Simultaneous Detection of Circulating Autoreactive CD8⁺ T-Cells Specific for Different Islet Cell–Associated Epitopes Using Combinatorial MHC Multimers

Samples from recent-onset type 1 diabetic patients were retrieved from the Kolibri type 1 diabetes cohort, which includes material from 350 patients with juvenile-onset type 1 diabetes (median age 8.7 years [range 1–17 years]). The cohort was collected consecutively after diagnosis by pediatricians in the southwestern part of the Netherlands between 1995 and 1999. The diagnosis was made according to International Society of Pediatric and Adolescent Diabetes and World Health Organization criteria. All patients were HLA-A2 positive. Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll-Isopaque density gradient centrifugation. PBMCs were frozen in a solution of 20% human pooled serum and 10% DMSO (5–10 × 10⁶ cells per vial) and kept in liquid nitrogen until use. Blood samples of islet transplant recipients were stored in liquid nitrogen for 12–36 months.

Seven patients were transplanted with islet cell grafts after signing informed consent and under appropriate ethical approval as reported previously (29). None of the patients presented alloantibodies against HLA alloantigen that was expressed on the donor cells before transplantation. Graft recipients were long-term type 1 diabetic patients without any earlier transplantation, with plasma C-peptide <0.09 ng/ml, large variation in blood glucose levels (coefficient of variation ≥25%), A1C concentration >7%, and one or more chronic diabetes lesions. Exclusion criteria were body weight >90 kg, active smoking, pregnancy, disturbed liver function tests, history of hepatic disease, presence of HLA antibodies, or negative Epstein-Barr virus serostatus.

Donor organs were procured from multiple heart-beating donors through the Eurotransplant Foundation (Leiden, the Netherlands) and processed at the Beta Cell Bank in Brussels, Belgium, to β-cell–enriched fractions that were cultured for 2–20 days (median 6 days). Immunosuppressive induction therapy consisted of antithymocyte globulin (ATG) (Fresenius HemoCare, Redmond, WA) with a single infusion of 9 mg/kg and subsequently with 3 mg/kg for 6 days except when the T-cell count was <50/mm³. Maintenance immunosuppression consisted of tacrolimus (Prograft; Fujisawa/Pharma Logistics; dose according to trough level: 8–10 ng/ml in the first 3 months post-transplantation, 6–8 ng/ml thereafter) and mycophenolate mofetil (2,000 mg/day; Roche, Basel, Switzerland). The HLA typing of the patients is depicted in the supplementary table (available in an online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db09-1486/DC1). Islet cell recipients were age-matched with the healthy control subjects.

Multimeric HLA-A2–peptide complexes were prepared essentially as previously described (30). Briefly, recombinant HLA-A2 and human β₂-macroglobulin were solubilized in urea and injected together with each synthetic peptide into a refolding buffer consisting of 100 mmol/l Tris (pH 8.0), 400 mmol/l arginine, 2 mmol/l EDTA, 5 mmol/l reduced glutathione, and 0.5 mmol/l oxidized glutathione. Refolded complexes were biotinylated by incubation for 2 h at 30°C with BirA enzyme (Avidity, Denver, CO). The biotinylated complexes were purified by gel filtration on a Superdex 75 column (Amersham Pharmacia Biotech, Piscataway, NJ). Multimeric HLA–peptide complexes were produced by addition of streptavidin-conjugated Qdots (28) (Invitrogen, Breda, the Netherlands) to achieve a 1:20 streptavidin-Qdot/biotinylated HLA class I ratio. Qdots used are Qdot-585, -605, -655, -705, and -800. Samples from HLA-A2–positive subjects were stained with a mixture containing six diabetes-associated epitopes, a HLA-A2 epitope expressed in HLA-A2, and a mix of viral antigens (Table 1).

PBMCs (2 × 10⁶) were stained simultaneously with all Qdot-labeled multimers (0.1 μg of each specific multimer) in 60 μl of PBS supplemented with 2% BSA and incubated for 15 min at 37°C (Table 1). Subsequently, 10 μl antigen-presenting cell (APC)-labeled anti-CD8 (stock 1:10) and 10 μl fluorescein isothiocyanate–labeled anti-CD4, -CD14, -CD16, -CD19, and -CD40 antibodies (Becton Dickinson, Franklin Lakes, NJ) were added for 30 min at 4°C. After being washed twice, cells were resuspended in PBS/2% BSA containing 7-aminoactinomycin D (7-AAD; eBioscience, San Diego, CA) to exclude dead cells, and samples were analyzed using the LSR II (Becton Dickinson).

