–0.47; p < 0.00001), HLA-DQB1*02 (genotype: OR 0.33; 95% CI 0.24–0.45; p < 0.00001; allele: OR 0.4; 95% CI 0.31–0.52; p < 0.00001) and HLA-DQB1*06 (genotype: OR 0.48; 95% CI 0.31–0.74; p = 0.0009; allele: OR 0.43; 95% CI 0.36–0.53; p < 0.00001).

Thrombotic thrombocytopenic purpura (TTP)

Seven TTP studies were analyzed, but due to a lack of data, quantitative analysis was only conducted on genotype frequency of alleles with data from at least three studies, and allele frequency was only analyzed qualitatively. We observed significantly increased genotype frequencies of HLA-DRB1*11 (genotype: OR 3.38; 95% CI 2.04–5.60; p < 0.00001), HLA-DRB1*12 (genotype: OR 2.52; 95% CI 1.32–4.84; p = 0.005), HLA-DRB1*15 (genotype: OR 1.67; 95% CI 1.11–2.51; p = 0.01) and significantly decreased genotype frequencies, and a trend towards a reduced allele frequency of HLA-DRB1*04 (genotype: OR 0.38; 95% CI 0.25–0.56; p < 0.00001) and HLA-DRB1*13 (genotype: OR 0.43, 95% CI 0.29–0.64; p < 0.0001) (supplementary Figs. S67–S84).

MuSK Myasthenia gravis

Five studies were available on MuSK MG, and quantitative analysis was only performed for alleles with a minimum of three studies per allele. MuSK MG patients had a strong, significant increase in genotype frequency of HLA-DRB1*14 (genotype: OR 6.36, 95% CI 2.75–14.75, p < 0.0001), HLA-DRB1*16 (genotype: OR 5.03, 95% CI 3.16–7.99; p < 0.00001) and HLA-DQB1*05 (genotype: OR 7.94, 95% CI 3.44–18.30, p < 0.00001). The haplotypes HLA-DRB1*14-DQB1*05 and HLA-DRB1*16-DQB1*05 showed an increased frequency in the two studies that defined the haplotype frequency like the genotype frequency (n), and a significant increase in the three studies that defined the haplotype frequency like the allele frequency (2n, HLA-DRB1*14-DQB1*05: OR: 4.78; 95% CI 2.65–8.62; p < 0.00001; HLA-DRB1*16-DQB1*05: OR 3.47, 95% CI 2.16–5.57; p < 0.00001). A tendency towards a decreased frequency of HLA-DQB1*06 was observed (supplementary Figs. S85–S105).

HLA alleles with increased frequency across IgG4-AID

To identify possible genetic risk factors that may be shared across diseases and that may predispose for the development of IgG4 autoantibodies, we analyzed HLA associations in all IgG4 patients (Fig. S19–28). Only figures with results that remained statistically significant after additional analysis (described below) are shown in the main manuscript, the remaining results are shown in the supplementary. We observed increased frequencies of HLA-DRB1*14 (Fig. 2, genotype: OR 4.31; 95% CI 2.82–6.59; p < 0.00001; allele: OR 4.78; 95% CI 3.52–6.49; p < 0.00001), HLA-DQB1*05 (Fig. 3, genotype: OR 3.8; 95% CI 2.44–5.9; p < 0.00001; allele: OR 2.54; 95% CI 1.82–3.55; p < 0.00001) as well as the HLA-DRB1*14-DQB1*05 haplotype (Fig. 4, n: OR 6.3; 95% CI 3.28–12.09; p < 0.00001, 2n: OR 4.98; 95% CI 3.8–6.53; p < 0.00001). Further associations were found in HLA-DQB1*03 (Fig. S2, genotype: OR 2.53; 95% CI 1.67–3.97; p < 0.0001; allele: OR 1.65; 95% CI 1.24–2.19; p = 0.0007) and HLA-DRB1*04 (Fig. S1, genotype: OR 2.72; 95% CI 1.81–4.10; p < 0.00001; allele: OR 2.72; 95% CI 1.94–3.81, p < 0.00001).

