Serum titres of anti-glutamic acid decarboxylase-65 and anti-IA-2 autoantibodies are associated with different immunoregulatory milieu in newly diagnosed type 1 diabetes patients


M. Andrade Lima Gabbay, Diabetes Center of Endocrinology Division from Department of Medicine of Federal University of São Paulo, Rua Pedro de Toledo 781/12, CEP: 04039-032, São Paulo, SP, Brazil. E-mail:


Several studies correlated genetic background and pancreatic islet-cell autoantibody status (type and number) in type 1A diabetes mellitus (T1AD), but there are no data evaluating the relationship among these markers with serum cytokines, regulatory T cells and β cell function. This characterization has a potential importance with regard to T1AD patients' stratification and follow-up in therapeutic prevention. In this study we showed that peripheral sera cytokines [interleukin (IL)-12, IL-6, II-1β, tumour necrosis factor (TNF)-α, IL-10] and chemokines (CXCL10, CXCL8, CXCL9, CCL2) measured were significantly higher in newly diagnosed T1AD patients when compared to healthy controls (P < 0·001). Among T1AD, we found a positive correlation between CXCL10 and CCL-2 (r = 0·80; P = 0·000), IL-8 and TNF-α (r = 0·60; P = 0·000); IL-8 and IL-12 (r = 0·57; P = 0·001) and TNF-α and IL-12 (r = 0·93; P = 0·000). Glutamic acid decarboxylase-65 (GAD-65) autoantibodies (GADA) were associated negatively with CXCL10 (r = −0·45; P = 0·011) and CCL2 (r = −0·65; P = 0·000), while IA-2A showed a negative correlation with IL-10 (r = −0·38; P = 0·027). Human leucocyte antigen (HLA) DR3, DR4 or DR3/DR4 and PTPN22 polymorphism did not show any association with pancreatic islet cell antibodies or cytokines studied. In summary, our results revealed that T1AD have a proinflammatory cytokine profile compared to healthy controls and that IA-2A sera titres seem to be associated with a more inflammatory peripheral cytokine/chemokine profile than GADA. A confirmation of these data in the pre-T1AD phase could help to explain the mechanistic of the well-known role of IA-2A as a more specific marker of beta-cell damage than GADA during the natural history of T1AD.


In human type 1A diabetes (T1AD), β cell destruction is related to autoreactive CD4+ helper and CD8+ cytotoxic T cells in response to at least four major antigens [insulin derivates, glutamic acid decarboxylase-65 (GAD-65), tyrosine phosphatase (IA2) and zinc transporter-8 (ZnT8)] present in the pancreatic β cells [1,2]. Antibodies against these cellular components enriched in islets are good markers for T1AD because they usefully explain the antibody-mediated tissue-specific inflammation in this disease. Preclinical follow-up studies in first-degree relatives of T1AD individuals have shown that insulinoma-associated antigen-2 autoantibodies (IA-2A) are associated with rapid progression to clinical diabetes [3] and may be indicative of more aggressive autoimmunity to β cells. Since the clinical diagnosis of this disease, these patients have been characterized by low serum C-peptide and more insulin dose requirement [4]. Recently the same association has been suggested for ZnT8A [5]. Conversely, it has been reported that a cellular immune response to GAD65 is related inversely to its humoral immune response. Individuals who only express anti-GAD65 autoantibodies (GADA) at high titres may also not progress to overt diabetes [6].

The important genetic regulating elements of the disease evolution process are polymorphisms in human leucocyte antigen (HLA), cytotoxic T lymphocyte antigen 4 (CTLA-4) and lymphoid tyrosine phosphatase-LYP;1p13 (PTPN22) [7]. The first one strengthens the interaction between the antigen-presenting cell and the T cell, facilitating the development of T helper type 1 (Th1) cells [8], while the second is considered a negative regulator of T cell receptor (TCR) signalling. A mutation in the last (PTPN22) results in failure to delete autoreactive T cells or in an insufficient activity of regulatory T cells (Tregs) [9]. Two polymorphisms of the PTPN22 gene have been associated with disease progression (Arg620trp) [10] and low residual β cell function (1858T) [11].

