High Frequency of Cytolytic 21-Hydroxylase–Specific CD8⁺ T Cells in Autoimmune Addison’s Disease Patients

Autoimmune Addison’s disease is caused by the destruction of the adrenal cortex, and can occur either in isolation or as part of autoimmune polyendocrine syndromes type 1 or 2 (1–3). The target Ag for Addison’s disease was identified over 20 years ago by Winqvist et al. (4) as steroid 21-hydroxylase (21-OH), a key intracellular steroidogenic enzyme exclusively expressed in the adrenal cortex. Rapidly, this seminal finding was translated into clinical practice; assay of 21-OH Abs is the most important biomarker for autoimmune Addison’s disease, present in >90% of patients (5). 21-OH Abs are typically present years before clinical disease is evident and may be found when testing individuals who are at risk for developing Addison’s disease, that is, patients who have another autoimmune disease or have a relative with Addison’s disease. Preclinical adult patients with 21-OH Abs, however, have a cumulative risk of only ∼20% of developing overt Addison’s disease if adrenal function is normal at the start of the observation (6). Thus, a significant proportion of individuals with 21-OH Abs continue to remain disease free.

Histological studies of adrenal glands from deceased Addison’s disease patients show significant mononuclear cell infiltration into the adrenal gland (7), and because 21-OH is an intracellular enzyme, autoantibodies are unlikely to directly mediate destruction of the adrenal gland. In addition, as demonstrated during pregnancy, 21-OH Abs transferred from a mother with Addison’s disease did not cause disease in the child (8). Instead, 21-OH Abs are more likely to be an indication of T cell–mediated destruction of the adrenal cortex, perhaps mediating or augmenting Ag presentation (9).

It was first shown by Freeman et al. (10) that PBMCs from Addison’s disease patients, but not controls, proliferated in response to adrenal proteins, and this observation was followed by a study demonstrating that PBMC proliferation and IFN-γ secretion occurred particularly in the presence of 21-OH (9). More recently, Rottembourg et al. (11) established that a significant proportion of HLA-B8⁺ patients have circulating T cells that are specific for a dominant 21-OH peptide. However, a need remains for comprehensive epitope mapping to show which epitopes on 21-OH are targeted by T cells in Addison’s disease patients with different haplotypes and, importantly, whether these cells are functionally capable of destroying the adrenal cortex.

To address these questions, we used 18-mer overlapping synthetic peptides spanning the entire 21-OH protein and demonstrated that T cells from Addison’s disease patients, unlike those from healthy controls, responded to the pool of 21-OH peptides. Such responses were mainly dominated by MHC class I–restricted CD8⁺ T cells and focused on immunodominant regions on 21-OH. We extended these findings by demonstrating that HLA-A2–restricted 21-OH–specific CD8⁺ T cell clones generated from an Addison’s patient are capable of lysing HLA-A2 target cells transduced with lentiviral vectors encoding the full-length 21-OH protein and HLA-A2⁺ 21-OH⁺ tumor cells, thus providing important insights into the pathogenesis and progression of Addison’s disease.

Blood samples were taken from 21-OH Ab–positive Addison’s disease patients attending hospital clinics in Sweden, Norway, Germany, and the U.K., and their characteristics are shown in Table I. Eight patients were analyzed from Bergen, Norway (E.H.); seven patients from Newcastle, U.K. (S.H.P.); four from Frankfurt am Main, Germany (K.B.); and one from Stockholm, Sweden (S.B.). Local ethical approval was verified and approved by the ethical committee for the EURADRENAL studies, and informed consent was obtained. Inclusion criteria for patients were a clinical diagnosis of primary adrenal insufficiency and the presence of 21-OH Ab at the onset. For comparison, blood samples from adult healthy volunteers from the U.K. and Norway were used, and all the volunteers were anonymized.

DNA was extracted from patients’ PBMCs for haplotyping, using the QIAGEN DNeasy Kit, and haplotyping was conducted by the sequencing facility at the Weatherall Institute for Molecular Medicine, Oxford, U.K.

Overlapping 18-aa peptides that span the whole 21-OH sequence were synthesized by facilities at the Weatherall Institute of Molecular Medicine and are listed below in Supplemental Table I. Peptides were dissolved in DMSO at a concentration of 40 μg/ml and were used either individually or as a combined pool.

