, Volume 56, Issue 3, pp 563–572 | Cite as

HLA-dependent autoantibodies against post-translationally modified collagen type II in type 1 diabetes mellitus

  • R. Strollo
  • P. Rizzo
  • M. Spoletini
  • R. Landy
  • C. Hughes
  • F. Ponchel
  • N. Napoli
  • A. Palermo
  • R. Buzzetti
  • P. Pozzilli
  • A. NissimEmail author



In this study the involvement of oxidative stress in type 1 diabetes mellitus autoimmunity and the possible association with rheumatoid arthritis (RA) was investigated. We tested the hypothesis that oxidative stress induced by chronic hyperglycaemia triggers post-translational modifications and thus the formation of neo-antigens in type 1 diabetes, similar to the ones found in RA.


Collagen type II (CII), a known autoantigen in RA, was treated with ribose and various reactive oxygen species (ROS). Levels of antibodies specific to native and ROS-modified CII (ROS-CII) were compared in type 1 diabetes, type 2 diabetes and healthy controls, and related to the HLA genotype.


Significantly higher binding to ROS-CII vs native CII was observed in type 1 diabetic patients possessing the HLA-DRB1*04 allele irrespective of variables of glucose control (blood glucose or HbA1c). Type 1 diabetic patients carrying a DRB1*04 allele with the shared epitope showed the highest risk for ROS-CII autoimmunity, while the DRB1*0301 allele was protective. Conversely, native CII autoimmunity was not associated with any specific DRB1 allele. Positive and inverse seroconversion rates of response to ROS-CII were high in DRB1*04-positive type 1 diabetic patients.


Hyperglycaemia and oxidative stress may trigger genetically controlled autoimmunity to ROS-CII and may explain the association between type 1 diabetes mellitus and RA.


Collagen type II Post-translational modification Reactive oxygen species Rheumatoid arthritis 



Anti-citrullinated peptide antibodies


Collagen type II


Horseradish peroxidase


Optical density


Rheumatoid arthritis


Reactive oxygen species


CII modified by ROS


Shared epitope


Thyroid peroxidase


Type 1 diabetes mellitus is characterised by chronic insulin deficiency and hyperglycaemia due to extensive destruction of insulin-producing beta cells. The autoimmune nature of this process is supported by the presence of a pool of autoantibodies against beta cell antigens as well as the association with genes controlling immune homeostasis [1].

The major genetic risk for type 1 diabetes resides within the HLA complex region, which contains alleles that encode molecules involved in the recognition and presentation of peptide antigens to T cells. This risk is conferred by a combination of alleles of the HLA class II super-types DRB1*03-DQB1*0201 and DRB1*04-DQB1*0302 [2]. The DQB1*0302-A1*0301 (DQ8) haplotype defines the greatest risk for the disease, with an increased diabetogenic effect if the DRB1*0401 allele is inherited as part of the haplotype. An effect predisposing to type 1 diabetes has also been found for other DRB1*04 alleles, *0402, *0404 and *0405 [3, 4]. Consistently, markers of beta cell autoimmunity such as GAD autoantibodies (GADA), tyrosine phosphatase autoantibodies (IA-2A) and insulin autoantibodies (IAA) are strongly associated with HLA-DRB1*04 and/or HLA-DRB1*03 not only in type 1 diabetes, but also in first-degree relatives of individuals with type 1 diabetes, as well as in the general population [5].

Hyperglycaemia is the hallmark of diabetes and is a major player in the chronic complications of the disease. Elevated blood glucose induces oxidative stress and changes in the cellular redox state. NADPH oxidase has been responsible for the formation of high levels of reactive oxygen species (ROS) in response to high glucose [6]. A second source of ROS production is the excessive production of AGEs. Two AGEs, namely carboxymethyl-lysine and pentosidine, are related to the severity of diabetic nephropathy and are known biomarkers for ‘carbonyl stress’ [7]. The main toxic effect of both ROS and AGEs is the induction of abnormal post-translational modifications of self-antigens and the generation of neo-antigens, thus bypassing immune tolerance and contributing to the development of autoimmune responses [8]. The involvement of oxidative stress in type 1 diabetes has been implied by the presence of autoantibodies against oxidised GAD [9, 10, 11]. In addition, experimental diabetes can be induced in rats by feeding with alloxan and streptozotocin, two substances that work by generating ROS and inducing selective damage to beta cells [12].

We have recently shown that post-translational modification by reactive oxidants results in the formation of neo-epitopes in rheumatoid arthritis (RA), an autoimmune disease where abnormal immune reactivity against a cartilage extracellular matrix protein, collagen type II (CII), is detected [13, 14]. We have demonstrated that ROS present in inflamed RA joints can generate CII neo-antigens, which stimulate autoimmunity to ROS-modified CII (ROS-CII), that have been detected in serum samples from individuals with RA [15]. Co-existence of type 1 diabetes and RA in the same individuals has been reported [16] and a large study of 3,093 individuals showed that type 1 diabetic patients are two to five times more likely than healthy individuals to develop RA [17]. Part of this association has been attributed to the presence of shared genetic susceptibility. Similarly to type 1 diabetes, the greatest risk for RA is restricted to DRB1*04 alleles containing a shared epitope (SE), namely DRB1*0401, *0404 and *0405, with an additional effect in the presence of the DRB1*0101 and DQB1*0302 alleles [18]. On the other hand, DRB1*0402, although conferring susceptibility to type 1 diabetes, is protective against RA [18].

The aim of this study was to assess the role of hyperglycaemia and oxidative stress in triggering neo-antigenicity to self-antigens in type 1 diabetes. We hypothesised that oxidative stress induced by chronic hyperglycaemia may trigger post-translational modifications, and thus the formation of neo-antigens in type 1 diabetes. Considering the existence of an association between type 1 diabetes and RA, we evaluated the presence of autoantibodies to ROS-CII, and the interplay with the HLA alleles conferring susceptibility in both diseases. Levels of native and ROS-CII antibodies were compared in type 1 diabetes, type 2 diabetes and healthy controls, and related to HLA genotype using both cross-sectional and longitudinal designs.



