Recently, a haplotype of three single-nucleotide polymorphisms (tri-SNP) in intron 1 of the HLA-DRA1 gene was found to be strongly associated with type 1 diabetes risk in HLA-DR3/3 individuals. The tri-SNP reportedly function as “expression quantitative trait loci,” modulating HLA-DR and -DQ expression. The aim was to investigate HLA-DRA1 tri-SNPs in relation to extended HLA class II haplotypes and human peripheral blood cell HLA-DQ cell-surface median fluorescence intensity (MFI), the first-appearing islet autoantibody, and autoimmunity burden. A total of 67 healthy subjects (10–15 y) at increased HLA risk for type 1 diabetes and with (n = 54) or without (n = 13) islet autoantibodies were followed longitudinally in the Diabetes Prediction in Skåne study. Among four tri-SNPs, AGG (n = 67), GCA (n = 47), ACG (n = 11), and ACA (n = 9), HLA-DQ cell-surface MFI on CD4+ T cells was lower in AGG than GCA (p = 0.030) subjects. Cumulative autoimmunity burden was associated with reduced HLA-DQ cell-surface MFI in AGG compared with GCA in CD16+ cells (p = 0.0013), CD4+ T cells (p = 0.0018), and CD8+ T cells (p = 0.016). The results suggest that HLA-DRA1 tri-SNPs may be related to HLA-DQ cell-surface expression and autoimmunity burden.

Autoimmune destruction of pancreatic β cells results in total loss of insulin and subsequently autoimmune type 1 diabetes (1, 2). Autoantibodies against islet cell autoantigens precede clinical onset of type 1 diabetes (36). HLA allele–encoded molecules involved in the presentation of Ag peptides to T cells influence genetic risk for type 1 diabetes (7, 8). HLA variants differ in the peptide-binding groove, which influences peptide binding and signal transduction after TCR engagement (9, 10). These HLA variants influence Ag presentation and are important for thymic selection processes and peripheral activation of immune responses (10, 11). HLA class I alleles, specifically HLA-A2, HLA-A24, HLA-B39, HLA-B57, and HLA-B18, have been found to contribute to type 1 diabetes risk (7, 12). HLA class II haplotypes, specifically HLA-DRB1*04 (DR4), DQA1*03:01-DQB1*03:02 (DQ8), HLA-DRB1*03:01 (DR3), and DQA1*05:01-DQB1*02:01 (DQ2), confer increased risk for islet autoantibodies and subsequent type 1 diabetes (6, 13). The type of first-appearing autoantibody has been shown to be associated with HLA-DR-DQ haplotypes (3, 13, 14). The highest risk for a first islet autoantibody was conferred by the HLA-DR3/4-DQ2/8 genotype, but only 7% of individuals with this genotype develop type 1 diabetes (15). The heterozygous genotype effectively presents autoantigen epitopes to T cells perhaps because of the formation of a trans-complementing HLA-DQ heterodimer consisting of the HLA-DQ2 α-chain (DQA1*05:01) and the HLA-DQ8 β-chain (DQB1*03:02) (16, 17).

The recently reported haplotype of three single-nucleotide polymorphisms (tri-SNP; rs3135394, rs9268645, and rs3129877) in intron 1 of the HLA-DRA1 gene was shown to be strongly associated with type 1 diabetes risk in HLA-DR3 homozygous individuals (18, 19). One tri-SNP, AGG, was found to increase the risk for type 1 diabetes in HLA-DR3 homozygous subjects, while another haplotype, GCA, was found to be protective (18). The individual SNPs were also found in the 1000 Genome database to be associated with HLA-DR and -DQ expression in EBV-transformed B cells. If the tri-SNP modulates class II expression, regulated expression of class II genes may be related to an increased risk for type 1 diabetes (18).

The aim of this study was to investigate HLA-DQ cell-surface median fluorescence intensity (MFI) in subjects with an increasing number of islet autoantibodies on selected peripheral blood cells (19) in relation to the tri-SNPs. We hypothesize that the HLA-DRA1 tri-SNP may be related to peripheral blood cell HLA-DQ cell-surface expression. We also hypothesize that autoimmunity burden defined as autoantibody exposure at (1) a cross-sectional analysis or (2) cumulative during long-term follow-up in the Diabetes Prediction in Skåne (DiPiS) study may affect MFI differently dependent on the tri-SNP. If the tri-SNPs were to be found to be associated with HLA-DQ cell-surface expression, it is also possible that they are associated with either the risk for autoantibodies, the subsequent type 1 diabetes, or both.

The DiPiS study complied with the ethical standards of the institutional and/or national research committee (the Regional Ethical Review Board in Lund: No. 2009/244, No. 2014/196, and No. 2015/861) for involving human participants, as well as the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. In the DiPiS study, initiated in 2000 (20, 21), application of a eutectic mixture of lidocaine cream preceded blood draws to reduce pain. Informed consent was obtained from the parents of the research subjects to participate in the present cross-sectional follow-up investigation.

The research subjects (n = 67) in this study cohort were all children, aged 10–15 y, randomly selected from their participation in the DiPiS study (Table I). In brief, DiPiS is a prospective population-based cohort study with the aim to investigate the genetic and environmental factors that might contribute to, or trigger, the development of type 1 diabetes. After screening at birth, subjects with increased risk for type 1 diabetes were asked to participate in a follow-up study from 2 to 15 y of age or until diabetes onset, whichever occurred first (2022). The subjects in the present cohort had an increased HLA genetic risk for type 1 diabetes and a varying number of autoantibodies as previously described (23). According to the current nomenclature, the subjects would be classified as stage 1 type 1 diabetes (23, 24). The timeline of autoantibody status (Supplemental Fig. 1) displays the autoantibody status at DiPiS visits and the time of sampling into our study.

β cell autoantibodies, also known as islet autoantibodies, were previously analyzed in plasma using in-house standard radiobinding methods (23). In brief, autoantibodies against insulin (IAA), glutamate decarboxylase (GADA), insulinoma-associated protein-2, and three variants of zinc transporter 8 autoantibodies to any of the amino acid variants at position 325 (ZnT8A; against either arginine, tryptophan, or glutamine at position 325 [R/W/Q], respectively) were measured annually or quarterly in sera or plasma throughout the DiPiS study, as well as in our cross-sectional sample (23, 25, 26).

In this study, the first-appearing autoantibody was identified from DiPiS follow-up data. In four subjects where multiple autoantibodies were identified in the first autoantibody-positive sample, the first-appearing autoantibody was attributed by applying prior knowledge of the first-appearing autoantibody relative to HLA-DR-DQ and age, as previously described (3, 14, 27, 28).

Peripheral blood cells were isolated and HLA-DQ cell-surface MFI was analyzed as previously described (23). In brief, a peripheral blood sample (30–50 ml) was obtained to isolate CD19+ B cells, CD16+ cells, CD14+CD16 monocytes, CD4+ T cells, CD8+ T cells, and CD16+CD66+ neutrophils from PBMCs and RBCs, respectively, using magnetic-associated cell sorting (Miltenyi Biotec, Bergisch Gladbach, Germany). HLA-DQ MFI was identified with flow cytometry using a gating strategy for identifying cell populations and HLA-DQ MFI on the isolated cell subsets adapted from Dang et al. (29).

Flow cytometry analyses had previously been carried out using the CyAnADP flow cytometer (Beckman Coulter, Brea, CA) and the Summit v4.3 software (DAKO, Copenhagen, Denmark) for each of the cell subsets. Quality control was performed once a week on the flow cytometer. Peripheral blood cell subsets were stained with titrated mAbs (Supplemental Table I) and analyzed with their own fluorescence minus one (FMO) controls. The titrated mAbs were used to stain cell subsets in a gating strategy (Supplemental Fig. 2) adapted from Dang et al. (29). First, duplicate events were removed using forward scatter height plotted against forward scatter area. Second, cell populations were identified in the initial gate using side scatter and forward scatter. Up to 10,000 events were recorded. The cell population was plotted in a histogram using the appropriate cell specific for HLA-DQ Ab for identification of cells and HLA-DQ MFI, respectively. Unstained, not sorted, PBMCs or erythrocyte pellets were used as negative controls in the plot. Data of HLA-DQ cell-surface MFI from six subjects, where an APC fluorescent marker was used, were not comparable with the FITC fluorescent marker. These six subjects were therefore excluded from the analyses.

Cells were fixed in 4% (v/v) formaldehyde and 0.01% (v/v) sodium azide in PBS and analyzed in the flow cytometer within 7 d (median, 3 d) of fixation. The instrument was calibrated before each run using UltraComp eBeads (catalog number 01–2222; Affymetrix eBioscience, Santa Clara, CA). Acquired data were analyzed with Kaluza Analysis Software 1.5a (Beckman Coulter).

The HLA-DQ Ab clone REA303 recognizes the same epitopes as the pan DQ-specific SPV-L3 (30). FMO controls were set up for analyzing CD16+CD66+ and CD14+CD16 cells. Stained and unstained PBMCs, as well as stained and unstained erythrocytes, were used as positive and negative controls in the flow analyses for each subject. For CD66b and HLA-DQ, two different markers, a second Ab was used because of order and delivery complications and the necessity to analyze fresh samples. For a few subjects, CD16+CD66+ were stained with CD66b Brilliant Blue 515 (BB515) instead of CD66b FITC; this should not affect the purity assessment because all samples are individually analyzed with their own FMO controls. Samples from a few subjects were stained with HLA-DQ BB515 instead of HLA-DQ FITC. Data from samples stained with HLA-DQ BB515 were excluded from analysis because of the difference in excitation levels of FITC and BB515.

Previously, HLA class II-DRB345, -DRB1, -DQA1, -DQB1, -DPA1, and -DPB1 were determined by next generation sequencing (NGS) (23). In brief, 6-mm punch-outs of dried blood spots were sent blinded to Cisco Systems (Seattle, WA) for NGS. Allelic information and the online database (Allele Frequencies in Worldwide Population; http://www.allelefrequencies.net) (31) were used to assemble the extended HLA haplotypes.

Genotyping HLA-DRA1 SNPs was performed using DNA isolated from previously obtained PBMCs (23). DNA was isolated using QIAamp DNA Blood mini kit (catalog number 51106; Qiagen, Hilden, Germany) as per the manufacturer’s instructions. Isolated DNA concentrations were established with Quant-iT PicoGreen dsDNA Assay Kit (catalog number P7589; Invitrogen, Carlsbad, CA) and CLARIOstar Plus (BMG LABTECH, Ortenberg, Germany). Polymorphisms of the three SNPs in intron 1 of the HLA-DRA1 gene were investigated with predesigned TaqMan SNP Genotyping Assays (catalog number 4351379; Thermo Fisher Scientific, Waltham, MA) for each SNP ID (rs3135394, rs9268645, and rs3129877). The dried-down DNA delivery method was applied, as described in the TaqMan SNP Genotyping Assays User Guide (Thermo Fisher Scientific). The PCR was run on a QuantStudio 7 system, and post-PCR data were analyzed as described in TaqMan SNP Genotyping Assay User Guide recommendations for genotyping experiment. Each genotyping assay consists of two sequence-specific primers and two TaqMan minor groove binder probes with nonfluorescent quenchers. Allele 1 and 2 sequences are detected by one probe each labeled with VIC and FAM dye, respectively. The SNP transition substitutions for each of the assays were automatically assigned by the software as Homozygous Allele 1, Homozygous Allele 2, or Heterozygous Allele 1/Allele 2. The context sequence [VIC/FAM], provided by the manufacturer, identified the transition substitution for rs3135394, rs9268645, and rs3129877 as [A/G], [C/G], and [A/G], respectively; hence the SNP typing results were decoded to the corresponding polymorphism. Tri-SNP (rs3135394, rs9268645, and rs3129877) haplotypes were assembled in association to HLA-DRB345-DRB1-DQA1-DQB1. First, SNPs homozygous for Alleles 1 and 2 enabled an interpretation of the association to the HLA-DR-DQ haplotype. HLA-DRA1 SNP rs3129877, closest to and perhaps in linkage disequilibrium (LD) with the extended HLA-DR-DQ haplotype, was estimated first, rs9268645 second, and rs3135394 third. Positions of the three SNPs are described in Supplemental Fig. 3.

