Identification of Tyrosine Phosphatase 2_(256–760) Construct as a New, Sensitive Marker for the Detection of Islet Autoimmunity in Type 2 Diabetic Patients: The Non–Insulin Requiring Autoimmune Diabetes (NIRAD) Study 2^*

During the Non–Insulin Requiring Autoimmune Diabetes (NIRAD) study, a nationwide survey supported by the Società Italiana di Diabetologia aimed at assessing prevalence and characteristics of adult autoimmune diabetes in Italy in patients attending diabetes clinics with a clinical diagnosis of type 2 diabetes, 4,250 type 2 diabetic patients with a disease duration <5 years, with no insulin requirement and no evidence of ketosis for at least 6 months from diagnosis, were enrolled and screened for GAD and IA-2_{IC (605–979)} autoantibodies. One hundred ninety-one (4.5%) and 39 (0.9%) type 2 diabetic patients were found to be GADA and IA-2_{IC (605–979)}A positive, respectively (18). Of the 4,250 NIRAD patients, 1,020 were collected at the “Sapienza” University of Rome (572 men and 406 women; age range 21.1–85.8 years [median 57.6 years]; mean disease duration 21.9 ± 18.0 months). Forty-two (4.1%) and 13 (1.3%) of these 1,020 patients were GADA and IA-2_{IC (605–979)}A positive, respectively, whereas 978 patients were GADA and IA-2_{IC (605–979)}A negative.

In the first step of the present study, we aimed at evaluating which IA-2 constructs among the seven studied here had the highest sensitivity to detect humoral IA-2 immunoreactivity in Italian LADA patients. To this end, we analyzed all GADA⁺ LADA patient sera available from the NIRAD study (n = 177: 92 men and 85 women; age range 23.7–86.6 years [median 54.3 years]; mean disease duration 24.0 ± 18.8 months).

In the second step of the study, the immunoreactivity of the IA-2_(256–760) construct, identified as the marker with the highest sensitivity for the detection of IA-2 autoantibodies in the group of GADA⁺ LADA patients, was analyzed in all abovementioned 1,020 samples (978 GADA⁻ and IA-2_ICA⁻ plus 42 GADA⁺ type 2 diabetic patients) screened at the “Sapienza” University of Rome during the NIRAD study.

In addition, among the abovementioned 1,020 patients, those found GADA or IA-2_(256–760)A positive were tested for thyroid peroxidase autoantibodies (TPO-As) and, when possible, typed for HLA DRB1-DQB1 polymorphisms; 114 (for TPO-A) and 64 (for HLA polymorphisms) GADA⁻/IA-2_(256–760)A⁻ type 2 diabetic patients of comparable age and sex were analyzed as well.

Finally, in another set of experiments, we compared the IA-2 epitope target domains recognized at disease diagnosis by LADA, adult-onset, and childhood-onset type 1 diabetic sera. To this end, by using the seven constructs represented in Fig. 1, the following groups of patients were studied at disease diagnosis: 33 GADA⁺ LADA patients (22 men and 11 women; median age 44.8 years; age range 25–73 years); 53 GADA⁺ type 1 diabetic children (28 boys and 25 girls; median age 7.6 years; age range 2–12 years); and 53 GADA⁺ type 1 diabetic adults (29 men and 24 women; median age 29.0 years; age range 18–48 years).

The 106 type 1 diabetic patients are part of 537 newly diagnosed patients recruited between 1990 and 2004 at “Sapienza” University of Rome. The 53 type 1 diabetic children represent all GADA⁺ patients aged <12 years consecutively recruited between 2001 and 2004, whereas the 53 type 1 diabetic adults represent all GADA⁺ patients aged >18 years available in this cohort of patients. All type 1 and type 2 diabetic patients analyzed in the present study were diagnosed according to American Diabetes Association criteria (19).

cDNAs encoding IA-2_(761–964) were amplified by PCR from full-length IA-2 using 5′-ACCATGAGCGATTACATCAACGCCA-3′ and 5′-TCAGCAGCTACAGTCAGAATT-3′ primers. Ligation and transformation of fresh PCR products were performed as previously described for IA-2_(256–760) construct (16). After purification, inserts were sequenced in both directions using an ABI 377 sequencer (Applied Biosystems, Foster City, CA). No deletions or truncations were found in this IA-2 construct. IA-2_BDC was prepared as reported previously (20) and was provided, as the IA-2_(256–760) construct, by Dr. G.S. Eisenbarth (University of Colorado, Denver, CO). IA-2_{FL (1–979)}, IA-2_{IC (605–979)}, IA-2_{JM (601–630)}, and IA-2_{PTP (687–979)} cDNAs were provided by Dr. E. Bonifacio (San Raffaele Scientific Institute, Milan, Italy).