Serum samples were obtained from participating centres of the Immunotherapy Diabetes (IMDIAB) group. Of the 151 individuals with type 1 diabetes, 81 were HLA-DRB1*04 positive and 70 HLA-DRB1*04 negative. Individuals with type 1 diabetes were selected according to the following criteria: (1) diagnosis of the disease according to American Diabetes Association criteria, with age at presentation less than 35 years—in most cases diagnosis was confirmed by positivity to glutamic acid decarboxylase, insulin or tyrosine phosphatase antibodies; (2) no other major chronic disease; and (3) less than 12 months’ duration of clinical disease (since the beginning of insulin therapy). The longitudinal study was conducted using sera from individuals recently diagnosed with type 1 diabetes (n = 21 HLA-DRB1*04 positive and n = 13 HLA-DRB1*04 negative) obtained within 6 months of diagnosis and 9–12 months after the first sampling. Sera from individuals with type 2 diabetes (n = 40) and healthy individuals (n = 85) were used as control groups. Within the control groups, HLA typing was obtained in individuals with type 2 diabetes (n = 20) and healthy individuals (n = 33). DRB1*04 alleles were carried by 0% and 15% of type 2 diabetes and healthy control individuals, respectively; the most common HLA-DRB1 types were DRB1*11 and/or DRB1*12, DRB1*15, DRB1*16 and DRB1*03. This project was approved by the Ethical Committee at University Campus Bio-Medico within the framework of the IMDIAB investigators type 1 study, with informed consent signed by patients or parents.

Chemical modification of CII and BSA

Bovine CII was prepared from bovine cartilage and chemically modified as described [15]. Briefly, CII (2 mg/ml) in PBS was incubated overnight at 37°C with the following systems: (1) 2 mol/l ribose (Sigma, Gillingham, UK); (2) 1 mmol/l HOCl (BDH, Oxford, UK); (3) 1 mmol/l CuCl2 (Sigma) and 2 mmol/l H2O2 (Sigma) which was used to produce hydroxyl radical (·OH) by the Fenton reaction; and (4) 2 mmol/l peroxynitrite (ONOO,Calbiochem, Beeston, UK). Modifications of CII were monitored by 8% SDS-PAGE and three-dimensional fluorescence, as described previously [15]. BSA was also modified as above.

HLA genotyping

HLA typing and subtyping were performed by PCR using specific primers and hybridisation with sequence specific oligonucleotides, as previously described [19]. DRB1*0401, *0404 and *0405 were grouped together because they encode a conserved amino-acid sequence, an HLA-DRB1 SE [20], depicting a genetic susceptibility shared by type 1 diabetes and RA [3, 4, 18].


An ELISA was performed using modified and native CII or BSA as targets as described previously [15]. Briefly, ELISA plates (Nunc, London, UK) were coated with 10 μg/ml of modified or native protein in PBS for incubation at 4°C overnight. Plates were then washed three times with PBS. After blocking for 1.5 h with 2% (wt/vol.) dry milk powder in PBS (Marvel-PBS), 100 μl of 1:100-diluted serum samples in 2% Marvel-PBS were added to each well, followed by 1.5 h incubation at 37°C. Plates were then washed with PBS plus 0.1% Tween, followed by three washes with PBS. Anti-human IgG-horseradish peroxidase (HRP)-conjugated antibodies (Sigma) were then added at 1:1,000 dilution in 2% Marvel-PBS for another 1.5 h incubation. The ELISA plates were washed, and 100 μg/ml 3,3′,5,5′-tetramethylbenzidine substrate (Sigma) in 100 mmol/l sodium acetate, pH 6.0, were added. Subsequently, the reaction was stopped with 1 mol/l sulphuric acid. The optical density (OD) was measured at 450 nm using a GENios plate reader and Magellan software (Tecan, Reading, UK).

Western blotting

For the western blots, 2 μg ROS-CII, native CII, ROS-BSA and native BSA were run on 8% denaturing SDS-PAGE and electroblotted onto a nitrocellulose membrane (BDH Chemicals). After blocking with 5% Marvel-PBS, membranes were incubated with a 1:100 dilution of serum samples in 5% Marvel-PBS for 1 h at room temperature, followed by incubation with 1:1,000 dilution of anti-human IgG-HRP (Sigma). Membranes were washed extensively with 0.1% Tween-PBS before development with enhanced chemiluminescence (Pharmacia, Milton Keynes, UK).

Statistical analysis

Statistical analyses were performed using Prism Software (GraphPad, San Diego, CA, USA) and Stata 12 (StataCorp. 2011, Stata Statistical Software: Release 12, College Station, TX, USA). Differences in levels of autoantibodies between groups were tested by the Mann–Whitney test. Fisher’s exact test was used to assess whether the reactivity toward ROS-CII or ROS-BSA was significantly different from native CII or native BSA. Levels of anti-CII and anti-BSA antibodies above the 95th percentile of the healthy individuals were defined as elevated. Pearson’s correlation test was used to assess the association between reactivity to ROS-CII and native CII and metabolic control. ORs were calculated [21] and reported with 95% CIs to assess the association between genotype and presence of elevated CII or BSA autoantibodies. Longitudinal change in antibody binding was evaluated using a Wilcoxon matched-pairs signed rank test.


