Activation of Human CD8⁺ T Cells with Nitroso Dapsone–Modified HLA-B*13:01–Binding Peptides

Dapsone is a widely used antibacterial agent and is frequently prescribed for the treatment of leprosy. However, exposure to dapsone is associated with the development of a hypersensitivity syndrome, characterized by fever, skin rash, hepatitis, and generalized lymphadenopathy, in 0.5–3.6% of treated patients (1, 2). Recent studies have shown that expression of HLA‐B*13:01 is associated with increased risk of dapsone hypersensitivity (3–6).

Dapsone contains aromatic amine groups that are susceptible to acetylation by N-acetyltransferase enzymes, a process of detoxification that limits the life span of the drug. The aromatic amine groups also undergo cytochrome P450–mediated hydroxylation yielding dapsone hydroxylamine (7). Nitroso dapsone is generated via auto-oxidation of the hydroxylamine metabolite. Nitroso dapsone is protein-reactive and covalently modifies cysteine residues on proteins (8–11).

PBMCs from hypersensitive patients expressing HLA-B*13:01 are stimulated to proliferate with both dapsone and its nitroso metabolites (12, 13). Through the generation of clones, HLA class II–restricted CD4⁺ and HLA class I– and II–restricted CD8⁺ T cells were shown to be stimulated with dapsone and nitroso dapsone via different pathways (13, 14). The T cell response to parent drug and metabolite was polyclonal with T cells displaying a variety of TCR sequences (14). Dapsone interacts directly with HLA to trigger TCRs, whereas nitroso dapsone triggers TCRs though a pathway dependent on Ag processing, presumably through the formation of protein adducts. Both dapsone and nitroso dapsone interact with multiple HLA class I proteins to activate CD8⁺ T cells, which makes it challenging to focus research on the drug HLA-B*13:01 interaction and specifically determine whether drug metabolite-modified peptides activate T cells in an HLA allele–restricted manner. Therefore, the aim of this work was to: 1) design HLA-B*13:01 binding peptides that contain a reactive cysteine residue; 2) generate nitroso dapsone–modified peptides that are free of dapsone and the nitroso metabolite; and 3) explore immunogenicity of the peptides using autologous APCs and APCs transfected with the single HLA-B allele HLA-B*13:01.

Three 9-mer peptides were identified as potential HLA-B*13:01 high-affinity binders using the MHC binding prediction tool obtainable at https://www.iedb.org (15). Peptides were designed to each contain a single cysteine residue at sites distal from the HLA-B*13:01 binding motifs (glutamine at P2 and leucine at P9), as previous studies have shown that nitroso dapsone binds covalently to cysteine (11). Polyalanine was chosen as peptide backbone to minimize the interaction between peptides and TCR. Cysteine was inserted within the binding motif to generate positional derivatives. To improve peptide solubility, glutamic acid and aspartic acid were also added. Aspartic acid is also a secondary anchor in position 3. All three peptides showed favorable binding to HLA-B*13:01 (a percentile rank of <1 as predicted by NetMHC, Immune Epitope Database). The strategy for peptide design and predicted HLA-B*13:01 binding affinity is summarized in Fig. 1A and 1B. Peptides were synthesized with a fluorenylmethyloxycarbonyl (Fmoc)–protecting group at the N terminus to avoid a nonspecific reaction between nitroso dapsone and the primary amine of the peptide. The protecting group was removed before any functional analyses. Characterization of the peptides was performed by mass spectrometry (MS) (16).

Fmoc–peptide 1 (Pep1, Fmoc-AQDCEAAAL), –peptide 2 (Pep2, Fmoc-AQDACEAAL), and –peptide 3 (Pep3, Fmoc-AQDAEACAL) were purchased from Synpeptide (Shanghai, China). Purity was initially confirmed by HPLC, and identity was confirmed by MS. Peptides were incubated with nitroso dapsone at a 4:1 molar ratio in 70% acetonitrile/30% H₂O for 16–24 h at 37°C. Nitroso dapsone–modified Fmoc-peptides were then analyzed by HPLC with a Gemini 5-μm NX-C18 110 Å, LC column 250 × 4.6 mm (Phenomenex, Cheshire, U.K.) connected to an Agilent 1260 Infinity Quaternary LC (Agilent Technologies, Cheshire, U.K.). The fractions containing modified peptides were then analyzed by MS using a TripleTOF 6600 (Sciex) mass spectrometer. Fmoc was removed by incubating of conjugated peptides with 20% piperidine in dimethylformamide at room temperature for 3 h. The nitroso dapsone–modified peptides were analyzed again by HPLC to confirm removal of the Fmoc-protecting group. Fractions were collected and analyzed by MS.