Each IA-2 fragment was in vitro transcribed and translated in the presence of [³⁵S]methionine (NEN) using the TNT-coupled rabbit reticulocyte system (Promega, Madison, WI) with Sp6 RNA polymerase. Autoantibodies against each single IA-2 construct were detected by a slightly modified quantitative radioimmunoprecipitation assay (13) using 50% protein A-Sepharose to separate free [³⁵S]methionine from antibody-bound labeled products. Results were expressed as an index defined as follows: (sample cpm − negative standard control cpm)/(positive standard control cpm − negative standard control cpm). Positive autoantibody indexes, defined as values >99th percentile of 211 healthy control sera (102 women and 109 men; median age 27 years; age range 3–77 years) were 0.094, 0.010, 0.073, 0.064, 0.138, 0.049, and 0.094 for IA-2_FL, IA-2_IC, IA-2_BDC, IA-2_(256–760), IA-2_JM, IA-2_PTP, and IA-2_(761–964), respectively. Intra- and inter-assay coefficients of variation were 5.7 and 10.3% for IA-2_(256–760), 6.2 and 10.9% for IA-2_JM, 5.8 and 10.0% for IA-2_(761–964), 5.1 and 9.2% for IA-2_PTP, 5.6 and 7.6% for IA-2_BDC, 4.8 and 9.9% for IA-2_IC, and 5.0 and 8.5% for IA-2_FL. In this study, IA-2_(761–964) and IA-2_{PTP (687–979)} constructs were used to evaluate IA-2 COOH-terminal immunoreactivities, whereas IA-2_{JM (601–630)} and IA-2_(256–760) constructs were used to detect IA-2 middle-domain immunoreactivities. The remaining three IA-2 fragments (IA-2_FL, IA-2_IC, and IA-2_BDC) are those most commonly used to evaluate IA-2 immunoreactivity in diabetes-related screenings. IA-2As in the NIRAD study were detected using IA-2_IC (19). IA-2_ICA assay obtained 72% sensitivity and 99% specificity at the 2007 4th assay proficiency evaluation (lab 155) of the Diabetes Antibody Standardization Program (DASP).

To evaluate the specificity of antibody binding to ³⁵S-labeled IA-2_(256–760) in comparison with ³⁵S-labeled IA-2_IC construct, the mutual inhibition activity of different concentrations of unlabeled IA-2_IC and/or IA-2_(256–760) fragments were tested. The unlabeled fragments were prepared by in vitro transcription and translation as described above but replacing [³⁵S]methionine with unlabeled methionine in the amino acid mixture. Unlabeled recombinant IA-2_(256–760) and/or IA-2_IC (0.5-, 1-, 2-, and 4-fold the amount of ³⁵S-labeled protein) were added to each tube and incubated overnight at 4°C with patient sera. The following day, after incubation with radiolabeled ³⁵S–IA-2_(256–760) or ³⁵S–IA-2_IC proteins, samples were processed with the usual radioimmunoprecipitation assay. In competition experiments, IA-2_(256–760)A⁺/IA-2_ICA⁻ (n = 6), IA-2_(256–760)A⁻/IA-2_ICA⁺ (n = 2), or IA-2_(256–760)A⁺/IA-2_ICA⁺ (n = 2) sera, respectively, were analyzed from 10 type 2 diabetic patients.

GADA in type 1 diabetic and LADA sera were detected by a slightly modified fluid-phase radioimmunoprecipitation assay (21) using a human recombinant full-length GAD65 construct furnished by Dr. Å. Lernmark (Department of Medicine, University of Washington, Seattle, WA). GADA assay obtained 80% sensitivity and 98% specificity at the 2007 4th DASP assay proficiency evaluation (lab 155).

TPO-As were measured by a commercial radioimmunoassay kit (cod.14752; Adaltis, Roma, Italy).

Genomic DNA was extracted using the salting-out method. High-resolution typing for DRB1*04 and DQB1 loci was performed using allele group-specific amplifications. A reverse line blot method, provided by H.A. Erlich and T. Bugawan (Roche Molecular System, Alameda, CA), was used as the detection system (22). HLA genotypes were classified in three risk categories (high, moderate, and low) based on the absolute risk values for type 1 diabetes previously estimated in the Italian population (23).

