Detection of four diabetes specific autoantibodies in a single radioimmunoassay: an innovative high-throughput approach for autoimmune diabetes screening

Authors


C. Tiberti, Department of Clinical Sciences, Policlinico Umberto I, University of Rome ‘Sapienza’, Viale del Policlinico 155, 00161 Rome, Italy. E-mail: claudio.tiberti@uniroma1.it

Summary

Highly sensitive and specific radioimmunoassays have been validated for autoantibodies reacting with the four major autoantigens identified so far in autoimmune diabetes. However, the analysis of this large number of autoantigens has increased the costs and time necessary for complete autoantibody screenings. Our aim was to demonstrate that it is possible to detect the immunoreactivity against a combination of four different autoantigens by a single assay, this representing a rapid, low-cost first approach to evaluate humoral autoimmunity in diabetes. By using this novel multi-autoantigen radioimmunoassay (MAA), in subsequent steps we analysed 830 sera, 476 of known and 354 of unknown diabetes-specific immunoreactivity, collected from various groups of individuals including type 1 and type 2 diabetes patients, autoantibody-positive patients with a clinical diagnosis of type 2 diabetes (LADA), prediabetic subjects, individuals at risk to develop autoimmune diabetes, siblings of type 1 diabetic patients, coeliac patients and healthy control subjects. All sera reacting with one or more of the four autoantigens by single assays also resulted positive with MAA, as well as eight of 24 type 1 diabetic patients classified initially as autoantibody-negative at disease onset based on single autoantibody assays. In addition, MAA showed 92% sensitivity and 99% specificity by analysing 140 blinded sera from type 1 diabetic patients and control subjects provided in the 2010 Diabetes Autoantibody Standardization Program. MAA is the first combined method also able to evaluate, in addition to glutamic acid decarboxylase (GAD) and tyrosine phosphatase (IA)-2, insulin and islet beta-cell zinc cation efflux transporter (ZnT8) autoantibodies. It appears to be particularly appropriate as a first-line approach for large-scale population-based screenings of anti-islet autoimmunity.

Introduction

Autoimmune diabetes (AD) is a T cell-dependent disease due to the chronically progressive destruction of insulin-producing islet β-cells [1,2]. During the last decade, a variety of genetic and molecular studies on human population and animal models, in addition to the refinement of islet autoantibody radioimmunoassays, contributed strongly to understanding the natural history of this complex disease [3,4]. To date, quantitative determination of serum autoantibodies directed against four well-characterized islet proteins [insulin (Ins), glutamic acid decarboxylase (GAD), tyrosine phosphatase 2 (IA-2) and islet beta-cell zinc cation efflux transporter (ZnT8)] represents an extremely valuable tool both for prediction and in diagnosis of AD. Detection of serum autoantibodies may help to identify subjects at risk to develop the disease, to distinguish among the different forms of the disease and to monitor the preclinical and clinical phases of the disease [5–8]. The increasing number of characterized autoantigens has also unfortunately contributed to an increase in the costs and time necessary to complete a comprehensive autoantibody analysis. Consequently, a simplification of the autoantibody assay procedures would allow a reduction of time and costs necessary for large-scale studies of the humoral anti-islet autoimmune response. With this purpose, and following the hypothesis that detection of the immunoreactivity against a panel of different autoantigens combined into a single assay may be possible, we set up and tested the sensitivity and specificity of a novel multi-autoantigen radioimmunoassay (MAA), employing a large panel of sera originating from different cohorts of diabetic patients, of subjects at risk to develop the disease and of healthy controls.

Materials and methods

Single autoantibody measurements

Insulin autoantibodies (InsA).  IAA were measured using a competitive fluid-phase radioimmunoassay [9], with minor modifications [10]. IAA detected with this assay obtained 100% sensitivity and specificity at the fourth Immunology of Diabetes Workshop (IDW) insulin autoantibody proficiency.

GAD autoantibodies (GADA).  GADA were detected by a slightly modified radioimmunosassay (RIA) [11] utilizing a human recombinant full-length GAD65 cDNA provided by Dr Å. Lernmark (Lund University, Malmö, Sweden). The GAD cDNA was transcribed in vitro and translated in the presence of [35S]-methionine (Perkin-Elmer Italia, Monza, Italy) using the Sp6 TNT-coupled rabbit reticulocyte system (Promega Italia, Milan, Italy). The GADA assay obtained 80% sensitivity and 98% specificity at the Fourth Diabetes Autoantibody Standardization Program (DASP) held in 2007.