Population study

Features of the studied cohort are reported in Table 1. No differences were detected between HLA-DRB1*04-positive and HLA-DRB1*04-negative patients in term of sex, age or metabolic variables. Individuals with type 1 diabetes were younger than type 2 diabetes and healthy controls, and had shorter disease duration than type 2 diabetes patients.
Table 1

Clinical and biochemical features of the studied populations


T1D DRB1*04 + (n = 81)

T1D DRB1*04 (n = 70)

T2D (n = 40)

Healthy controls (n = 85)

Male sex, n (%)

42 (52)

37 (53)

19 (48)

31 (36)

Age, years

12.09 ± 0.79

14.02 ± 0.95

56.88 ± 1.86

40.12 ± 1.3

BMI, kg/m2

18.2 ± 0.43

18.19 ± 0.49

30.92 ± 1.39


Disease duration, years



6.52 ± 1.19


Blood glucose, mmol/l

8.46 ± 0.47

8.26 ± 0.40

8.14 ± 0.27


HbA1c, % (mmol/mol)

6.61 ± 0.18 (48.73 ± 1.96)

7.05 ± 0.34 (53.57 ± 3.74)

6.87 ± 0.19 (51.54 ± 2.06)


Data are mean ± standard error

Sex, age and metabolic variables of type 1 diabetes HLA-DRB1*04 positive and HLA-DRB1*04 negative individuals were similar but they were younger than type 2 diabetic patients and healthy controls and had shorter disease duration than patients with type 2 diabetes

T1D, type 1 diabetes; T2D, type 2 diabetes

Antibody binding to ROS-CII and ROS-BSA in type 1 diabetes and in control individuals

Samples were grouped into type 1 diabetes HLA-DRB1*04 positive and HLA-DRB1*04 negative. Binding to native and ROS-CII was significantly higher in type 1 diabetes HLA-DRB1*04-positive compared with type 1 diabetes HLA-DRB1*04-negative samples, type 2 diabetes controls and healthy individuals (p < 0.001, Fig. 1a). An increased percentage of binders to native and ROS-CII was observed in the HLA-DRB1*04-positive samples, 25.93% vs 62.96%, respectively (p < 0.0001). Binding of the HLA-DRB1*04-negative group was generally low, with only 2.90% and 14.29% binders to native and ROS-CII, respectively. The lowest binding was seen in samples taken from individuals with type 2 diabetes, with 0% and 27.5% binders to native and ROS-CII, respectively. Nevertheless, in both type 1 diabetes (HLA-DRB1*04 positive and negative) and type 2 diabetes groups, binding to ROS-CII was higher than to native CII (p < 0.001, Fig. 1a). Binding to native BSA was higher in type 1 diabetes groups compared with healthy controls (p < 0.001, Fig. 1b), but no significant differences occurred between DRB1*04-positive or DRB1*04-negative type 1 diabetes and type 2 diabetes. We observed 8% and 12% binders to native BSA and ROS-BSA in type 1 diabetes DRB1*04 positive, respectively. Similarly in type 1 diabetes DRB1*04 negative, 3% and 10% bound to native BSA and ROS-BSA, respectively. In both type 1 diabetes and type 2 diabetes, binding to ROS-BSA was weaker than to native BSA (p < 0.001, Fig. 1b), except for peroxynitrite-modified BSA. Only 3.85% of healthy control individuals responded to either native or ROS-CII regardless of HLA-DRB1*04 positivity, while 2.7% responded to BSA or ROS-BSA.
Fig. 1

Binding to CII (a) and BSA (b) by serum samples from patients with type 1 or type 2 diabetes and healthy controls. Binding was assessed by ELISA in individuals according to HLA-DRB1*04 genotype. (a) Binding to native and ROS-CII was significantly higher in type 1 diabetes DRB1*04 + compared with the other groups: type 1 diabetes DRB1*04 , type 2 diabetes and healthy controls (p < 0.001). A significant increase in the binding to ROS-CII was detected in type 1 and type 2 diabetes serum when compared with binding to native CII (p < 0.001), except for CII modified by hydroxyl radical (·OH). Lane 1, native CII; lane 2, glycated CII; lane 3, HOCl-modified CII; lane 4, ·OH-modified CII; lane 5, peroxynitrite-modified CII. (b) In contrast, a significant decrease in binding to ROS-BSA was found in all individuals with type 1 diabetes (p < 0.001). Lane 1, native BSA; lane 2, glycated BSA; lane 3, HOCl-modified BSA; lane 4, ·OH-modified BSA; lane 5, peroxynitrite-modified BSA. AU, arbitrary units; HC, healthy controls; T1D, type 1 diabetes; T2D, type 2 diabetes

Analysis of serum reactivity to ROS-CII and ROS-BSA by western blotting

Western blot analysis was performed on a range of serum samples with high or low binding to modified CII in ELISA. A correlation between ELISA binders and samples showing strong binding at the western blot analysis was found. A typical pattern of binding is shown in Fig. 2a for a DRB1*04-positive individual with type 1 diabetes and in Fig. 2b for a DRB1*04-negative individual. The DRB1*04-positive type 1 diabetes serum sample bound to native CII and ROS-CII as follows. For native CII, binding to a fragment in the region of 100 kDa, which correlates with the mobility of the CII α chain (lane 1). For glycated CII, binding was: (1) to a higher mobility fragment in the region of 250 kDa; (2) 100 kDa fragment corresponding to the α chain; (3) a slightly higher band that corresponds to CII with ribose condensation (shift base); and (4) fragmented CII in the region of 50–100 kDa (lane 2). A weaker pattern of binding to fragmented CII and high-molecular-weight aggregates was observed for CII modified by HOCl (lane 3). For CII modified by ·OH and modified by ONOO, there was strong binding to high-molecular-weight aggregates, with some stuck in the slots of the SDS-PAGE (lanes 4 and 5). Binding of the DRB1*04-negative serum sample was generally low or absent. No specific binding was found for native and ROS-BSA (data not shown).
Fig. 2