HLA-DRA1 tri-SNPs were investigated in relation to extended HLA class II haplotypes and autoimmunity burden, previously conceptualized as sampling autoimmunity burden (sAB) and cumulative autoimmunity burden (cAB) (23). The number of autoantibodies present at the time of sampling was used to estimate the sAB (negative, single, or multiple = 0, 1, or 2+, respectively). ZnT8A was counted as 1 if any of the three types, tryptophan 325 zinc transporter 8 autoantibody, arginine 325 zinc transporter 8 autoantibody, or glutamine 325 zinc transporter 8 autoantibody, were detected. The cAB was estimated as the area under the trajectory of autoantibodies over time during DiPiS follow-up and stratified into tertiles (low, medium, high = [0, 2.79], [2.79, 8.43], and [8.43, 30], respectively). In this study, cAB included DiPiS follow-up from 2 y of age until the time of sampling (Supplemental Fig. 1).

The sample analysis plan was to analyze samples individually by flow cytometry and group the subjects by tri-SNPs and genotypes to evaluate whether there was a difference in HLA-DQ cell-surface expression on the individual cell subsets related to the autoimmunity burden and HLA.

Analyzing the first-appearing autoantibody, data were excluded from analysis if there was a gap in the DiPiS follow-up before the first autoantibody-positive sample. Out of the autoantibody-positive subjects (n = 54), 14 subjects were excluded because of a gap in their DiPiS follow-up before the first-appearing autoantibody. Furthermore, in four subjects, we applied prior knowledge of HLA-DR-DQ relative to the type of first-appearing autoantibody because of multiple autoantibodies in the first autoantibody-positive sample. IAA and ZnT8A were found in subjects with DQ2/DQ8 (n = 2) and DQ2.1/DQ8.4 (n = 1). Insulinoma-associated protein-2 autoantibody and ZnT8A were found in relation to DQ6.4/DQ8 (n = 1). Published data helped attribute either IAA or ZnT8A, respectively, as the first-appearing autoantibody (3, 14, 32).

Boxplots were used to examine peripheral blood cell HLA-DQ cell-surface MFI stratified by HLA-DRA1 tri-SNPs and autoimmunity burden (sAB and cAB as factors). Analysis of tri-SNP genotypes was not possible due to the limited number of subjects. Kruskal–Wallis and likelihood ratio tests were applied to examine the association of HLA-DQ cell-surface expression, tri-SNPs, and autoimmunity burden.

The association between HLA-DQ cell-surface MFI and tri-SNPs was further examined with linear models, using HLA-DQ cell-surface MFI as the outcome and tri-SNP as the predictor. The models were adjusting for age at sampling, sex, and HLA-DQ2/8. To determine whether autoimmunity was a mediator of the association between the HLA-DQ cell-surface MFI and tri-SNP, additional models were fitted adjusting for autoimmunity burden (sAB and cAB). Distribution and possible outliers for each of the isolated cell subsets were identified using histograms of HLA-DQ cell-surface MFI. SEs were estimated using robust and generalized linear methods.

R version 3.6.1 (http://www.r-project.org) was used for the analysis. The p values presented are nominal and considered suggestive of an association if <0.05. False discovery rate at 5% was controlled by Benjamini–Hochberg procedure (23). An asterisk indicates the p values that remained significant after adjusting for multiple comparisons and that can be considered statistically significant.

Four tri-SNPs were identified in the present cohort (Tables I, II). Among the haplotypes, AGG was the most common, representing 50.0% (n = 67) of the entire cohort of 67 individuals, followed by GCA (35.1%, n = 47), ACG (8.2%, n = 11), and ACA (6.7%, n = 9). The tri-SNP genotypes and the extended class II HLA-DRB345-DRB1-DQA1-DQB1 haplotypes are summarized in Table III.

Table I.

Study sample characteristics

CharacteristicsStudy subjects (n = 67)
Female, n (%) 34 (50.7) 
Age (y), mean (SD) 13.0 (1.20) 
Follow-up time (y), mean (SD) 9.84 (2.38) 
Follow-up visits, mean (SD) 11.6 (4.25) 
HLA-DQ2/8, n (%) 41 (61.2) 
T1D diagnosis after sampling, n (%) 5 (7.5) 
Autoimmunity burden  
 sAB, n (%)  
  0 26 (38.8) 
  1 23 (34.3) 
  2 18 (26.9) 
 cAB, n (%)  
  Low 25 (37.3) 
  Medium 24 (35.8) 
  High 18 (26.9) 
CharacteristicsStudy subjects (n = 67)
Female, n (%) 34 (50.7) 
Age (y), mean (SD) 13.0 (1.20) 
Follow-up time (y), mean (SD) 9.84 (2.38) 
Follow-up visits, mean (SD) 11.6 (4.25) 
HLA-DQ2/8, n (%) 41 (61.2) 
T1D diagnosis after sampling, n (%) 5 (7.5) 
Autoimmunity burden  
 sAB, n (%)  
  0 26 (38.8) 
  1 23 (34.3) 
  2 18 (26.9) 
 cAB, n (%)  
  Low 25 (37.3) 
  Medium 24 (35.8) 
  High 18 (26.9) 

The subjects (n = 67), screened for type 1 diabetes (T1D) risk at birth, were followed annually or quarterly from 2 y of age in the DiPiS study. The subjects were followed annually if they had no or one autoantibody (AAb) or quarterly if they had multiple (AAb). Autoimmunity burden defines the sAB or cAB during follow-up in DiPiS.

Table II.

Study sample characteristics of the 67 subjects stratified by the HLA-DRA1 tri-SNPs

HLA-DRA tri-SNPs (n = 134) in 67 study subjectsACA (n = 9)ACG (n = 11)AGG (n = 67)GCA (n = 47)
Female, n (%) 5 (55.6) 6 (54.5) 33 (49.3) 24 (51.1) 
Age (y), mean (SD) 12.5 (1.48) 12.6 (0.97) 13.1 (1.16) 13.0 (1.25) 
HLA-DQ2/8, n (%) 0 (0) 0 (0) 46 (68.7) 36 (76.6) 
T1D diagnosis after sampling, n (%) 2 (22.2) 1 (9.1) 6 (9.0) 1 (2.1) 
sAB, n (%)     
 0 1 (11.1) 0 (0) 27 (40.3) 24 (51.1) 
 1 3 (33.3) 6 (54.5) 27 (31.3) 16 (34.0) 
 2+ 5 (55.6) 5 (45.5) 19 (28.4) 7 (14.9) 
cAB, n (%)     
 Low 1 (11.1) 1 (9.1) 26 (38.8) 22 (46.8) 
 Medium 4 (44.4) 5 (45.5) 23 (34.3) 16 (34.0) 
 High 4 (44.4) 5 (45.5) 18 (26.9) 9 (19.1) 
HLA-DRA tri-SNPs (n = 134) in 67 study subjectsACA (n = 9)ACG (n = 11)AGG (n = 67)GCA (n = 47)
Female, n (%) 5 (55.6) 6 (54.5) 33 (49.3) 24 (51.1) 
Age (y), mean (SD) 12.5 (1.48) 12.6 (0.97) 13.1 (1.16) 13.0 (1.25) 
HLA-DQ2/8, n (%) 0 (0) 0 (0) 46 (68.7) 36 (76.6) 
T1D diagnosis after sampling, n (%) 2 (22.2) 1 (9.1) 6 (9.0) 1 (2.1) 
sAB, n (%)     
 0 1 (11.1) 0 (0) 27 (40.3) 24 (51.1) 
 1 3 (33.3) 6 (54.5) 27 (31.3) 16 (34.0) 
 2+ 5 (55.6) 5 (45.5) 19 (28.4) 7 (14.9) 
cAB, n (%)     
 Low 1 (11.1) 1 (9.1) 26 (38.8) 22 (46.8) 
 Medium 4 (44.4) 5 (45.5) 23 (34.3) 16 (34.0) 
 High 4 (44.4) 5 (45.5) 18 (26.9) 9 (19.1) 

Autoimmunity burden is defined as the sAB and cAB during follow-up in the DiPiS study.

T1D, type 1 diabetes.

Table III.

The tri-SNP genotypes and class II HLA-DRB345-DRB1-DQA1-DQB1 haplotypes identified in the cohort of 67 subjects with varying number of β cell AAb