Patient cells were stimulated at 6 × 10⁶ per well with 1 μg/ml 21-OH peptide pool in RH-10 (RPMI 1640, pH 7.4, supplemented with 10% human serum, 1% penicillin/streptomycin, 1% L-glutamine, 1% nonessential amino acids, 1% sodium pyruvate, 1% HEPES, and 0.1% 2-ME). Cells were initially pulsed in 200 μl for 1 h before the volume was increased to 2 ml and IL-7 was added at 25 ng/ml. On day 3, cells were fed with RH-10 containing 1000 IU/ml IL-2 (Pharmacia) and continued to be fed and/or split as necessary with IL-2–containing medium until day 13, when they were washed to remove IL-2 and rested overnight in RH-10, in preparation for the intracellular staining (ICS) assay.

Cells were stained for intracellular cytokine secretion after 14 d in culture. A total of 200,000 cells were pulsed with individual peptides or the pool of peptides at 10 μg/ml for 5 h, with 10 μg/ml brefeldin A (eBioscience) added after the first hour to prevent cytokine secretion into the supernatant. Cells were washed and stained for the surface markers CD8 APC-H7, CD4 PerCP, and CD3 V500 (BD) together with LIVE/DEAD viability marker (Invitrogen) diluted in PBS for 30 min on ice, then washed and fixed with fixation solution (eBioscience) for a further 30 min on ice or left overnight at 4°C. Cells were permeabilized by washing with permeabilization buffer (eBioscience) and stained with the intracellular Ab IFN-γ FITC (eBioscience) diluted in the permeabilization buffer for 30 min on ice. Cells were washed with PBS and resuspended in FACS buffer for acquisition by flow cytometry. The presence of CD107 on the T cell surface was assessed by the addition of PE-conjugated CD107 (BD) Ab at the same time as brefeldin A. CD4⁺ and CD8⁺ T cell responses to the peptide diluent DMSO were negligible and <0.001%.

Cells were acquired by flow cytometry on FACSCanto II and samples were analyzed using FlowJo (TreeStar). Cells were gated to remove doublets, then separated into Live CD3⁺ CD4⁺ or CD8⁺ cells and analyzed separately for their cytokine positivity.

Plates were coated overnight at 4°C with 4 μg/ml capture Ab (MabTech) in coating buffer (Sigma-Aldrich). Plates were washed twice with RPMI 1640, then incubated for 1 h at 37°C with 200 μl blocking buffer (RPMI 1640 + 10% human serum). After washing plates three times, cells were added in triplicate at 500,000 per well with 10 μg/ml peptide in X-Vivo 15 serum-free medium (Lonza) and incubated for 18 h at 37°C. Plates were washed with wash buffer (distilled water + 0.05% Tween-20) followed by distilled water before adding 0.2 μg/ml biotin detection Ab (MabTech) diluted in PBS, then incubated for 2 h at 37°C. Plates were washed again, and 1 μg/ml Streptavidin-ALP (MabTech) diluted in PBS was added, and plates were then incubated for a further hour at room temperature. Finally, plates were washed once more before developing by adding 100 μl substrate solution and incubated for 10 min in the dark at room temperature. Development was stopped by rinsing plates thoroughly with cold water and the plates were left to dry completely before counting spots using an AID EliSpot Reader.

Monomers were made by refolding the HLA-A2 or HLA B8 H chain and β₂-microglobulin proteins with the 21-OH_342–350 peptide (LLNATIAEV) or the 21-OH_428–435 peptide (EPLARLEL) together at 4°C for 40 h (12). The refold mixture was then concentrated to ∼8 ml using a nitrogen gas–pressured stir cell with a presoaked ultrafiltration 150-mm membrane (Millipore) before being biotinylated with a BirA enzyme overnight, followed by fast protein liquid chromatography separation to remove aggregates. The concentration of the eluted monomer was measured by BCA assay, and aliquots were stored at −80°C. Monomers were tetramerized using Streptavidin-APC (eBioscience) before being used to stain T cells.

Granzyme B secretion was measured by ELISA in the supernatant after overnight coculture of 20,000 CTLs with 40,000 target cells. ELISA plates were precoated overnight with anti-Granzyme B Ab before being washed and loaded with supernatants overnight at 4°C. Plates were washed and incubated with the biotin Ab followed by ExtrAvidin Peroxidase and developed using TMB (3,3′,5,5′-tetramethylbenzidine) solution (Sigma-Aldrich).