Figure 2
figure 2

Forest plot depicting allele and genotype frequency of HLA-DRB1*14 in patients with class I IgG4-AID. Cumulative meta-analysis with a random-effects model demonstrated a significant increased frequency in patients compared to controls. ‡ Study did not differentiate between disease subgroups of pemphigus or CIDP. * Study was included after discussion with W.B. **Study in which the same control group was used for pemphigus foliaceus and pemphigus vulgaris, here data was pooled for analysis.

Full size image
Figure 3
figure 3

Forest plots of the allele and genotype frequency for HLA-DQB1*05 in patients with class I IgG4-AID. Meta-analysis using a random-effects model demonstrated a significant increased frequency in patients compared to controls. ‡ Study did not differentiate between disease subgroups of pemphigus or CIDP. * Study was included after discussion with W.B. ** Study in which the same control group was used for pemphigus foliaceus and pemphigus vulgaris, here data was pooled for analysis.

Full size image
Figure 4
figure 4

Forest plots of haplotype frequencies for HLA-DRB1*14-DQB1*05 in patients with class I IgG4-AID. Meta-analysis using a random-effects model showed a significant positive association in patients throughout all six diseases. (n): haplotype frequency was defined as the number of individuals with a specific haplotype out of the number of total individuals; (2n): haplotype frequency was defined as the total number of a specific haplotype out of the total number of alleles of all study participants. ** Study in which the same control group was used for pemphigus foliaceus and pemphigus vulgaris, here data was pooled for analysis.

Full size image

Since the predominance of pemphigus studies (36/52 studies) may have skewed the data towards pemphigus-specific risk alleles, the data was re-analyzed after excluding the pemphigus studies to validate the findings (Figs. S3–S7 and S29–S45).

While we could confirm the positive association with HLA-DRB1*14, HLA-DQB1*05 and the HLA-DRB1*14-DQB1*05 haplotype after exclusion of pemphigus patients (Fig. S5–S7), the frequency of HLA-DRB1*04 (Fig. S3) was significantly decreased, suggesting this association is specific for pemphigus. Further positive associations after exclusion of pemphigus were observed in HLA-DRB1*11, *12, *15 and *16 (Fig. S36–S39).

Reduced frequency of HLA alleles in IgG4-AID

Several HLA variants were significantly decreased in patients with IgG4-AID, which is interesting as these may potentially contribute to a protection from IgG4 autoimmunity (Figs. 5; Suppl Fig. S8–S12). Reduced frequencies were observed for HLA-DRB1*03 (genotype: OR 0.54; 95% CI 0.35–0.83; p = 0.005; allele: OR 0.46; 95% CI 0.25–0.84; p = 0.01), HLA-DRB1*07 (genotype: OR 0.49; 95% CI 0.34–0.69; p < 0.00001; allele: OR 0.52; 95% CI 0.37–0.74; p = 0.0003), HLA-DRB1*09 (genotype: OR 0.62; 95% CI 0.47–0.82; p = 0.0008; allele: OR 0.70; 95% CI 0.56–0.89; p = 0.003), HLA-DRB1*13 (genotype: OR 0.48; 95% CI 0.34–0.68; p < 0.0001; allele: OR 0.46; 95% CI 0.34–0.62; p < 0.00001), HLA-DQB1*02 (genotype: OR 0.5; 95% CI 0.28–0.89; p = 0.02; allele: OR 0.51; 95% CI 0.36–0.71; p < 0.0001) and HLA-DQB1*06 (genotype: OR 0.61; 95% CI 0.44–0.84; p = 0.003; allele: OR 0.59; 95% CI 0.38–0.9; p = 0.01).