Allele 1Allele 2
Tri-SNP allele 1/2DRB345DRB1DQA1DQB1DRB345DRB1DQA1DQB1DQ2/8 (n = 41)Non-DQ2/8 (n = 26)Subjects (n = 67)
ACA/ACG DRB3*03:01:01 DRB1*13:02:01 DQA1*01:02:01 DQB1*06:04:01 DRB345*Not_present DRB1*01:02:01 DQA1*01:01:02 DQB1*05:01:01 
ACA/AGG DRB345*Not_present DRB1*01:01:01 DQA1*01:01:01 DQB1*05:01:01 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 
 DRB3*03:01:01 DRB1*13:02:01 DQA1*01:02:01 DQB1*06:04:01     
      DRB1*04:04:01   
ACA/GCA DRB345*Not_present DRB1*01:01:01 DQA1*01:01:01 DQB1*05:01:01 DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 
  DRB1*01:03:01 DQA1*05:05:01 DQB1*03:01:01     
  DRB1*10:01:01 DQA1*01:05:01 DQB1*05:01:01     
 DRB3*03:01:01 DRB1*13:02:01 DQA1*01:02:01 DQB1*06:04:01     
ACG/ACG DRB3*02:02:01 DRB1*12:01:01 DQA1*05:05:01 DQB1*03:01:01 DRB3*01:01:02 DRB1*13:03:01 DQA1*05:05:01 DQB1*03:01:01 
ACG/AGG DRB3*02:02:01 DRB1*14:54:01 DQA1*01:04:01 DQB1*05:03:01 DRB4*01:03:02 DRB1*09:01:02 DQA1*03:02:01 DQB1*03:03:02 
 DRB4*01:01:01 DRB1*07:01:01 DQA1*02:01:01 DQB1*02:02:01 DRB4*01:03:01 DRB1*04:08:01 DQA1*03:03:01 DQB1*03:04:01 
      DRB1*04:04:01 DQA1*03:01:01 DQB1*03:02:01 
 DRB4*01:03:01 DRB1*04:05:01 DQA1*03:03:01 DQB1*03:03:02  DRB1*04:01:01   
  DRB1*07:01:01 DQA1*02:01:01      
ACG/GCA DRB345*Not_present DRB1*08:01:01 DQA1*04:02:01 DQB1*04:02:01 DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 
 DRB3*02:02:01 DRB1*12:01:01 DQA1*05:05:01 DQB1*03:01:01     
AGG/AGG DRB3*02:02:01 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 
      DRB1*04:02:01   
      DRB1*04:04:01   
 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 
     DRB4*01:03:02 DRB1*09:01:02 DQA1*03:02:01  
  DRB1*04:05:01 DQA1*03:03:01  DRB4*01:03:01 DRB1*04:03:01 DQA1*03:01:01  
  DRB1*07:01:01 DQA1*02:01:01 DQB1*02:02:01  DRB1*04:01:01   
       DQA1*03:03:01 DQB1*03:01:01 
GCA/AGG DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 DRB3*02:02:01 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 
     DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 19 19 
       DQA1*03:03:01  
      DRB1*04:02:01 DQA1*03:01:01  
      DRB1*04:04:01   10 10 
      DRB1*04:05:01 DQA1*03:03:01  
      DRB1*04:08:01  DQB1*03:04:01  
     DRB4*01:03:02 DRB1*04:04:01 DQA1*03:01:01 DQB1*03:02:01 
GCA/GCA DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 
Allele 1Allele 2
Tri-SNP allele 1/2DRB345DRB1DQA1DQB1DRB345DRB1DQA1DQB1DQ2/8 (n = 41)Non-DQ2/8 (n = 26)Subjects (n = 67)
ACA/ACG DRB3*03:01:01 DRB1*13:02:01 DQA1*01:02:01 DQB1*06:04:01 DRB345*Not_present DRB1*01:02:01 DQA1*01:01:02 DQB1*05:01:01 
ACA/AGG DRB345*Not_present DRB1*01:01:01 DQA1*01:01:01 DQB1*05:01:01 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 
 DRB3*03:01:01 DRB1*13:02:01 DQA1*01:02:01 DQB1*06:04:01     
      DRB1*04:04:01   
ACA/GCA DRB345*Not_present DRB1*01:01:01 DQA1*01:01:01 DQB1*05:01:01 DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 
  DRB1*01:03:01 DQA1*05:05:01 DQB1*03:01:01     
  DRB1*10:01:01 DQA1*01:05:01 DQB1*05:01:01     
 DRB3*03:01:01 DRB1*13:02:01 DQA1*01:02:01 DQB1*06:04:01     
ACG/ACG DRB3*02:02:01 DRB1*12:01:01 DQA1*05:05:01 DQB1*03:01:01 DRB3*01:01:02 DRB1*13:03:01 DQA1*05:05:01 DQB1*03:01:01 
ACG/AGG DRB3*02:02:01 DRB1*14:54:01 DQA1*01:04:01 DQB1*05:03:01 DRB4*01:03:02 DRB1*09:01:02 DQA1*03:02:01 DQB1*03:03:02 
 DRB4*01:01:01 DRB1*07:01:01 DQA1*02:01:01 DQB1*02:02:01 DRB4*01:03:01 DRB1*04:08:01 DQA1*03:03:01 DQB1*03:04:01 
      DRB1*04:04:01 DQA1*03:01:01 DQB1*03:02:01 
 DRB4*01:03:01 DRB1*04:05:01 DQA1*03:03:01 DQB1*03:03:02  DRB1*04:01:01   
  DRB1*07:01:01 DQA1*02:01:01      
ACG/GCA DRB345*Not_present DRB1*08:01:01 DQA1*04:02:01 DQB1*04:02:01 DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 
 DRB3*02:02:01 DRB1*12:01:01 DQA1*05:05:01 DQB1*03:01:01     
AGG/AGG DRB3*02:02:01 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 
      DRB1*04:02:01   
      DRB1*04:04:01   
 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 
     DRB4*01:03:02 DRB1*09:01:02 DQA1*03:02:01  
  DRB1*04:05:01 DQA1*03:03:01  DRB4*01:03:01 DRB1*04:03:01 DQA1*03:01:01  
  DRB1*07:01:01 DQA1*02:01:01 DQB1*02:02:01  DRB1*04:01:01   
       DQA1*03:03:01 DQB1*03:01:01 
GCA/AGG DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 DRB3*02:02:01 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 
     DRB4*01:03:01 DRB1*04:01:01 DQA1*03:01:01 DQB1*03:02:01 19 19 
       DQA1*03:03:01  
      DRB1*04:02:01 DQA1*03:01:01  
      DRB1*04:04:01   10 10 
      DRB1*04:05:01 DQA1*03:03:01  
      DRB1*04:08:01  DQB1*03:04:01  
     DRB4*01:03:02 DRB1*04:04:01 DQA1*03:01:01 DQB1*03:02:01 
GCA/GCA DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 DRB3*01:01:02 DRB1*03:01:01 DQA1*05:01:01 DQB1*02:01:01 

The three SNPs in intron 1 of HLA-DRA1 were imputed based on homozygosity and as the frequency of alleles co-occurring with flanking HLA markers, typed by NGS, in subjects homozygous for one marker but not the other. Alleles of the tri-SNP closest to the extended HLA-DRB1-DRB345-DQA1-DQB1 were imputed first, then the second and third SNP alleles.

AAb, autoantibodies.

The four tri-SNPs were associated with 13 different extended HLA-DR-DQ haplotypes (Fig. 1). In AGG subjects, 13 HLA-DR-DQ haplotypes were identified, distributed over three different DRB345 types where DRB4*01:03:01 (n = 57) was the most common followed by DRB3*02:02:01 (n = 6) and DRB4*01:03:02 (n = 4). In GCA, only a single HLA-DR-DQ haplotype was identified, DRB3*01:01:02-DRB1*03:01:01-DQA1*05:01:01-DQB1*02:01:01. In ACA and ACG subjects, four and eight HLA-DR-DQ haplotypes were identified, distributed over two and six DRB345 variants, respectively (Fig. 1). Furthermore, three prominent HLA-DRB1-DQA1-DQB1 haplotypes were identified with AGG and GCA tri-SNPs: DRB1*03:01:01-DQA1*05:01:01-DQB1*02:01:01 (AGG, n = 6; GCA, n = 47), DRB1*04:01:01-DQA1*03:01:01-DQB1*03:02:01 (AGG, n = 31), and DRB1*04:04:01-DQA1*03:01:01-DQB1*03:02:01 (AGG, n = 15). Furthermore, the AGG and GCA tri-SNPs were found in subjects with HLA-DQ2/8 genotype (n = 32 and n = 21, respectively) and non–HLA-DQ2/8 genotype (n = 24 and n = 11, respectively). ACA and ACG tri-SNPs were present only in subjects with non–HLA-DQ2/8 genotype (n = 9 and n = 11, respectively).

FIGURE 1.

Schematic of the HLA-DRA1 tri-SNPs and the respective class II extended HLA haplotypes.

For the n = 67 subjects in this study, the three SNPs in intron 1 of HLA-DRA1 were imputed based on homozygous alleles, as well as the frequency of alleles co-occurring with flanking HLA markers in subjects homozygous for one marker but not the other. Alleles of the tri-SNP closest to the extended HLA-DRB1-DRB345-DQA1-DQB1 were imputed first, then the second and third SNP alleles.

FIGURE 1.

Schematic of the HLA-DRA1 tri-SNPs and the respective class II extended HLA haplotypes.

For the n = 67 subjects in this study, the three SNPs in intron 1 of HLA-DRA1 were imputed based on homozygous alleles, as well as the frequency of alleles co-occurring with flanking HLA markers in subjects homozygous for one marker but not the other. Alleles of the tri-SNP closest to the extended HLA-DRB1-DRB345-DQA1-DQB1 were imputed first, then the second and third SNP alleles.

Close modal

Next, we wanted to test whether there was an association between the tri-SNPs and autoimmunity burden. In this study, 13 subjects are autoantibody negative in their follow-up in DiPiS, and all have the HLA-DQ2/8 genotype. These subjects also have the AGG (n = 14) and GCA (n = 12) tri-SNPs. We found no association between the tri-SNPs and the presence of autoantibodies (p = 0.110) (Table IV). However, there was an association between HLA-DRA1 tri-SNPs and sAB (p = 0.004) (Table IV), but not with cAB (p = 0.126) (Table IV). To test whether there was an association between the tri-SNPs and the first-appearing autoantibody, subjects (n = 14) who had missed a visit before the first autoantibody-positive sample were excluded from analysis. The remaining autoantibody-positive subjects (n = 40) were represented by all four tri-SNPs (Table V). In four cases, the first-appearing autoantibody was attributed as two autoantibodies were detected simultaneously in the first autoantibody-positive sample. In these subjects, literature of HLA and age relative to the first-appearing autoantibody helped attribute an autoantibody as the first-appearing autoantibody. Subjects (n = 40) without a break in their DiPiS follow-up before developing a first autoantibody had glutamate decarboxylase autoantibody (GADA; n = 20, 50.0%), IAA (n = 18, 45.0%), or ZnT8A (n = 2, 5.0%) as the first-appearing autoantibody. Due to HLA-DQ2/8 being a strong confounder for the risk for type 1 diabetes, we tested whether there was an association between the HLA-DRA1 tri-SNPs and the first-appearing autoantibody for subjects with (1) HLA-DQ2/8 genotype and (2) non–HLA-DQ2/8 genotypes. In this cohort we observed no indication that the HLA-DRA1 tri-SNPs were impacting the type of first-appearing autoantibody, be it GADA or IAA, in subjects with HLA-DQ2/8 or non-DQ2/8 genotypes (Table VI).

Table IV.

HLA-DRA1 tri-SNPs were investigated in subjects (n = 67) with autoimmunity burden defined as the presence of AAb, AAb at sAB, or cAB during follow-up in DiPiS

AAb presenceACA (n = 9)ACG (n = 11)AGG (n = 67)GCA (n = 47)Total (n = 134)pa
AAb, n 14 12 26 0.110 
AAb+, n 11 53 35 108 0.110 
sAB, n      0.004 
 AAb 27 24 52  
 1 AAb 21 16 46  
 2+ AAb 19 36  
cAB, n      0.126 
 Low 26 22 50  
 Medium 23 16 48  
 High 18 36  
AAb presenceACA (n = 9)ACG (n = 11)AGG (n = 67)GCA (n = 47)Total (n = 134)pa
AAb, n 14 12 26 0.110 
AAb+, n 11 53 35 108 0.110 
sAB, n      0.004 
 AAb 27 24 52  
 1 AAb 21 16 46  
 2+ AAb 19 36  
cAB, n      0.126 
 Low 26 22 50  
 Medium 23 16 48  
 High 18 36  

There is an association between the tri-SNPs and sAB, but not with the presence of autoantibodies (AAbs) at any time during follow-up in DiPiS or cAB.

a

Fisher's exact test was used to test whether there was an association between the tri-SNPs and AAb burden.

AAb, absence of AAb; AAb+, presence of AAb.

Table V.

The HLA-DRA tri-SNPs of the AAb-positive study subjects (n = 40)

First-appearing AAbACA (n = 8)ACG (n = 8)AGG (n = 39)GCA (n = 25)
GADA, n 18 11 
IAA, n 19 13 
ZnT8A, n 
First-appearing AAbACA (n = 8)ACG (n = 8)AGG (n = 39)GCA (n = 25)
GADA, n 18 11 
IAA, n 19 13 
ZnT8A, n 

Autoantibody (AAb)-positive subjects with annual follow-up until developing the first-appearing AAb were included. Subjects with a gap in the follow-up before developing a first AAb were excluded.

Table VI.

HLA-DRA1 tri-SNPs in subjects (n = 67) with either GADA or IAA as the first-appearing AAb and HLA-DQ2/8 or non–HLA-DQ2/8

ACAACGAGGGCATotalpa
HLA-DQ2/8       
 GADA first, n NA NA 14 22 0.363 
 IAA first, n NA NA 13 11 24 0.363 
Non–HLA-DQ2/8       
 GADA+, n 18 0.343 
 GADA, n 16 0.343 
 IAA+, n 12 0.615 
 IAA, n 22 0.615 
ACAACGAGGGCATotalpa
HLA-DQ2/8       
 GADA first, n NA NA 14 22 0.363 
 IAA first, n NA NA 13 11 24 0.363 
Non–HLA-DQ2/8       
 GADA+, n 18 0.343 
 GADA, n 16 0.343 
 IAA+, n 12 0.615 
 IAA, n 22 0.615 

Subjects with a gap in their follow-up before developing a first autoantibody (AAb) were excluded. Subjects with HLA-DQ2/8 had either AGG or GCA. The other two tri-SNPs, ACA and ACG, were observed in subjects with a non–HLA-DQ2/8 genotype only.

a

Fisher's exact test was used to compare tri-SNPs stratified by HLA-DQ genotype (HLA-DQ2/8 or non-DQ2/8) and the first-appearing AAb.

NA, not applicable.