Subsequent studies have attempted to identify cytokines and chemokines of the inflammatory process that are critical to the development of T1AD based on the knowledge that many autoimmune diseases are accompanied by high serum levels of inflammatory cytokine diseases such as thyroiditis, coeliac disease and Addison's disease. It has been assumed that Th1 and Th2 imbalance is an important factor for inflammation in these diseases. However, Th1/Th2 imbalance does not necessarily mean that Th1 is only proinflammatory and Th2 is only anti-inflammatory. Recent studies have shown variable cytokine and chemokine results when comparing patients to controls [12–14]. The systemic levels of circulating cytokines and chemokines may be a surrogate of the immunoregulatory milieu.

Few studies have evaluated the association among genetic markers, autoantibody status, immunoregulatory markers and disease activity in new-onset T1AD.

The genetic, antibody and cytokine profile characterization at the time of clinical T1AD manifestation may be important for stratification of immunological intervention studies as well as for prediction of the natural course of the disease after exogenous insulin treatment initiation.

The present investigation was performed in order to verify difference in circulating immunoregulatory markers between health and new-onset T1AD young individuals and whether GADA and IA2A sera titres were associated with different peripheral cytokines, with a focus on Th1 and Th2 responses in newly diagnosed T1AD patients.

Subjects and methods

Patients and health controls

Thirty-five newly diagnosed T1AD patients (American Diabetes Association standards, diagnosed within 3 months of enrolment) and 25 healthy individuals paired for sex, age and body mass index (BMI) were studied. Serum peripheral blood mononuclear cells (PBMC) (isolated by Ficoll-Paque density gradient centrifugation) were obtained from venous blood samples of controls and patients after an overnight fasting state. In the diabetic patients we collected a new blood sample 90 min after they ingested a standardized liquid meal (6 ml/kg of body weight of Sustagen®, maximum 360 ml; Mead Johnson, São Paulo, Brazil). None of the study participants had signs of acute infection, allergy or used anti-inflammatory drugs at the time of blood drawing.

Peripheral cytokines and chemokines

Cytokines and chemokines (pg/ml) from the undiluted serum of patients and controls were measured using inflammatory cytokine and chemokine bead array (CBA) kits (Becton Dickinson, San Diego, CA, USA). Briefly, latex particles with six different fluorescence intensities and coated with antibodies against specific cytokines were incubated with the samples. The particles were then labelled with a mixture of phycoerythrin (PE)-conjugated cytokine-specific antibodies and analysed by flow cytometry [fluorescence activated cell sorter (FACS)Calibur; Becton Dickinson]. Respective isotype control antibodies were also used.

The intra- and interassay coefficients of variation (CV) were, respectively, for interleukin (IL)-12 (4–6% and 5–8%), IL-6 (4–7% and 2–5%), IL-1β (5–6% and 6–10%), IL-8 (6–9% and 8–10%), IL-10 (8–13% and 4–7%), tumour necrosis factor (TNF)-α (8–11% and 8–13%), CXCL10 (10–11% and 8·7–11·5%), CXCL9 (3·5–5·3% and 3·3–9·8%), CXCL8 (3·8–13·9% and 5·2–15·9%) and CCL2 (3·4–8·1% and 6·9–9·2%), according to the manufacturer's instructions.

Analysis of CD4+CD25+forkhead box P3 (FoxP3+) (Treg)

Analysis of Tregs was performed on PBMC. Briefly, after washing with cold wash buffer, cells were stained with fluorescein isothiocyanate (FITC) anti-CD4 and PE-Cy5 anti-CD25 at 4°C for 20 min. Cells were then washed and fixed and permeabilized at 4°C for 30 min using the fix/permeabilization solution and incubated with PE anti-FoxP3 (eBioscience San Diego, CA, USA) for 30 min. Appropriate fluorochrome-conjugated isotype control antibodies were also used. Lymphocytes were gated based on forward-scatter/side-scatter (FSC/SSC) and Treg cell markers CD25 and FoxP3 were examined on cells within the CD4 gate. Samples were acquired on a BD FACSCalibur flow cytometer and data were analysed using FlowJo computer software (Tree Star, Ashland, OR, USA). A minimum of 105 events was acquired for each sample.