Binding of type 1 diabetes serum samples to native CII and ROS-CII as determined by western blotting. Representative binding pattern to native CII and ROS-CII using two serum samples from type 1 diabetes-DRB1*04-positive (a) and DRB1*04-negative (b) individuals, respectively. (a) Type 1 diabetes-DRB1*04-positive sample showed binding to native CII and to a range of fragmented CII and high-molecular-weight aggregates following treatment with any of the systems for generating ROS. (b) Binding to native and ROS-CII was very low in the type 1 diabetes DRB1*04-negative sample. The position of molecular-weight markers (in kDa) is shown on the left. Lane 1, native CII; lane 2, glycated CII; lane 3, HOCl-modified CII; lane 4, ·OH-modified CII; lane 5, peroxynitrite-modified CII. AU, arbitrary units

Effect of HLA genotype on antibody binding to ROS-CII and ROS-BSA in type 1 diabetes

The association between specific DRB1 genotype and reactivity to native or modified CII was analysed in 151 individuals by calculating the OR of responders vs non-responders for each genotype. Native CII and ROS-CII autoimmunity was positively associated with DRB1*04 (OR 11.9 [95% CI 2.68, 52.88], p = 0.0001) and (OR 10.2 [95% CI 4.55, 22.86], p < 0.0001), respectively. DRB1*03, however, was protective against native and ROS-CII autoimmunity in the absence of DRB1*04 (OR 0.04 [95% CI 0.003, 0.75], p = 0.0006) and (OR 0.1 [95% CI 0.05, 0.37], p = 0.0001), respectively (Table 2). Complete HLA-DRB1 subtyping was available only for 57 individuals. Phenotype analysis showed that type 1 diabetic patients carrying SE-containing DRB1*04 alleles, defined by the presence of at least one DRB1*0401 *0404 or *0405, had the highest risk for ROS-CII autoimmunity (OR 3.62 [95% CI 1.12, 11.74], p = 0.027), while DRB1*0301 provided protection from ROS-CII reactivity (OR 0.23 [95% CI 0.07, 0.75], p = 0.013). However, other DRB1 phenotypes, including the type 1 diabetes DRB1*04 susceptibility allele *0402, were not associated with anti-ROS-CII reactivity (OR 2.1 [95% CI 0.39, 11.23] p = 0.378), suggesting a more selective relationship between DBR1*04 and ROS-CII autoimmunity in type 1 diabetes. Native CII autoimmunity was, however, not associated with any specific DRB1 phenotype (Table 2).
Table 2

HLA-DRB1 genotypes and phenotypes associated with anti-CII reactivity


Native CII antibodies

ROS-CII antibodies

Positive n

OR (95% CI)

p value

Positive n

OR (95% CI)

p value

HLA-DRB1 genotypes, T1D, n = 151

DRB1*04/DRB1*03 or DRB1*X (n = 81)


11.9 (2.68, 52.88)



10.2 (4.55, 22.86)


DRB1*03/not DRB1*04 (n = 41)


0.04 (0.003, 0.75)



0.1 (0.05, 0.37)


DRX/DRX (n = 29)


0.34 (0.075, 1.553)



0.35 (0.01, 0.98)


HLA-DRB1 phenotypes, T1D, n = 57

DRB1*0301 (n = 27)


0.22 (0.042, 1.15)



0.23 (0.07, 0.75)


SE (n = 18)


1.57 (0.38, 6.44)



3.62 (1.12, 11.74)


DRB1*X (n = 48)


5.18 (0.28, 96.5)



2.1 (0.39, 11.23)


The associated risk between autoreactivity to CII modified by reactive oxidants (ROS-CII) and HLA-DRB1 genotypes and phenotypes was analysed

DRX represents all except DRB1*04 and DRB1*03

SE refers to at least one HLA-DRB1*0401, *0404 or *0405

DRB1*X equates to HLA-DRB1 phenotypes other than those indicated in the table

T1D, type 1 diabetes; T2D, type 2 diabetes

To define the reciprocal effect of SE and DRB1*0301 on CII autoimmunity, type 1 diabetic individuals were divided into four groups according to the presence/absence of SE and DRB1*0301 on both alleles at the DRB1 locus (SE+/DRB1*0301 +; SE+/DRB1*0301 ; SE/DRB1*0301 +; and SE/DRB1*0301 ). The strongest association was between HLA-DRB1 SE and ROS-CII in the absence of DRB1*0301. This association, however, disappeared when SE and DRB1*0301 coexisted in the same individual (Fig. 3). In addition, DQB1 analysis, conducted on 98 individuals (data not shown), showed a positive association between DQB1*0302 and ROS-CII reactivity (OR 3.73 [95% CI 1.28, 10.91]), as expected from the high degree of linkage disequilibrium with the DRB1*04 alleles (DQB1*0302 was carried by 100% of our DRB1*04-positive patients). On the other hand, native and ROS-BSA were not associated with any specific HLA super type or allele.
Fig. 3

Association between HLA-DRB1 SE and DRB1*0301 alleles with autoimmunity to native CII and ROS-CII in type 1 diabetic individuals. The association between HLA-DRB1 SE and native CII autoimmunity was modest but became high when CII was chemically modified by reactive oxidants and in the absence of DRB1*0301 (p = 0.01; OR 5.5 [95% CI 1.39, 21.64]). HLA-DRB1*0301 contrasted the effect of SE (OR 0.92 [95% CI 0.15, 5.50]) for patients carrying DRB1*0301 and SE, as well as protective against ROS-CII autoimmunity in the absence of SE (p = 0.02; OR 0.18 [95% CI 0.05, 0.75]). SE refers to at least one DRB1*0401, *0404 or *0405 allele. Black bars, native CII; white bars, ROS-CII

Absence of relationship between metabolic control and antibody binding to ROS-CII and ROS-BSA