A pattern of decreased HLA-DQ cell-surface MFI on peripheral blood cells was observed with the AGG relative to the GCA tri-SNP (Fig. 2). HLA-DQ cell-surface MFI was found to be lower on CD4+ T cells (p = 0.030) with the AGG tri-SNP. Due to a limited number of subjects, ACA and ACG tri-SNPs were excluded from analysis of HLA-DQ cell-surface MFI analysis. Further stratifying HLA-DQ cell-surface MFI by autoimmunity burden revealed a pattern of decreasing HLA-DQ cell surface MFI with increasing autoimmunity burden in both AGG and GCA tri-SNPs (Fig. 3), For the AGG tri-SNP only, lower HLA-DQ cell-surface MFI was found on CD16+ (p = 0.0013*), CD4+ T (p = 0.0018*), and CD8+ T (p = 0.016) cells for cAB (Fig. 3) and on CD16+ (p = 0.0012*), CD14+CD16 (p = 0.019), CD4+ T (p = 0.00018*), and CD8+ T (p = 0.0075*) cells for sAB (data not shown). Observations that remained significant after adjusting for multiple comparisons with the Benjamini–Hochberg method are marked with an asterisk (*).

FIGURE 2.

A pattern of decreased HLA-DQ cell-surface MFI on isolated peripheral blood cells was observed with the AGG relative to the GCA tri-SNP.

Low, medium, and high cAB describe the burden of autoantibodies during follow-up in DiPiS. A pattern of decreased HLA-DQ MFI was observed with the AGG haplotype in all isolated cell types (n = 67 subjects). Lower HLA-DQ MFI was observed with the AGG haplotype in CD4+ T cells. Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. No observations remained significant after adjusting for multiple comparisons using the Benjamini–Hochberg method.

FIGURE 2.

A pattern of decreased HLA-DQ cell-surface MFI on isolated peripheral blood cells was observed with the AGG relative to the GCA tri-SNP.

Low, medium, and high cAB describe the burden of autoantibodies during follow-up in DiPiS. A pattern of decreased HLA-DQ MFI was observed with the AGG haplotype in all isolated cell types (n = 67 subjects). Lower HLA-DQ MFI was observed with the AGG haplotype in CD4+ T cells. Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. No observations remained significant after adjusting for multiple comparisons using the Benjamini–Hochberg method.

Close modal
FIGURE 3.

HLA-DQ cell-surface MFI on isolated peripheral blood cells, stratified by DRA1 tri-SNPs and autoimmunity burden as measured by cumulative burden of autoantibodies during follow-up in DiPiS (cAB).

A pattern of decreased HLA-DQ MFI was observed with both tri-SNPs in all isolated cell types (n = 67 subjects) with increasing autoimmunity burden. Lower HLA-DQ MFI was observed with increasing autoimmunity burden with the AGG haplotype in CD16+ cells, CD14+CD16 monocytes, and CD4+ and CD8+ T cells. Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. Observations that remained significant after adjusting for multiple comparisons with the Benjamini–Hochberg method are marked with an asterisk (*).

FIGURE 3.

HLA-DQ cell-surface MFI on isolated peripheral blood cells, stratified by DRA1 tri-SNPs and autoimmunity burden as measured by cumulative burden of autoantibodies during follow-up in DiPiS (cAB).

A pattern of decreased HLA-DQ MFI was observed with both tri-SNPs in all isolated cell types (n = 67 subjects) with increasing autoimmunity burden. Lower HLA-DQ MFI was observed with increasing autoimmunity burden with the AGG haplotype in CD16+ cells, CD14+CD16 monocytes, and CD4+ and CD8+ T cells. Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. Observations that remained significant after adjusting for multiple comparisons with the Benjamini–Hochberg method are marked with an asterisk (*).

Close modal

In this cohort, HLA-DRB3 01:01:02 and DRB3*02:02:01 are the only alleles separating subjects with the GCA (n = 47) and AGG (n = 6) tri-SNPs (Fig. 1). We investigated whether the DRA tri-SNP or HLA-DRB3 impacted HLA-DQ MFI between subjects with the GCA (n = 47) and AGG (n = 61) tri-SNPs by removing the subjects (n = 6) with HLA-DRB3*02:02:01-DRB1*03:01:01-DQA1*05:01:01-DQB1*02.01:01 from the AGG haplotype group. The HLA-DQ MFI did not differ between GCA and AGG subjects (Fig. 4A), as previously seen on CD4+ T cells in (Fig. 2. However, HLA-DQ MFI still decreases with increasing autoimmunity burden cAB on CD16+ (p = 0.0022*), CD4+ (p = 0.0024*), and CD8+ (p = 0.015) cells (Fig. 4B). Similarly, HLA-DQ MFI decreases with increasing autoimmunity burden sAB on CD16+ (p = 0.0017), CD4+ (p = 0.00022), and CD8+ (p = 0.006) cells (data not shown). Observations that remained significant after adjusting for multiple comparisons with the Benjamini–Hochberg method are marked with an asterisk (*).

FIGURE 4.

Investigation of HLA-DQ MFI on isolated peripheral blood cells stratified by HLA-DRA1 tri-SNPs, excluding subjects (n = 6) with AGG-DRB3*02:02:01.

(A) GCA versus AGG tri-SNPs. (B) Autoimmunity burden within GCA and AGG tri-SNPs. We investigated whether the DRA tri-SNP or HLA-DRB3 impacted HLA-DQ MFI between subjects with the GCA (n = 47) and AGG (n = 61) tri-SNPs by removing the subjects (n = 6) with HLA-DRB3*02:02:01-DRB1*03:01:01-DQA1*05:01:01-DQB1*02.01:01 from the AGG haplotype group. The HLA-DQ MFI did not differ between GCA and AGG subjects (A), as previously seen on CD4+ T cells in (Fig. 2. However, HLA-DQ MFI still decreased with increasing autoimmunity burden (cAB) on CD16+, CD4+, and CD8+ cells (B). Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. Observations that remained significant after adjusting for multiple comparisons with the Benjamini–Hochberg method are marked with an asterisk (*).

FIGURE 4.

Investigation of HLA-DQ MFI on isolated peripheral blood cells stratified by HLA-DRA1 tri-SNPs, excluding subjects (n = 6) with AGG-DRB3*02:02:01.

(A) GCA versus AGG tri-SNPs. (B) Autoimmunity burden within GCA and AGG tri-SNPs. We investigated whether the DRA tri-SNP or HLA-DRB3 impacted HLA-DQ MFI between subjects with the GCA (n = 47) and AGG (n = 61) tri-SNPs by removing the subjects (n = 6) with HLA-DRB3*02:02:01-DRB1*03:01:01-DQA1*05:01:01-DQB1*02.01:01 from the AGG haplotype group. The HLA-DQ MFI did not differ between GCA and AGG subjects (A), as previously seen on CD4+ T cells in (Fig. 2. However, HLA-DQ MFI still decreased with increasing autoimmunity burden (cAB) on CD16+, CD4+, and CD8+ cells (B). Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. Observations that remained significant after adjusting for multiple comparisons with the Benjamini–Hochberg method are marked with an asterisk (*).

Close modal

The number of subjects was too few to analyze all of the tri-SNP genotypes identified in this cohort relative to the HLA-DQ MFI on isolated peripheral PBMCs. The preciously suggested tri-SNP risk genotypes were identified in this cohort where 10 subjects had the AGG/AGG, 1 had the GCA/GCA, and 38 subjects had the heterozygous GCA/AGG genotype (Table III). Subjects with the high-risk HLA-DQ2/8 genotype had either heterozygous GCA/AGG (n = 36) or homozygous AGG/AGG (n = 5), while subjects with other HLA-DQ genotypes, not HLA-DQ2/8, had heterozygous GCA/AGG (n = 2), AGG/AGG (n = 5), or homozygous GCA/GCA (n = 1), as well as different combinations of the four tri-SNP alleles (ACA, ACG, AGG, and GCA) (Table III). The subjects with a non–HLA-DQ2/8 genotype and heterozygous GCA/AGG (n = 2) had either DQ2/2 or DQ2/8.4; one subject with homozygous GCA/GCA had DQ2/2. The subjects with non–HLA-DQ2/8 and homozygous AGG/AGG (n = 5) had DQ2.1/7v, DQ2.1/8, DQ8/8, DQ8/8v, or DQ8/8.3.

Stratifying the tri-SNP genotypes by autoimmunity burden and the HLA-DQ2/8 or non–HLA-DQ2/8 genotype, we found that most subjects with HLA-DQ2/8 and AGG/AGG or GCA/AGG had no or low autoimmunity burden (Tables VII, VIII). One subject with a non–HLA-DQ2/8 genotype had ACA/GCA and no autoimmunity burden (Table VII), and two subjects had either ACA/GCA or ACG/AGG and low autoimmunity burden (Table VIII).

Table VII.

Overview of the tri-SNP genotypes and autoimmunity burden defined as AAb at sAB

HLA-DQ2/8Non–HLA-DQ2/8
Tri-SNP genotype0 (n = 25)1 (n = 12)2+ (n = 4)0 (n = 1)1 (n = 11)2+ (n = 14)
AGG/AGG 
AGG/ACA 
AGG/ACG 
GCA/AGG 23 
GCA/GCA 
GCA/ACA 
GCA/ACG 
ACA/ACG 
ACG/ACG 
HLA-DQ2/8Non–HLA-DQ2/8
Tri-SNP genotype0 (n = 25)1 (n = 12)2+ (n = 4)0 (n = 1)1 (n = 11)2+ (n = 14)
AGG/AGG 
AGG/ACA 
AGG/ACG 
GCA/AGG 23 
GCA/GCA 
GCA/ACA 
GCA/ACG 
ACA/ACG 
ACG/ACG 

Subjects with the HLA-DQ2/8 genotype had either AGG/AGG or GCA/AGG, while additional tri-SNP genotypes were found in subjects with non–HLA-DQ2/8 genotypes.

AAb, autoantibodies.

Table VIII.

Overview of the tri-SNP genotypes and autoimmunity burden defined as cAB during DiPiS follow-up (cAB)

HLA-DQ2/8Non–HLA-DQ2/8
Tri-SNP genotype0 (n = 23)1 (n = 11)2+ (n = 7)0 (n = 2)1 (n = 13)2+ (n = 11)
AGG/AGG 
AGG/ACA 
AGG/ACG 
GCA/AGG 21 
GCA/GCA 
GCA/ACA 
GCA/ACG 
ACA/ACG 
ACG/ACG 
HLA-DQ2/8Non–HLA-DQ2/8
Tri-SNP genotype0 (n = 23)1 (n = 11)2+ (n = 7)0 (n = 2)1 (n = 13)2+ (n = 11)
AGG/AGG 
AGG/ACA 
AGG/ACG 
GCA/AGG 21 
GCA/GCA 
GCA/ACA 
GCA/ACG 
ACA/ACG 
ACG/ACG 

Subjects with the HLA-DQ2/8 genotype had either AGG/AGG or GCA/AGG, while additional tri-SNP genotypes were found in subjects with non–HLA-DQ2/8 genotypes.

HLA-DQ MFI was investigated for subjects with heterozygous GCA/AGG and homozygous AGG/AGG (Fig. 5). HLA-DQ MFI was lower in subjects with homozygous AGG/AGG than heterozygous GCA/AGG on CD19+, CD16+, CD4+, and CD8+ cells (Fig. 5A). The pattern of decreasing HLA-DQ MFI with increasing cAB was identified for subjects heterozygous for GCA/AGG and homozygous for AGG/AGG (Fig. 5B).

FIGURE 5.

HLA-DQ MFI was investigated in subjects with heterozygous GCA/AGG (n = 38) and homozygous AGG/AGG (n = 10).

HLA-DQ MFI was lower in subjects with homozygous AGG/AGG than heterozygous GCA/AGG, on CD19+, CD16+, CD4+, and CD8+ cells. The pattern of decreasing HLA-DQ MFI with increasing autoimmunity burden [(A) sAB, (B) cAB] was identified for both subjects heterozygous for GCA/AGG and homozygous for AGG/AGG. Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. Observations that remained significant after adjusting for multiple comparisons with the Benjamini-Hochberg method are marked with an asterisk (*).