GAD65 and IA2 autoantibodies

The levels of GADA and IA-2A were determined by radioimmunoassay (RIA) (RSR Ltd, Cardiff, UK), with normal cut-off levels of 1·72 units and 0·97 units, respectively (99% of normal individuals of our laboratory data) [15]. Sensitivity for the GADA and IA-2A assays was 71% and 49% with a specificity of 100% and 99%, respectively, to detect T1AD.

Serum C-peptide and A1c

Serum C-peptide was measured using an immunofluorimetric assay (autoDefia, Turku, Finland) with a detection limit of 0·015 ng/ml. The intra-assay variation was 4·2% (to 0·52 and 6·11 ng/ml) and the interassay variation was 1·1% and 3·4% (to 0·52 ng/ml and 6·11 ng/ml, respectively).

The HbA1c in whole blood was measured by high-performance liquid chromatography (HPLC) (normal range 3·5–6·0%).

Genotyping for HLA and PTPN22

HLA class II genotyping was determined by conventional polymerase chain reaction (PCR)-sequence specific oligonucleotide hybridization technique (RSSO; Luminex, Canoga Park, CA, USA).

PTPN22 C1858T polymorphism (rs2476601) was detected by real-time PCR, using the Rotor-Gene Q 6 plex Platform (Qiagen, Valencia, CA, USA). Commercially available Taqman primers and probes for the PTPN22 C1858T polymorphism (Applied Biosystems®, Carlsbad, CA, USA) were used (C_16021387_20). Assays were performed with the Taqman Universal Master Mix (Applied Biosystems), with 50 ng of DNA per reaction. PCR conditions were performed as recommended by the manufacturer: initial denaturation at 95°C (15 min), followed by 40 denaturation cycles at 95°C (15 s), and a final annealing/extension cycle at 60°C (1 min).

Statistical analysis

Statistical analysis was performed with spss version 17 for Windows. Values were expressed as mean ± standard deviation (s.d.). Student's t-test was used to determine the significance of differences between two continuous variables and the non-parametric Mann–Whitney U-test was used to compare results obtained from the controls and T1AD groups. Cytokines and chemokines were also evaluated after logarithmic transformation of their levels to achieve a roughly normal distribution. A simple linear correlation analysis was performed using Pearson's method to assess the correlation between age, BMI, A1c, C-peptide and transformed cytokine, chemokine levels and Treg numbers. To investigate associations between cytokines, Tregs and C-peptide, a multivariate regression analysis was applied. Statistical significance was defined as P < 0·05.


Patient and healthy control groups

The patient and healthy control groups did not differ significantly in their mean age (13·0 ± 5·0 versus 13·6 ± 4·6 years, respectively), gender distribution (female: 42% versus 34%, respectively) or BMI (19·8 ± 3·3 versus 19·8 ± 2·7 kg/m2, respectively) (Table 1).

Table 1.  Clinical and laboratory characteristics of type 1A diabetes and control subjects included in the study.
Age at diagnosis in years; median (range)13·0 ± 5·013·6 ± 5·40·26
Body mass index (kg/m2)18·6 ± 2·619·8 ± 3·40·21
HbA1c (%)8·4 ± 2·25·4 ± 0·3<0·001
Fasting serum C-peptide (ng/ml)0·79 ± 0·111·37 ± 0·930·012

Serum cytokine and chemokine levels

The peripheral sera cytokines (IL-12, IL-6, IL-1β, TNF-α, IL-10) and chemokines (CXCL10, CXCL8, CXCL9, CCL2) measured were significantly higher in newly diagnosed T1AD patients when compared to healthy controls (P < 0·001) (Table 2). Among T1AD, we found a positive correlation between CXCL10 and CCL2 (r = 0·80; P = 0·000), IL-8 and TNF-α (r = 0·60; P = 0·000); IL-8 and IL-12 (r = 0·57; P =  0·001) and TNF-α and IL-12 (r = 0·93; P = 0·000) (Fig. 1).