Statistical analyses were performed using SPSS software, version 13 (SPSS, Chicago, IL). Frequency differences were calculated by χ² test with Yates’ correction, whenever appropriate, or by Fisher's exact test. A P value <0.05 was considered significant.

Of 177 LADA patients, 59 (33.3%) were positive for at least one of the seven IA-2 constructs analyzed. IA-2_(256–760) was the fragment showing the highest sensitivity, identifying IA-2 immunoreactivity in 29.4% (52 of 177) of LADA sera (Table 1), significantly more frequently than IA-2_PTP (11.3%, 20 of 177, P < 0.0001), IA-2_(761–964) (9.6%, 17 of 177, P < 0.0001), IA-2_JM (9.6%, 17 of 177, P < 0.0001), IA-2_FL (18.1%, 32 of 177, P = 0.017), IA-2_BDC (13.0%, 23 of 177, P = 0.0002), and IA-2_IC (19.8%, 35 of 177, P = 0.048). Of 59 IA-2A⁺ sera, 20 (33.9%) reacted with only one of the seven constructs (17 with IA-2_(256–760), 2 with IA-2_IC, and 1 with IA-2_BDC). All IA-2_PTPA⁺ (n = 20) and IA-2_(761–964)A⁺ (n = 17) sera reacted with IA-2_IC. Thirteen sera reacted with IA-2_PTP and IA-2_(761–964). All IA-2_JMA⁺ sera reacted with IA-2_(256–760) construct. No significant difference in terms of sex, age, disease duration, and BMI was detected between the 59 GADA/IA-2A⁺ and the 118 GADA⁺/IA-2A⁻ LADA patients.

Autoantibodies to IA-2_(256–760) were detected in 33 of 978 (3.4%) GADA/IA-2_ICA⁻ type 2 diabetic patients. Table 2 reports GAD, IA-2_IC, and IA-2_(256–760) immunoreactivities of the whole group of type 2 diabetic patients (N = 1,020) analyzed. In this group of patients, 4.9% (50 of 1,020) of sera were found to be IA-2_(256–760)A positive, a higher percentage not only versus IA-2_ICA (1.3%, 13 of 1,020, P < 0.0001) but also versus GADA (4.1%, 42 of 1,020). The IA-2_IC construct detected as positive three IA-2_(256–760)A⁻ patients and on the other hand detected as positive seven GADA⁺ LADA and 33 GADA⁻ type 2 diabetic individuals negative for IA-2_IC antibodies.

Table 3 reports sex, age at disease diagnosis, BMI, and fasting glucose of the 1,020 type 2 diabetic patients classified according to their different GAD or IA-2_(256–760) immunoreactivities. The GADA⁺ but not the IA-2_(256–760)A⁺ LADA patients showed significantly lower BMI and mean age at diagnosis and higher fasting glucose values than antibody-negative type 2 diabetic patients. Table 4 reports the TPO antibody frequencies of GAD or IA-2_(256–760)A⁺ LADA in comparison with 114 GADA/IA-2_(256–760)A⁻ type 2 diabetic patients of comparable age and sex. TPO antibody frequencies were significantly higher in both groups of GADA⁺ or IA-2_(256–760)A⁺ patients versus type 2 diabetic GADA/IA-2_(256–760)A⁻ patients (P = 0.001 and 0.032, respectively). LADA patients positive for GADA or IA-2_(256–760)A showed a significantly higher frequency of high–and moderate–HLA risk genotypes compared with type 2 diabetic patients negative for GADA and IA-2_(256–760)A (P < 0.05) (Table 5).

Fig. 2 shows the results of representative competition experiments performed with sera single positive for IA-2_(256–760) or IA-2_IC and double positive for IA-2_(256–760)/IA-2_IC autoantibodies. An IA-2 fragment–specific, dose-dependent reduction of antibody binding was observed in each of the three sera analyzed according to their relative IA-2_(256–760)/IA-2_IC autoantibody pattern. Specific results are detailed in the Fig. 2 legend. Similar data were found for the other seven sera investigated, according to the corresponding autoantibody pattern (data not shown).

Altogether, at disease diagnosis, immunoreactivity against COOH-terminal containing IA-2_(761–964) and/or IA-2_{PTP (687–979)} constructs was found in 58.5% (31 of 53) of childhood-onset type 1 diabetic patients, a significantly higher percentage versus adult-onset type 1 diabetic (35.8%, 19 of 53, P = 0.032) and LADA patients (24.2%, 8 of 33, P = 0.003). In particular, in childhood-onset type 1 diabetic patients, IA-2_(761–964)A frequency (47.2%) was significantly higher (P < 0.01) versus LADA patients (18.2%), whereas IA-2_{PTP (687–979)}A frequency (52.5%) was significantly higher (P < 0.01) versus both adult-onset type 1 diabetic (20.8%) and LADA (21.2%) patients (Fig. 3A).