FIGURE 5.

HLA-DQ MFI was investigated in subjects with heterozygous GCA/AGG (n = 38) and homozygous AGG/AGG (n = 10).

HLA-DQ MFI was lower in subjects with homozygous AGG/AGG than heterozygous GCA/AGG, on CD19+, CD16+, CD4+, and CD8+ cells. The pattern of decreasing HLA-DQ MFI with increasing autoimmunity burden [(A) sAB, (B) cAB] was identified for both subjects heterozygous for GCA/AGG and homozygous for AGG/AGG. Kruskal–Wallis test was used to compare HLA-DQ cell-surface MFI in the isolated cell subsets stratified by the GCA and AGG tri-SNPs. Observations that remained significant after adjusting for multiple comparisons with the Benjamini-Hochberg method are marked with an asterisk (*).

Close modal

The association between HLA-DQ cell-surface MFI for HLA-DRA1 tri-SNPs AGG and GCA on six subtypes of isolated peripheral blood cells was analyzed in a model (1) without adjusting for autoimmunity burden (Fig. 6A, Table IX), (2) adjusting for sAB (Fig. 6B, Table IX), and (3) adjusting for cAB (Fig. 6C, Table IX). Estimates and 95% confidence intervals from the model were based on linear regression with robust SEs and were additionally adjusted for age at sampling and sex. Without adjusting for autoimmunity burden, HLA-DQ cell-surface MFI measured in subjects with the HLA-DQ2/8 genotype compared with non–HLA-DQ2/8 subjects was 0.10 (−0.17, −0.04; p = 0.001) lower on CD14+CD16 classical monocytes, 0.05 (−0.07, −0.03; p = 0.000) lower on CD4+ T cells, 0.06 (−0.08, −0.03; p = 0.000) lower on CD8+ T cells, and marginally lower on CD19+ B cells and CD16+ cells (Fig. 6A, Table IX). Additionally adjusting for autoimmunity burden, HLA-DQ cell-surface MFI measured in subjects with the HLA-DQ2/8 genotype compared with non–HLA-DQ2/8 subjects was 0.04 (−0.07, −0.01; p = 0.005) lower in CD8+ T cells, marginally lower in CD4+ T cells for sAB (Fig. 6B, Table IX) and 0.03 (−0.05, −0.01; p = 0.003) lower in CD4+ T cells, 0.06 (−0.10, −0.02; p = 0.002) lower in CD8+ T cells, and marginally lower on CD16+ cells and CD14+CD16 classical monocytes for cAB (Fig. 6C, Table IX). An association was observed between age and HLA-DQ cell-surface MFI on CD4+ T and CD8+ T cells without adjusting for autoimmunity burden (Fig. 6A, Table IX). The association with age remained after adjusting for autoimmunity burden, not sAB and cAB, but was also observed on CD16+ cells (Fig. 6B, 6C, Table IX).

FIGURE 6.

Estimates and 95% confidence intervals (Est, 95% CI) of the association between HLA-DQ cell-surface MFI on isolated peripheral blood cells in subjects with the AGG relative to the GCA HLA-DRA1 tri-SNPs.

Linear models of HLA-DQ cell-surface MFI as the outcome and tri-SNP as the predictor were adjusted for age, sex, HLA-DQ2/8 (A), and sAB (B) or cAB (C). Distribution and possible outliers were identified using histograms of HLA-DQ cell-surface MFI. Standard errors were estimated using robust linear methods. See Table IX for detailed results corresponding to these plots.

FIGURE 6.

Estimates and 95% confidence intervals (Est, 95% CI) of the association between HLA-DQ cell-surface MFI on isolated peripheral blood cells in subjects with the AGG relative to the GCA HLA-DRA1 tri-SNPs.

Linear models of HLA-DQ cell-surface MFI as the outcome and tri-SNP as the predictor were adjusted for age, sex, HLA-DQ2/8 (A), and sAB (B) or cAB (C). Distribution and possible outliers were identified using histograms of HLA-DQ cell-surface MFI. Standard errors were estimated using robust linear methods. See Table IX for detailed results corresponding to these plots.

Close modal
Table IX.

Estimates and 95% confidence intervals of the association between HLA-DQ cell-surface MFI on isolated peripheral blood cells and DRA tri-SNPs AGG and GCA

B cells, Est (95% CI), pNeutrophils, Est (95% CI), pCD16+ cells, Est (95% CI), pClassical monocytes, Est (95% CI), pCD4 T cells, Est (95% CI), pCD8 T cells, Est (95% CI), p
Model 1       
 AGG versus GCA tri-SNP −0.08 (−0.22, 0.05), −0.01 (−0.08, 0.07), −0.01 (−0.03, 0.02), 0.01 (−0.04, 0.07), −0.01 (−0.02, 0.01), −0.01 (−0.03, 0.01), 
 0.227 0.881 0.681 0.710 0.520 0.328 
 Age (y) 0.04 (−0.02, 0.10), 0.01 (−0.03, 0.04), 0.01 (0.00, 0.02), 0.02 (0.00, 0.05), 0.02 (0.01, 0.02), 0.02 (0.01, 0.03), 
 0.162 0.670 0.025 0.067 0.000* 0.000* 
 Female 0.03 (−0.10, 0.17), −0.05 (−0.12, 0.03), 0.00 (−0.02, 0.03), 0.02 (−0.03, 0.07), 0.00 (−0.02, 0.02), 0.01 (−0.01, 0.03), 
 0.624 0.203 0.747 0.452 0.922 0.481 
 HLA-DQ2/8 −0.18 (−0.34, −0.02), −0.06 (−0.15, 0.03), −0.04 (0.06,0.01), −0.10 (−0.17, −0.04), −0.05 (−0.07, −0.03), −0.06 (−0.08, −0.03), 
Model 2 0.029 0.180 0.012 0.001* 0.000* 0.000* 
 AGG versus GCA tri-SNP −0.05 (−0.20, 0.11), −0.01 (−0.07, 0.05), 0.00 (−0.02, 0.02), 0.01 (−0.06, 0.09), 0.00 (−0.02, 0.01), −0.01 (−0.02, 0.01), 
 0.556 0.681 0.840 0.718 0.706 0.484 
 Age (y) 0.04 (−0.02, 0.11), 0.01 (−0.02, 0.03), 0.01 (0.00, 0.02), 0.02 (−0.01, 0.06), 0.02 (0.01, 0.02), 0.02 (0.01, 0.03), 
 0.205 0.613 0.004* 0.169 0.000* 0.000* 
 Female 0.00 (−0.14, 0.15), −0.05 (−0.11, 0.01), 0.00 (−0.02, 0.02), 0.02 (−0.06, 0.09), 0.00 (−0.02, 0.01), 0.00 (−0.02, 0.02), 
 0.985 0.098 0.778 0.633 0.603 0.848 
 HLA-DQ2/8 −0.01 (−0.27, 0.25), −0.06 (−0.15, 0.03), −0.02 (−0.05, 0.01), −0.07 (−0.18, 0.05), 0.03 (0.06,0.01), −0.04 (0.07, −0.01), 
 0.943 0.164 0.122 0.267 0.017 0.005* 
 1 versus 0 AAb −0.01 (−0.18, 0.17), −0.04 (−0.11, 0.03), −0.01 (−0.03, 0.01), −0.06 (−0.14, 0.03), 0.02 (0.04, 0.00), −0.01 (−0.03, 0.01), 
 0.935 0.256 0.299 0.217 0.037 0.292 
 2+ versus 0 AAb −0.30 (−0.62, 0.02), 0.03 (−0.08, 0.14), −0.03 (−0.06, 0.00), −0.06 (−0.20, 0.07), −0.03 (−0.06, 0.00), −0.02 (−0.05, 0.01), 
 0.073 0.654 0.076 0.361 0.096 0.125 
Model 3       
 AGG versus GCA tri-SNP −0.09 (−0.26, 0.08), 0.00 (−0.07, 0.06), 0.00 (−0.02, 0.01), 0.01 (−0.04, 0.06), −0.01 (−0.02, 0.01), −0.01 (−0.04, 0.02), 
 0.299 0.970 0.646 0.671 0.450 0.434 
 Age (y) 0.05 (−0.03, 0.12), 0.01 (−0.02, 0.04), 0.01 (0.01, 0.02), 0.02 (0.00, 0.05), 0.02 (0.01, 0.02), 0.02 (0.01, 0.03), 
 0.247 0.419 0.002* 0.062 0.000* 0.002* 
 Female 0.04 (−0.14, 0.21), −0.04 (−0.10, 0.03), 0.00 (−0.02, 0.01), 0.02 (−0.04, 0.07), −0.01 (−0.02, 0.01), 0.01 (−0.02, 0.04), 
 0.659 0.267 0.738 0.570 0.517 0.410 
 HLA-DQ2/8 −0.21 (−0.45, 0.03), 0.00 (−0.09, 0.09), −0.02 (−0.04, 0.00), −0.08 (−0.15, −0.01), −0.03 (−0.05, −0.01), −0.06 (−0.10, −0.02), 
 0.085 0.983 0.037 0.022 0.003* 0.002* 
 Medium versus low cAB 0.06 (−0.15, 0.27), −0.05 (−0.13, 0.03), −0.01 (−0.03, 0.01), −0.03 (−0.09, 0.04), −0.02 (−0.04, 0.00), −0.01 (−0.04, 0.02), 
 0.578 0.251 0.412 0.412 0.066 0.595 
 High versus low cAB 0.05 (−0.22, 0.33), −0.08 (−0.18, 0.02), 0.03 (0.06,0.01), −0.05 (−0.12, 0.03), −0.02 (−0.04, 0.01), 0.01 (−0.03, 0.05), 
 0.704 0.130 0.007 0.245 0.172 0.699 
B cells, Est (95% CI), pNeutrophils, Est (95% CI), pCD16+ cells, Est (95% CI), pClassical monocytes, Est (95% CI), pCD4 T cells, Est (95% CI), pCD8 T cells, Est (95% CI), p
Model 1       
 AGG versus GCA tri-SNP −0.08 (−0.22, 0.05), −0.01 (−0.08, 0.07), −0.01 (−0.03, 0.02), 0.01 (−0.04, 0.07), −0.01 (−0.02, 0.01), −0.01 (−0.03, 0.01), 
 0.227 0.881 0.681 0.710 0.520 0.328 
 Age (y) 0.04 (−0.02, 0.10), 0.01 (−0.03, 0.04), 0.01 (0.00, 0.02), 0.02 (0.00, 0.05), 0.02 (0.01, 0.02), 0.02 (0.01, 0.03), 
 0.162 0.670 0.025 0.067 0.000* 0.000* 
 Female 0.03 (−0.10, 0.17), −0.05 (−0.12, 0.03), 0.00 (−0.02, 0.03), 0.02 (−0.03, 0.07), 0.00 (−0.02, 0.02), 0.01 (−0.01, 0.03), 
 0.624 0.203 0.747 0.452 0.922 0.481 
 HLA-DQ2/8 −0.18 (−0.34, −0.02), −0.06 (−0.15, 0.03), −0.04 (0.06,0.01), −0.10 (−0.17, −0.04), −0.05 (−0.07, −0.03), −0.06 (−0.08, −0.03), 
Model 2 0.029 0.180 0.012 0.001* 0.000* 0.000* 
 AGG versus GCA tri-SNP −0.05 (−0.20, 0.11), −0.01 (−0.07, 0.05), 0.00 (−0.02, 0.02), 0.01 (−0.06, 0.09), 0.00 (−0.02, 0.01), −0.01 (−0.02, 0.01), 
 0.556 0.681 0.840 0.718 0.706 0.484 
 Age (y) 0.04 (−0.02, 0.11), 0.01 (−0.02, 0.03), 0.01 (0.00, 0.02), 0.02 (−0.01, 0.06), 0.02 (0.01, 0.02), 0.02 (0.01, 0.03), 
 0.205 0.613 0.004* 0.169 0.000* 0.000* 
 Female 0.00 (−0.14, 0.15), −0.05 (−0.11, 0.01), 0.00 (−0.02, 0.02), 0.02 (−0.06, 0.09), 0.00 (−0.02, 0.01), 0.00 (−0.02, 0.02), 
 0.985 0.098 0.778 0.633 0.603 0.848 
 HLA-DQ2/8 −0.01 (−0.27, 0.25), −0.06 (−0.15, 0.03), −0.02 (−0.05, 0.01), −0.07 (−0.18, 0.05), 0.03 (0.06,0.01), −0.04 (0.07, −0.01), 
 0.943 0.164 0.122 0.267 0.017 0.005* 
 1 versus 0 AAb −0.01 (−0.18, 0.17), −0.04 (−0.11, 0.03), −0.01 (−0.03, 0.01), −0.06 (−0.14, 0.03), 0.02 (0.04, 0.00), −0.01 (−0.03, 0.01), 
 0.935 0.256 0.299 0.217 0.037 0.292 
 2+ versus 0 AAb −0.30 (−0.62, 0.02), 0.03 (−0.08, 0.14), −0.03 (−0.06, 0.00), −0.06 (−0.20, 0.07), −0.03 (−0.06, 0.00), −0.02 (−0.05, 0.01), 
 0.073 0.654 0.076 0.361 0.096 0.125 
Model 3       
 AGG versus GCA tri-SNP −0.09 (−0.26, 0.08), 0.00 (−0.07, 0.06), 0.00 (−0.02, 0.01), 0.01 (−0.04, 0.06), −0.01 (−0.02, 0.01), −0.01 (−0.04, 0.02), 
 0.299 0.970 0.646 0.671 0.450 0.434 
 Age (y) 0.05 (−0.03, 0.12), 0.01 (−0.02, 0.04), 0.01 (0.01, 0.02), 0.02 (0.00, 0.05), 0.02 (0.01, 0.02), 0.02 (0.01, 0.03), 
 0.247 0.419 0.002* 0.062 0.000* 0.002* 
 Female 0.04 (−0.14, 0.21), −0.04 (−0.10, 0.03), 0.00 (−0.02, 0.01), 0.02 (−0.04, 0.07), −0.01 (−0.02, 0.01), 0.01 (−0.02, 0.04), 
 0.659 0.267 0.738 0.570 0.517 0.410 
 HLA-DQ2/8 −0.21 (−0.45, 0.03), 0.00 (−0.09, 0.09), −0.02 (−0.04, 0.00), −0.08 (−0.15, −0.01), −0.03 (−0.05, −0.01), −0.06 (−0.10, −0.02), 
 0.085 0.983 0.037 0.022 0.003* 0.002* 
 Medium versus low cAB 0.06 (−0.15, 0.27), −0.05 (−0.13, 0.03), −0.01 (−0.03, 0.01), −0.03 (−0.09, 0.04), −0.02 (−0.04, 0.00), −0.01 (−0.04, 0.02), 
 0.578 0.251 0.412 0.412 0.066 0.595 
 High versus low cAB 0.05 (−0.22, 0.33), −0.08 (−0.18, 0.02), 0.03 (0.06,0.01), −0.05 (−0.12, 0.03), −0.02 (−0.04, 0.01), 0.01 (−0.03, 0.05), 
 0.704 0.130 0.007 0.245 0.172 0.699 