Table 2.  Comparison between cytokines and chemokine sera levels from healthy control and newly diagnosed type 1A diabetes (T1AD).
  1. IL, interleukin; TNF, tumour necrosis factor.

IL-12 (pg/ml)0·18 ± 0·4817·0 ± 29·00·009
IL-6 (pg/ml)0·50 ± 0·652·0 ± 1·1<0·001
IL-1β (pg/ml)0·01 ± 0·011·0 ± 3·0<0·001
IL-8 (pg/ml)4·40 ± 3·9014·6 ± 19·6<0·001
IL-10 (pg/ml)0·52 ± 0·591·32 ± 0·59<0·001
TNF-α (pg/ml)0·11 ± 0·3813·8 ± 26·9<0·001
CXCL10 (pg/ml)32·0 ± 12·0120·7 ± 81·9<0·001
CXCL9 (pg/ml)43·8 ± 135·0105·3 ± 167·0<0·001
CXCL8 (pg/ml)5·62 ± 4·9118·1 ± 23·0<0·001
CCL2 (pg/ml)56·2 ± 42·8180·4 ± 89·2<0·001
Figure 1.

Correlation between serum cytokines and chemokines in new-onset type 1A diabetes and control group studied. (a) CXCL10 (pg/ml) versus CCL2 (pg/ml); n = 60, r = 0·80, P = 0·000. (b) Interleukin (IL)-8 (pg/ml) versus tumour necrosis factor (TNF) (pg/ml); n = 60, r = 0·60, P = 0·000. (c) IL-8 (pg/ml) versus IL-12 (pg/ml); n = 60, r = 0·57, P = 0·001. (d) TNF-α (pg/ml) versus IL-12 (pg/ml); n = 60, r = 0·93, P = 0·000.

Treg cells

We did not find any significant difference in Treg cell numbers between T1AD and healthy controls (Figs 2 and 3) or any correlation between GADA, IA-2A titres and Treg cells.

Figure 2.

A representative flow cytometry plots CD4+forkhead box P2 (FoxP3)+-labelled peripheral blood mononuclear cells in one control (a) and one type 1A diabetes (T1AD) (b) individual studied.

Figure 3.

Values of regulatory T cell levels measured in the peripheral blood of each patient for the type 1A diabetes (T1AD) or control group studied.

GADA and IA-2A serum levels

All healthy controls were negative for GADA and IA-2A. The GADA and IA-2A mean serum levels in the new-onset diagnosed T1AD were, respectively, 19·3 ± 28·8 U/ml and 5·7 ± 9·3 U/ml. Of the 35 newly diagnosed T1AD subjects, 28 (50% female) had both positive GADA and IA-2A, while seven (15% female) had only one AAb (four for GADA and three for IA-2A).

Association of antibody with cytokines/chemokines and β cell function

In the T1AD group, we found a negative correlation between GADA and CXCL10 sera levels (r = −0·452; P = 0·011) and GADA and CCL2 (r = −0·656; P = 0·000) (Fig. 4).

Figure 4.

Correlation between humoral response [glutamic acid decarboxylase-65 (GAD-65) or IA2 antibodies] and peripheral sera levels of cytokine [interleukin (IL-10)] and chemokine (CXCL10 and CCL2) in type 1A diabetes (T1AD) patients studied. (a) GAD65 autoantibodies (GADA) (U/ml) versus CXCL10 (pg/ml); n = 35, r = −0·45, P = 0·011. (b) GADA (U/ml) versus CCL2 (pg/ml); n = 35, r = −0·65, P = 0·000. (c) IA2A (U/ml) versus IL-10 (pg/ml); n = 35, r = −0·38, P = 0·027.

It is interesting that IA-2A titres showed a negative correlation with IL-10 sera levels (r = −0·385; P = 0·027) (Fig. 4). Because we found that IL-10 sera was associated negatively with A1c (r = −0·441; P = 0·040) in the healthy control group, we adjusted IL-10 for A1c and observed that IA-2A sera levels were still correlated negatively with IL-10 (r = −0·442; P = 0·011).

