Association of High-Affinity Autoantibodies With Type 1 Diabetes High-Risk HLA Haplotypes

Abstract

Objective

Electrochemiluminescence (ECL) assays are high-affinity autoantibody (Ab) tests that are more specific than Abs detected by traditional radiobinding assays (RBA) for risk screening and prediction of progression to type 1 diabetes. We sought to characterize the association of high-risk human leukocyte antigen (HLA) haplotypes and genotypes with ECL positivity and levels in relatives of individuals with type 1 diabetes.

Methods

We analyzed 602 participants from the TrialNet Pathway to Prevention Study who were positive for at least 1 RBA diabetes-related Ab [glutamic acid decarboxylase autoantibodies (GADA) or insulin autoantibodies (IAA)] and for whom ECL and HLA data were available. ECL and RBA Ab levels were converted to SD units away from mean (z-scores) for analyses.

Results

Mean age at initial visit was 19.4 ± 13.7 years; 344 (57.1%) were female and 104 (17.3%) carried the high-risk HLA-DR3/4*0302 genotype. At initial visit 424/602 (70.4%) participants were positive for either ECL-GADA or ECL-IAA, and 178/602 (29.6%) were ECL negative. ECL and RBA-GADA positivity were associated with both HLA-DR3 and DR4 haplotypes (all Ps < 0.05), while ECL and RBA-GADA z-score titers were higher in participants with HLA-DR3 haplotypes only (both Ps < 0.001). ECL-IAA (but not RBA-IAA) positivity was associated with the HLA-DR4 haplotype (P < 0.05).

Conclusions

ECL-GADA positivity is associated with the HLA-DR3 and HLA-DR4 haplotypes and levels are associated with the HLA-DR3 haplotype. ECL-IAA positivity is associated with HLA-DR4 haplotype. These studies further contribute to the understanding of genetic risk and islet autoimmunity endotypes in type 1 diabetes.

Type 1 diabetes is characterized by the development of islet autoantibodies that can be present for years before onset of clinical type 1 diabetes (1). Development of autoantibodies can vary among individuals and progression to clinical diabetes differs markedly (2). Relatives of individuals with type 1 diabetes are at an increased risk for developing type 1 diabetes, particularly after the development of islet autoantibodies (3). The high-risk human leukocyte antigen (HLA) genotype DR3 DQA1*05:01 DQB1*02:01 DR4 DQA1*03:01 DQB1*03:02 (abbreviated here DR3/4*0302) is associated with progression to type 1 diabetes (4), but not all individuals with this genotype progress to diabetes. The presence of additional known non-HLA high-risk genes can influence the development of initiation of autoimmunity in those with high-risk HLA genotypes (5). The development of autoantibodies and their persistence or progression to multiple autoantibodies can be heterogenous (6). Given the degree of variability in genetic susceptibility and variable rates of progression to type 1 diabetes, endotypes have been proposed as a way to better categorize individuals at risk for developing type 1 diabetes(7). Given this heterogeneity within disease, additional criteria are needed to further characterize potential type 1 diabetes distinct endotypes and help develop personalized treatments and prevention trials.

The TrialNet Pathway to Prevention study (previously the TrialNet Natural History study) (8) has been screening relatives of patients with type 1 diabetes since 2004 and follows these participants with serial autoantibody testing for the development of islet autoantibodies and type 1 diabetes. The study offers longitudinal monitoring for autoantibody-positive participants through hemoglobin A1c testing and oral glucose tolerance tests as well as prevention trials. Participants are screened by radiobinding assays (RBAs) for type 1 diabetes–related autoantibodies. Recently, there has been emerging work in further refining these autoantibody assays to improve prediction for progression to type 1 diabetes. High-affinity electrochemiluminescence (ECL) assays have been shown to be more disease specific and sensitive compared to RBA (9-11). Further evaluation and validation are needed to characterize immunologic and HLA associations and determine potential type 1 diabetes endotypes. We sought to characterize the association of high-risk HLA haplotypes and genotypes with ECL and RBA positivity and levels in relatives of individuals with type 1 diabetes at baseline screening. We hypothesized that ECL and high-risk HLA haplotypes and genotypes could be used to further stratify type 1 diabetes endotypes in relatives at risk for the development of type 1 diabetes. This study adds to the current literature to further stratify those who may be at risk of progression to type 1 diabetes and potentially benefit from intervention or prevention.