Statistical analysis on recent-onset diabetic patients versus their siblings was performed using a Wilcoxon matched-pairs test. Differences between recent-onset diabetic patients and control subjects were tested using the unpaired t test with Welch correction (for HLA-A2 peptide and PPI_15–24; passed normality test) or the Mann-Whitney test (all other epitopes). Changes in epitope reactivity of islet cell transplant recipients over time were tested using the Friedman test followed by a Dunn multiple comparisons test. All statistical analyses were performed using GraphPad Prism software.

Recently, the use of multidimensional encoded MHC multimers was reported as a powerful tool to allow the parallel detection of multiple antigen-specific T-cell populations within a single sample (28). This technology opens the possibility of designing kits of defined peptide-MHC multimers that may be used to report on disease state or vaccine response. To test this concept, a Qdot-based combinatorial approach was developed to simultaneously monitor multiple islet epitopes associated with the development of type 1 diabetes. HLA-A2 molecules were loaded with identified diabetes peptides, and MHC multimers were created by labeling with Qdots such that T-cells specific for each of these epitopes are defined by binding of MHC multimers with a unique combination of two Qdots (Table 1).

During flow cytometric analysis, single-cell lymphocytes were gated that stained positive for CD8 (on average 60,000 CD8 T-cells were gated per blood sample) but negative for the “dump”-channel (combination of CD4, CD14, CD16, CD19, and CD40) and negative for the exclusion (viability) dye 7-AAD (Fig. 1,A). Staining of PBMCs of healthy control subjects (Fig. 1,B, typical example shown) with a mixture of three virus-derived epitopes resulted in a clearly distinguishable population characterized by a positive signal for both fluorescent signals used to encode the peptide-MHC multimers. A clear population of CD8⁺ T-cells reactive to insulin B_10–18 was seen in the sample of the recent-onset diabetic patient. These insulin B_10–18-reactive CD8 T-cells were also found in an islet cell transplantation recipient, in which virus-specific CD8⁺ T-cells also were seen. No CD8⁺ T-cells reactive to insulin B_10–18 were found in the healthy control subjects (Fig. 1), whereas few virus-specific CD8⁺ T-cells were detectable in samples of recent-onset diabetic patients (Fig. 1). In terms of reproducibility, the coefficient of variation between experiments was 9.5% across Qdot multimers. For the separate epitopes, the variation varied (HLA-A2 peptide, 10.8%; virus mix, 34.9%; InsB, 15.9%; IA-2, 0.0%; IGRP, 0.0%; PPI, 6.3%; GAD65, 4.5%; and ppIAPP, 6.9%) (supplementary Figs. 1–3, available in the online appendix).

To determine whether the simultaneously measured islet epitopes were sensitively and specifically detected in the circulation of recent-onset type 1 diabetic patients, we determined the frequency of CD8⁺ T-cells specific for all currently known epitopes in recent-onset diabetic patients, their siblings (when available), and healthy control subjects.

Unfortunately, of the 20 recent-onset diabetic patients studied, material from just 5 HLA-A2–expressing siblings was available, corresponding to 3 recent-onset diabetic patients, allowing a direct comparison of the presence of CD8⁺ T-cells specific for the islet epitopes (Fig. 2, left panels). Generally, the frequency of β-cell antigen-reactive CD8⁺ T-cells was higher in recent-onset diabetic patients than in their siblings, but because of low numbers, no statistically significant differences were observed. Also, we analyzed the frequency of CD8⁺ T-cells in the circulation of all examined recent-onset diabetic patients (n = 20) and matched control blood donors (n = 15). Clearly, the frequencies of CD8⁺ T-cells recognizing the islet cell–derived epitopes were all significantly higher in recent-onset diabetic patients than in control subjects. Conversely, a higher frequency of virus-reactive CD8⁺ T-cells was seen in the control subjects. These data point to higher frequencies of islet-reactive CD8⁺ T-cells within the circulation of recent-onset diabetic patients.