There was a significant positive correlation between fasting or stimulated serum C-peptide (r = 0·716; P = 0·000). However, neither GADA nor IA-2A sera levels showed association with fasting or stimulated serum C-peptide.

Genes (HLA and PTPN 22) and autoantibodies (GADA and IA-2A)

The prevalence of high-risk HLA-DQ genotypes in our patients was preferentially HLA-DR4-DQ8 (45%), followed by HLA-DR3-DQ2 (29%), and 19·5% showed both genotypes. Conversely, the frequency of PTPN22 (C1858T) polymorphism (CT) was found in 57% of the patients. There was no significant correlation between the prevalence of the HLA phenotype (DR3, DR4, DR3/DR4 and others) or PTPN22 gene polymorphism (1858T allele) with GADA or IA-2A levels.


The peripheral levels of cytokines and chemokines may be a reflex of the immunoregulatory milieu, and in this study we observed that newly diagnosed T1AD patients had elevated circulating pro- as well anti-inflammatory cytokines and chemokines compared to healthy controls.

Specifically, in these T1AD patient within the first 6 months of clinical diagnosis the GADA and IA-2A serum levels correlated differently with systemic cytokine/chemokine concentrations. While we found that GADA correlated negatively with CXCL10 and CCL2, IA-2A titres were correlated negatively with circulating IL-10 levels.

The pancreatic AAb have been used as marker of T1AD diagnosis, but the mechanistic role has been discussed recently [16].

The inflammatory cytokine profile in T1AD individuals was similar to that found in other autoimmune diseases such as rheumatoid arthritis, systemic lupus erythematosus, psoriasis and multiple sclerosis, where the inflammatory cytokines and Th1 cells appeared to be a common modulator of the adaptive immune response towards self-tissue in disease [17]. However, peripheral blood cytokines might not represent the microenvironment of the regional lymph nodes or pancreas, and conflicting results regarding immune mediators in T1AD have already been described (Th1, Th1/Th3 or both Th1 and Th2 cytokine profiles) [15,18–20].

There is no one clear consensus and the relationship between those peripheral cytokine levels and pancreatic AAb status in T1AD is not well studied [21]. Multiple islet-cell AAb positivity has been proposed as a predictor of the development of disease and is associated with a faster loss of β cell function [3,22,23]. Additionally, a particular AAb type or titre can also influence this prediction [24]. Recently, the Diabetes Prevention Trial (DPT-1) demonstrated that those first-degree relatives who present with IAA ≥ 80 nU/ml might gain some benefit from oral insulin treatment, demonstrating a link between AAb titres and patient selection for intervention [25]. Whether the isolated AAb titre should be considered an additional marker of the natural T1AD history is not known.

Among pancreatic AAb, IA-2A titre had been demonstrated to be a particularly strong prediction of T1AD even in the absence of GADA and insulin autoantibody while GADA, in contrast, in the absence of IA-2A was associated with a relatively low risk or slow progression of this disease [26,27]. Furthermore, in the model of islet transplantation, IA-2A is activated only when there is clear evidence of autoimmunity, and studies of diabetic patients treated with cyclosporin have shown that the IA-2A-positive group were more resistant to immunosuppression, supporting the hypothesis that IA-2A is a marker of active β cell lesions [28,29]. In spite of considering that positivity for a single AAb usually reflects harmless, non-progressive β cell autoimmunity, isolated IA-2A positivity was more common among newly diagnosed T1AD than isolated IAA or GADA and was associated with the strongest HLA allele of susceptibility. This might indicate that IA-2A is associated with the progressive disease process [30]. It was also demonstrated recently for ZnT8A, which showed an increased T1AD risk in single AAb-positive relatives [31].

In our group of new-onset T1AD we found that the GADA titres were about three times higher than in IA-2A.

Considering that IA-2A may be more a specific marker of active disease in patients with newly diagnosed T1AD, we found a negative association between IA-2A and IL-10. This cytokine is considered to have a potent anti-inflammatory and immunomodulatory property [32], is associated with recovery of β cell function in T1AD patients [33] and also presents a positive correlation with insulin sensitivity in healthy individuals [34]. Moreover, this cytokine has been considered valuable in monitoring the outcome of immunosuppressant trials in T1AD [28].