Binding to native CII and ROS-CII, as well as native BSA and ROS-BSA, were not correlated with indices of metabolic control, such as fasting plasma glucose and HbA1c (−0.22851 < ρ < 0.236, Fig. 4), nor to fasting C-peptide or insulin requirement (not shown). Although not significantly related to any other clinically relevant variables, autoimmunity to ROS-CII tended to be more prevalent among females (p = 0.079).
Fig. 4

Absence of relationship between metabolic control and antibody binding to ROS-CII in type 1 diabetes DRB1*04-positive (a, c) and DRB1*04-negative (b, d) individuals. Binding of samples from type 1 diabetic individuals to native CII and ROS-CII was not related to indices of metabolic control, such as fasting plasma glucose and HbA1c (−0.228 < ρ < 0.236). White squares, ROS-CII; black circles, native CII; black square, overlapping of ROS-CII and native CII. AU, arbitrary units

Longitudinal evaluation of antibody binding to ROS-CII and ROS-BSA in type 1 diabetes

Change in binding to native and ROS-CII was evaluated in 34 recent-onset type 1 diabetes individuals (21 DRB1*04 positive and 13 DRB1*04 negative) over a time frame of 9 months. Positive and negative seroconversion rate to ROS-CII, but not to native CII, was high in DRB1*04-positive type 1 diabetic individuals, with significant changes in ROS-CII antibodies in 8/21 (38%) individuals. In the DRB1*04-positive group, 24% of individuals who were negative to ROS-CII at the first evaluation became strong binders after 9–12 months’ follow-up, while 14% of those who were positive to ROS-CII became low binders; 19% of individuals remained positive to ROS-CII antibodies during the time frame and 43% remained negative. No relevant changes in ROS-CII reactivity were detected in the DRB1*04-negative group (Fig. 5). Change in binding was not related to any variable of metabolic control, such as blood glucose, HbA1c, C-peptide or insulin requirement. No significant changes in native or ROS-BSA reactivity were detected in the DRB1*04-positive and DRB1*04-negative groups, although both native BSA and ROS-BSA tended to be lower after 9 months in the DRB1*04-negative group (data not shown, p = 0.08 for both).
Fig. 5

Longitudinal evaluation of antibody binding to native- (a, c) and ROS-CII (b, d) in type 1 diabetes. Change in binding to native CII and ROS-CII was evaluated in 34 individuals with recent onset of type 1 diabetes (21 DRB1*04 positive [a, b] and 13 DRB1*04 negative [c, d]) within 6 months from the diagnosis (time 0) and 9–12 months after the first evaluation (time 1). In contrast to native CII (a), positive and negative seroconversion rates to ROS-CII (b) were high in DRB1*04-positive type 1 diabetic individuals. Of DRB1*04-positive individuals who were negative to ROS-CII at the first evaluation, 5/21 (24%) became strong binders after 9–12 months’ follow-up, while 3/21 (14%) of those who were positive to ROS-CII became low binders. No relevant changes in native CII (c) and ROS-CII (d) reactivity were detected in the DRB1*04-negative group. AU, arbitrary units


In the present study we reported on the combined effect of oxidative stress and HLA genotype in generating autoimmunity to ROS-CII in type 1 diabetes. We demonstrated for the first time that the exposure to reactive oxidants in diabetes can elicit autoimmune reactivity against ROS-CII specifically in type 1 diabetic patients, irrespective of clinical or biochemical features. In contrast to the specific association of ROS-CII autoreactivity with type 1 diabetes, antibodies to BSA were detected in both type 1 and type 2 diabetes regardless of the genotype, and hence were significantly reduced by ROS modification.

We found that autoimmunity to ROS-CII was positively associated with DRB1*04, while DRB1*03 was protective against native CII and ROS-CII autoimmunity in the absence of DRB1*04. When autoimmune responses to CII and ROS-CII were analysed according to HLA-DRB1 phenotype, we found that ROS-CII but not native CII autoimmunity was restricted to SE-containing DRB1*04 alleles, known to confer the greatest risk for developing RA (OR 3–13) [20] and commonly associated with type 1 diabetes [1]. Besides the positive effect of the SE, we found that HLA-DRB1*03 was negatively associated with the presence of reactivity against modified CII. The effect of each allele was also nullified when the SE and DRB1*0301 co-existed within the same individual.

Our study may imply that autoimmune reactivity in type 1 diabetes is ubiquitous and not necessarily directed to the diseased tissue, the pancreas. Autoantibodies to CII and ROS-CII, a specific extracellular matrix protein of the articular cartilage, were common in type 1 diabetes, even in the absence of RA. This phenomenon has been previously demonstrated by the presence of anti-thyroid autoantibodies, such as thyroid peroxidase (TPO) antibodies, that develop in over 20% of individuals with type 1 diabetes, even in absence of clinical thyroid disease [22, 23, 24]. TPO is an enzyme specifically produced in the thyroid and is a major target of autoimmune thyroid diseases. The same has been shown in RA: for example anti-citrullinated peptide antibodies (ACPA) directed against peptides modified by a post-translational modification of arginine residues to citrulline. ACPA is a known serological test for RA diagnosis, and is detected several years before the onset of RA [25]. It was demonstrated that this group of antibodies respond to filaggrin, an epithelial cell protein not present in the synovium, which plays a key role in joint inflammation during RA pathogenesis [26]. Therefore, the presence of anti-ROS-CII antibodies might be part of the autoimmune response in type 1 diabetes that does not necessarily imply that those individuals will develop RA. Nevertheless, the similar genetic constrain of the anti-ROS-CII of both diseases may lead to the hypothesis that the linkage is due to the anti-ROS-CII autoimmunity. This, however, needs to be established by further longitudinal studies.