Figure 5
figure 5

Forest plots of allele and genotype frequency of HLA-DRB1*13 in patients with class I IgG4-AID. Meta-analysis using a random-effects model demonstrated a significant decreased frequency in patients compared to controls. ‡ Study did not differentiate between disease subgroups of pemphigus or CIDP. *Study was included after discussion with W.B. **Study in which the same control group was used for pemphigus foliaceus and pemphigus vulgaris, here data was pooled for analysis.

Full size image

The negative associations were less pronounced, and after exclusion of pemphigus (Figs. S30, S32, S34, S40, S43), only HLA-DRB1*13 (Fig. S4) was found at reduced frequency (genotype: OR 0.41; 95% CI 0.28–0.61; p < 0.00001, allele: OR: 0.49, 95% CI 0.20–1.21, p = 0.12).

Analysis of higher resolution data

We were interested to know whether the association was due to specific alleles, but high-resolution data was only available for a fraction of studies, as most studies only reported one-field resolution data (supplementary Table S3 and S4). We analyzed the available datasets with higher resolution data, which was mostly derived from pemphigus studies. Data of the available variants (Fig. S13–S15) was analyzed and positive associations with HLA-DRB1*14:01, HLA-DRB1*14:04, HLA-DRB1*04:02 and HLA-DQB1*05:03 were observed.

Within-ancestry analysis

To study the potential effect of ancestry, we conducted a within-ancestry analysis from the three countries with the highest number of datasets (Brazil: 6 studies, Turkey and Japan: each 5 studies) separately (Fig. S16–S18). A trend for similar outcomes could be observed in all three populations where enough data was available, but there was variation in the strength of the association, e.g. the OR for HLA-DRB1*14 was higher in Japan than in Brazil or Turkey. An across-ancestry analysis was not considered feasible with the available data.

Evaluation of heterogeneity and publication bias

The heterogeneity was assessed by Tau2, X2 and I2 tests (supplementary Table S5), whereas potential publication bias was assessed by funnel plots (Figs. 6; Suppl Figs.S106–112).

Figure 6
figure 6

Funnel plot analysis of genotype and allele frequency data for HLA-DQB1*05, DRB1*13 and HLA-DRB1*14 in all IgG4 patients. A funnel plot analysis was undertaken to assess publication bias. Odds ratios were plotted against the standard error and the studies demonstrated symmetrical scattering along the funnel axis (pooled effect estimate from meta-analysis).

Full size image

There was substantial heterogeneity for most of the alleles with the exception of HLA-DRB1*14, which showed a low level of heterogeneity only in the pemphigus allele frequency, but was highly heterogenic otherwise. HLA-DRB1*13 showed low heterogeneity in TTP and IgG4-AID excluding pemphigus, but moderate heterogeneity in all IgG4 AID collectively.

Due to the high level of heterogeneity between the studies, the publication bias was assessed only by funnel plots. We found a low to moderate and mostly symmetrical publication bias in HLA-DQB1*05 and HLA-DRB1*14, with very few outliers in both directions, while 1–2 outliers towards lower ORs were found for HLA-DRB1*13.

Discussion

We conducted a systematic review and meta-analysis on the genotype, haplotype and allele frequency of reported HLA class II alleles across IgG4-AID and found that HLA- DQB1*05, an allele that is not typically associated with autoimmunity, is significantly more frequent in patients with IgG4-AID. This suggests it may be a genetic susceptibility factor for the production of IgG4 subclass antibodies. In addition, HLA-DRB1*14, a known genetic susceptibility factor for autoimmunity, is also associated with IgG4 autoimmunity, as is the HLA- DQB1*05-DRB1*14 haplotype. HLA-DRB1*13, which is considered as protective for autoimmunity in general, is also negatively associated with IgG4-AID. HLA-DRB1*03 and *04, which are often associated with autoimmunity, did not correlate with IgG4-AID, with the notable exception of pemphigus, which showed a strong association with HLA-DRB1*04.

Therefore, HLA-DRB1*14 and HLA-DQB1*05 may be genetic risk factors for IgG4 AID, and HLA-DRB1*13 may have a protective effect.