The models were adjusted for age, sex, HLA-DQ2/8 (model 1) and autoimmunity burden measured as the number of AAbs detected at sAB (model 2) or cAB measured as area under the trajectory of AAbs over time (model 3). The models were fit using linear regression with robust SEs. The p values presented are nominal, and those that remain significant after adjusting for multiple comparisons (corrected using the Benjamini–Hochberg procedure assuming 192 comparisons, in robust and nonrobust models, and a 5% false discovery) are indicated by an asterisk.

AAb, autoantibody; CI, confidence interval; Est, estimates.

Adjusting for sAB, HLA-DQ cell-surface MFI was marginally lower on CD4+ T cells for subjects with one autoantibody compared with no autoantibodies (Fig. 6B, Table IX). The association was shifted adjusting for cAB, and HLA-DQ cell-surface MFI was 0.03 (−0.06, −0.01; p = 0.007) lower on CD16+ cells for subjects with high compared with low autoimmunity burden (Fig. 6C, Table IX).

The results based on linear regression with robust SE estimation shown earlier were comparable with a generalized linear model based on the linear regression with model-based SE estimation (data not shown). The p values presented in Table IX are nominal, and those that remain significant after adjusting for multiple comparisons (corrected using the Benjamini–Hochberg procedure assuming 192 comparisons, in a robust and a generalized linear model, and a 5% false discovery) are indicated by an asterisk.

In this study, we investigated the HLA-DRA1 tri-SNPs in relation to HLA-DQ cell-surface expression in the unique cross-sectional cohort of 67 subjects randomly selected from the DiPiS study. We used previously published HLA-DQ flow cytometric analysis to test the hypothesis that the tri-SNPs may be related to HLA-DQ expression. The hypothesis was reasonable because a previous study, in a cohort of DR3 homozygous subjects, identified a protective and a risk element for type 1 diabetes with the tri-SNP (18). We identified four HLA-DRA1 tri-SNPs and showed that they were related to 26 class II extended HLA haplotypes (Fig. 1). The major finding in our analysis of the extended HLA haplotypes was that among the four tri-SNPs, DR3-DQ2 was in LD with both AGG (n = 6) and GCA (n = 47). Indeed, the latter showed a unique extended haplotype with DRB3*01:01:02, while DR3-DQ2 on the AGG-containing haplotype was linked to DRB3*02:02:01 (Fig. 1). Further studies will therefore be needed to explore the role of DRB3 to explain a possible difference of DR3-DQ2 on the AGG- and GCA-containing haplotypes. The second major finding was the relationship between the tri-SNP and the HLA-DQ MFI in the CD4+ T cells indicating that subjects with AGG had a reduced expression compared with subjects with GCA. This is of interest because the AGG-containing haplotype includes HLA-DQ haplotypes that confer risk, such as DRB1*04:01-DQA1*03:01:01-DQB1*03:02:01, but also DRB1*04:03-DQA1*03:01:01-DQB1*03:02:01 and DRB1*07:01:01-DQA1*02:01:01-DB1*02:02:01 that confer protection. These data suggest that the tri-SNP may contribute to HLA-DQ expression on the cell surface of at least some peripheral blood cell subsets.

The haplotypes for the three markers were imputed based on homozygous alleles and the frequency of alleles co-occurring with flanking HLA markers in subjects homozygous for one marker but not the other. With LD in mind, the alleles of the SNP closest to the extended HLA-DRB1-DRB345-DQA1-DQB1 were imputed first, then the second and third SNP alleles. Computational methods and large datasets that might increase confidence in the assignment of these haplotypes were not applicable in this study. In this study, we have obtained HLA-DR-DR haplotypes and genotypes through NGS, which is also the basis for computational methods in large datasets.

Because HLA-DR-DQ is strongly associated with increased risk for type 1 diabetes, it is important to further investigate HLA for different phenotypes that may have protective properties or increase risk for type 1 diabetes. Because only 7% of individuals with the HLA-DR3/4-DQ2/8 genotype, conferring highest risk for type 1 diabetes, develop the disease (15), it is probable that there are other risk elements that distinguish different HLA-DR3/4-DQ2/8. In a previous study, we observed tri-SNPs and genotypes that were protective against (GCA) or increased risk (AGG) for type 1 diabetes (18). The addition of the tri-SNP and genotype to standard HLA typing and NGS could be useful in enrolling subjects for longitudinal studies and stratifying subjects being studied for risk for type 1 diabetes relative to other risk factors such as autoantibodies.

The GCA tri-SNP was identified with the HLA-DR3-DQ2 haplotype only. This could be explained by a small sample set or coincidence. AGG, predominantly found in subjects with the HLA-DR4-DQ8 haplotype, was the most common haplotype. Due to small sample sizes, ACA and ACG tri-SNPs were included in only a few analyses.

The main finding of this study is the pattern of decreased HLA-DQ cell-surface MFI, on all cell subsets, in subjects with the AGG tri-SNP compared with subjects with the GCA tri-SNP (Fig. 2) and with increasing autoimmunity burden (Fig. 3). This is consistent with previous findings of lower HLA-DQ cell-surface MFI with increasing autoimmunity burden (23). The pattern also confirms previous observations of lower intensity of HLA class II expression in monocytes for type 1 diabetes patients compared with siblings and controls (33). Our observation that subjects with certain HLA-DRA1 tri-SNPs have decreased HLA-DQ cell-surface expression with increasing autoimmunity burden suggest that HLA-DRA1 tri-SNPs could be considered an additional risk element to the more conventional HLA-DR-DQ haplotypes considered to date. Previously, the AGG tri-SNP was found to increase risk for type 1 diabetes in DR3 homozygous subjects, while the GCA haplotype was found to be protective (18). The inability to maintain peripheral tolerance has previously been speculated to be affected by a reduction in HLA-DQ class II Ag expression (34, 35). Abnormalities in the regulation of HLA class II expression in type 1 diabetes subjects could reflect the ongoing autoimmune process. If HLA-DRA1 tri-SNPs are associated with extended HLA-DR-DQ haplotypes, additional risk may present as a variation in the possibility for heterodimers to bind Ag on the cell surface, thus affecting the ability of peripheral blood cells to present Ag.

Due to limited data, only a limited investigation of tri-SNP genotypes (heterozygous GCA/AGG and homozygous AGG/AGG) relative to HLA-DQ MFI was possible. Future studies must include homozygous AGG to determine which of the tri-SNP or HLA-DRB3 is more important for HLA-DQ to be decreased.

In a previous study, our in silico HLA class II gene RNA expression in EBV-transformed B lymphocytes stratified by tri-SNP was analyzed using both sequence and expression data available through the 1000 Genomes Project. We found that for subjects carrying the putative protective homozygous haplotype (GCA/GCA), class II gene expression in general was lower, and significantly so for HLA-DQB1, compared with that observed in individuals homozygous for the risk haplotype (AGG/AGG). In the GCA/AGG heterozygotes, the expression of class II genes was intermediate between those values seen in either homozygote. In this study, we observed the opposite pattern from the DQB1 RNA levels reported in the previous paper (AGG = higher DQB1 RNA levels versus GCA). We observed a trend toward decreased HLA-DQB1 MFI in subjects carrying the AGG haplotype compared with the GCA haplotype primarily in T cells.

The order of autoantibody appearance has been found to be related to HLA-DR-DQ and differential risk for type 1 diabetes (3, 27). Two endotypes have previously been identified in regard to the development of islet autoimmunity by the first-appearing autoantibody, whether it be IAA, usually associated with HLA-DR4-DQ8, or GADA, usually associated with HLA-DR3-DQ2 (3, 14, 36). Another intermediate phenotype has been suggested to be the simultaneous appearance of IAA and GADA. The endotypes might have different environmental determinants and risk for type 1 diabetes (14, 28). IAA as the first-appearing autoantibody was found earlier in life than GADA (14, 27). There is a difference in incidence relative to age where IAA has been observed to peak at around 1 y of age, while GADA reaches a plateau around 2 y of age (3).

In this study, we investigated whether there was an association between the tri-SNPs, autoimmunity burden, and the first-appearing autoantibody. This was possible because the subjects had been followed annually from birth, with a first β cell autoantibody test performed at 2 y of age, until 10–15 y of age (Supplemental Fig. 1). The concept of autoimmunity burden was previously proposed to investigate the presence of autoantibodies at cross-sectional sampling (sAB) and during follow-up in DiPiS (cAB) (23). A standard approach to measure cAB remains to be fully established. The burden of autoantibodies over time may put a strain on the child and could possibly contribute to T cell exhaustion just as HLA-DQ2/8 contributes to the highest risk for type 1 diabetes, and the risk for progression to clinical onset increases with increasing number of autoantibodies (6, 13). T cell exhaustion is thought to allow partial containment of chronic infections by persistence of T cells, without causing immunopathy (37, 38). It has been suggested that T cell exhaustion may be of importance to limit immunopathology or autoreactivity (38, 39), and it cannot be excluded that our observation of decreasing HLA-DQ cell-surface MFI by HLA-DRA1 tri-SNP and autoimmunity burden is related to T cell exhaustion.

