Antibodies to post-translationally modified insulin as a novel biomarker for prediction of type 1 diabetes in children

Aims/hypothesis We have shown that autoimmunity to insulin in type 1 diabetes may result from neoepitopes induced by oxidative post-translational modifications (oxPTM). Antibodies specific to oxPTM-insulin (oxPTM-INS-Ab) are present in most newly diagnosed individuals with type 1 diabetes and are more common than autoantibodies to native insulin. In this study, we investigated whether oxPTM-INS-Ab are present before clinical onset of type 1 diabetes, and evaluated the ability of oxPTM-INS-Ab to identify children progressing to type 1 diabetes. Methods We used serum samples collected longitudinally from the ‘All Babies in Southeast Sweden (ABIS)’ cohort tested for the gold standard islet autoantibodies to insulin (IAA), GAD (GADA), tyrosine phosphatase 2 (IA-2A) and zinc transporter 8 (ZnT8A). We studied 23 children who progressed to type 1 diabetes (progr-T1D) and 63 children who did not progress to type 1 diabetes (NP) after a median follow-up of 10.8 years (interquartile range 7.7–12.8). Of the latter group, 32 were positive for one or more islet autoantibodies (NP-AAB+). oxPTM-INS-Ab to insulin modified by •OH or HOCl were measured by our developed ELISA platform. Results Antibodies to at least one oxPTM-INS were present in 91.3% of progr-T1D children. oxPTM-INS-Ab co-existed with GADA, IA-2A, IAA or ZnT8A in 65.2%, 56.5%, 38.9% and 33.3% progr-T1D children, respectively. In addition, oxPTM-INS-Ab were present in 17.4%, 26.1%, 38.9% and 41.6% of progr-T1D children who were negative for GADA, IA-2A, IAA and ZnT8A, respectively. •OH-INS-Ab were more common in progr-T1D children than in NP-AAB+ children (82.6% vs 19%; p < 0.001) and allowed discrimination between progr-T1D and NP-AAB+ children with 74% sensitivity and 91% specificity. None of the NP-AAB− children were positive for oxPTM-INS-Ab. Conclusions/interpretation oxPTM-INS-Ab are present before the clinical onset of type 1 diabetes and can identify children progressing to type 1 diabetes.


Introduction
Type 1 diabetes is characterised by insulin deficiency and hyperglycaemia due to extensive destruction of insulinproducing beta cells. The autoimmune nature of the disease is suggested by the presence of a pool of circulating autoantibodies against beta cell proteins even years before the clinical onset, such as autoantibodies to insulin (IAA), GAD (GADA), tyrosine phosphatase (IA-2A) and zinc transporter 8 (ZnT8A) [1]. The mechanism underlying the breach of immune tolerance to beta cell antigens is still unclear. Neoepitopes that are post-translationally modified from the native antigens have been recently described in type 1 diabetes [2][3][4][5][6][7]. Being different from the native proteins, the post-translational modification (PTM) of self-antigens may be recognised as foreign and result in breakdown of tolerance [8][9][10]. Oxidative stress is a key feature of many autoimmune diseases and results in an excess of reactive oxidants able to generate oxidative PTM (oxPTM) [4,11]. Products of oxidative stress are increased in type 1 diabetes [12] and also in individuals at risk [13]. During insulitis, the beta cells that are under stress and the high influx of metabolically active immune cells generate large quantities of reactive oxidants, including the superoxide radical, hydroxyl radical ( • OH), hypochlorous acid (HOCl) and peroxynitrate [14]. We have shown that autoimmunity to insulin may result from neoepitopes induced by oxPTM [11]. We found that antibodies specific to oxidative posttranslationally modified insulin (oxPTM-INS) are present in the majority of newly diagnosed individuals with type 1 diabetes and are significantly more abundant than autoantibodies to native insulin (NT-INS) [11].