IA-2 autoantibodies (IA-2A).  IA-2A were detected by a slightly modified RIA [12] utilizing a human recombinant IA-2IC(a.a.605–979) cDNA provided by Dr E. Bonifacio (University of Dresden, Dresden, Germany). As for GADA, the IA-2 cDNA was transcribed in vitro and translated in the presence of [35S]-methionine using the Sp6 TNT-coupled rabbit reticulocyte system. The IA-2A autoantibody assay obtained 72% sensitivity and 99% specificity at the Fourth DASP.

ZnT8 autoantibodies (ZnT8A).  ZnT8A were detected according to a modification of standard fluid-phase RIA procedures for GADA and IA-2A, in which a human ZnT8 probe (pJH4·1) was substituted [6]. The pJH4·1 probe, provided by Dr J. Hutton (University of Colorado, Denver, CO, USA), is a synthetic molecule fusion that combines cytoplasmic ZnT8 C-terminal domains with an immunoglobulin Cγ3 hinge sequence in a single-chain construct. It accounts for a ZnT8 dimer containing the protein epitopes 325Trp and 325Arg. The ZnT8 cDNA was transcribed in vitro and translated in the presence of [35S]-methionine using the T7 TNT-coupled rabbit reticulocyte system. The ZnT8A assay obtained 51·0% sensitivity and 98% specificity at the Fourth DASP.

Multi-autoantigen assay (MAA)

Constructs utilized for MAA.  GAD65, IA-2IC(605–979) and ZnT8(pJH4·1) cDNAs were transcribed separately in vitro and translated in the presence of [35S]-methionine, as for single autoantibody measurements. 125I radiolabelled insulin was purchased from Adaltis Italia (Casalecchio di Reno, Italy).

Preliminary studies.  The serum volumes and radiolabelled autoantigen counts per minute (cpm) utilized in the MAA were established preliminarily by analysing, in multiple experiments, the immunoreactivity of different amounts of 12 serum samples (six diabetes-specific autoantibody-positive type 1 diabetic patients and six autoantibody-negative healthy subjects) with different amounts of each of the four radiolabelled autoantigens (range 15 000–50 000 cpm). Specifically, before setting up the optimal MAA conditions for GAD, IA-2 and ZnT8 autoantigens, the ratio of [125I] amount/serum volume able to discriminate more effectively between autoantibody-positive and -negative subjects was evaluated carefully. These preliminary studies on insulin immunoreactivity led to the possibility of providing, in addition to a MAA ‘global autoantibody index’ related to the presence of autoantibodies against four autoantigens, information on the specific presence of InsA in each serum investigated.

MAA method.  The combined immune response to Ins, GAD, IA-2 and ZnT8 proteins was detected by using a fluid-phase radioimmunoprecipitation assay. Each serum sample was tested in duplicate. Briefly, 125 µl of radiolabelled insulin was added to 25 µl of human serum. After 8 h of incubation at 4°C in a rotating platform, 50 µl of a phosphate-buffered solution containing 30 000 cpm of each single radiolabelled autoantigen ([35S]-GAD, [35S]-IA-2 and [35S]-ZnT8) was added. After incubation overnight at 4°C in a rotating platform, 50 µl of a 50% solution of protein A/G Sepharose (GE Healthcare, NJ, USA) (volume ratio 4:1) was used to separate free from antibody-bound labelled products. Several washings and centrifugations, aimed at eliminating non-specific radioactivity, were performed subsequently and residual pellets were transferred to a scintillation vial, resuspended in scintillation liquid (Ultima Gold, Perkin-Elmer Italia, Monza, Italy) and counted in a β-counter. Results were expressed as a Global Autoantibody Index (GAbI) calculated by using GAD/IA-2/ZnT8/Ins autoantibody-positive and GAD/IA-2/ZnT8/Ins autoantibody-negative reference standard sera according to the following formula: (mean cpm sample – mean cpm negative control)/(mean cpm positive control – mean cpm negative control). Sera from 158 GAD, insulin, IA-2 and ZnT8 autoantibody-negative healthy subjects with no history of diabetes (CTRL: 85 females, 73 males, age range 3·0–54·0 years, median age 15·1 years) was utilized to establish the MAA limit of positivity. The overall performance of MAA was evaluated in three further steps, analysing 318, 214 and 140 sera, respectively. In each step, sample sera from the following groups of patient and control sera were analysed.