In the multivariate analysis, the impact of the tri-SNP on HLA-DQ expression is nonsignificant, and it appears that the presence or absence of the type 1 diabetes high-risk HLA haplotype DQ2/8 is the major influence on cell-surface HLA-DQ MFI in primarily T cells. This is not completely surprising given the high LD in the HLA region. The novelty of the results presented in this study is that no one has previously reported on the tri-SNPs in a cohort of healthy subjects (followed from 2 y until 10–15 y of age) relative to cell-surface expression of HLA-DQ MFI on isolated peripheral blood cells, varying number of autoantibodies, and HLA. It is known that HLA-DQ2/8 increase risk for type 1 diabetes, but it is not known whether a tri-SNP could be considered an additional risk element.

Numerous studies have previously indicated “expression quantitative trait loci” (eQTLs) within the HLA region that modulate HLA gene expression. For example, eQTLs analysis of type 1 diabetes subjects suggested that five eQTL genes (TAP2, HLA-DOB, HLA-DQB1, HLA-DQA1, and HLA-DRB5) were differentially expressed in type 1 diabetes–related cells (40). It will be of interest to further analyze the blood cell subtypes reported in this study for combined HLA class II gene expression using RNA sequencing and cell-surface MFI in individuals with the AGG and CGA tri-SNPs. It remains to be clarified to what extent gene expression (mRNA) of HLA class II a and b chains is always coordinated with the cell-surface expression of the class II heterodimer.

A weakness of the study is the relatively small sample size and that despite random selection of the research subjects from the DiPiS cohort, autoantibody-negative subjects with high-risk HLA-DQ2/8 are overrepresented in the cohort. A possible explanation is that autoantibody-positive subjects with high-risk HLA already had been diagnosed with type 1 diabetes and therefore dropped out of DiPiS. However, this is a cohort of subjects with an incredible annual to quarterly follow-up and monitoring of autoantibodies and well-being from 2 y until 10–15 y of age. We believe that the investigation and comparisons are important for the research community and to shed light on the combination of HLA cell-surface expression and HLA haplotypes and genotypes combined with autoantibodies.

The strengths of the study are the well-characterized prospective cohort of subjects identified by early screening as being of high risk for developing type 1 diabetes. It is of interest to note that as of June 2022, there are seven subjects who have subsequently developed type 1 diabetes. Again, using their 14 tri-SNPs, in AGG-positive subjects, there were n = 3 with medium and n = 5 with high cAB compared with AGG-negative subjects with n = 1 with medium and n = 5 with high. The development of the tri-SNP as a possible additional genetic risk marker for type 1 diabetes may improve disease prediction and trial design for preventive therapies.

Variation in HLA-DQ cell-surface MFI on six types of peripheral blood cells with different HLA-DRA1 tri-SNPs in subjects with an increased genetic risk for type 1 diabetes at different stages of autoimmunity were investigated. Aggregating the HLA-DRA1 tri-SNPs, autoimmunity burden, and HLA-DQ cell-surface MFI provides an indication of a trend of lower HLA-DQ cell-surface MFI in subjects with the AGG haplotype. Future studies will be necessary to clarify if and when HLA-DRA1 tri-SNPs are a risk element for type 1 diabetes independent of either HLA DR3-DQ2 or DR4-DQ8.

Findings from this study could contribute to identifying a novel risk element in the HLA-DRA1 gene for autoantibodies and enhance the possibility to prevent and treat type 1 diabetes. We hypothesized that the tri-SNP, by being associated with HLA, contributes to the risk for autoantibodies and subsequently type 1 diabetes. This tri-SNP in the intron of HLA-DRA1 gene may regulate the expression of HLA class II molecules and thereby be associated with the first-appearing autoantibody, increasing the risk of developing additional autoantibodies and thereby increasing the risk for clinical onset of type 1 diabetes. The significance of identifying genetic high-risk factors for type 1 diabetes will improve diabetes risk assessment and narrow clinical trials within high-risk groups. Further investigation of the DRA1 tri-SNP should help to clarify the role of HLA in type 1 diabetes susceptibility and improve diabetes risk assessment.

We are indebted to Alexander Lind, Anita Ramelius, Gertie Hansson, Ida Jönsson, Maria Ask, and Rasmus Bennet for technical support and advice.

This work was supported by grants from Anna och Edwin Bergers Stiftelse, Filip Lundbergs Stiftelse, Blekinge Diabetesförening: Fogelstöms Fond, Fredrik och Ingrid Thurings Stiftelse, H.K.H. Kronprinsessan Lovisas Förening för Barnasjukvård, Kungliga Fysiografiska Sällskapet i Lund: Edla och Eric Smedbergs Forskningsdonation, LIONS Research Fund Skåne, Maggie Stephens Stiftelse, Stiftelsen Samariten, Sydvästra Skånes Diabetesförening, Stiftelsen Till Minne av Personalföreningarna i Holmia Försäkring AB, Sven Mattssons Stiftelse, Svenska Diabetesstiftelsen, the Swedish Research Council (2016–01792), the Swedish Child Diabetes Foundation, Tage Blüchers Stiftelse för Medicinsk Forskning, Gyllenstiernska Krapperupsstiftelsen, Wera Ekströms Stiftelse för Pediatrisk Forskning, the Strategic Research Area Exodiab (2009–1039), and the Swedish Foundation for Strategic Research (IRC15-0067).

The funding sources had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.

The online version of this article contains supplemental material.

A.A.S.: conceptualization, methodology, validation, formal analysis, investigation, data curation, writing – original draft, writing – review and editing, visualization, project administration, and funding acquisition. E.B.: methodology, validation, and writing – review and editing. M.L.: resources and writing – review and editing. Å.L.: conceptualization, investigation, resources, writing – original draft, writing – review and editing, supervision, and funding acquisition. M.M.: formal analysis, data curation, and writing – review and editing. H.E.L.: conceptualization, investigation, resources, writing – original draft, writing – review and editing, supervision, and funding acquisition.

Abbreviations used in this article

     
  • BB515

    Brilliant Blue 515

  •  
  • cAB

    cumulative autoimmunity burden

  •  
  • DiPiS

    Diabetes Prediction in Skåne

  •  
  • eQTL

    expression quantitative trait locus

  •  
  • FMO

    fluorescence minus one

  •  
  • GADA

    glutamate decarboxylase autoantibody

  •  
  • IAA

    autoantibody against insulin

  •  
  • LD

    linkage disequilibrium

  •  
  • MFI

    median fluorescence intensity

  •  
  • NGS

    next generation sequencing

  •  
  • sAB

    sampling autoimmunity burden

  •  
  • SNP

    single-nucleotide polymorphism

  •  
  • tri-SNP

    haplotype of three single-nucleotide polymorphisms

  •  
  • ZnT8A

    zinc transporter 8 autoantibodies to any of the amino acid variants at position 325