Materials and Methods

Study Participants

Nondiabetic relatives of patients with type 1 diabetes were recruited to the TrialNet Pathway to Prevention Study (ClinicalTrials.gov identifier: NCT00097292), as previously described (8). Race/ethnicity of participants was self-reported. Participants who were single autoantibody positive at the screening visit were required to have a confirmatory positive result for the same autoantibody. In the TrialNet protocol, individuals with single confirmed autoantibody or multiple islet autoantibodies were offered longitudinal follow-up (8). We analyzed data from 602 participants who were positive for at least 1 RBA diabetes-related autoantibody and for whom ECL and HLA data were available at baseline screening. Only participants who were RBA positive and had ECL and HLA data were included in this analysis. ECL autoantibody testing for insulin autoantibodies (IAA) and GAD65 autoantibodies (GADA) were available for this cohort, as these are the only ECL assays tested in the TrialNet Pathway to Prevention Study. All study participants gave informed consent, and the ethics committee responsible for each clinical site approved the study.

Autoantibody and HLA Testing

All participants in the TrialNet Pathway to Prevention study are screened for RBA GADA, insulinoma-associated protein 2 autoantibody (IA2A), and IAA measured by RBA in the TrialNet Core Laboratory at the Barbara Davis Center for Childhood Diabetes, Aurora, Colorado, as previously described (12,13). In the 2020 Islet Autoantibody Standardization Program Workshop, sensitivities and specificities for the RBA were 62% and 99%, respectively, for micro-IAA, 78% and 99%, respectively for GADA (14). ECL assays were performed at baseline for ECL-IAA and ECL-GADA (10,15). In the 2020 Islet Autoantibody Standardization Program Workshop, sensitivities and specificities for ECL were 66% and 99% respectively, for IAA and 78% and 100%, respectively, for GADA. Participants were typed for HLA class II DRB1, DQA1, and DQB1 alleles using DNA-based typing with oligonucleotide probes, as previously described (16,17). Participants were categorized by the presence of the high-risk haplotypes HLA DR3 or HLA DR4*0302 (DR4) or by high-risk genotypes. For genotype analyses we divided the cohorts by the presence of the highest-risk HLA genotype DR 3/4*0302, or the high-risk genotypes HLA DR 3/3, or HLA DR 3/x (where x is neither DR3 nor DR4*0302) and HLA DR4/4 or HLA DR4/x. Participants without either of these high-risk alleles were categorized as DR x/x.

Statistical analysis

Statistical analyses were performed using PRISM (GraphPad Software, Inc., La Jolla, CA) and R version 4.0 (R Core Team, Vienna). For all samples ECL and RBA autoantibody levels were converted to standard deviation (SD) units away from mean participant (z-scores) for analyses. We performed cross sectional analyses of ECL and RBA autoantibodies and HLA genotype/haplotypes. Autoantibody z-scores were not normally distributed and were compared using the Kruskal-Wallis test. Autoantibody positivity was compared using the chi-square test. Results were considered statistically significant with P-value < 0.05.