Genetic associations with individual IgG4 autoimmune diseases

This is to the best of our knowledge the first systematic review and meta-analysis investigating a potential association of HLA class II alleles with IgG4-AID. Systematic reviews on individual IgG4-AID (Pemphigus, MuSK MG) agree with our findings39,40. A significant positive association of MuSK MG with HLA-DRB1*14, HLA-DRB1*16 and HLA-DQB1*05 could be confirmed in our study40. In contrast, a significant negative association for HLA-DQB1*03 reported in the MuSK MG study could not be reproduced in our analysis, and the reported negative association with HLA-DQB1*06 did not reach significance in our study. Possible reasons for this might be 1) the exclusion of one Italian study48 from our analysis that was included in the Hong study40 as it did not match our inclusion/exclusion criteria and 2) the use of different statistical methodology (random- vs fixed- effects model).

Our analysis of pemphigus data is in line with previous meta-analyses. Increased frequencies of HLA-DRB1*04 and HLA-DRB1*14 and decreased frequencies of HLA-DRB1*03, HLA-DRB1*07 and HLA-DRB1*15 were observed in the pemphigus patients39. In contrast to the latter study, we found HLA-DRB1*09, HLA-DRB1*11 and HLA-DRB1*13 also to be significantly decreased in pemphigus patients, but with a very broad 95% CI. In contrast to the Yan study39, there was no positive association with HLA-DRB1*08 and pemphigus, but analysis of pemphigus vulgaris studies only (data not shown) could reproduce the positive association for the genotype frequency. In a different study49 HLA-DQB1*05 and HLA-DQB1*03 were positively associated with pemphigus vulgaris, which is in line with our findings.

There were only few studies with haplotype data in IgG4-AID available, but the increased frequency of the HLA-DQB1*05-DRB1*14 haplotype may be due to linkage disequilibrium between the two genes.

Interestingly, while MuSK MG and pemphigus seem to have very similar genetic associations, TTP showed opposite effects for several alleles, and in HLA-DRB1*04 and HLA-DRB1*11 these were significant. Perhaps the different type and location of the antigen play a role: MuSK MG and pemphigus antibodies target antigens of the cell surface/extracellular matrix (type II hypersensitivity reactions, Gell and Coombs classification50), while ADAMTS13 is a soluble antigen (type III hypersensitivity reactions). Another explanation could be that there are shared sequence motifs between e.g. MuSK and desmoglein 1/3 that facilitate binding to the peptide binding groove that are not present in ADAMTS13, causing a decreased affinity of ADAMTS13 derived peptides to specific HLA alleles.

Systematic reviews on genetic associations of the HLA with TTP or CIDP were not available. Although antibodies against CNTN1 and NF155 are known since the early 2000s51,52, possible associations with HLA polymorphisms have only recently been determined and investigated. A (non-systematic) review53 also reports a handful of individual papers with genetic associations of neurological IgG4-AID with HLA-DQB1*05, namely MuSK MG and anti-IgLON5 disease, but different alleles for IgG4-AID with antibodies against LGI1 (HLA-DRB1*07:01), Caspr2 (HLA-DRB1*11:01) or neurofascin (HLA-DRB1*15). Whether these diseases are not associated with HLA-DRB1*14 and HLA-DQB1*05 cannot be concluded without further studies, as these were few studies with a low number of participants. HLA-DRB1*11 and 15 were also positively associated with IgG4-AID after exclusion of pemphigus (in addition to HLA-DRB1*12 and 16), these could play a role in a different subset of patients, perhaps in neurological IgG4-AID.

Furthermore, the DQB1 locus was not investigated in all studies. Nevertheless, it is very likely that several different genetic associations may exist that may predispose for the production of IgG4 autoantibodies in different forms of IgG4-AID, also depending on the structure of the autoantigens.

