Antigen-Specific T Cell Analysis Reveals That Active Immune Responses to β Cell Antigens Are Focused on a Unique Set of Epitopes Free

Type 1 diabetes (T1D) is a chronic inflammatory autoimmune disease in which islet β cells are selectively destroyed (1, 2). Autoreactive T cells that recognize islet β cell Ags (βCAs) (3–5) play a major role in disease pathogenesis (6–9). Susceptibility to disease incidence is highly associated with HLA class II genes, particularly the DR4-DQ8 and DR3-DQ2 haplotypes (10, 11). CD4 T cells specific to βCAs are speculated to be the major players in initiating the pathogenic process (6, 12). Many studies have demonstrated that CD4⁺ T cells, isolated from T1D patients or from the islets and pancreata of NOD mice, recognize antigenic peptides derived from βCAs. These include preproinsulin (13, 14), glutamic acid decarboxylase (GAD) 65 and 67 isotypes (15), islet-specific glucose-6-phosphatase catalytic subunit–related protein (IGRP) (16, 17), chromogranin A (ChgA) (18, 19), and zinc transporter 8 (ZnT8) (20, 21). Numerous antigenic epitopes derived from these Ags have been identified (3, 4). Experiments in NOD mice, which carry a single MHC class II allele and unique genetic background, demonstrated that the insulin B:9–23 epitope is the major autoantigenic epitope in T1D pathogenesis (14). However, in humans who carry multiple class II alleles and a diverse genetic background, Ags and epitopes critical to the disease process are not fully defined.

We hypothesized that each individual T1D subject has a unique repertoire of βCA-specific T cells. Identification of the T cell epitopes of these Ags and detection of these epitope-specific T cells could facilitate the understanding of T1D pathogenesis and subsequent development of Ag-specific immunomodulation therapies.

We also hypothesized that destruction of β cells in islets is an active process that continues even after onset of T1D, and that T1D subjects years after disease onset could have recently activated βCA-specific T cells. We investigated the use of biomarkers that track activated βCA-specific T cells for identification of target Ags and T cell epitopes critical to the disease process. CD38 is a type II glycoprotein that is expressed on surface naive T cells and recently activated memory T cells (22–24), but it is not expressed in resting memory T cells (23, 24). Thus, the presence of Ag-specific CD38⁺ memory T cells for a given Ag is an indication of active immunity directed against that Ag (23, 24).

We exploited the specificity of CD38 expression on memory T cells to identify βCA-specific CD4⁺ T cells from PBMCs that were immune active in vivo. We used a CD154 upregulation assay to track overall βCA-specific CD4⁺ T cells ex vivo. With the combination of CD38 and CD45RA/RO markers, we were able to detect, enumerate, and compare the frequency of immune-active (CD45RO⁺CD38⁺) βCA-specific T cells in T1D subjects and healthy controls. Furthermore, we developed a novel CD154 upregulation epitope mapping assay to define the epitope specificity of these βCA-specific T cells. This study provides a novel strategy to investigate the role of CD4 T cells in T1D pathogenesis. In the future, this approach can potentially be used to tailor personalized Ag-specific immunomodulation therapy.

The healthy donors and T1D patients were recruited locally after approval by the Benaroya Research Institute Institutional Review Board and with the written consent of participants. A total of 15 subjects were recruited for each group. Age, duration of disease, and serum C-peptide status for the T1D patients is shown in Supplemental Table I. The healthy control group was age matched with the T1D group. All subjects in both the T1D group and healthy control group had the T1D disease susceptible HLA-DRB1*0401-DQB1*0302 haplotype.

For T1D-associated autoantigens, peptide libraries containing peptides of 20 aa in length with 12 aa overlapping of adjacent peptides to cover each protein were synthesized (Mimotopes, Clayton, VIC, Australia). GAD65 protein (access no. NP_000809) had 72 peptides; ZnT8 protein (access no. NP_776250.2) had 39 peptides; IGRP (access no. NP_066999) had 43 peptides; ChgA protein (access no. AAP36245.1) had 56 peptides; and Ins B_11–23 preproinsulin protein (PPI; access no. NP_001008996) had two modified peptides, PPI_78–90^K88S (25) and PPI_35–47^R46E (Ins B_11–23^R22E) (26). Peptide library for influenza A matrix protein (MP) (access no. ACP41108) had 30 peptides.