Results

Of the 602 participants who were positive for at least 1 RBA autoantibody, 424 (70%) were positive for at least 1 ECL autoantibody, and of these individuals, 34% (206/602) were ECL-IAA positive and 61% (366/602) were ECL-GADA positive. One hundred forty-eight participants (25%) were positive for both ECL-IAA and ECL-GADA. Of those who were RBA-IAA positive (n = 206), 160 (78%) were also ECL-IAA positive. Of the participants who were RBA-GADA positive (n = 433), 364 (80%) were also ECL-GADA positive. Baseline characteristics for all participants are shown in Table 1. There was no significant difference between ECL-positive and ECL-negative participants regarding the distribution of sex or race and ethnicity. ECL-positive individuals were younger at initial visit (17.8 ± 13.2 years) compared to ECL-negative participants (23.1 ± 14.4 years; P < 0.001). Mean age of initial visit for these participants (19.4 years) is similar to the mean age of participants screened in TrialNet overall (20.3 years). ECL-positive participants were more likely to have the high-risk HLA haplotype DR3 (195/424, 46%) compared to the ECL-negative participants (59/178, 33%; P = 0.005). Similarly, ECL-positive participants were more likely to have the HLA haplotype DR4 (257/424, 61%) compared to the ECL-negative participants (72/178, 40%; P < 0.001). ECL-positive participants also had lower frequency (14%) of the low-risk HLA-genotype DR x/x compared to the ECL-negative participants (35%) (Table 1).

Table 1.

Baseline characteristics of participants by ECL autoantibody status

	All (n = 602)	ECL Ab negative (n = 178)	ECL Ab positive (n = 424)	P-value
Age at visit	19.4 ± 13.7	23.1 ± 14.4	17.8 ± 13.2	<0.001
Sex, female)	344 (57)	111 (62)	233 (55)	0.24
Race/ethnicity				0.18
Non-Hispanic white	468 (82)	131 (78)	337 (84)
Hispanic	61 (11)	23 (14)	38 (10)
Other	41 (7)	15 (9)	26 (7)
DR3 Present	254 (42)	59 (33)	195 (46)	0.005
DR4 Present	329 (55)	72 (40)	257 (61)	<0.001
Genotype
DR 3/4	104 (17)	16 (9)	88 (21)	<0.001
DR 3/3 or DR 3/x	149 (25)	43 (24)	106 (25)
DR 4/4 or DR 4/x	225 (37)	56 (32)	169 (40)
DR x/x	124 (21)	63 (35)	61 (14)

Data presented as mean ± SD or n (%).

Abbreviations: Ab, autoantibody; ECL, electrochemiluminescence; DR3, DR3 DQA1*05:01 DQB1*02:01; DR4, DR4 DQA1*03:01 DQB1*03:02; DR x/x, neither DQA1*05:01 DQB1*02:01 nor DQA1*03:01 DQB1*03:02.

Participants were more likely to be positive for ECL-GADA with the high-risk HLA genotypes HLA DR 3/4 (80/104, 77%), HLA DR 3/3 or HLA DR 3/x (96/149, 64%), and HLA DR 4/4 or HLA DR 4/x (143/225, 64%) compared to participants who did not have a high-risk HLA genotype (47/124, 38%; P < 0.001) (Table 2). Similarly, participants who were positive for RBA-GADA were more likely to have the high-risk HLA genotypes HLA DR 3/4 (86/104, 83%), HLA DR 3/3 or HLA DR 3/x (111/149, 76%), and HLA DR 4/4 or HLA DR 4/x (164/225, 73%) compared to those without high-risk HLA genotypes (72/124, 58%; P < 0.001). There was no association between ECL-IAA positivity and HLA DR genotypes. Participants with HLA DR 3/3 or HLA DR 3/x genotypes had lower rates of RBA-IAA positivity (34/149, 23%) compared to the other genotypes (P = 0.003), but these numbers are relatively small.

Table 2.

Percentage of participants positive for ECL assay or RBA by HLA genotypes

	HLA DR3/4 (n = 104)	HLA DR 3/3 or HLA DR 3/x (n = 149)	HLA DR 4/4 or HLA DR 4/x (n = 225)	HLA X/X (n = 124)	P-value
ECL-IAA positive	42 (40.4)	41 (27.5)	83 (36.9)	40 (32.3)	0.13
RBA-IAA positive	41 (39.4)	34 (22.8)	78 (34.7)	53 (42.7)	0.003
ECL-GADA positive	80 (76.9)	96 (64.4)	143 (63.6)	47 (37.9)	<0.001
RBA-GADA positive	86 (82.7)	111 (74.5)	164 (72.9)	72 (58.1)	<0.001

Abbreviations: DR3, DR 3 DQA1*05:01 DQB1*02:01; DR4, DR4 DQA1*03:01 DQB1*03:02; DRX, neither DQA1*05:01 DQB1*02:01 nor DQA1*03:01 DQB1*03:02; ECL, electrochemiluminescence; GADA, GAD65 autoantibodies; HLA, human leukocyte antigen; IAA, insulin autoantibody; RBA, radiobinding assay.