Islet autoantibodies (IAA, GADA, IA-2A and ZnT8A) represent the most robust approach to identify individuals at risk and to predict progression to clinical disease in those genetically at risk [15] and in the general population [16]. Children with any two autoantibodies may have a risk of 80% for developing type 1 diabetes during childhood or adolescence [15]. However, multiple antibody testing is required for the best prediction. An additional problem is that radiobinding assays (RBAs), the gold standard for islet autoantibodies, have expensive regulatory requirements. These elements introduce additional complexity that may limit the implementation of autoantibody screening. Finally, a significant number of individuals test negative to these markers [17]. Therefore, the development of alternative technologies remains an unmet need. To address this we have developed an ELISA that detects auto-reactivity to oxPTM-INS. We showed that the assay detecting antibodies to oxPTM-INS (oxPTM-INS-Ab) is highly accurate (84% sensitivity, 99% specificity), may detect over 30% of individuals who are negative to the RBA for IAA and, when combined with IAA, identifies 95% of individuals with newly diagnosed type 1 diabetes [11]. However, the predictive potential of oxPTM-INS-Ab is not known yet.
We hypothesise that oxPTM-INS-Ab are present before the clinical onset of the disease. To test this hypothesis, considering the high prevalence of oxPTM-INS reactivity in newly diagnosed type 1 diabetes, we evaluated the ability of this novel autoantibody specificity to identify children progressing to clinical disease. We used serum samples from the ABIS (All Babies in Southeast Sweden) study, a large prospective study in which unselected children from the general population born during 1997-1999 have been followed prospectively with regular evaluation of islet autoantibodies for the development of type 1 diabetes [18,19].

Methods
Participants and serum samples Serum samples from the ABIS study were obtained and analysed in a blinded fashion for detection of oxPTM-INS-Ab. The ABIS study is a prospective population-based follow-up study which included 17,055 unselected children born between 1 October 1997 and 1 October 1999 in southeast Sweden [18]. Of the screened children, 116 developed type 1 diabetes during the follow-up. In the present study, we tested 51 samples from the 23 children progressing to type 1 diabetes (progr-T1D), collected longitudinally before diagnosis at three different time points (at the ages of 5, 7 and 11 years). Samples were selected where sufficient serum and autoantibody data were available for this study. Only one time point was available from seven progr-T1D children before the collection at the second time point. As controls, we used samples from 63 children of similar age and sex who did not progress to type 1 diabetes over time, including 64 samples from 32 autoantibody-positive children (autoantibody-positive, nonprogressing to type 1 diabetes [NP-AAB + ]) and 31 samples from autoantibody-negative children (autoantibody-negative, nonprogressing to type 1 diabetes [NP-AAB − ]). NP-AAB + children were defined as positive to at least one islet-antibody marker (IAA, GADA, IA-2A or ZnT8A). Informed consent was obtained from parents prior to the collection of blood. The study was approved by the Research Ethics Committees of the Medical Faculties of Linköping University, Linköping and Lund University, Lund, Sweden and by the Ethics Committee of the Università Campus Bio-Medico di Roma.
ELISA for detection of oxPTM-INS-Ab Insulin was chemically modified as previously described to generate oxPTM-INS modified by HOCl, • OH and glycation using ribose [11]. Hen egg lysozyme (HEL; Sigma-Aldrich, Milan, Italy) was similarly modified and used as a control antigen. oxPTM-INS encompasses glycated (GLY-INS), • OH-modified ( • OH-INS) and HOCl-modified insulin (HOCl-INS). An ELISAwas performed using NT-INS, oxPTM-INS, control native HEL or control oxidative post-translationally modified HEL (oxPTM-HEL) as targets. Development and calibration of the ELISA is described in our previous publication [11]. ELISA plates (Nunc, London, UK) were coated with 10 μg/ml of modified or native protein in 0.05 mol/l carbonate/bicarbonate buffer (pH 9.6) at 4°C overnight. Plates were then washed three times with PBS. After blocking for 2 h with 5% BSA in 0.5% Tween PBS, 100 μl of 1:200-diluted serum samples in 5% BSA in 0.5% Tween PBS were added to each well, followed by 2 h incubation at room temperature. Plates were then washed with PBS plus 0.1% Tween, followed by three washes with PBS. Antihuman IgG-horseradish peroxidase-conjugated antibodies (Sigma-Aldrich) were then added at 1:1000 dilution in 5% BSA in 0.5% Tween PBS for another 2 h incubation. The ELISA plates were washed, and 100 μg/ml 3,3′,5,5′tetramethylbenzidine substrate (Sigma-Aldrich) in 100 mmol/l sodium acetate (pH 6.0), was added. Subsequently, the reaction was stopped with 1 mol/l sulphuric acid. The absorbance was measured at 450 nm using a GENios plate reader and Magellan software (Tecan, Reading, UK). The ELISA absorbance values obtained for HEL and oxPTM-HEL were used as background controls that were subtracted from the absorbance values of NT-INS and oxPTM-INS, respectively. In addition, to account for assay fluctuation, binding to insulin, oxPTM-INS, HEL and oxPTM-HEL was tested for each individual sample on the same plate. Each assay included known positive or negative reference control samples. Longitudinal samples obtained from the same individuals were tested on the same plate. Levels of oxPTM-INS-Ab above the 99th percentile of 88 healthy individuals were defined as ELISA cut-off. Intra-assay CV was <8% (n = 10). Inter-assay CVs were <10% and <13% for NT-INS and oxPTM-INS-Ab (n = 12), respectively.