Next, we determined the sensitivity and specificity of T-cell responses to each epitope, defining a frequency of one cell reactive in 10,000 CD8⁺ T-cells (i.e., 0.010%) as a cutoff, with the exception of the IGRP_265–273, where a frequency of 1 in 20,000 (i.e., 0.005%) was used as a cutoff. Three epitopes were found to be 100% specific because no relevant frequencies were seen in the control subjects: insulin B_10–18 showed a sensitivity of 65% and was the most specific and considerably sensitive epitope detected; IA-2_797–805 and IGRP_265–273 provided a specificity of 100% but a sensitivity of 25% (Table 2). The epitope with the highest sensitivity was PPI_15–24, as epitope-specific CD8⁺ T-cell reactivity was detectable in 17 of 20 recent-onset diabetic patients (85%). Yet, 4 of 15 control subjects also exhibit CD8⁺ T-cells against this epitope, affecting the specificity (73%). Qdot stainings of HLA-A2–negative patients or control subjects with HLA-A2 multimers were always below the detection limit (n = 27), regardless of the peptide epitopes tested, supporting the specificity of these reagents. Overall, our data indicate that we can discretely monitor the specific presence of multiple epitope reactive CD8⁺ T-cells simultaneously using the newly developed Diab-Q-kit.

Next, the Diab-Q-kit was used to monitor the presence of HLA-A2 epitope-specific CD8⁺ T-cells in seven islet cell transplant recipients, at four different time points: before transplantation, 6 weeks thereafter (reconstitution of the T-cell compartment after ATG-induction treatment), 26 weeks after transplantation, and 52 weeks after transplantation. Examination of the frequency of CD8⁺ T-cells recognizing endogenously processed HLA-A2 peptide presented in the HLA-A2 molecule showed reactivity in two recipients at two different time points (Fig. 3). Viral reactivity was seen in two of seven HLA-A2–positive islet cell transplantation recipients before transplantation (Fig. 3, viral mix), but clearly the induction therapy with ATG strongly reduced their frequency in circulation. One year after transplantation, virus-specific T-cells had reappeared, albeit to a lower level than before transplantation.

Despite disease durations up to several decades, islet autoreactive CD8⁺ T-cells were still detectable in the majority of type 1 diabetic patients at the time of islet transplantation (0 weeks; Fig. 4). No particular islet epitope or pattern of reactivity dominated. After reconstitution of the T-cell compartment under maintenance immunosuppression after ATG induction therapy (4 weeks; Fig. 4), the cumulative numbers of autoreactive CD8⁺ T-cells waned in a minority of recipients. In four of seven patients, islet autoreactivity increased after islet transplantation, one patient displayed stable frequencies of circulating islet reactive T-cells, and, in two case subjects, a stable declined was observed.

The dynamics of T-cells specific for the (pre-)(pro-)insulin epitopes (insulin B_10–18 and PPI_15–24) were most pronounced (Fig. 3): considerable frequencies of these cells were detected before islet cell infusion and (re-)emerged at later time points thereafter. In contrast, T-cell frequencies against GAD65_114–123 and IGRP_265–273 were infrequently seen, with only a single increase late after islet implantation in patients 6 and 7, respectively (Fig. 3). These two case subjects also displayed the most epitope spreading at that time point (Fig. 4). Of note, patient βct #7 cytomegalovirus-converted after transplantation.

Subsequently, the patterns of autoreactive CD8⁺ T-cell frequencies were correlated with clinical outcome. All immune parameters were defined and interpreted without prior knowledge of clinical outcome. Because reactivity to single epitopes was sometimes observed in healthy control subjects (Fig. 2), only the presence of two or more epitope-specific autoreactive CD8⁺ T-cells at any time after transplantation was interpreted as being detrimental; to predict the clinical outcome based on the full autoimmune spectrum, the presence of CD8⁺ T-cells reactive to the recently uncovered PPI epitopes PPI_76–84 and PPI_79–88 in HLA-A3 as well as PPI_4–13 in HLA-B7 were also considered (W.W.U., J.H.V. et al., unpublished observations) as indicated in Table 3. Consequently, the data obtained with the novel Diab-Q-kit predicted that six of seven recipients studied would not reach insulin independence.