Step 1

MAA analysis of 318 sera with known diabetes-specific immunoreactivity.  The characteristics of these individuals and the distribution of their autoantibody specificities are reported in Table 1a.

Table 1a.  Step 1: characteristics and distribution of autoantibody specificities in the 318 subjects investigated.
 No.Female/maleAge range (years)Median age (years)No. antibody+ (%)No. antibody- (%)No. GADA+ (%)No. IA-2A+ (%)No. InsA+ (%)ZnT8A+ (%)
  1. DM1: newly-diagnosed type 1 diabetic patients; PDM: prediabetic subjects; HR: autoantibody positive non-diabetic subjects at risk to develop type 1 diabetes; LADA: autoantibody-positive patients with a clinical diagnosis of type 2 diabetes; T2DM: type 2 diabetes mellitus patients; IA-2A: tyrosine phosphatase 2 autoantibodies; GADA: glutamic acid decarboxylase autoantibodies; InsA: insulin autoantibodies; ZnT8A: islet beta-cell zinc cation efflux transporter autoantibodies.

DM113653/830·7–69·213·0112 (82·4%)24 (17·6%)82 (60·3%)58 (42·6%)68 (50·0%)38 (27·9%)
PDM5226/261·0–49·210·146 (88·5%)6 (11·5%)43 (82·7%)30 (57·7%)27 (51·9%)12 (23·1%)
HR2912/174·0–63·113·529 (100%)025 (86·2%)11 (37·9%)9 (31·0%)9 (31·0%)
LADA2411/1325·5–71·547·524 (100%)023 (95·8%)10 (41·7%)8 (33·3%)4 (16·7%)
T2DM2513/1235·3–76·958·5025 (100%)0000
Neg Sib5228/242·3–29·310·0052 (100%)0000
  • • 136 newly diagnosed type 1 diabetic patients (DM1), representing a subset of 627 DM1 sera collected between 1990 and 2008 at ‘La Sapienza’ University of Rome. Of these DM1 patients, 100 were selected randomly among those positive for one or more diabetes-specific autoantibodies, 24 represented the entire cohort of DM1 patients which resulted autoantibody negative for GAD, Ins, IA-2 and ZnT8 autoantibodies, 12 were DM1 patients positive only for insulin autoantibodies, but negative for GAD, IA-2 and ZnT8 autoantibodies;
  • • 52 prediabetic subjects (PDM): samples collected during the preclinical phase of individuals that subsequently developed AD;
  • • 29 GAD and/or Ins and/or IA-2 and/or ZnT8 autoantibody-positive non-diabetic subjects at risk to develop type 1 diabetes (HR);
  • • 52 GAD, Ins, IA-2 and ZnT8 autoantibody-negative siblings of type 1 diabetic patients (NegSib);
  • • 24 GAD and/or Ins and/or IA-2 and/or ZnT8 autoantibody-positive patients with a clinical diagnosis of type 2 diabetes (LADA); and
  • • 25 GAD, Ins, IA-2 and ZnT8 autoantibody-negative type 2 diabetes patients (T2DM).

Step 2

Detection of MAA sensitivity and specificity by analysis of 214 sera of unknown diabetes-specific immunoreactivity.  The characteristics of the individuals investigated in this step are reported in Table 1b. At the time of MAA analysis, single diabetes-specific immunoreactivities of these sera were still unknown and investigated only in a second analysis.