In a subanalysis stratified by age, overall trends of antibody positivity were similar by genotype for participants above and below 18 years old. Participants above and below 18 years were more likely to be positive for ECL-GADA and RBA-GADA for the high-risk HLA genotypes compared to participants who did not have a high-risk HLA genotype. However, there was a higher frequency of ECL-IAA and RBA-IAA positivity in younger participants for all genotypes and a lower frequency of ECL-IAA and RBA-IAA positivity in older participants for all genotypes. Participants younger than 18 years vs older than 18 years had higher rates of ECL-IAA positivity (49% vs 15% for DR3/4 participants, 37 vs 15% for DR3/3 or DR3/x participants, 48% vs 22% for DR4/4 or DR4/x participants, and 43% vs 17% for DR x/x participants, respectively). RBA-IAA positivity rates were similarly higher in younger participants vs older participants (48% vs 15% for DR3/4 participants, 33% vs 9% for DR3/3 or DR3/x participants, 44% vs 22% for DR4/4 or DR4/x participants, and 53% vs 29% for DR x/x participants, respectively).

Antibody z-score levels were assessed by high-risk HLA haplotypes (Table 3). ECL-GADA z-scores were higher (0.36 ± 1.33) in participants with the presence of the high-risk HLA haplotype DR3 compared to those without HLA DR3 (−0.02 ± 0.90; P < 0.001). This was similar when stratified by age. In those younger than 18 years, ECL-GADA z-scores were higher (0.39 ± 1.35) in participants with HLA DR3 compared those without HLA DR3 (0.06 ± 0.98; P = 0.008). In those older than 18 years of age, ECL-GADA z-scores were higher (0.32 ± 1.30) in participants with HLA DR3 compared those without HLA DR3 (−0.13 ± 0.76; P = 0.001). There was no difference in ECL-IAA z-score in those with or without HLA DR3. This was the same when stratified by age (above and below 18 years).

Table 3.

ECL and RBA z-scores by presence of HLA haplotype

	HLA-DR3 present (n = 254)	HLA-DR3 absent (n = 348)	P-value	HLA-DR4 present (n = 329)	HLA-DR4 absent (n = 273)	P-value
ECL-IAA z-score mean ± SD	0.01 ± 0.50	0.06 ± 0.64	0.23	0.08 ± 0.62	−0.01 ± 0.53	0.07
RBA-IAA z-score mean ± SD	−0.06 ± 0.60	−0.02 ± 0.44	0.31	−0.04 ± 0.44	−0.03 ± 0.59	0.85
ECL-GADA z-score mean ± SD	0.36 ± 1.33	−0.02 ± 0.90	<0.001	0.17 ± 1.12	0.12 ± 1.11	0.58
RBA-GADA z-score mean ± SD	0.19 ± 1.09	−0.14 ± 0.89	<0.001	0.04 ± 1.00	−0.06 ± 0.99	0.20

Data presented as mean ± SD.

Abbreviations: DR3, DR 3 DQA1*05:01 DQB1*02:01; DR4, DR4 DQA1*03:01 DQB1*03:02; ECL, electrochemiluminescence; GADA, GAD65 autoantibodies; HLA, human leukocyte antigen; IAA, insulin autoantibody; RBA, radiobinding assay.

For RBA testing, participants with HLA DR3 had higher RBA-GADA z-scores (0.19 ± 1.09) compared to those without HLA DR3 (−0.14 ± 0.89; P < 0.001). This was similar when stratified by age. In those younger than 18 years, RBA-GADA z-scores were higher (0.17 ± 1.08) in participants with HLA DR3 compared those without HLA DR3 (−0.10 ± 0.92) (P = 0.012). In those older than 18 years of age, RBA-GADA z-scores were higher (0.23 ± 1.13) in participants with HLA DR3 compared those without HLA DR3 (−0.21 ± 0.86; P = 0.001). There was no difference in the z-score levels of RBA-IAA by HLA DR3 haplotype. This was the same when stratified above and below 18 years of age.