Statistical analysis Statistical analyses were performed using Prism Software version 6.01 (GraphPad, San Diego, CA, USA). Differences in antibody levels between groups were tested by the Mann-Whitney test. Longitudinal changes in antibody binding were evaluated by the Wilcoxon paired test.
To determine predictive discrimination between progr-T1D and control groups, we used the 99th percentile of the healthy individuals as cut-off point absorbance units to construct a contingency table of positive oxPTM-INS-Ab against clinical diagnosis and tested it by Fisher's Exact Test.

Results
Features of the studied population The characteristics of the study population are shown in Table 1. At the earliest time point studied, the progr-T1D, NP-AAB + and NP-AAB − groups were comparable in terms of sex, while age was slightly higher in the NP-AAB + group (p = 0.04). The majority of progr-T1D children (n = 18, 78%) had multiple positive islet autoantibodies; two were single-positive. Two children were negative for IAA, GADA and IA-2A; one child was negative for GADA and IA-2A (IAA was not assessed). Only 25% (n = 8) of NP-AAB + children had multiple positive islet autoantibodies. HLA typing was available in 43 children; HLA susceptible phenotypes were more common in progr-T1D than in NP-AAB + and NP-AAB − children (p < 0.007), while the prevalence of protective phenotypes was similar between the three groups (p > 0.07) ( Table 2). In the progr-T1D group, median follow-up from the first sample analysed to diagnosis was 5.  Figure 3 shows the degree of overlap between oxPTM-INS-Ab ( • OH-INS-Ab) and other islet autoantibodies in progr-T1D and NP-AAB + children. IAA data were available in 18 progr-T1D and 27 NP-AAB + children and ZnT8A data were available in 12 progr-T1D and 10 NP-AAB + children. While binding and prevalence of oxPTM-INS-Ab were significantly higher in progr-T1D children than in NP-AAB + children (p < 0.001), both titres and prevalence of GADA, IAA and ZnT8A were similar between the two groups (p > 0.257). IA-2A was the most specific Data are presented as means ± SD or n (%)

Distribution of oxPTM-INS-Ab and islet autoantibodies
The three groups were comparable in terms of sex, although age was slightly higher in the NP-AAB + group than in the progr-T1D group (p = 0.04). Eighteen (78%) of progr-T1D and 8 (25%) of NP-AAB + children had multiple positive islet autoantibodies Table 2 HLA-genotypes prevalence in the study population  (Fig. 4). In the NP-AAB + group, oxPTM-INS-Ab reactivity was similar in those who were positive to one or more standard islet autoantibodies (p > 0.148).

Discussion
We have recently shown that oxPTM-INS-Ab are very common in newly diagnosed type 1 diabetes, being detected in 84% of individuals [11]. In this study, we found that oxPTM-INS auto-reactivity is present before diabetes diagnosis in over 90% of individuals. oxPTM-INS-Ab could discriminate between prog-T1D children and those who did not progress to type 1 diabetes regardless of positivity to other islet autoantibodies. Therefore, these data may suggest a potential role for oxPTM-INS-Ab as a predictive biomarker of type 1 diabetes.