Table 1b.  Step 2: characteristics of the 214 subjects investigated (at the time of multi-autoantigen radioimmunoassay (MAA) analysis, single diabetes-specific immunoreactivities of these sera were still unknown).
 No.Female/maleAge range (years)Median age (years)No. antibody+ (%)No. Antibody- (%)No. GADA+ (%)No. IA-2A+ (%)No. InsA+ (%)No. ZnT8A+ (%)
  1. DM1: newly-diagnosed type 1 diabetic patients; CRTL: healthy control subjects; CD: coeliac disease patients; GADA: glutamic acid decarboxylase autoantibodies; IA-2A: tyrosine phosphatase 2 autoantibodies; InsA: insulin autoantibodies; ZnT8A: islet beta-cell zinc cation efflux transporter autoantibodies.

CTRL11448/6619·0–68·043·00114 (100%)0000
DM15019/310·3–69·312·548 (96%)2 (4%)41 (82%)26 (52%)21 (42%)16 (32%)
CD5028/220·9–15·27·31 (2%)49 (98%)1 (2%)1 (2%)1 (2%)0
  • • 114 healthy subjects with no history for diabetes;
  • • 50 DM1 patients at disease diagnosis; and
  • • 50 coeliac disease (CD) patient sera at diagnosis.

All step 2 sera were collected between 2008 and 2010 at ‘La Sapienza’ University of Rome. Diabetes patients analysed in steps 1 and 2 were diagnosed according to American Diabetes Association Criteria [13].

Step 3

MAA analysis of DASP 2010 workshop sera.  Finally, in order to further evaluate specificity and sensitivity of the MAA method, we participated (as laboratory no. 155) in the 2010 DASP Workshop, analysing, blinded, 140 coded sera (50 from type 1 diabetic patients and 90 from control individuals).

Statistical analysis

Statistical analyses were performed using spss statistical software, version 13 (SPSS, Inc., Chicago, IL, USA). Data are expressed as frequencies or as means ± standard deviation or median values. Frequency differences were calculated by Fisher's exact test. A P-value <0·05 was considered significant.

Results

MAA limit of positivity

The limit of positivity of MAA, calculated according to the 99th percentile of 158 CTRL subjects, was 0·053. Intra- and interassay coefficients of variation of the assay were 6·3% and 9·2%, respectively.

Step 1

MAA analysis of 318 sera of known diabetes-specific immunoreactivity.  DM1, PDM, HR, NegSib, LADA and T2DM serum reactivities to single autoantigens are shown in Table 1a. Overall, 211 of 318 (66·4%) sera reacted with one or more of the four autoantigens investigated. Table 2 shows the 16 possible combinations of results virtually detectable by analysing the immunoreactivity of one serum with four distinct autoantigens. The table also shows the total number of sera per single combination observed in the study.

Table 2.  Patterns of autoantibody positivity in step 1 sera.
GADA+IA-2A+InsA+ZnT8A+No. antibodies+ 
  1. Sera from 318 HR, PD, DM1, LADA, T2DM and NegSib patients showed all 16 possible combinations of autoantibody immunoreactivity. IA-2A: tyrosine phosphatase 2 autoantibodies; InsA: insulin autoantibodies; GADA: glutamic acid decarboxylase autoantibodies; HR: autoantibody-positive subjects at risk to develop type 1 diabetes; PD: prediabetic subjects; LADA: autoantibody-positive patients with a clinical diagnosis of type 2 diabetes; T2DM: type 2 diabetes mellitus patients; ZnT8A: islet beta-cell zinc cation efflux transporter autoantibodies.

0107
+146
+14
+113
+14
++221
++223
++23
++26
++27
++22
+++333
+++313
+++311
+++32
++++423

As shown in Fig. 1, 100% of the 211 sera reacting in single assays with one or more of the four autoantigens were found positive using the MAA method, including all sera positive for one single autoantibody (n = 67), even when their positivity was at low titres (20 of 67). In addition, eight of 24 (33·3%) of DM1 patients found individually GAD, Ins, IA-2 and ZnT8-autoantibody-negative, were positive using the MAA method. The six PDM, 25 T2DM and 52 NegSib, found to be autoantibody-negative by single assays (Table 1a), also resulted negative using MAA. A significant correlation was found between the sum of the single GAD, Ins, IA-2 and ZnT8 autoantibody indices detected in each of the 317 subjects and the corresponding MAA GAbI (r = 0·752, P < 0·0001) (Fig. 2).

Figure 1.