1.
Lindbladh
I.
,
A.
Andersson Svärd
,
Å.
Lernmark
.
2020
.
Autoimmune (type 1) diabetes.
In
The Autoimmune Diseases
, 6th Ed.
N. R.
Rose
,
I. R.
Mackay
.
Academic Press
,
Amsterdam
, p.
769
787
.
2.
Atkinson
M. A.
,
G. S.
Eisenbarth
,
A. W.
Michels
.
2014
.
Type 1 diabetes.
Lancet
383
:
69
82
.
3.
Krischer
J. P.
,
K. F.
Lynch
,
D. A.
Schatz
,
J.
Ilonen
,
Å.
Lernmark
,
W. A.
Hagopian
,
M. J.
Rewers
,
J. X.
She
,
O. G.
Simell
,
J.
Toppari
, et al
TEDDY Study Group
.
2015
.
The 6 year incidence of diabetes-associated autoantibodies in genetically at-risk children: the TEDDY study.
Diabetologia
58
:
980
987
.
4.
Krischer
J. P.
,
X.
Liu
,
Å.
Lernmark
,
W. A.
Hagopian
,
M. J.
Rewers
,
J. X.
She
,
J.
Toppari
,
A. G.
Ziegler
,
B.
Akolkar
;
TEDDY Study Group
.
2017
.
The influence of type 1 diabetes genetic susceptibility regions, age, sex, and family history on the progression from multiple autoantibodies to type 1 diabetes: a TEDDY Study Report.
Diabetes
66
:
3122
3129
.
5.
Ziegler
A. G.
,
G. T.
Nepom
.
2010
.
Prediction and pathogenesis in type 1 diabetes.
Immunity
32
:
468
478
.
6.
Ziegler
A. G.
,
M.
Rewers
,
O.
Simell
,
T.
Simell
,
J.
Lempainen
,
A.
Steck
,
C.
Winkler
,
J.
Ilonen
,
R.
Veijola
,
M.
Knip
, et al
2013
.
Seroconversion to multiple islet autoantibodies and risk of progression to diabetes in children.
JAMA
309
:
2473
2479
.
7.
Noble
J. A.
,
A. M.
Valdes
,
M. D.
Varney
,
J. A.
Carlson
,
P.
Moonsamy
,
A. L.
Fear
,
J. A.
Lane
,
E.
Lavant
,
R.
Rappner
,
A.
Louey
, et al
Type 1 Diabetes Genetics Consortium
.
2010
.
HLA class I and genetic susceptibility to type 1 diabetes: results from the Type 1 Diabetes Genetics Consortium.
Diabetes
59
:
2972
2979
.
8.
Farina
F.
,
S.
Picascia
,
L.
Pisapia
,
P.
Barba
,
S.
Vitale
,
A.
Franzese
,
E.
Mozzillo
,
C.
Gianfrani
,
G.
Del Pozzo G
.
2019
.
HLA-DQA1 and HLA-DQB1 alleles, conferring susceptibility to celiac disease and type 1 diabetes, are more expressed than non-predisposing alleles and are coordinately regulated.
Cells
8
:
751
.
9.
Pugliese
A.
2017
.
Autoreactive T cells in type 1 diabetes.
J. Clin. Invest.
127
:
2881
2891
.
10.
Shiina
T.
,
K.
Hosomichi
,
H.
Inoko
,
J. K.
Kulski
.
2009
.
The HLA genomic loci map: expression, interaction, diversity and disease.
J. Hum. Genet.
54
:
15
39
.
11.
Cruz-Tapias
P.
,
J.
Castiblanco
,
J.-M.
Anaya
.
2013
.
Major histocompatibility complex: Antigen processing and presentation.
In
Autoimmunity. From Bench to Bedside.
J.-M.
Anaya
,
Y.
Shoenfeld
,
A.
Rojas-Villarraga
,
R.
Levy
,
R.
Cervera
.
El Rosario University Press
,
Bogota, Columbia
, p.
169
183
.
12.
Nejentsev
S.
,
J. M. M.
Howson
,
N. M.
Walker
,
J.
Szeszko
,
S. F.
Field
,
H. E.
Stevens
,
P.
Reynolds
,
M.
Hardy
,
E.
King
,
J.
Masters
, et al
Wellcome Trust Case Control Consortium
.
2007
.
Localization of type 1 diabetes susceptibility to the MHC class I genes HLA-B and HLA-A.
Nature
450
:
887
892
.
13.
Krischer
J. P.
,
X.
Liu
,
K.
Vehik
,
B.
Akolkar
,
W. A.
Hagopian
,
M. J.
Rewers
,
J. X.
She
,
J.
Toppari
,
A. G.
Ziegler
,
Å.
Lernmark
;
TEDDY Study Group
.
2019
.
Predicting islet cell autoimmunity and type 1 diabetes: an 8-year TEDDY Study Progress Report.
Diabetes Care
42
:
1051
1060
.
14.
Rewers
M.
,
H.
Hyöty
,
Å.
Lernmark
,
W.
Hagopian
,
J.-X.
She
,
D.
Schatz
,
A.-G.
Ziegler
,
J.
Toppari
,
B.
Akolkar
,
J.
Krischer
;
TEDDY Study Group
.
2018
.
The Environmental Determinants of Diabetes in the Young (TEDDY) Study: 2018 update.
Curr. Diab. Rep.
18
:
136
.
15.
Rewers
M.
,
T. L.
Bugawan
,
J. M.
Norris
,
A.
Blair
,
B.
Beaty
,
M.
Hoffman
,
R. S.
McDuffie
Jr.
,
R. F.
Hamman
,
G.
Klingensmith
,
G. S.
Eisenbarth
,
H. A.
Erlich
.
1996
.
Newborn screening for HLA markers associated with IDDM: diabetes autoimmunity study in the young (DAISY).
Diabetologia
39
:
807
812
.
16.
van Lummel
M.
,
P. A.
van Veelen
,
A.
Zaldumbide
,
A.
de Ru
,
G. M. C.
Janssen
,
A. K.
Moustakas
,
G. K.
Papadopoulos
,
J. W.
Drijfhout
,
B. O.
Roep
,
F.
Koning
.
2012
.
Type 1 diabetes-associated HLA-DQ8 transdimer accommodates a unique peptide repertoire.
J. Biol. Chem.
287
:
9514
9524
.
17.
Zhao
L. P.
,
G. K.
Papadopoulos
,
A. K.
Moustakas
,
G. P.
Bondinas
,
A.
Carlsson
,
H. E.
Larsson
,
J.
Ludvigsson
,
C.
Marcus
,
M.
Persson
,
U.
Samuelsson
, et al
2021
.
Nine residues in HLA-DQ molecules determine with susceptibility and resistance to type 1 diabetes among young children in Sweden.
Sci. Rep.
11
:
8821
.
18.
Aydemir
Ö.
,
J. A.
Noble
,
J. A.
Bailey
,
Å.
Lernmark
,
P.
Marsh
,
A.
Andersson Svärd
,
F.
Bearoff
,
E. P.
Blankenhorn
,
J. P.
Mordes
;
Better Diabetes Diagnosis (BDD) Study Group
.
2019
.
Genetic variation within the HLA-DRA1 gene modulates susceptibility to type 1 diabetes in HLA-DR3 homozygotes.
Diabetes
68
:
1523
1527
.
19.
Nygård
L.
,
A. P.
Laine
,
M.
Kiviniemi
,
J.
Toppari
,
T.
Härkönen
,
M.
Knip
,
R.
Veijola
,
J.
Lempainen
,
J.
Ilonen
.
2021
.
Tri-SNP polymorphism in the intron of HLA-DRA1 affects type 1 diabetes susceptibility in the Finnish population.
Hum. Immunol.
82
:
912
916
.
20.
Larsson
H. E.
,
K.
Lynch
,
B.
Lernmark
,
A.
Nilsson
,
G.
Hansson
,
P.
Almgren
,
A.
Lernmark
,
S. A.
Ivarsson
;
DiPiS Study Group
.
2005
.
Diabetes-associated HLA genotypes affect birthweight in the general population.
Diabetologia
48
:
1484
1491
.
21.
Larsson
K.
,
H.
Elding-Larsson
,
E.
Cederwall
,
K.
Kockum
,
J.
Neiderud
,
S.
Sjöblad
,
B.
Lindberg
,
B.
Lernmark
,
C.
Cilio
,
S.-A.
Ivarsson
,
A.
Lernmark
.
2004
.
Genetic and perinatal factors as risk for childhood type 1 diabetes.
Diabetes Metab. Res. Rev.
20
:
429
437
.
22.
Elding Larsson
H.
2016
.
A Swedish approach to the prevention of type 1 diabetes.
Pediatr. Diabetes
17
(
Suppl. 22
):
73
77
.
23.
Andersson Svärd
A.
,
M.
Maziarz
,
A.
Ramelius
,
M.
Lundgren
,
Å.
Lernmark
,
H.
Elding Larsson
;
DiPiS study group
.
2020
.
Decreased HLA-DQ expression on peripheral blood cells in children with varying number of beta cell autoantibodies.
J. Transl. Autoimmun.
3
:
100052
.
24.
Insel
R. A.
,
J. L.
Dunne
,
M. A.
Atkinson
,
J. L.
Chiang
,
D.
Dabelea
,
P. A.
Gottlieb
,
C. J.
Greenbaum
,
K. C.
Herold
,
J. P.
Krischer
,
Å.
Lernmark
, et al
2015
.
Staging presymptomatic type 1 diabetes: a scientific statement of JDRF, the Endocrine Society, and the American Diabetes Association.
Diabetes Care
38
:
1964
1974
.
25.
Williams
A. J.
,
P. J.
Bingley
,
E.
Bonifacio
,
J. P.
Palmer
,
E. A.
Gale
.
1997
.
A novel micro-assay for insulin autoantibodies.
J. Autoimmun.
10
:
473
478
.
26.
Vaziri-Sani
F.
,
A. J.
Delli
,
H.
Elding-Larsson
,
B.
Lindblad
,
A.
Carlsson
,
G.
Forsander
,
S. A.
Ivarsson
,
J.
Ludvigsson
,
C.
Marcus
,
Å.
Lernmark
.
2011
.
A novel triple mix radiobinding assay for the three ZnT8 (ZnT8-RWQ) autoantibody variants in children with newly diagnosed diabetes.
J. Immunol. Methods
371
:
25
37
.
27.
Krischer
J. P.
,
X.
Liu
,
Å.
Lernmark
,
W. A.
Hagopian
,
M. J.
Rewers
,
J.-X.
She
,
J.
Toppari
,
A.-G.
Ziegler
,
B.
Akolkar
;
TEDDY Study Group
.
2021
.
Characteristics of children diagnosed with type 1 diabetes before vs after 6 years of age in the TEDDY cohort study.
Diabetologia
64
:
2247
2257
.
28.
Bauer
W.
,
R.
Veijola
,
J.
Lempainen
,
M.
Kiviniemi
,
T.
Härkönen
,
J.
Toppari
,
M.
Knip
,
A.
Gyenesei
,
J.
Ilonen
.
2019
.
Age at seroconversion, HLA genotype, and specificity of autoantibodies in progression of islet autoimmunity in childhood.
J. Clin. Endocrinol. Metab.
104
:
4521
4530
.
29.
Dang
M. N.
,
C. M.
Bradford
,
P.
Pozzilli
,
R. D.
Leslie
.
2016
.
Methylation analysis in distinct immune cell subsets in type 1 diabetes.
In
Type-1 Diabetes: Methods and Protocols.
K. M.
Gillespie
.
Springer New York
,
New York, NY
, p.
143
151
.
30.
Bergseng
E.
,
S.
Dørum
,
M. O.
Arntzen
,
M.
Nielsen
,
S.
Nygård
,
S.
Buus
,
G. A.
de Souza
,
L. M.
Sollid
.
2015
.
Different binding motifs of the celiac disease-associated HLA molecules DQ2.5, DQ2.2, and DQ7.5 revealed by relative quantitative proteomics of endogenous peptide repertoires.
Immunogenetics
67
:
73
84
.
31.
González-Galarza
F. F.
,
L. Y.
Takeshita
,
E. J.
Santos
,
F.
Kempson
,
M. H.
Maia
,
A. L.
da Silva
,
A. L.
Teles e Silva
,
G. S.
Ghattaoraya
,
A.
Alfirevic
,
A. R.
Jones
,
D.
Middleton
.
2015
.
Allele frequency net 2015 update: new features for HLA epitopes, KIR and disease and HLA adverse drug reaction associations.
Nucleic Acids Res.
43
(
D1
):
D784
D788
.
32.
Andersson
C.
,
F.
Vaziri-Sani
,
A.
Delli
,
B.
Lindblad
,
A.
Carlsson
,
G.
Forsander
,
J.
Ludvigsson
,
C.
Marcus
,
U.
Samuelsson
,
S.
Ivarsson
, et al
BDD Study Group
.
2013
.
Triple specificity of ZnT8 autoantibodies in relation to HLA and other islet autoantibodies in childhood and adolescent type 1 diabetes.
Pediatr. Diabetes
14
:
97
105
.
33.
Parkkonen
P.
,
H.
Hyöty
,
T.
Huupponen
,
P.
Leinikki
,
E.
Tuomilehto-Wolf
,
M.
Knip
;
The Childhood Diabetes in Finland Study Group
.
1993
.
Defective HLA class II expression in monocytes of type 1 diabetic patients.
APMIS
101
:
395
402
.
34.
Manczinger
M.
,
L.
Kemény
.
2018
.
Peptide presentation by HLA-DQ molecules is associated with the development of immune tolerance.
PeerJ
6
:
e5118
.
35.
Jurewicz
M. M.
,
L. J.
Stern
.
2019
.
Class II MHC antigen processing in immune tolerance and inflammation.
Immunogenetics
71
:
171
187
.
36.
Krischer
J. P.
,
K. F.
Lynch
,
Å.
Lernmark
,
W. A.
Hagopian
,
M. J.
Rewers
,
J. X.
She
,
J.
Toppari
,
A. G.
Ziegler
,
ad B.
Akolkar
;
TEDDY Study Group
.
2017
.
Genetic and environmental interactions modify the risk of diabetes-related autoimmunity by 6 years of age: The TEDDY Study.
Diabetes Care
40
:
1194
1202
.
37.
Blank
C. U.
,
W. N.
Haining
,
W.
Held
,
P. G.
Hogan
,
A.
Kallies
,
E.
Lugli
,
R. C.
Lynn
,
M.
Philip
,
A.
Rao
,
N. P.
Restifo
, et al
2019
.
Defining ‘T cell exhaustion’.
Nat. Rev. Immunol.
19
:
665
674
.
38.
McLane
L. M.
,
M. S.
Abdel-Hakeem
,
E. J.
Wherry
.
2019
.
CD8 T cell exhaustion during chronic viral infection and cancer.
Annu. Rev. Immunol.
37
:
457
495
.
39.
McKinney
E. F.
,
K. G.
Smith
.
2016
.
T-cell exhaustion: understanding the interface of chronic viral and autoinflammatory diseases.
Immunol. Cell Biol.
94
:
935
942
.
40.
Qiu
Y. H.
,
F. Y.
Deng
,
Z. X.
Tang
,
Z. H.
Jiang
,
S. F.
Lei
.
2015
.
Functional relevance for type 1 diabetes mellitus-associated genetic variants by using integrative analyses.
Hum. Immunol.
76
:
753
758
.

Anita Ramelius, Cecilia Andersson, Rasmus Bennet, Ida Jönsson, Maria Ask, Jenny Bremer, Charlotte Brundin, Corrado Cilio, Carina Hansson, Gertie Hansson, Sten Ivarsson, Berglind Jonsdottir, Bengt Lindberg, Barbro Lernmark, and Jessica Melin (Department of Clinical Sciences, Lund University, Malmö, Sweden)

Annelie Carlsson (Department of Clinical Sciences, Lund University, Lund, Sweden)

Elisabeth Cedervall (Department of Paediatrics, Ängelholm Hospital, Ängelholm, Sweden)

Björn Jönsson (Department of Paediatrics, Ystad Hospital, Ystad, Sweden)

Karin Larsson (Department of Paediatrics, Kristianstad Hospital, Kristianstad, Sweden)

Jan Neiderud (Department of Paediatrics, Helsingborg Hospital, Helsingborg, Sweden)

The authors have no financial conflicts of interest.

This article is distributed under the terms of the CC BY-NC 4.0 Unported license.

Supplementary data