Only ECL-IAA z-scores were slightly higher, but not statistically different, in those with the DR4 haplotype (0.08 ± 0.62) compared to those without HLA DR4 (−0.01 ± 0.53; P = 0.07). When stratified by age, participants older than 18 years of age with the DR4 haplotype had slightly higher ECL-IAA z-scores (−0.08 ± 0.26) compared to those without HLA DR4 (−0.12 ± 0.10; P = 0.07). There was no difference in ECL-IAA z-scores in those younger than 18 years. There was no difference by HLA DR4 haplotype in the z-scores of RBA-IAA, ECL-GADA, or RBA-GADA, overall and when stratified by age.

In those with the haplotype HLA DR3 (n = 254) compared those without the HLA DR 3 haplotype (n = 348), participants were less likely to be RBA-IAA positive (75/254, 30%) with DR3 than without DR3 (131/348, 38%; P = 0.047) (Fig. 1A). There was no difference in the frequency of ECL-IAA positivity and the presence of the DR3 haplotype. Participants with the HLA DR3 haplotype were more likely to be both ECL-GADA positive (70%) and RBA-GADA positive (78%) than those without HLA DR3 haplotype [54% (P < 0.001) and 68% (P = 0.007), respectively].

Figure 1.

(A) The percentage of electrochemiluminescence (ECL) or radiobinding assays (RBA) antibody positivity in participants with and without the high-risk human leukocyte antigen (HLA) haplotype DR3 DQA1*05:01 DQB1*02:01. (B) The percentage of ECL or RBA positivity in participants with and without the high-risk HLA haplotype DR4 DQA1*03:01 DQB1*03:02. * P < 0.05, ** P < 0.01. Abbreviations: DR3, DR 3 DQA1*05:01 DQB1*02:01; DR4, DR4 DQA1*03:01 DQB1*03:02; ECL, electrochemiluminescence; GADA, GAD65 autoantibodies; IAA, insulin autoantibody; RBA, radiobinding assay.

In those with the haplotype HLA DR4 (n = 329) compared those without the HLA DR4 haplotype (n = 273), participants were more likely to be ECL-IAA positive (125/329, 32%) with DR4 than without DR4 (81/273, 30%; P = 0.04) (Fig. 1B). There was no difference in the frequency of RBA-IAA positivity and the presence of the DR4 haplotype. Participants with the HLA DR4 haplotype were more likely to be both ECL-GADA positive (68%) and RBA-GADA positive (77%) than those without HLA DR4 haplotype [52% (P < 0.001) and 67% (P = 0.02), respectively].

Discussion

ECL assays have been shown to be more specific for characterizing risk for type 1 diabetes (18). To our knowledge, this is the first study to analyze the association of type 1 diabetes susceptibility HLA haplotypes and genotypes with ECL positivity and levels. In relatives of individuals with type 1 diabetes, we have shown that ECL-IAA (but not RBA-IAA) positivity is associated with the HLA-DR4 haplotype, while ECL-IAA (but not RBA-IAA) levels showed a trend toward an association with the HLA-DR4 haplotype (P = 0.07). Both ECL-GADA and RBA-GADA positivity and levels were associated with the HLA-DR3 haplotype. These findings were also seen when stratified by age above and below 18 years old. By performing these genetic and immune characterizations, we have further clarified how ECL assays, especially ECL-IAA, may be used to describe endotypes within the development of type 1 diabetes. Endotype characterization is an important aspect in tailoring future prevention and intervention trials.