A modified CD154 upregulation assay was used to identify βCA-specific T cells ex vivo. Briefly, 20 million freshly isolated PBMCs were stimulated in vitro for 3 h with a set of βCA peptide libraries containing GAD65, IGRP, ZnT8, ChgA, and two modified preproinsulin peptides, a total of 212 peptides, at the final concentration of 0.5 μg/ml for each peptide in the presence of anti-CD40 (1 μg/ml) (clone HB-14; Miltenyi Biotec, San Diego, CA). Anti-CD40 Ab was added during the in vitro stimulation to prevent the downregulation of CD154 molecules through CD40/CD154 interaction on newly activated T cells (27). After stimulation, PBMCs were collected and stained with anti–CD154-PE Ab (clone 5C8; Miltenyi Biotec) followed by labeling with anti-PE microbeads (clone PE4-14D10; Miltenyi Biotec). The Ag-responsive T cells with upregulated CD154 were enriched on a magnetic bar according to the manufacturer’s protocol. The enriched cells were further labeled with Abs (all from BD Biosciences), including anti–CD3-V500 (clone SP34-2), anti–CD4-allophycocyanin-H7 (clone RPA-T4), anti–CD45RA-PE-Cy7 (clone HI 100), anti–CD45RO-FITC (clone UCHL1), anti–CD38-V450 (clone HB7), anti–CD69-allophycocyanin (clone L78), anti–CD14-PerCP (clone MΦ9), and anti–CD19-PerCP (clone Leu-12). Via-Probe (BD Biosciences), together with anti-CD14 and anti-CD19, was used to dump CD14⁺, CD19⁺, and dead cells. βCA-responsive T cells were identified by upregulation of CD154 and CD69 on CD4 T cells. CD154⁺CD69⁺CD45RO⁺CD38⁺ cells were identified as in vivo immune active islet βCA-specific T cells. Anti–CD278-Brilliant Violet 786 (clone DX29) was used to evaluate ICOS expression. In samples with significant numbers of CD45RO⁺CD38⁺ T cells, the T cells were sorted out and expanded as oligoclones.

To set the flow cytometric gate to define positivity of CD45RO and CD38 markers, total CD4 T cells collected from a pre-enrichment tube were used. The gate was set at a position such that the sum of CD45RO⁻CD38⁺ and CD45RO⁺CD38⁻ populations was at a maximum, and the CD45RO⁺CD38⁺ population was at a minimum. Using this strategy, the activated memory T cells were set at the lowest possible level to avoid an overestimation of the activated memory T cell population.

Sorted Ag-specific T cells were seeded into a round-bottom 96-well plate at 10 cells per well, with 1.5 × 10⁵ irradiated PBMCs as feeder cells, in 200 μl of T cell culture medium (TCM: RPMI 1640 containing 10% [v/v] of pooled human serum, 1% [v/v] of l-glutamine, sodium pyruvate, and penicillin/streptomycin; all Invitrogen, Carlsbad, CA) and 1 μg/ml PHA. On the next day, each well was supplemented with 40 IU (in 10 μl of TCM) of recombinant human IL-2 (Sigma-Aldrich, St. Louis, MO). After 7–10 d culture at 37°C, 5% CO₂, expanded T cells became visible pellets in the 96-well plate. These T cell pellets were transferred to the flat-bottom 96-well plate and fed with 100 μl of fresh TCM supplemented with 200 IU/ml IL-2. When the T cells became confluent in the plate, the cells were split and fed with fresh TCM and IL-2, and eventually placed in a 48-well plate to obtain 5–10 × 10⁶ T cells for CD154 epitope mapping assay.