The main strength of our study is the prospective evaluation of oxPTM-INS-Ab in a well-characterised cohort of children from the general population, tested for the standard diabetes autoantibodies. The uniqueness of the ABIS study resides in the possibility of applying its findings to the general population, since participants were not selected according to predetermined diabetes risk (genetic or familial). An additional strength was that the long follow-up in some progr-T1D children allowed us to discover whether oxPTM-INS-Ab develop very early in the natural history of the disease. Our study has limitations. The earliest time point tested in the progr-T1D children was at 5 years of age. Data from birth cohorts suggest that peak age of seroconversion in high-risk individuals is around 2 years of age [25]. Therefore, we cannot exclude  the possibility that children defined as negative for the standard islet autoantibodies have not previously been seropositive for one or more islet autoantibodies. A second limitation is that the determination of ZnT8A was performed on a limited number of children, as ZnT8A assays became available only after the follow-up for many children in the ABIS study.
Our findings shed further light on type 1 diabetes pathogenesis. Detection of oxPTM-INS reactivity before clinical onset of  type 1 diabetes is consistent with evidence that the unbalanced redox state takes place early in the natural history of the disease. Individuals with a short disease duration have an early impairment in antioxidant capacity and over threefold increased levels of lipid peroxidation regardless of blood glucose control [12]. An impaired oxidation status anticipates dysglycaemia; it has been shown that plasma malondialdehyde and erythrocyte malondialdehyde, two markers of oxidative stress, are abnormal in euglycaemic individuals at increased risk for type 1 diabetes compared with individuals not at risk [13]. These data, together with our findings, may support the involvement of oxidative stress in type 1 diabetes pathogenesis.
To our knowledge, this is the first study investigating antibody reactivity to oxPTM/PTM of a beta cell antigen as a predictive biomarker of type 1 diabetes. Therefore, a direct comparison with prediction potential of other modified antigens is not possible. Our data are consistent with previous finding by our group in newly diagnosed individuals with type 1 diabetes. Similar to our previous study [11], oxPTM-INS-Ab identified over 80% of individuals with the disease and were able to detect more than one-third of those who tested negative to the IAA assay. Additional neoepitopes derived from PTM of beta cell autoantigens have been described in humans and in animal models of type 1 diabetes, including antibody response to oxidised GAD [26] and T cell reactivity to GAD citrullination [7], C-peptide deamination [5] and hybrid fused peptides [2,27]. The response to modified antigens may also involve proteins that are not proper to beta cells [4,28], especially in the presence of defined genetic background. An example is the identification of antibodies to oxPTM-collagen type II in a large proportion of individuals carrying HLA-DRB1*04 [4]. In this regard, the extracellular matrix surrounding beta cells [29], or other tissues attacked by autoimmune response in type 1 diabetes (thyroid, gut, joints, etc) [30], may become additional potential targets of oxPTM [4,31]. Often such PTM forms induce a more pronounced immune reactivity than the native antigen.
Our results may also have implications for disease prediction and staging. Consistent with previous findings [15,32], we found low predictive accuracy of GADA and IAA when analysed as a single test, while IA-2A showed a highly specific association with type 1 diabetes progression [15]. A main finding of our study is the predictive accuracy of oxPTM-INS-Ab. Of note, oxPTM-INS-Ab identified individuals with preclinical disease otherwise classified as antibody-negative or single-positive. As highlighted by a recent statement by the JDRF, Endocrine Society and ADA, islet autoimmunity (as defined by the presence of two or more islet autoantibodies) represents the earliest stage of type 1 diabetes and identifies a target population for prevention trials and future preventive strategies [33]. If confirmed in larger studies, oxPTM-INS-Ab may be adopted as an additional biomarker to further redefine disease taxonomy, allowing better prediction and therefore better stratification into eligibility trials.
In conclusion, we showed that immune reactivity to oxPTM-INS is present before clinical onset of type 1 diabetes and that measurement of oxPTM-INS-Ab may identify children likely to progress to type 1 diabetes. This is the first evidence suggesting that oxPTM of a beta cell autoantigen precedes diabetes onset in humans and that auto-reactivity to oxPTM may act as a predictive biomarker of the disease. Additional studies with larger cohorts are required to confirm the predictive potential of oxPTM-INS-Ab in type 1 diabetes.
Data availability The data are available on request from the authors. Duality of interest The authors declare that there is no duality of interest associated with this manuscript.
Contribution statement RS, PP, JL and AN conceived the study. CV contributed to acquisition and analysis of data. NN contributed to data acquisition. RS analysed the data and wrote the first draft. All authors critically revised the manuscript for intellectual content. All authors have seen and approved the final draft. RS is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Open Access This article is distributed under the terms of the Creative Comm ons Attribution 4.0 International License (http:// creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.