Altogether, at disease diagnosis, immunoreactivity against IA-2_(256–760) and/or IA-2_{JM (601–630)} constructs was found in 35.8% (19 of 53) of childhood-onset type 1 diabetic patients, a lower but not significantly different percentage versus adult-onset type 1 diabetic (45.3%, 24 of 53) and LADA (42.4%, 14 of 33) patients. However, in childhood-onset type 1 diabetic patients, single IA-2_{JM (601–630)}A frequency (5.7%) was significantly lower (P < 0.01) versus adult-onset type 1 diabetic (26.4%) and LADA (24.2%) patients. IA-2_(256–760) immunoreactivities in LADA (39.4%), adult-onset (39.6%), and childhood-onset type 1 diabetic patients (30.2%) were not significantly different (Fig. 3B).

At disease diagnosis, IA-2_IC, IA-2_FL, and IA-2_BDC autoantibody frequencies (69.8, 62.3, and 62.3%, respectively) in childhood-onset type 1 diabetic patients (Fig. 3C) were significantly higher in comparison with adult-onset type 1 diabetic (45.3%, P < 0.02; 39.6%, P = 0.03; and 30.2%, P < 0.002, respectively) and LADA patients (27.3%, P < 0.001; 27.3%, P = 0.002; and 21.2%, P < 0.001, respectively).

IA-2 is one of the major autoantigens in type 1 diabetes, a target of both humoral (24–28) and T-cell reactivity (29,30). IA-2As have also been detected in small percentages of type 2 diabetic patients but only in a few cases in addition to GADAs (18,13). IA-2A presence, in addition to GADA, increases the risk of LADA patients to require future insulin therapy (13). The first aim of the present study was to identify the IA-2 construct able to detect IA-2 immunoreactivity with the highest sensitivity in Italian LADA GADA⁺ patients. We found that IA-2_{IC (605–979)}, at present considered as the most sensitive construct for IA-2A detection in autoimmune diabetes (16), reacted with 19.8% of LADA patients investigated, a significantly lower percentage in comparison with the 29.4% of IA-2_(256–760), a construct used for the first time in LADA to investigate IA-2 middle-domain immunoreactivity. By the use of competition experiments, we showed that these two fragments represent distinct IA-2 immunoreactive epitopes. The different immunoreactivities of IA-2_{IC (605–979)} and IA-2_(256–760) fragments seem to be related to the age at diagnosis; COOH-terminal domain of the IA-2 protein was found to be the major IA-2 autoantigenic region in childhood-onset type 1 diabetes, whereas in adult-onset type 1 diabetic and LADA GADA⁺ patients, there is a lower IA-2 COOH-terminal immunoreactivity that results in a significant decrease of diagnostic sensitivity of all IA-2 constructs containing a COOH-terminal residue. The finding that IA-2_(256–760) construct reacts with more LADA patients than IA-2_{FL (1–979)}, even if the former is a portion of the latter, might be due to a different conformation of the 256–760 amino acidic residues of the two constructs or to steric hindrances, ultimately leading to variable autoantibody binding affinities. Several studies demonstrated that IA-2As in type 1 diabetic patients are directed against multiple epitopes of the intracellular cytoplasmic portion of the protein (amino acids 601–979), located in the juxtamembrane region (JM, amino acids 605–682), and in the protein (PTP)-like COOH-terminal domain (amino acids 630–979) (31–36). To date, an immune response against the extracellular domain of the IA-2 protein has not been demonstrated in type 1 diabetes. In our cohort of LADA patients, immunoreactivity against the IA-2_{JM (601–630)} fragment was almost three times less frequent than IA-2_(256–760) construct, thus suggesting that the main epitopes target of IA-2A in LADA may be located either between amino acids 631 and 760 of the protein and/or, more interestingly, between amino acids 256 and 600, a domain comprising the extracellular portion of the protein.