Once the T cells were successfully expanded, T cells from each oligoclone were washed and suspended at 0.5 × 10⁶/ml in TCM containing 1 μg/ml CD40-blocking Ab. T cells (10⁵) in 200 μl from each oligoclone were stimulated with 43 different pools of five T1D Ag peptides and with DMSO only as a negative control. The cells were stimulated for 3 h and then stained with Abs against CD3-FITC, CD4-PerCP, CD69-allophycocyanin, and CD154-PE for 10 min. After washing, the upregulation of CD154 after Ag stimulation was analyzed by flow cytometry. If a well contained a subpopulation of T cells responding to the T1D Ag peptide pool stimulation, a second round of CD154 epitope mapping assay was performed with individual peptides contained in that T1D Ag peptide pool.

Ex vivo tetramer staining and phenotype analysis were performed as described previously (17).

CD4⁺ T cells from PBMCs were sorted into three fractions: CD45RO⁻ (5.6 × 10⁶ naive T cells), CD45RO⁺CD38⁻ (3.2 × 10⁶ resting memory T cells), and CD45RO⁺CD38⁺ (0.23 × 10⁶ activated T cells). Genomic DNA from each fraction was isolated and subjected to TCR Vβ CDR3 region deep sequencing by Adaptive Biotechnologies (Seattle, WA) (28). Preproinsulin-specific T cell oligoclones were also subjected to survey sequencing by Adaptive Biotechnologies.

We used a Student t test and Pearson correlation analyses in this study as described in each figure legend.

In our previous influenza vaccination and infection study, we noticed that the expression of CD38 on memory T cells was dynamically regulated (24). To confirm that CD38 could be a useful marker in identifying recent in vivo–activated T cells, the kinetics of CD38 expression on influenza A MP-specific T cells from subject pre- and post- Fluzone vaccination were examined. For these experiments, PBMCs were stimulated with a peptide library covering MP for 3 h in vitro. MP peptide–activated T cells in total PBMCs were stained with anti–CD154-PE and enriched using anti-PE magnetic beads. The enriched cells were stained with anti-CD4, anti-CD45RA, anti-CD38, and anti- CD14/CD19/Via-Probe. Cells were gated on CD4⁺CD45RA⁻ memory T cells, and CD38 expression on CD154⁺ T cells was examined at different time points (Fig. 1A). MP-specific memory CD4⁺ T cells were CD38⁻ or very dim before vaccination. This result also suggested that CD38 was not upregulated by 3 h of peptide stimulation. A substantial proportion of these cells expressed CD38 at day 14 postvaccination. The proportion of MP-specific cells that expressed CD38 declined at 4 wk after vaccination, and gradually returned to the prevaccination level after 2–3 mo (Fig. 1). This result implies that CD38 expression on memory T cells can be used as a marker for recently activated T cells and suggests that the duration of CD38 expression on Flu specific memory CD4⁺ T cells lasted for 2–3 mo after activation (Fig. 1).

For further evaluation of whether CD38 expression could be used to identify recently activated βCA-specific CD4⁺ T cells, PBMCs from T1D subjects were stimulated with 210 overlapping peptides derived from four different βCAs (GAD65, IGRP, ZnT8, ChgA) with the addition of two modified preproinsulin peptides, that is, DR0401-restricted PPI_78–90^K88S (25) and DQ8-restricted PPI_35–47^R46E (Ins B_11–23^R22E) (26). βCA-specific CD4 T cells were identified by the coexpression of CD154 and CD69 surface markers on CD4⁺ T cells. Fig. 2A shows the representative staining profiles from a T1D subject. The CD45RO and CD38 phenotypes of these CD154⁺CD69⁺ βCA-specific T cells were analyzed (right panels of Fig. 2A). These panels show three major subsets, the naive CD45RO⁻CD38⁺ subset and the CD45RO⁺CD38⁻ and CD45RO⁺CD38⁺ memory subsets of βCA-specific CD4 T cells. The phenotype of islet-specific cells within these different CD38 populations in T1D subjects was further evaluated by examining another T cell activation marker, ICOS. An example of ICOS expression in one subject is shown in Fig. 2B, and results from five T1D subjects are shown in Fig. 2C. We observed that significant higher percentage of islet-specific cells within the CD45RO⁺CD38⁺ subset expressed ICOS compared with CD45RO⁺CD38⁻ and CD45RO⁻CD38⁺ subsets. These data support the premise of CD45RO⁺CD38⁺ as the activated memory subset and CD45RO⁺CD38⁻ as the less activated or resting memory subset.