However, we did not find any correlation between either IA-2 or GADA and Tregs in spite of these regulatory cells being able to secrete IL-10. One possible explanation is that we measure Tregs on peripheral blood and not at the site of autoimmune inflammatory process, and we did not access their functional capacity. Most data reported suggested no differences in the peripheral blood frequency of these cells between T1AD and controls subjects, but less suppressive activity of Tregs or autoreative T cells become resistant to suppression [35]. Notably, we do not know if the IL-10/IA-2A relationship found in our T1AD patients is the same during the preclinical phase of this disease.

In contrast, we found an inverse correlation of GADA sera titres with periphery sera CXCL10 levels. It is well known that GADA titres are not able to stratify diabetes risk and are associated with a slow progression to diabetes [36], while CXCL10 is considered an IFN-γ-induced chemokine expressed almost exclusively by distressed β cells, and not only recruits Th1 cells to the inflamed tissue but also drives T cell proliferation to antigenic stimulation [17,37,38]. Interestingly, we also found that this chemokine was associated negatively with C-peptide, suggesting that at the time of T1AD diagnosis there is a relationship between the inflammation pattern and residual β cell function. Supporting this idea, it was demonstrated in newly diagnosed T1AD that low levels of IFN-γ (which induces CXCL10) are associated with later remission [12] and in high-risk first-degree relatives the spontaneous secretion of IFN-γ showed an inverse relation with GADA [39]. Conversely, CCL2, also known as monocyte chemoattractant protein-1 (MCP-1), produced in response to inflammatory stimuli to promote the recruitment of monocytes, macrophages, dendritic cells, and activated T cells at nanomolar levels, is a key inflammatory chemokine while at picomolar levels it can exert global suppressive effects on T cell trafficking to inflamed lymph nodes.

Taken together, the negative correlation of GADA with CXCL10 and CCL2 could represent a less aggressive profile.

A recent study has shown an association among PTPN22 polymorphisms, T1AD HLA risk and GADA [9,40]. We observed that 94% of our patients presented the high-risk DRB1*03 and *04 haplotypes, similar to what was also found in a later southeastern Brazilian study [41], supported by reports from several populations [42]. Fifty-seven per cent of our patients also carried the PTPN22 CT genotype, with no TT homozygotes. The genotypic frequencies observed in the present study of a Brazilian population sample are similar to those described for T1AD in other populations [43]. In that sample of new-onset T1AD of our study, we did not find any association between GADA and IA2-A levels with HLA risk alleles or PTPN22 polymorphisms. These data are in accordance with a recent study of a large sample of child-onset T1AD (2500 individuals) that did not show any evidence of association of either GADA or IA-2A with the highly type 1 diabetes-predisposing genotype, HLA-DRB1*03/04 [44].

One of the main limitations in our study was that we analysed very few patients and our data were based on measurements performed exclusively in the peripheral blood, and not at the site of local inflammation. Nevertheless, we attempted to identify any relation between genetic, immunomediators and islet-cell AAb at the time of T1AD diagnosis.

In summary, our results revealed that T1AD have a proinflammatory cytokine profile compared to healthy controls and that IA-2A sera titres seem to be associated with a more inflammatory peripheral cytokine/chemokine profile than GADA. Confirmation of these data in pre-T1AD phase could help to explain the mechanistic of the well-known role of IA-2A as a more specific marker of β cell damage than GADA during the natural history of T1AD.


This study was supported by FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo) and CAPES-PROEX (Coordenadoria de Aperfeiçoamento do Pessoal de Ensino Superior- Programas de Excelencia)- Ministry of Education of Brazil. We are grateful to Soraya Ogusuku for help with cytokine and chemokine analyses, Felipe Crispim for DNA extraction, Walkiria L. Miranda and Aparecida Filomena P. F. Machado for AAb analyses, Maria do Rosário D. O. Latorre for statistical analyses, and all parents and patients who participated in the study.