RNA was extracted from 21-OH–expressing NCI H295 cells using the QIAGEN RNeasy Kit and cDNA synthesized using the RETROSCRIPT kit. The 21-OH fragment was amplified from prepared cDNA using the forward primer 5′-CGCGGATCCACCATGCTGCTCCTGGGCCTGCTG-3′ and the reverse primer 5′-CCGCTCGAGCTGGCTCTGGCCCGGGCTGTG-3′, containing the BamHI and XhoI restriction sites. The insert was digested using BamHI and Xho1 and ligated into a GFP lentiviral vector, which was then used to transform competent DH5alpha bacteria.

Lentiviral particles were made by combining the GFP–21-OH lentivector, pMDG, and the Gag-pol expressor with Fugene 6 transfection reagent (Promega), then adding dropwise to 293T cells, which had been grown to 90% confluency. Supernatants containing viral particles were collected after 2 d and used to transduce B cells. Cells were sorted for GFP positivity and cultured for a further 2 wk before being used in functional assays.

The VITAL assay was performed as previously described (13). Briefly, 150,000 21-OH–GFP transduced cells were placed together with 150,000 untransduced B cells that had been labeled with CellTracker Orange CMTMR (5-(and -6)-(((4-chloro-methyl)benzoyl)amino)tetramethylrhodamine; Invitrogen) in flat-bottomed 96-well plates and cocultured with 21-OH–specific T cell clones overnight at different effector:target ratios. Cells were then washed and stained with Invitrogen LIVE/DEAD dye and acquired by flow cytometry to determine killing of target cells. Equal numbers of CMTMR-labeled untransduced B cells were placed in a well containing the 21-OH–GFP B cells and cocultured with 21-OH_342–350 T cell clone at different effector:target ratios. The target cell survival was calculated as the percentage of GFP-positive cells compared with the CMTMR-positive cells, and the target cell lysis was calculated as 100 minus the cell survival percentage.

To assess the overall T cell responses to 21-OH in Addison’s disease patients, PBMCs were obtained from Addison’s disease patients with a high titer of 21-OH–Ab. The HLA haplotype revealed a high proportion of patients expressing HLA-A2, HLA-A1, and HLA-B8 (Table I). The frequency of 21-OH–specific CD4⁺ and CD8⁺ T cells was then measured by ICS for IFN-γ in response to restimulation with the individual peptides or with the pool of peptides for 13 d.

We found that most of the 20 Addison’s disease patients analyzed in this study had high numbers of 21-OH–specific CD8⁺ and CD4⁺ T cells, compared with the baseline values found in healthy controls (Fig. 1). Of note, although the magnitude of 21-OH–specific CD8⁺ T cell responses appeared greater than the magnitude of 21-OH–specific CD4⁺ T cell responses (Fig. 1) because these experiments were performed after stimulating PBMCs once in vitro with a pool of overlapping 18-mer peptides, the relative frequency of 21-OH–specific CD8⁺ and CD4⁺ T cell responses cannot be compared.

To assess the longevity of 21-OH–specific T cell responses, we compared the frequency of 21-OH–specific T cell responses in patients who had been diagnosed with Addison’s disease for up to 20 y (Fig. 2). We observed that although the magnitude of the 21-OH–specific T cell response decreased over time, we were still able to detect responses in the majority of our samples from 1 to 2 y since diagnosis.

Having established the presence of high frequency CD8⁺ and CD4⁺ T cell responses to 21-OH in patients with Addison’s disease, we next mapped the peptide epitopes recognized by the 21-OH–specific T cells. PBMCs that had been expanded with the pool of 21-OH peptides were tested against each individual overlapping peptide before measuring the IFN-γ secretion by ICS and flow cytometry. The majority of patients revealed CD8⁺ T cells capable of recognizing 21-OH_337–354 and/or 21-OH_428–445 (Fig. 3A, Supplemental Fig. 1A). In contrast, a number of CD4⁺ T cells recognized the peptide 21-OH _207–224 (Fig. 3B, Supplemental Fig. 1B).