Previously, we reported that proliferation of CD4⁺ T-cells to whole IA-2 and GAD65 before transplantation predicted transplantation outcome (14). For comparison, the prediction of clinical outcome based on this proliferation assay was also taken into consideration (Table 3). Prediction based on proliferation of islet-specific CD4 T-cells pointed toward four patients not reaching insulin independence. For three patients, there was agreement in the prediction by both immunological end points.

The Diab-Q-kit predicted that six recipients would remain insulin requiring, of which four actually required exogenous insulin injection. Thus, this method showed a prediction accuracy of 66%. The prediction of clinical outcome based on proliferation (14) predicted insulin requirement in four case subjects, of which three were correct (accuracy of 75%). However, when both methods agreed in their prediction, all transplant recipients remained insulin requiring after transplantation, and thus the combined method provided an accuracy of 100%, underlining the importance of monitoring CD8⁺ T-cell frequencies.

Our study is the first to implement simultaneous detection of multiple islet cell–specific CD8⁺ T-cell responses, by development of the Diab-Q-kit. To this purpose, multidimensional encoded MHC multimers were used. Recently, their use was extensively validated and shown to be a powerful tool to parallel detect antigen-specific T-cells (28). Here, we determined the sensitivity and specificity of previously reported HLA-A2 restricted epitopes (15,,–18,20) in recent-onset type 1 diabetic patients, their siblings, control blood donors, and islet cell transplant recipients. Insulin B_10–18 was found to be 100% specific, as no CD8⁺ T-cell frequencies of >1 in 10,000 cells were seen in PBMCs from healthy control subjects. Although this also holds true for IGRP_265–273 and IA-2_797–805, CD8⁺ T-cells recognizing insulin B_10–18 showed the highest sensitivity of these epitopes. The relevance of insulin B_10–18 in type 1 diabetes is underlined by previously published observations that PBMCs from a type 1 diabetic patient produced interferon-γ in response to this peptide (31), that expression of insulin B_10–18 renders target cells sensitive to killing by CTL lines (32), and, most importantly, that the presence of insulin B_10–18-specific CD8⁺ T-cells correlates with destruction of β-cells (15). As the transplantation of isolated islet cells can result in reactivation of CD8⁺ T-cell–mediated autoreactivity toward islet cell–specific epitopes, peripheral blood from islet cell transplantation recipients can be used to monitor the factors important in β-cell destruction. Also in this cohort, the presence of insulin B_10–18-reactive CD8⁺ T-cells after transplantation correlated with a poor clinical outcome, with the exception of the cytomegalovirus-converted recipient.

Recently, we reported PPI_15–24 as a naturally produced and presented HLA-A2 epitope (16). Cytotoxic CD8⁺ T-cells against this peptide could be cloned that killed β-cells in vitro in a glucose concentration–dependent fashion, linking β-cell immunogenicity with its functional activity (16). This study underscores the relevance of CD8 islet autoreactivity in the pathogenesis of type 1 diabetes, and it indicates that β-cells are actively involved in their own demise. Interestingly, this PPI epitope provided the highest sensitivity (85%) combined with a specificity of 73%. All but one HLA-A2–positive islet cell recipient exhibited increased frequencies against PPI_15–24 at a time point after transplantation. Of these patients, half did not reach insulin independence after transplantation.

The HLA-A2–restricted islet epitopes IA-2_797–805, GAD65_114–123, IGRP_265–273, and ppIAPP_5–13 also exhibited a highly specific staining, with only incidental reactivity in healthy control subjects. In contrast to previously reported results (18), we did not observe any CD8⁺ T-cells reactive to IA-2_797–805 in healthy control subjects. However, we performed a direct assessment of the CD8⁺ T-cell frequency in PBMCs, rather than functional assays, whereas Takahashi et al. (18) cultured CD8⁺ cells for 14 days with autologous APC expressing the IA-2 peptide and a cocktail of cytokines. Their observations may therefore be influenced by in vitro phenomena.

Reactivity to IA-2_797–805, GAD65_114–123, and IGRP_265–273 were almost exclusively seen in combination with reactivities to insulin B_10–18 and/or PPI_15–24, suggesting that PPI epitopes may comprise the primary epitopes after reactivation of autoimmunity upon islet cell transplantation, whereas reactivity to the other may result from epitope spreading.