Our results may suggest that the DRB1-dependent autoimmune reactivity is associated with post-translationally modified CII neo-epitopes rather than with native non-modified CII epitope. This was previously demonstrated in RA using CII epitope 263–270 that needed to be glycosylated to breach the T cell tolerance in DRB1*0401 transgenic mice and RA patients carrying the DRB1 SE alleles [27]. Similarly, DRB1*04 alleles bind to insulin (A1–13 peptide) in individuals with type 1 diabetes, but they are able to trigger pathogenic T cells only when the presented peptide has been post-translationally modified [28]. Brorsson et al showed in type 1 diabetes that DRB1*0401 is functionally related (co-expressed, interacting or structurally similar) to genes and proteins involved in antigen processing and proteasome activity [29] which are potentially involved in the regulation of post-translational modifications of proteins [30]. The negative association of DRB1*0301 with the presence of reactivity against modified CII is reminiscent of recent findings by two large studies on RA patients, showing a contrasting effect of SE and HLA-DRB1*03 [31, 32]. In an RA population, HLA-DRB1*03 has been significantly associated with reduced titres of ACPA. These results highlight the complexity of the HLA system in controlling antibody production and immune response in general.

We did not find any correlation between antibodies to native CII or ROS-CII and levels of blood glucose or HbA1c, nor was change in ROS-CII level associated with change in metabolic or clinical features. This is not completely unexpected and other studies assessing different ROS-modified proteins have yielded variable results. For example, Vay et al found that antibodies against glycated human albumin were not associated with metabolic control or diabetes duration [33]. On the other hand, levels of oxidised LDL (ox-LDL) antibodies have been correlated with HbA1c levels and the presence of chronic complications [34]. The absence of a relationship between hyperglycaemia and ROS-autoimmunity in our study may be due to the fact that once the redox state has been unbalanced the oxidative reactions can continue for longer after the initial stimulus. In addition, it is possible that the relationship between antigen oxidation and autoantibody production is not continuous, rather it is an ‘all-or-none’ process. Accordingly, once hyperglycaemia and oxidative stress have triggered or amplified the immune response against ROS-CII, the presence of the native molecule might then be sufficient to sustain the immune response through a mechanism of epitope spreading [35]. Second, the majority of individuals with type 1 diabetes included in this study had relatively short disease duration presented with blood glucose control close to normal and a small inter-individual variation at the time of the analysis; therefore, it is possible that our sample size was too small for the detection of a significant association between dysglycaemia and antibody to ROS-CII. Finally, although hyperglycaemia leads to increased oxidative stress, neither blood glucose nor HbA1c are reliable indices of oxidative stress.

The longitudinal assessment of ROS-CII antibody binding also did not show any relationship between clinical or biochemical features and ROS-CII autoimmunity, despite the presence of relevant changes in antibody binding during the time frame of 9 months soon after the diagnosis of type 1 diabetes. Changes of ROS-CII binding were variable, with high rates of positive and inverse seroconversion. This is generally different from what has been described for diabetes-related autoantibodies, which are already detectable before the diagnosis and rarely appear afterwards [36, 37]. Conversely, autoantibodies directed to other organs, such as anti-endomysium antibodies, which are not specific to diabetes, can appear also after the diagnosis [38]. This suggests the possibility that ROS-CII autoantibodies belong to the ubiquitous autoimmune response in type 1 diabetes, exacerbated after the onset, and possibly are not primarily involved in the pathogenesis of diabetes. The presence of such antibodies in the long term may result in RA in the presence of susceptible genes. This, however, needs to be studied longitudinally in a large diabetes cohort.

Larger longitudinal follow-up studies of people with impaired fasting glucose in which anti-ROS-CII autoreactivity will be measured together with markers of oxidative stress, for example measuring levels of carboxymethyl-lysine and pentosidine in parallel, may better explain the link between oxidative stress and antibodies to ROS-modified autoantigen in type 1 diabetes. A similar approach was recently used to study the importance of carboxymethyl-lysine as an environmentally determined diabetes risk factor. Islet cell antibodies were measured in parallel to levels of carboxymethyl-lysine in a classic twin study that demonstrated that carboxymethyl-lysine is an environmentally determined diabetes risk factor, in addition to HLA genetic risk [39].

In conclusion, our finding supports the hypothesis that type 1 diabetes-related hyperglycaemia and oxidative stress might trigger neo-antigenicity to extrapancreatic self-antigens, as shown for CII in this study. This process is under the genetic control of the HLA system, highlighting the complex interplay between genetic susceptibility and non-genetic factors in the genesis of autoimmunity. However, larger genotyped cohorts and longer follow-up studies are required to establish this current proof of principle study for further understanding the role of oxidative stress in the pathogenesis of type 1 diabetes mellitus and the possible relevance of ROS-CII autoantibodies as a biomarker in type 1 diabetes mellitus and its association with the development of RA.



We would like to thank Università Campus Bio-Medico di Roma for the support of this study. Special thanks to P. Sasieni (Wolfson Institute of Preventive Medicine, Queen Mary, University of London, UK) for his help with the statistical analysis of our data, and L. Marandola and S. Zampetti (Internal Medicine and Medical Specialties, ‘Sapienza’ University of Rome, Italy) for the help with the HLA typing.


This study was supported by Università Campus Bio-Medico di Roma and ‘Il Circolo’ Scholarship to R. Strollo.

Duality of interest

The authors confirm there is no duality of interest associated with this manuscript.

Contribution statement

RS was responsible for the conception and design of the study, acquisition of data, analysis and interpretation of data, writing the manuscript and revising it critically for important intellectual content. PR, MS, CH, FP, NN, AP and RB made substantial contributions to the acquisition of data and revised the manuscript critically for important intellectual content. RL made a substantial contribution to the analysis of the data and critical revision of the manuscript for important intellectual content. PP made a substantial contribution to the conception and design of the study and critical revision of the manuscript for important intellectual content. AN made substantial contributions to the conception and design of the study, analysis and interpretation of data, writing of the manuscript and revising the manuscript critically for important intellectual content. All authors approved the final version.