Similar results were observed by ex vivo ELISPOT assays using patients' PBMCs, including both 21-OH–specific CD4⁺ and CD8⁺ T cell responses (Figs. 4, 5). Although the recall assay response is indicative of a memory T cell response to 21-OH in Addison’s patients, this assay remains semiquantitative and does not represent the actual frequency of 21-OH–specific T cells in Addison’s disease patients. To gain a better understanding of the magnitude of 21-OH–specific T cells, we performed ex vivo ELISPOT assays in duplicate with the immunodominant peptide identified during the recall assay (Figs. 4, 5). We found that the patients showed responses in the ex vivo assay that were specific to the same peptides recognized in the recall assay (Fig. 5), demonstrating that the T cell responses were not being biased to particular peptides during the bulk culturing with the pool of peptides. The ex vivo frequency of 21-OH–specific T cell responses (ranging from ∼0.001–0.01%) was similar to melanoma-specific CD8⁺ T cell responses observed in cancer patients (12). As shown above, we confirmed the longevity of 21-OH–specific responses, as ex vivo ELISPOT assays demonstrated the presence of 21-OH–specific responses many years after diagnosis (Fig. 4A). Cumulative ex vivo ELISPOT data are shown in Fig. 4B. To further characterize the relative frequency of CD8⁺ and CD4⁺ 21-OH–specific T cell responses in ex vivo assays, CD4⁺ and CD4^- T cells were sorted from PBMCs from patient 5 and patient 15 and tested in ex vivo ELISPOT assays for their ability to recognize the pool of 21-OH peptides (Fig. 4C). The results in Fig. 4A showed that PBMCs from patient 15 have a detectable ex vivo ELISPOT response against the pool of 21-OH peptides, whereas PBMCs from patient 5 have a much lower frequency of 21-OH–specific T cells. Separation of CD4⁺ and CD4⁻ T cells showed that the ex vivo 21-OH–specific T cell response seen in patient 15 was mainly dominated by CD4⁻ 21-OH–specific T cells, as 21-OH–specific CD4⁺ T cell responses could be detected only after enriching T cell cultures for CD4⁺ T cells (Fig. 4C).

Specificity and HLA restriction of 21-OH–specific T cell responses were confirmed by staining 21-OH–specific T cell lines and clones with HLA A2 and HLA B8 tetramers loaded with the peptides 21-OH_342–350 and 21-OH_428–435, respectively (Supplemental Fig. 2). HLA-A2 tetramer sorted 21-OH_342–350–specific CD8⁺ T cells were assayed for their ability to recognize 21-OH–expressing cells. First, we used overlapping 9 mers spanning the immunodominant peptide 21-OH_337–354 to identify the epitope. The peptide 21-OH_342–350 (LLNATIAEV) elicited the strongest response (Fig. 6). Then, using HLA-A2 tetramers loaded with the peptide 21-OH_342–350, we were able to sort a panel of CD8⁺ T cell lines and clones (Supplemental Fig. 2) and tested their specificity and functional activity (Figs. 7, 8, and Supplemental Fig. 3).

HLA-A2/21-OH_342–350 tetramer⁺ CD3⁺CD8⁺ T cells recognized targets pulsed with nanomolar concentrations of the 21-OH_342–350 peptide, but not with irrelevant peptides (Supplemental Fig. 3 and data not shown). These data provide direct evidence that HLA-A2/21-OH_342–350 peptide tetramer⁺ CD3⁺CD8⁺ T cell populations derived from PBMCs of Addison’s disease patients contain HLA-A2–restricted CD8⁺ T cells recognizing the peptide 21-OH_342–350. In addition, T cells were able to lyse HLA A2⁺ target cells transduced with a lentiviral vector encoding the full-length 21-OH, thus confirming their cytotoxic capacity (Fig. 7).

To further confirm the functional capability of 21-OH_342–350–specific T cells, we cocultured the 21-OH_342–350–specific HLA-A2–restricted CD8⁺ T cell clone with the HLA-A2–expressing tumor line NCI H295 (14) and found an increase in CD107 expression and granzyme B secretion in the supernatant, as measured by ELISA. NCI H295 recognition was reduced in the presence of the HLA-A2–blocking Ab BB7.2 (Fig. 8).

In this study we show strong and long-lasting T cell responses against a few specific peptides derived from 21-OH, the major adrenal cortex cell autoantigen. We have measured the frequency, phenotype, and functional activity of 21-OH–specific CD8⁺ and CD4⁺ T cells in the PBMCs of Addison’s patients directly ex vivo and after in vitro culture. Our results show that 21-OH–specific CD8⁺ and CD4⁺ T cells are often present in high numbers in PBMCs, are Ag experienced, and are capable of massive expansion when exposed to the appropriate cytokines, generating highly cytolytic T cell populations. The ability of 21-OH–specific CD8⁺ T cell clones to recognize endogenous 21-OH protein may therefore be directly linked to the progression of Addison’s disease.