Global Autoantibody Index (GAbI) according to the number of single autoantibody positivities. The y-axis indicates the GAbI detected in the various groups of patients investigated. The horizontal dashed line represents the GAbI limit of positivity of the multi-autoantigen radioimmunoassay (MAA) method. The x-axis indicates the various groups of patients investigated, subdivided according to the number of autoantibody positivities. HR: autoantibody-positive subjects at risk to develop type 1 diabetes; PD: prediabetic subjects; DM1: type 1 diabetes patients at disease onset; LA: autoantibody-positive patients with a clinical diagnosis of type 2 diabetes. The x-axis also indicates the median GAbI relative to the number of autoantibody positivities.

Figure 2.

Correlation between the sum of the single autoantibody indices and Global Autoantibody Index (GAbI). The quantitative correlation between the sum of the single glutamic acid decarboxylase (GADA), insulin (InsA), tyrosine phosphatase (IA-2A) and islet beta-cell zinc cation efflux transporter (ZnT8A) autoantibody indices and the corresponding multi-autoantigen radioimmunoassay (MAA) GAbI in 318 subjects.

Step 2

Detection of MAA sensitivity and specificity by analysis of 214 sera of non-known diabetes-specific immunoreactivity.  Results are shown in Fig. 3. Using the MAA, 114 of 114 healthy subject sera were found to be autoantibody-negative (MAA specificity 100%), whereas 48 of 50 DM1 patient sera were autoantibody-positive (MAA sensitivity 96·0%). Only one of 50 (2·0%) coeliac patients' sera was found to be MAA-positive.

Figure 3.

Multi-autoantigen radioimmunoassay (MAA) immunoreactivity in patients with or at risk to develop type 1 diabetes and in healthy controls. Sensitivity of the MAA method in 50 type 1 diabetes (DM1) and 50 coeliac patients at diagnosis (CD), and specificity in 114 healthy controls. The y-axis indicates the Global Autoantibody Index (GAbI) detected in the various groups of patients investigated. The horizontal line represents the GAbI limit of positivity of the MAA method. The x-axis indicates the various groups of patients investigated. Under each group of patients, in brackets, is indicated the median GAbI index.

Single autoantibody measurements

The subsequent study of single autoantigen immunoreactivities in 214 sera showed that none of the 114 healthy subject sera was positive for immunoreactivity to single autoantigens, whereas 48 of 50 (96·0%) DM1 had autoantibodies directed against at least one of the four autoantigens investigated. Specifically, 18, 11, 12 and seven DM1 patients were positive for one, two, three and four single autoantibodies, respectively. The two DM1 MAA-negative were also negative for each single autoantibody. Among DM1 sera, 82·0%, 52·0%, 42·0% and 32·0% were positive for GAD, IA-2, Ins and ZnT8 autoantibodies, respectively. The only CD patient serum found MAA-positive was negative only for ZnT8 autoantibodies.

Step 3

MAA specificity and sensitivity were evaluated further by participating in the 2010 DASP workshop and analysing, blinded, 140 coded sera. MAA positivity was detected in 46 of 50 DM1 sera and in one of 90 control sera, thus providing a 92% sensitivity and 98·9% specificity of such method (personal communication of the authors with the permission of DASP organizers, who confirm that the results are as given).

MAA and insulin autoantibodies.  As reported previously, the MAA method may also provide information on the specific presence of serum InsA. This is possible because insulin utilized in the assay is radiolabelled with [125I] and not with [35S] (as for GAD, IA-2IC and Znt8 proteins), thus giving the opportunity, in the final step of the assay, to count in a γ-counter the specific insulin autoantibody binding before the autoantibody-containing pellet is transferred to the scintillation vial. The limit of positivity of InsA (0·054), detected by using InsA-positive and InsA-negative reference standard sera, was calculated according to the 99th percentile of the same 158 subjects utilized for MAA. Intra- and interassay coefficients of variation of the assay were 7·9% and 11·4%, respectively. Figure 4 shows the immunoreactivity of 28 DM1 sera (with different patterns of autoantibody immunoreactivity) analysed according to the MAA procedure but using three distinct combinations of radiolabelled antigens [125I]-Ins only; [35S]-(GAD, IA-2IC and Znt8) only and [125I]-Ins/[35S]-(GAD, IA-2IC and Znt8). Of note, all 13 sera (12DM1 patients and one HR subject) positive for insulin, but negative for GAD, IA-2 and ZnT8 autoantibodies (Table 2), were found positive for InsA as well MAA. A significant correlation between InsA titres detected by competitive fluid-phase radioimmunoassay (CIAA) [11] and by cpm γ-counts of the MAA method was found (r = 0·678, P < 0·0001). Among DM1 sera analysed in step 1 of the study, we found three DM1 sera showing discordant results when analysed with the two different InsA assays (two samples low-positive with CIAA, but negative with InsA assay; one sample negative with CIAA, but low-positive with InsA assay). However, taking into account that the three DM1 discordant sera were also positive for other autoantibodies, MAA results indicated an autoantibody positivity, as expected.