Frequencies of total (CD154⁺CD69⁺), naive (CD45RO⁻), memory (CD45RO⁺), and immune active memory (CD45RO⁺CD38⁺) βCA-specific T cells in both T1D and healthy subjects were then examined (Fig. 2D–G). Statistically significant differences were observed for immune-active memory Ag-specific T cells between T1D and healthy subjects (Student t test, p = 0.0043), but not for total (Student t test, p = 0.0612), naive (Student t test, p = 0.5274), or memory (Student t test, p = 0.0826) Ag-specific T cells (Fig. 2B–E). Percentages of CD45RO⁺ βCA-specific T cells that were CD38⁺ were also higher in T1D subjects compared with controls (data not shown). A correlation between the frequency of islet-specific cells and duration of disease was not observed within the T1D group.

Next we sought to identify the antigenic epitopes recognized by the CD38⁺ βCA-specific cells. To achieve this, CD154⁺CD69⁺CD38⁺CD45RO⁺ T cells were sorted into a 96-well plate with 10 cells per well and expanded as oligoclones with PHA. After 1 mo of expansion, each oligoclone was aliquoted into 43 different wells in a 96-well plate and stimulated with 43 pools of peptides (five peptides per pool for a total of 212 peptides). Ag-specific T cells upregulated CD154 and CD69 upon peptide stimulation (Fig. 3A). When a positive response was identified, the responding oligoclone was restimulated with individual peptides from the positive peptide pool to define the antigenic peptide (Fig. 3B). In this way, specific epitopes of the T cell lines were identified. Fig. 3C shows multiple antigenic peptides that were identified from one oligoclone. Table I lists peptides that contain antigenic epitopes to elicit active immune responses in four T1D patients. Supplemental Table II lists the frequency of peptide identified in all CD38⁺CD45RO⁺ oligoclones. Interestingly, CD154⁺CD69⁺CD38⁺CD45RO⁺ T cells from each patient recognized a unique set of antigenic peptides. The responses to the GAD65 and IGRP were diverse, with no identical epitopes identified across patients. In contrast, responses to PPI_78–90^K88S and ZnT8_266–285 were detected in all four subjects tested. Response to ChgA was not observed. To determine whether the epitopes recognized by these CD38⁺ T cells were presented by HLA-DR or HLA-DQ, we analyzed HLA restrictions for 10 oligoclones for which we had cells available (Fig. 4). We found that the reactivities of all these oligoclones were blocked by anti–HLA-DR Ab but not by anti–HLA-DQ Ab, including the reactivity to GADp32-specific peptide, which has been shown to contain an HLA-DQ–restricted epitope (29) (Fig. 4). These results suggest that all of these oligoclones recognize peptides presented on HLA-DR.

Because CD38⁺ PPI_78–90^K88S T cells were prevalent in the T1D subjects, we evaluated whether PPI-reactive T cells of a specific clonotype could be maintained in the periphery. For that purpose, CD38⁺ oligoclones from subject K276T1D (isolated as described above) were stimulated with PPI_78–90^K88S. CD154⁺CD69⁺ cells were sorted and expanded. Specificity of these cells to PPI_78–90^K88S was confirmed by upregulation of CD154 upon reactivation. Deep TCR sequencing was carried out with these oligoclonal populations (Supplemental Table III shows the top 10 functional Vβ sequences).

To determine whether PPI_78–90^K88S-specific T cells of identical clonotype were being retained in the periphery, the subject K276T1D was recalled 4 mo after the oligocloning assay was performed. PBMCs were sorted into naive (CD45RO⁻, 5.6 × 10⁶ cells), resting memory (CD45RO⁺CD38⁻, 3.2 × 10⁶ cells), and immune-active memory (CD45RO⁺CD38⁺, 0.23 × 10⁶ cells) CD4⁺ T cells. The TCR repertoire of each population was subjected to genomic TCR Vβ deep sequencing analysis and 163,518, 110,264, and 66,484 functionally rearranged TCR Vβ sequences were obtained from each population, respectively. When the DNA sequences of the top 10 rearranged Vβ VDJ regions of PPI_78–90^K88S-specific T cells were compared with the TCR repertoire from the above three populations, the PPI_78–90^K88S-specific TCR VDJ sequence (5′-TGTGCCAGCAGTTACGGAATAGGGAGGGCAGATACGCAGTATTTT-3′) was detected as one of the productive TCR VDJ regions in the memory population (1 out of 110,264 unique TCR Vβ sequences). This Vβ sequence is equivalent to 3.7 cells/million (30) (Supplemental Fig. 1). This suggests that PPI-specific T cells can divide and persist in the periphery for at least 4 mo.