RS and AN are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.


  1. 1.
    Bluestone JA, Herold K, Eisenbarth G (2010) Genetics, pathogenesis and clinical interventions in type 1 diabetes. Nature 464:1293–1300PubMedCrossRefGoogle Scholar
  2. 2.
    Concannon P, Rich SS, Nepom GT (2009) Genetics of type 1A diabetes. N Engl J Med 360:1646–1654PubMedCrossRefGoogle Scholar
  3. 3.
    Petrone A, Battelino T, Krzisnik C et al (2002) Similar incidence of type 1 diabetes in two ethnically different populations (Italy and Slovenia) is sustained by similar HLA susceptible/protective haplotype frequencies. Tissue Antigens 60:244–253PubMedCrossRefGoogle Scholar
  4. 4.
    Thomson G, Valdes AM, Noble JA et al (2007) Relative predispositional effects of HLA class II DRB1-DQB1 haplotypes and genotypes on type 1 diabetes: a meta-analysis. Tissue Antigens 70:110–127PubMedCrossRefGoogle Scholar
  5. 5.
    Pihoker C, Gilliam LK, Hampe CS, Lernmark A (2005) Autoantibodies in diabetes. Diabetes 54(Suppl 2):S52–S61PubMedCrossRefGoogle Scholar
  6. 6.
    West IC (2000) Radicals and oxidative stress in diabetes. Diabet Med 17:171–180PubMedCrossRefGoogle Scholar
  7. 7.
    Brownlee M (2001) Biochemistry and molecular cell biology of diabetic complications. Nature 414:813–820PubMedCrossRefGoogle Scholar
  8. 8.
    Griffiths HR (2008) Is the generation of neo-antigenic determinants by free radicals central to the development of autoimmune rheumatoid disease? Autoimmun Rev 7:544–549PubMedCrossRefGoogle Scholar
  9. 9.
    Trigwell SM, Radford PM, Page SR et al (2001) Islet glutamic acid decarboxylase modified by reactive oxygen species is recognized by antibodies from patients with type 1 diabetes mellitus. Clin Exp Immunol 126:242–249PubMedCrossRefGoogle Scholar
  10. 10.
    Khan MW, Banga K, Mashal SN, Khan WA (2011) Detection of autoantibodies against reactive oxygen species modified glutamic acid decarboxylase-65 in type 1 diabetes associated complications. BMC Immunol 12:19PubMedCrossRefGoogle Scholar
  11. 11.
    Khan MW, Sherwani S, Khan WA, Ali R (2009) Characterization of hydroxyl radical modified GAD65: a potential autoantigen in type 1 diabetes. Autoimmunity 42:150–158PubMedCrossRefGoogle Scholar
  12. 12.
    Szkudelski T (2001) The mechanism of alloxan and streptozotocin action in B cells of the rat pancreas. Physiol Res 50:537–546PubMedGoogle Scholar
  13. 13.
    Kim WUCM, Jung YO, Min SY, Park SW, Min DJ, Yoon JH, Kim HY (2004) Type II collagen autoimmunity in rheumatoid arthritis. Am J Med Sci 327:202–211PubMedCrossRefGoogle Scholar
  14. 14.
    Kim WUYW, Park W, Kang YM, Kim SI, Park JH, Lee SS, Joo YS, Min JK, Hong YS, Lee SH, Park SH, Cho CS, Kim HY (2000) IgG antibodies to type II collagen reflect inflammatory activity in patients with rheumatoid arthritis. J Rheumatol 3:575–581Google Scholar
  15. 15.
    Nissim A, Winyard PG, Corrigall V et al (2005) Generation of neoantigenic epitopes after posttranslational modification of type II collagen by factors present within the inflamed joint. Arthritis Rheum 52:3829–3838PubMedCrossRefGoogle Scholar
  16. 16.
    Cooper GS, Bynum ML, Somers EC (2009) Recent insights in the epidemiology of autoimmune diseases: improved prevalence estimates and understanding of clustering of diseases. J Autoimmun 33:197–207PubMedCrossRefGoogle Scholar
  17. 17.
    Liao KP, Gunnarsson M, Kallberg H et al (2009) Specific association of type 1 diabetes mellitus with anti-cyclic citrullinated peptide-positive rheumatoid arthritis. Arthritis Rheum 60:653–660PubMedCrossRefGoogle Scholar
  18. 18.
    Newton JL, Harney SM, Wordsworth BP, Brown MA (2004) A review of the MHC genetics of rheumatoid arthritis. Genes Immun 5:151–157PubMedCrossRefGoogle Scholar
  19. 19.
    Petrone A, Giorgi G, Galgani A et al (2005) CT60 single nucleotide polymorphisms of the cytotoxic T-lymphocyte-associated antigen-4 gene region is associated with Graves' disease in an Italian population. Thyroid 15:232–238PubMedCrossRefGoogle Scholar
  20. 20.
    Imboden JB (2009) The immunopathogenesis of rheumatoid arthritis. Annu Rev Pathol 4:417–434PubMedCrossRefGoogle Scholar
  21. 21.
    Bland JM, Altman DG (2000) Statistics notes. The odds ratio. BMJ 320:1468PubMedCrossRefGoogle Scholar
  22. 22.
    Rosario PW, Reis JS, Borges MA et al (2005) Extrapancreatic autoimmunity in patients with latent autoimmune diabetes of adults. Diabetes Care 28:496–497PubMedCrossRefGoogle Scholar
  23. 23.
    Murao S, Kondo S, Ohashi J et al (2008) Anti-thyroid peroxidase antibody, IA-2 antibody, and fasting C-peptide levels predict beta cell failure in patients with latent autoimmune diabetes in adults (LADA)–a 5-year follow-up of the Ehime study. Diabetes Res Clin Pract 80:114–121PubMedCrossRefGoogle Scholar
  24. 24.