In the context of type 1 diabetes, endotype refers to a distinct functional and/or pathobiological mechanism with potential disease modification therapy targeting the pathobiological mechanism (19). These endotypes could be helpful in categorizing type 1 diabetes heterogeneity and therefore benefit clinical treatments and research endeavors (19). Genotypic and phenotypic subsets within the disease spectrum have emerged that characterize HLA-specific timing of autoimmunity, such as an early peak incidence of IAA linked to the HLA-DR4 haplotype in first-degree relatives of patients with type 1 diabetes (20). In the Environmental Determinants of Diabetes in the Young (TEDDY) study, the HLA DR4 haplotype was associated predominately with early IAA peak while the HLA-DR3/3 genotype was associated with GADA positivity (5). In this cross-sectional study, we show that GADA (both ECL-GADA and RBA-GADA) positivity and levels are associated with DR3 haplotypes and that ECL-IAA (but not RBA-IAA) positivity and associated trending levels are associated with DR4 haplotypes. When stratified by age, younger patients (<18 years) had higher rates of ECL-IAA and RBA-IAA positivity, as expected with IAA being mainly present in young children. Age of type 1 diabetes onset has been shown to strongly correlate with the appearance of IAA positivity (21). We did not confirm the association of RBA-IAA and DR4 haplotypes, and this could be due to the older age of the TrialNet cohort. In comparison, TEDDY is a birth cohort that follows the development of autoantibodies over childhood, whereas TrialNet participants are screened after a proband relative is diagnosed with type 1 diabetes and thus are older. In addition, the RBA-IAA assay is less specific, and it is possible there be may some variability in these results. Based on these genetic and immunologic phenotypes incorporating both autoantibody positivity and levels, this study further supports that there are 2 main type 1 diabetes endotypes: GADA-HLA DR3 and IAA-HLA DR4.

By characterizing subsets within type 1 diabetes we may precisely target therapies that preserve beta cell function (22). A precision-medicine approach targeted to those who may benefit from intervention or prevention is more likely to be successful. Several clinical trials in type 1 diabetes have reported subgroups of responders and nonresponders to treatment (23-26). In individuals with the HLA DR4*0302 haplotype, methyldopa has been shown to block major histocompatibility complex class II binding and thereby decrease inflammatory T cell responses to insulin in participants recently diagnosed with type 1 diabetes (27). This finding supports the goal of patient-specific targeted therapies based on HLA risk (28). Phase 1b trials are underway to assess the pharmacokinetics of orally administered targeted medications in patients with type 1 diabetes and the HLA DR4*0302 haplotype (NCT04625595). In the TrialNet anti-CD3 teplizumab prevention study in stage 2 type 1 diabetes, subgroups stratified by HLA and islet autoantibody status showed differential responsiveness to treatment (23). The response to teplizumab compared to placebo was greater in participants with the presence of the HLA DR4*0302 haplotype and the absence of the HLA DR3 haplotype. Participants without anti-zinc transporter 8 antibodies also had a more robust response to teplizumab compared to those who were positive for this autoantibody. The success of future prevention and intervention trials will be improved by considering and recruiting individuals with specific diabetes endotypes.

Limitations of this study include the relatively small sample size, as ECL assays were only measured in a subset of participants followed in TrialNet. In addition, only ECL-IAA and ECL-GADA have been measured at baseline, and further testing should be considered in the larger TrialNet cohort as ECL antibody tests become more widely available. Future work should also be considered in characterizing HLA genes and additional islet autoimmune markers such as zinc transporter 8 and insulinoma-associated protein 2 autoantibody ECL assays. While TrialNet only measures type 1 diabetes associated autoantibodies, GADA levels have been associated with autoimmune thyroid disease in the setting of type 1 diabetes (29). Additional testing of thyroid peroxidase autoantibodies and thyroid function tests in this cohort could elucidate further heterogeneity within this cohort.

In summary we have shown that ECL and RBA-GADA positivity are associated with both HLA-DR3 and DR4 haplotypes, while ECL and RBA-GADA z-scores were higher in participants with HLA-DR3 haplotypes. Only ECL-IAA positivity, but not RBA-IAA was associated with HLA-DR4 haplotype. These studies further contribute to the understanding of genetic risk and islet autoimmunity endotypes in type 1 diabetes.

Additional Information

Disclosure Statement: The authors have no conflicts of interest to disclose.

Data Availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.