This study demonstrates the presence of in vivo immune-active βCA-specific T cells in T1D subjects. Although activation markers such as CD69 and CD25 can be used to define T cell activation status, their upregulation is relatively transient (31, 32). Expression of CD38 in memory cells after activation can last for ∼2–3 mo (23, 24), making it an ideal biomarker to track T cells that are actively involved in autoimmune responses. Our results show that activated βCA-specific T cells are present at a much higher frequency in T1D subjects compared with healthy controls. These data suggest that βCA-specific T cells were likely activated by islet βCAs in vivo, and they might be involved in T1D pathogenesis. Indeed, C-peptide could also be detected in serum of our cohort (Supplemental Table I). Presence of C-peptide in the serum indicates the presence of viable β cells, and these β cells act as a source of autoantigen for the activation of βCA-specific T cells.

To identify the epitope specificity of activated βCA-specific T cells, we developed a CD154 upregulation epitope mapping assay. We analyzed samples from four T1D subjects, and we detected a unique population of βCA-specific T cells with different epitope specificities in each subject. Among the five βCAs we tested, immune-active T cells recognizing preproinsulin, ZnT8, GAD65, and IGRP were detected; T cells that recognized ChgA were not detected. The epitope specificities of GAD65- and IGRP- specific T cells among the four subjects tested were quite diverse. T cells from these subjects did not share the same epitope specificity toward these two proteins. However, responses to modified preproinsulin epitope (PPI_78–90^K88S) and ZnT8 epitope (ZnT8_266–285) were shared by all four subjects tested. Furthermore, T cells that responded to these two epitopes were frequently contained in multiple oligoclones (Supplemental Table II), which suggests a high frequency of these T cells in the immune-active T cell fraction. These results indicate that the PPI_78–90^K88S epitope and the ZnT8_266–285 epitope are dominant antigenic epitopes that could potentially be targets for Ag-specific therapies. The relative abundance of PPI_78–90^K88S-reactive T cells in the periphery was also supported by the longitudinal detection of PPI memory T cells of identical clonotype in the periphery.

The CD154 upregulation epitope mapping approach also led to the discovery of new epitopes, including ZnT8_266–285 epitopes, which have not been previously reported. Although T cell epitopes for GAD65 and IGRP have been studied extensively, our present study also identified novel epitopes of these two proteins. These results illustrated the robustness of this mapping approach and emphasize the high degree of diversity among individual autoimmune repertoires.

It is clear that T1D incidence is highly associated with HLA-DQ8. Thus, there is great interest in identifying DQ8 restricting T cells. To our surprise, the responses of all the oligoclones to their corresponding peptides we tested were blocked by HLA-DR, not HLA-DQ–blocking Ab, including an oligoclone specific to GADp32 peptide, which contains HLA-DQ8 eptitope by tetramer staining (29). One interpretation of this finding is that in this subject, HLA-DR– and HLA-DQ8–restricted T cells may play different roles in the pathogenesis: DQ8-restricted T cells may be contributing to disease susceptibility or initiation (14), wereas DR-restricted T cells may become dominant during disease progression and after disease onset.

In summary, we demonstrate that a small fraction of βCA-specific memory T cells show evidence of recent Ag stimulation and persist in the circulation, even after onset of T1D. This type of targeted detection method can be used to study the roles of Ag-specific T cells in the pathogenesis of T1D, but also has potential application in the development of personalized Ag-specific vaccines.

We thank Cynthia Cousens-Jacobs for assistance in the preparation of this manuscript.

The authors have no financial conflicts of interest.