The second aim of our study was to evaluate the frequency of IA-2_(256–760) immunoreactivity in a large cohort of type 2 diabetic patients negative for GADA and IA-2_{IC (605–979)}A. Surprisingly, 33 of these patients were IA-2_(256–760)A positive. To date, IA-2A presence in the absence of GADA has been considered a rare phenomenon in type 2 diabetes (13,37). However, our results clearly demonstrate that most IA-2 autoantibody screenings performed so far in type 2 diabetic patients underestimated the frequency of immunoreactivity against the IA-2 autoantigen and, as a consequence, the number of patients with autoimmunity. The double positivity for GAD and IA-2 autoantibodies in type 2 diabetes (11,37) suggests a pathogenetic link more consistent with type 1 rather than with type 2 diabetes, identifying patients who progress toward insulin dependency within a relatively brief period of time. In the present study, we found that the LADA patients positive for GADA have significantly lower mean age and BMI and higher fasting glucose levels compared with type 2 diabetic patients negative for GADA and IA-2_(256–760)A. IA-2_(256–760)A positivity does not seem to determine a similar metabolic phenotype; however, genetic analysis demonstrates that LADA patients positive for GADA or IA-2_(256–760)A have a significantly higher frequency of autoimmune diabetes HLA susceptible genotypes compared with type 2 diabetic patients negative for both GADA and IA-2_(256–760)A.

GADA positivity in type 2 diabetic patients also identifies those at high risk for developing thyroid autoimmunity (38). Interestingly, in our cohort of Caucasian patients, a similar result was found not only for GADA⁺ (18) but also for IA-2_(256–760)A⁺ type 2 diabetic patients. TPO-A occurred significantly more frequently in IA-2_(256–760)A⁺ (regardless of GADA positivity) than in islet-related autoantibody-negative type 2 diabetic patients, thus supporting the hypothesis that some of these patients may represent a phenotypic expression of a complex autoimmune polyendocrine syndrome.

Recently, it was shown that the islet proteins recognized by T-cells and autoantibodies in type 1 diabetic and LADA patients may be in part different (39) and that measures of multiple islet protein T-cell responses in type 2 diabetic patients may improve the identification of patients with autoimmune diabetes compared with autoantibody assessment alone (40). It was hypothesized that the higher sensitivity of T-cell responses could be due to the capacity of T-cells to react with unknown islet antigens (40). In that study, however, IA-2 immunoreactivity was evaluated with IA-2_{IC (605–979)} and not with IA-2_(256–760) construct. It is possible that IA-2_(256–760)A detection might contribute to reduce the bias of sensitivity between T-cell response and autoantibody analysis for identification of LADA patients. It is also of potential interest that IA-2_(256–760) construct contains in its sequence a number of IA-2 T-cell epitopes recognized by human CD4 T-cells (41).

Finally, our data, related to Caucasian patients, confirm and extend those reported in Japanese diabetic populations (17), in which an association between IA-2 autoantibody epitope specificities and age at onset was found. These results suggest that the mechanisms responsible for the generation of IA-2A in Caucasian and Japanese LADA patients are similar and that different genetic backgrounds probably do not influence IA-2A epitope specificity, in contrast with what has been reported for GAD epitope immunoreactivity (12).

In summary, by analyzing IA-2 epitope immunoreactivity in Caucasian LADA and type 1 diabetic patients, we found that 1) IA-2_(256–760), an IA-2 construct lacking the COOH-terminal portion of the protein, may represent a new sensitive marker for the study of the humoral IA-2 immunoreactivity in LADA patients, by being able to identify IA-2 immunoreactivity also among GADA⁻ type 2 diabetic patients; 2) IA-2_(256–760)A presence is associated with increased risk for developing thyroid autoimmunity and higher frequency of autoimmune diabetes HLA susceptible genotypes; and 3) the specificity of IA-2 humoral immune response in autoimmune diabetic patients is related to the age at diagnosis, with an increased IA-2 COOH-terminal immunoreactivity in childhood-onset type 1 diabetes.

In conclusion, the results of the present study suggest that IA-2 immunoreactivity in type 2 diabetic and LADA patients is more frequent than previously demonstrated and that the analysis of IA-2_(256–760) immunoreactivity in type 2 diabetic patients may represent an additional, important diagnostic tool for a more appropriate classification of diabetes.

The NIRAD study is supported by Foundation for the Research of the Società Italiana di Diabetologia (FoRiSID) based on an unconditioned research grant from Novo Nordisk Italy.

We are indebted to the past President of FoRiSID, Prof. Riccardo Giorgino, for his effort in promoting and supporting this study and to Prof. Riccardo Vigneri, present advisor of the study, for his invaluable continuous support.

This article is dedicated to the memory of Prof. Umberto Di Mario, who greatly contributed to the design and implementation of the study.