Comparison of HLA associations between classical and IgG4 autoimmune diseases

We wanted to compare genetic HLA associations with classical autoimmune diseases (i.e. autoimmune diseases that are not caused by IgG4 autoantibodies) with the associations observed in IgG4-AID. In our study, HLA-DQB1*05 was associated strongly with IgG4-AID, and where higher resolution data was available, it was the HLA-DQB1*05:03 allele that was associated with IgG4-AID. Only few autoimmune diseases were reported to be associated with HLA-DQB1*05, and these are mostly IgG4-AID, including MuSK MG, pemphigus and anti-IgLON5 disease. In other autoimmune diseases, negative associations were found with the HLA-DQB1*05:02 in T1D54,55 and Sjögren’s syndrome56. One single study reported HLA-DQB1*05:02 to be positively associated with myelin oligodendrocyte glycoprotein-associated disorders (MOGAD), a rare neurological autoimmune disease57. Overall this suggests that HLA-DQB1*05 may be specifically associated with IgG4-AID. 

HLA-DRB1*14 also is strongly associated with IgG4-AID in our study, and was also reported as increased in patients with rheumatoid arthritis, Guillain-Barré syndrome and MuSK MG22, suggesting it may be a genetic risk factor to develop autoimmune diseases. HLA-DRB1*13 was found to be less frequent in IgG4-AID in our study, and this was also observed in classical AID, including T1D and autoimmune hepatitis58,59,60.

The HLA-DRB1*03 allele frequency is increased in classical AID, including diabetes mellitus type 154,55, multiple sclerosis61, neuromyelitis optica62, systemic lupus erythematosus63, Graves’ disease64 and Sjögren’s syndrome56, but we observed no association across IgG4-AID, only a decrease in studies on pemphigus. A similar difference could be found for HLA-DRB1*04, which is increased in classical AID diabetes mellitus type 1, rheumatoid arthritis and autoimmune hepatitis patients58,65, but decreased in IgG4-AID (MuSK, TTP and CIDP)—with the exception of pemphigus, where a strong association was observed.

HLA polymorphisms and the induction of IgG4 autoantibodies

Autoimmune diseases are thought to have a multifactorial etiology with a cumulative effect of genetic predispositions and environmental triggers. The shared pathophysiology indicates a common origin, leading to the investigation of common genetic factors in AIDs66. One genetic compound suggested for this susceptibility are the HLA class II genes, which encode proteins required for antigen presentation to CD4 + T-cells in the thymus and the periphery, thereby affecting central tolerance development and T-cell activation in the periphery. HLA-DRB1, the most polymorphic gene with over 1800 alleles, is frequently associated with autoimmune diseases22. Different HLA alleles present distinct peptide repertoires, and may directly affect T-cell fate by inducing Tconv or Tregs and a proinflammatory or tolerogenic cytokine environment24,25. The latter also includes IL-10, which is an important regulator for IgG4 production26,27,28,29,30,31. MuSK MG patients with the HLA-DRB1*14 allele were found to have higher autoantibody titers and higher levels of the cytokine IL-10 than patients with other HLA alleles33, and elevated IL-10 levels were found in several IgG4-AID, including pemphigus34,35, MuSK-MG36 and thrombotic thrombocytopenic purpura37. We hypothesize that HLA-DRB1*14, HLA-DQB1*05 and/or other HLA alleles may have therefore a direct effect on T-cell fate, favoring IL-10 producing Tregs and the production of tolerogenic cytokines which then induce class switch of B-cells to IgG4 and the production of IgG4 autoantibodies. How may HLA variants affect T-cell fate? The mechanism could depend for example on characteristics of the peptide repertoires presented in the different MHC II variants25,67,68,69, by differential interactions between the TCR and the HLA that may affect peptide recognition70 or by different cell type specific expression levels of the MHC II depending on the HLA variant71.