    Triolo TM, Armstrong TK, McFann K et al (2011) Additional autoimmune disease found in 33% of patients at type 1 diabetes onset. Diabetes Care 34:1211–1213PubMedCrossRefGoogle Scholar
  25. 25.
    Nielen MM, van Schaardenburg D, Reesink HW et al (2004) Specific autoantibodies precede the symptoms of rheumatoid arthritis: a study of serial measurements in blood donors. Arthritis Rheum 50:380–386PubMedCrossRefGoogle Scholar
  26. 26.
    Baeten D, Peene I, Union A et al (2001) Specific presence of intracellular citrullinated proteins in rheumatoid arthritis synovium: relevance to antifilaggrin autoantibodies. Arthritis Rheum 44:2255–2262PubMedCrossRefGoogle Scholar
  27. 27.
    Backlund J, Carlsen S, Hoger T et al (2002) Predominant selection of T cells specific for the glycosylated collagen type II epitope (263-270) in humanized transgenic mice and in rheumatoid arthritis. Proc Natl Acad Sci USA 99:9960–9965PubMedCrossRefGoogle Scholar
  28. 28.
    Mannering SI, Harrison LC, Williamson NA et al (2005) The insulin A-chain epitope recognized by human T cells is posttranslationally modified. J Exp Med 202:1191–1197PubMedCrossRefGoogle Scholar
  29. 29.
    Brorsson C, Tue Hansen N, Bergholdt R, Brunak S, Pociot F (2010) The type 1 diabetes-HLA susceptibility interactome—identification of HLA genotype-specific disease genes for type 1 diabetes. PLoS One 5:e9576PubMedCrossRefGoogle Scholar
  30. 30.
    Queisser MA, Yao D, Geisler S et al (2010) Hyperglycemia impairs proteasome function by methylglyoxal. Diabetes 59:670–678PubMedCrossRefGoogle Scholar
  31. 31.
    Irigoyen P, Lee AT, Wener MH et al (2005) Regulation of anti-cyclic citrullinated peptide antibodies in rheumatoid arthritis: contrasting effects of HLA-DR3 and the shared epitope alleles. Arthritis Rheum 52:3813–3818PubMedCrossRefGoogle Scholar
  32. 32.
    Verpoort KN, van Gaalen FA, van der Helm-van Mil AH et al (2005) Association of HLA-DR3 with anti-cyclic citrullinated peptide antibody-negative rheumatoid arthritis. Arthritis Rheum 52:3058–3062PubMedCrossRefGoogle Scholar
  33. 33.
    Vay D, Vidali M, Allochis G et al (2000) Antibodies against advanced glycation end product Nepsilon-(carboxymethyl)lysine in healthy controls and diabetic patients. Diabetologia 43:1385–1388PubMedCrossRefGoogle Scholar
  34. 34.
    Festa A, Kopp HP, Schernthaner G, Menzel EJ (1998) Autoantibodies to oxidised low density lipoproteins in IDDM are inversely related to metabolic control and microvascular complications. Diabetologia 41:350–356PubMedCrossRefGoogle Scholar
  35. 35.
    Doyle HA, Mamula MJ (2002) Posttranslational protein modifications: new flavors in the menu of autoantigens. Curr Opin Rheumatol 14:244–249PubMedCrossRefGoogle Scholar
  36. 36.
    Lipponen K, Gombos Z, Kiviniemi M et al (2010) Effect of HLA class I and class II alleles on progression from autoantibody positivity to overt type 1 diabetes in children with risk-associated class II genotypes. Diabetes 59:3253–3256PubMedCrossRefGoogle Scholar
  37. 37.
    Knip M, Korhonen S, Kulmala P et al (2010) Prediction of type 1 diabetes in the general population. Diabetes Care 33:1206–1212PubMedCrossRefGoogle Scholar
  38. 38.
    Barera G, Bonfanti R, Viscardi M et al (2002) Occurrence of celiac disease after onset of type 1 diabetes: a 6-year prospective longitudinal study. Pediatrics 109:833–838PubMedCrossRefGoogle Scholar
  39. 39.
    Beyan H, Riese H, Hawa MI et al (2012) Glycotoxin and autoantibodies are additive environmentally determined predictors of type 1 diabetes: a twin and population study. Diabetes 61:1192–1198PubMedCrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • R. Strollo
    • 1
    • 2
  • P. Rizzo
    • 1
    • 3
  • M. Spoletini
    • 4
  • R. Landy
    • 5
  • C. Hughes
    • 1
  • F. Ponchel
    • 6
  • N. Napoli
    • 2
  • A. Palermo
    • 2
  • R. Buzzetti
    • 4
  • P. Pozzilli
    • 2
    • 7
  • A. Nissim
    • 1
    Email author
  1. 1.Bone and Joint Research Unit, William Harvey Research Institute, Barts and The London School of Medicine and DentistryQueen Mary, University of LondonLondonUK
  2. 2.Department of Endocrinology and DiabetesCampus Bio-Medico University of RomeRomeItaly
  3. 3.Diabetes Care Unit, Internal MedicineCatholic UniversityRomeItaly
  4. 4.Department of Internal Medicine and Medical Specialties‘Sapienza’ University of RomeRomeItaly
  5. 5.Centre for Cancer Prevention, Wolfson Institute of Preventive MedicineQueen Mary, University of LondonLondonUK
  6. 6.Division of Musculoskeletal disease and NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds Institute of Molecular MedicineUniversity of Leeds, Chapel Allerton HospitalLeedsUK
  7. 7.Centre for DiabetesQueen Mary, University of LondonLondonUK

Personalised recommendations