Figure 4.

Immunoreactivity of 28 type 1 diabetes-related sera with different patterns of autoantibody positivity analysed according to the multi-autoantigen radioimmunoassay (MAA) procedure but using three distinct combinations of radiolabelled antigens: [125I]-Ins only [column (a)]; [35S]- glutamic acid decarboxylase (GAD), tyrosine phosphatase 2 (IA-2) and islet beta-cell zinc cation efflux transporter (ZnT8) only [column (b)] and [125I]-Ins/[35S]-(GAD, IA-2IC and ZnT8A) [column (c)]. Columns (a), (b) and (c) values are expressed as counts per minute (cpm) counted in β-counter, whereas column (d) values are related to the same data of column (a), but expressed as cpm counted in a γ-counter. Column (c) values (β cpm) are utilized to calculate GAbI index of the MAA method, whereas column (d) values (cpm γ) are utilized to calculate the insulin autoantibody (InsA) titre. The horizontal dashed lines represent the limit of positivity of the various assays.

Discussion

In the present study we demonstrated that it is possible, by utilizing a combination of four diabetes autoantigens into a single fluid-phase radioimmunoassay, to simplify, speed up and reduce remarkably the costs necessary for screening diabetes-associated humoral autoimmunity regarding separate detection of the four autoantibodies in a single assay. In the recent past, the idea of combining different diabetes autoantigens into a single detection attracted several investigators, and a number of assays have been developed over the years [14–21]. Lampasona et al. utilized three autoantigens, combining GAD and IA-2 in a single assay with tissue transglutaminase C, a CD-associated autoantigen, with the aim of identifying coeliac as well type 1 diabetic patients [14]. Some investigators measured combined GAD and IA-2 autoantibodies by using the two autoantigens labelled with the same radioisotope ([35S]-methionine) [15–17], others developed an assay with GAD and IA-2 radiolabelled with two distinct isotopes, [35S] and [3H], respectively [18]. Rickert et al. combined GAD and IA-2 into a fusion protein unifying the immunodominant epitopes of the two autoantigens [19]; a radioimmunoassay based on this principle achieved high sensitivity, specificity and the highest area under the receiver operating characteristic (ROC) curve overall in the first DASP proficiency evaluation [20]. Three laboratories reported results for GAD/IA-2 combined assays in DASP 2005; however, among them, only one enzyme-linked immunosorbent assay (ELISA) assay achieved high sensitivity and specificity [5]. More recently, Yu et al. analysed diabetes-related immunoreactivities by a chimeric autoantibody assay where IA-2 and a ZnT8 heterodimer were linked together to form a chimeric protein [21]. However, despite all these different methodological approaches, the assays reported so far have utilized not more than two diabetes-specific autoantigens, thus missing those patients positive only for insulin autoantibodies and/or for autoantibodies directed against other autoantigen(s) eventually not included in the assay. This bias probably represented one of the main reasons of the lack of utilization of combined assays in large-scale screening programmes aimed at evaluating risk assessment for autoimmune diabetes. The difficulty of including insulin autoantigen into a combined assay was due essentially to the relatively large amount of serum necessary for reliable insulin autoantibody detection. The recent description of novel micro-radiobinding assays able to detect insulin autoantibodies with lower amounts of serum [22,23], in addition to the introduction of ZnT8 as an additional diabetes-associated autoantigen [6], prompted us to investigate the possibility of using the four main diabetes autoantigens into a single fluid-phase radioimmunoassay. The new assay was applied to the analysis of DM1, LADA and individuals with ongoing islet autoimmunity who had not yet developed diabetes. All cases investigated and previously found individually to be serum GADA- and/or InsA- and/or IA-2A- and/or ZnT8A-positive, were confirmed positive by using the MAA method. These cases included the 16 possible combinations of autoantibody positivity detectable by analysing the immunoreactivity with four distinct autoantigens, including all the sera positive for one single autoantibody, even when their positivity was at low autoantibody titres. In addition, the MAA method was