Intriguingly, in two patients without islet epitope reactive CD8⁺ T-cells after the transplantation induction therapy (i.e., at 6 weeks), CD8⁺ T-cells specific for insulin B_10–18 and PPI_15–24 were among the first CD8⁺ T-cells to occur. Their occurrence after ATG treatment may have resulted from homeostatic proliferation (33,–35), indicating that these CD8⁺ T-cells, although undetectable in the peripheral blood after ATG induction, remain as a memory population in lymphatic organs. Because in one of these patients these cells were undetectable because transplantation, this may indicate that these cells persist many years after destruction of the islet cells. From these data, we speculate that homeostatic proliferation of autoreactive T-cells, including CD8-expressing cells, may be detrimental to islet cell transplantation, similar as homeostatic proliferation of alloreactive cells can be to solid organ transplantation (36,37). We cannot conclude on any order of reactivity among islet autoantigens, because there may be technical explanations for differences in precursor frequencies of the corresponding islet autoreactive CD8 T-cells, such as avidity of the Qdots for the TCR or affinity of the peptide epitope to HLA-A2, that contribute to differences in the detection limit.

Virus-specific CD8 T-cells showed frequencies in some control subjects that were higher than those in recent-onset diabetic subjects, an inverse pattern compared with frequencies of autoreactive T-cells in patient and healthy subjects. Yet, the difference was moderate and was lost completely if frequencies of virus-specific T-cells of patients before islet transplantation were combined with those of patients with newly diagnosed diabetes. We have no clear explanation for this trend, other than patients being slightly younger than control subjects. Yet, this finding suggests that the increases in antigen-specific CD8 T-cells in type 1 diabetes are not a general phenomenon reflecting hyperimmune reactivity per se but seem more specific for islet autoreactive T-cells.

Even though the gate settings will be largely similar among individuals and among experiments, analyzing (auto)antigen T-cell specificities is subject to subtle differences in background stainings between individual subjects that require adjusting the gate settings accordingly. This may partly result from the need to compensate the light channels on each day that blood samples are analyzed on the FACS LSR II. Importantly, the minor differences in background staining did not distinguish patients from healthy subjects. We recommend that a longitudinal series of blood samples from a given subject be analyzed on the same day to minimize interassay variation, as we pursued for the analyses of blood samples of islet transplant recipients. We contend that the availability of a second dimension of staining in our combinatorial approach (each epitope being represented by two different colors) facilitates setting the gates and to a great extent copes with the difficulties of distinguishing background from positive staining of low-avidity TCR.

It will be interesting to use our new technology to assess functional and phenotypic differences in islet autoreactive T-cells between type 1 diabetic patients and other subjects (siblings, healthy unrelated subjects, and patients with other diseases). Preliminary results suggest that CD8 T-cells recognizing the PPI_15–24 epitope in an islet cell–transplanted patient are largely of a memory phenotype (supplementary Fig. 2).

In conclusion, our novel, highly sensitive detection system allows for direct assessment of circulating autoreactive CD8⁺ T-cells against an array of islet epitopes simultaneously. Another major advance over the current procedures to determine islet-specific epitopes is the freedom from in vitro culture and expansion that is otherwise required in most studies on novel diabetes-associated antigens to be able to detect responses (15,,–18). Testing multiple epitope specificities in the same sample further reduces the blood volumes required for analysis and extends opportunities for testing for additional and novel immune reactivities. Finally, our methods provided applicable and informative data on thawed blood samples that had been stored for up to 15 years, for the first time allowing assessment of cellular islet autoreactivity retrospectively, and enabling use in the context of large cohorts and multicenter immune intervention studies. It is conceivable that other or yet to be discovered epitopes may provide a stronger correlation with clinical outcome. Even though the currently used Qdot-MHC multimer technique allows a highly sensitive, combinatorial assessment of multiple islet epitope-specific CD8⁺ T-cell populations in type 1 diabetes to study pathogenesis, prediction, progression, and intervention of the disease, it can be modified and extended to up to 25 different epitopes in the future.

These studies were supported by the Juvenile Diabetes Research Foundation, the Dutch Diabetes Research Foundation, and a VICI award to B.O.R. from The Netherlands Organisation for Health Research and Development.

No potential conflicts of interest relevant to this article were reported.