IgG4-AID, IgG4-related diseases and IgG4 subclass

IgG4-related disease (IgG4-RLD) is the umbrella term for a distinct group of diseases associated with the IgG4 subclass, that is unrelated to IgG4-AID14. IgG4-RLD are clinically distinct from IgG4-AID, their pathogenic mechanism is unknown, the role of IgG4 in these diseases is unclear, and clinical characteristics of IgG4-RLD include fibrosis, IgG4 + plasma cell infiltrates in the tissue, organ swelling and increased serum IgG4 concentrations, which are not characteristic for IgG4-AID14. In line with these findings, HLA associations also differ for IgG4-RLD, which was found to be associated with HLA-DRB1*04 allele72.The pathogenic mechanisms of IgG4 and the regulatory mechanisms that lead to the production of pathogenic IgG4 in IgG4-AID are not well understood, and are subject of an ongoing review series13,15,17.

Study limitations

The main limitation of the study was owed to the low prevalence of IgG4-AID, including (1) small numbers of patient per individual study (mostly between 30 and 100 patients), and (2) a low number of available studies, leading to (3) substantial heterogeneity, which was especially pronounced in studies on TTP. Pooling of data was not always possible due to different types of analysis and the differential use of nomenclature (e.g. genotype, haplotype, allele and phenotype frequency). Lack of information on homozygosity or heterozygosity in studies with genotype frequencies prevented a combined analysis for allele and genotype frequency, and since the HLA genes are in linkage disequilibrium73, homo- and heterozygosity cannot be “re-calculated” by using the Hardy–Weinberg equilibrium. Therefore, we only included studies where the frequency was given in absolute and relative numbers and data for allele and genotype frequency was analyzed individually. Several studies used a single control group for two different datasets, and to avoid overestimating the number of controls, data of these studies was pooled where possible74,75,76,77 (exception: two studies from Serbia78,79). All studies included in the meta-analysis reported that the controls and patients derived from the same geographic location or that the controls were ethnically matched to the patients, but most studies did not provide further details on the ethnical matching.

Furthermore, high-resolution data was only available from a subset of studies, mostly on pemphigus, therefore the observed associations with the specific HLA-DQB1*05:03, HLA-DRB1*14:01 and DRB1*14:04 alleles need to be validated in further studies. Heterogeneity in ancestries across countries was addressed by only including studies with patients and controls that were ethnically matched and/or derived from the same population and use of the random-effects model for the meta-analysis.

Our understanding of the proposed kinship between individual IgG4-AID is very limited1,12,16, and it is likely that there are different true effects of the HLA alleles in the distinct diseases. To account for this possibility, we used a random-effects model and also analyzed the diseases individually. Since there was a predominance of pemphigus studies (37/52 studies), we re-analyzed the data after exclusion of the pemphigus studies and could reproduce the associations with the HLA-DRB1*13, HLA-DRB1*14 and HLA-DQB1*05 alleles and the HLA-DRB1*14-DQB1*05 haplotype. In contrast, the HLA-DRB1*04 allele, which was more frequent in pemphigus patients, was not associated with the other diseases.

Antibody tests were not described in a substantial number of studies on pemphigus, but histopathologic diagnosis implicates the presence of the relevant IgG4 autoantibodies (mostly desmoglein 1 and desmoglein 3, < 0.5% of patients desmocollin), the inclusion criteria were changed during the second round of screening to include the pemphigus studies in the quantitative analysis. The PRISMA statement acknowledges this iterative process and accepts that modifications in the review protocol during the synthesis may sometimes be inevitable45.

Conclusions

With the limitations of this study in mind, we observed an increased frequency of HLA-DRB1*14 and HLA-DQB1*05 alleles as well as the HLA -DQB1*05 -DRB1*14 haplotype in patients with IgG4 AID. These findings agree with the literature, where these alleles are also associated with individual IgG4-AIDs. Thus HLA-DRB1*14 and HLA-DQB1*05 individually—or in combination as haplotype—might pose a genetic risk factor for the susceptibility to develop IgG4 AID. HLA-DRB1*13 seems to be consistently less frequent in patients, indicating a possible protective effect. Nevertheless, the low number of individual studies and the relatively small patient cohorts contributed to the substantial heterogeneity, therefore further HLA association studies are needed to validate the findings.