able to increase the number of DM1 patients with signs of islet autoimmunity, based on the results obtained with single autoantibody assays. This finding indicates that by combining various autoantigen constructs into a single assay it is possible to unmask, in addition to the known autoantibody positivities, additional borderline immunoreactivities otherwise not identifiable by single autoantigen detection. A significant correlation was found between the sum of the single GADA, InsA, IA-2A and ZnT8A indices detected in each patient and the corresponding MAA GAbI index. The strong sensitivity of MAA is accompanied by high specificity, as demonstrated by results found in autoantibody-negative DM1 siblings, T2DM patients and additional healthy subjects analysed in step 2 of the study. Such high specificity and sensitivity of MAA were demonstrated further by results obtained in the blinded analysis of 140 coded sera in the 2010 DASP workshop. MAA appears to be considerably versatile, as novel diabetes-associated autoantigens that may become available and/or various fragments of the same autoantigen might be potentially included in the assay, in addition to the four autoantigens utilized so far. Accordingly, we have demonstrated previously [12] that by combining into a single assay three different IA-2 fragments, it is possible to detect IA-2 autoantibodies in an additional 30% of prediabetes subjects, compared to the single IA-2 construct evaluation. This IA-2 combined assay resulted, in DASP 2009, as the most sensitive and specific method for detection of IA-2 autoantibodies. Another relevant peculiarity of the MAA method is that it may provide information not only on the global diabetes immunoreactivity of each serum analysed, but also on the specific presence of insulin autoantibodies. This distinction is possible because the insulin utilized in the multi-autoantigen assay is radiolabelled with [125I] (and not with [35S], such as GAD, IA-2 and ZnT8 proteins), thus being able to provide a separate gamma-counter cpm response relative to the MAA beta-counter cpm response. In this context, we showed that InsA results are not affected by the aspecific background of MAA, and that all patients single-positive for insulin autoantibodies were positive for InsA as well the MAA method, even in the cases at low-antibody titre. A significant correlation was found between InsA titres detected by a competitive fluid-phase radioimmunoassay and γ-counts of the MAA method. Despite the performance of our InsA assay, appearing to be very similar to those of the competitive RIA of reference (at least for the sera analysed in the present study), we do not claim that such an assay may be as sensitive as the competitive one. Nevertheless, we have shown here that (i) radiolabelled insulin indeed contributes to the total immune response detected by the MAA method; (ii) it is possible, through a simple cpm γ-count, to gain additional, inexpensive, non-time-consuming information on the eventual presence of insulin autoantibodies in the samples analysed. In this regard, we note that the MAA format is not intended to replace the analysis of single diabetes-specific humoral immunoreactivities; rather, we consider it as the first step of a two-step strategy aimed to contribute to a more correct classification of the various diabetes forms (i.e. type 1a/type 1b diabetes; LADA/type 2 diabetes) or for screening large serum sample collections with an expected low prevalence of positive sera (e.g. relatives of patients with type 1 diabetes; children from the general population). In such a setting, using the MAA, the first step will permit identification of the limited number of anti-islet autoantibody-positive samples. The second step, consisting in the determination of single-diabetes specific immunoreactivities, will be useful to define the presence of single autoantibody specificities and their titres. In this way expensive and labour-intensive laboratory testing will be confined only to a low percentage of the total samples.

Acknowledgements

This paper is dedicated to the memory of Edmondo and Cecilia Tiberti. This work was supported by the European Union (Collaborative Project NAIMIT in the Framework Program 7), the Italian Ministry of Research and the Italian Ministry of Health. The excellent technical assistance of Donata Masotti and Lucia Pallotta has been greatly appreciated.

Disclosure

None.

Ancillary