It has previously been established that T cells target autoantigens and cause destruction in other organ-specific autoimmune diseases, such as type 1 diabetes, myasthenia gravis, and multiple sclerosis (15, 16). However, it has often been difficult to characterize such autoreactive T cells, as they are present at a very low frequency in peripheral blood. Addison's disease offers an opportunity to better understand why the immune system in organ-specific autoimmunity often targets intracellular proteins with a tissue-specific expression and often key enzymatic functions in the tissues involved—for example, thyreoperoxidase in autoimmune thyroiditis, glutamic acid decarboxylase in type 1 diabetes, or steroid side-chain cleavage enzyme in autoimmune oophoritis (17).

The high frequency of 21-OH–specific T cells, comparable to the frequency of tumor-specific T cells observed in melanoma patients (12), as well as their long-term persistence after diagnosis is therefore unexpected and consistent with the possibility that 21-OH–specific T cells are continuously stimulated in vivo by mechanisms that need to be further investigated. It is tempting to speculate that this may be caused by nonadrenal expression of 21-OH (18) or that some remnants of the adrenal cortex persist within Addison’s disease patients, perhaps owing to continued proliferation from residual adrenal progenitor or stem cells, either of which could continue to stimulate the 21-OH–specific T cells (19, 20). Alternatively, the long-term persistence of 21-OH–specific CD8⁺ and CD4⁺ T cells may be due to cross-reactive viral or bacterial epitopes, as suggested for other autoimmune disorders (21). It is not possible to draw firm conclusions about the relative difference in frequency of 21-OH–specific CD8⁺ and CD4⁺ T cells, as the majority of the experiments to compare the frequency of 21-OH–specific T cells were performed after stimulating PBMCs in vitro with a pool of overlapping 18-mer peptides spanning the full-length 21-OH protein. The apparent higher frequency of CD8⁺ 21-OH–specific T cells is consistent with previous observations demonstrating higher frequency of Ag-specific CD8⁺ T cells specific for self (22), virus (23), and cancer epitopes (24). However, a higher frequency of virus-specific CD4⁺ T cells has also been documented (25).

Of interest, we found that T cell responses in Addison’s patients are clustered to just a few immunodominant 21-OH epitopes. A previous report had identified the HLA-B8–restricted epitope 21-OH_431–438 (11). Our results have confirmed and extended these findings by demonstrating that the response to peptide 21-OH_431–438 is immunodominant, as it was detected in a large proportion of HLA B8⁺ patients. In addition, we identified a further HLA-A2–restricted dominant epitope at position 21-OH_342–350.

Previous studies have indicated an association between Addison’s disease and the HLA DRB1*04- DQA1* 0301- DQB1* 0302 haplotype (as reviewed in Ref. 26) and the HLA DRB1*0404–restricted recognition of a peptide spanning the region 21-OH_342–361 (9). The results of our study, although confirming the presence of 21-OH–specific CD4⁺ T cell responses, highlighted a significantly larger frequency of HLA-class I–restricted CD8⁺ T cell response, with a strong bias toward T cell responses restricted by HLA-A2, HLA-A1, and HLA-B8.

Because only a minority of individuals with 21-OH Ab develop disease within an observation period of ≤5 y (6), future investigations of T cell responses in these individuals and the ability to predict disease progression are of major importance, as diagnostic and therapeutic opportunities could result. A better understanding of the initiation of T cell responses at the molecular level in Addison’s disease may also open up a way to induce responses against adrenal cortex cancer, a 21-OH–positive tumor with very poor prognosis often afflicting younger people.

In conclusion, we have identified immunodominant regions of 21-hydroxylase and have shown that 21-OH–specific T cells recognize endogenous 21-OH protein and may therefore be directly linked to the progression of Addison’s disease. Unexpectedly, the persistence for many years of 21-OH T cell responses in the face of destroyed adrenal glands without remaining endogenous steroid production indicates the presence of remaining Ag stimulation perhaps by ectopic expression of 21-OH in other tissues or by pathogens expressing a cross-reactive epitope continuously stimulating the 21-OH–specific T cells.

We thank Tim Rostron for haplotyping of patient samples, Zhanru Yu for synthesizing all the peptides used in this article, and Hemza Ghadbane, Uzi Gileadi, and Yanchun Peng for technical support and reagents.

The authors have no financial conflicts of interest.