Nitroso dapsone–modified peptides were characterized using a TripleTOF 6600 (Sciex) mass spectrometer. Briefly, samples were delivered into the MS by automated in-line reversed-phase liquid chromatography using an Eksigent NanoLC 400 System (Sciex) mounted with a trap and analytical column (15 cm × 75 µm). A NanoSpray III source was fitted with a 10-µm inner diameter Picotip emitter (New Objective). A gradient of 2–50% (v/v) acetonitrile/0.1% (v/v) formic acid over 90 min was applied to the column at a flow rate of 300 nl/min. Spectra were acquired automatically in positive ion mode using information-dependent acquisition using mass ranges of 400–1600 Da in MS and 100–1400 Da in tandem MS (MS/MS). Up to 25 MS/MS spectra were acquired per cycle (∼10 Hz) using a threshold of 100 counts/s, with dynamic exclusion for 12 s and rolling collision energy.

Free dapsone/nitroso dapsone remaining in the drug-modified peptide fractions were quantified by MS. Calibration standards were prepared at the following concentrations: 5–500 nM. All samples were diluted in 0.1% formic acid prior to analysis and spiked with the internal standard sulfamethoxazole (250 nM). Samples and standards were analyzed immediately by a QTRAP 5500 mass spectrometer (Sciex) coupled with an UltiMate 3000 LC system (Dionex, Sunnyvale, CA). The multiple reaction monitoring transitions for each analyte were as following: dapsone, 249.1/156.1; dapsone nitro, 263.1/156.1; and the internal standard, 254.1/156.1. Other MS parameters, such as voltage potential and collision energy, were optimized to achieve great sensitivity. Data acquisition and quantification were performed using Analyst 1.5 software.

A homology model of HLA-B*13:01 was built from HLA-B*52:01 (96% sequence similarity; Protein Data Bank: 3W39) using SWISS-MODEL. GOLD 5.2 (Cambridge Crystallographic Data Centre software) was used to dock the peptides AQDCEAAAL, AQDWEAAAL, and AQDC(DDS)EAAAL within the binding groove, with the binding site defined as 10 Å around the binding point. The binding point was further refined with key amino acid residues within B pocket (Tyr⁷, Tyr⁹, Phe²², Thr²⁴, Val³⁴, Met⁴⁵, Glu⁶³, Ile⁶⁶, Ser⁶⁷, Thr⁶⁹, and Asn⁷⁰) and F pocket (Tyr⁷⁴, Asn⁷⁷, Thr⁸⁰, Tyr⁸⁴, Ile⁹⁵, Arg⁹⁷, Asp¹¹⁴, Leu¹¹⁶, Tyr¹²³, Ile¹²⁴, Thr¹⁴³, Lys¹⁴⁶, and Trp¹⁴⁷). A generic algorithm with ChemPLP as the fitness function was used to generate 10 binding modes per ligand. Default settings were retained for the “ligand flexibility,” “fitness and search options,” and “GA settings.”

PBMCs were cultured in R9 medium (RPMI 1640 medium supplemented with 100 U/ml streptomycin, 100 mg/ml penicillin, 25 mg/ml transferrin, 2 mM l-glutamine 10% human AB serum, and 25 mM HEPES buffer). Medium was supplemented with IL-2 to maintain T cell lines and clones. RPMI 1640 supplemented with FBS, penicillin (100 U/ml), streptomycin (0.1 mg/ml), 25 mM HEPES buffer, and 2 mM l-glutamine was used for EBV-transformed B cell cultures.

Venous blood samples (50 ml) were taken from two dapsone-hypersensitive patients with a: 1) positive dapsone patch test; and 2) dapsone and nitroso dapsone lymphocyte transformation test as reported previously (14). Patient 8 (female, 28 y old at time of adverse event) displayed fever and abnormal liver function tests following 17-d exposure to dapsone. Patient 14 (male, 24 y old at time of adverse event) displayed fever, skin rash (erythema), and abnormal liver function tests following 21-d exposure to dapsone. Both individuals expressed HLA-B*13:01; the full patient HLA profiles are available in Zhao et al. (14). Approval for the study was acquired from Shandong Provincial Institute of Dermatology and Venereology, and informed written consent was obtained. A material transfer agreement was signed prior to transport of PBMCs to Liverpool, U.K.

OVA–dapsone conjugates were prepared by the reaction of nitroso dapsone with OVA at a molar ratio of 10:1 (drug to protein) using methods as previously described (17). Ab production was performed by Kaneka Eurogentec (Liège, Belgium) using a speedy 28-polyclonal package. Detailed information is available online (https://www.eurogentec.com).

PBMCs (1 × 10⁶/well) were cultured in a 48-well plate with nitroso dapsone–modified Pep1 (patient 8 and 14) and Pep3 (patient 14 only due to limitations on the availability of the modified peptide) for 14 d to enrich the number of responsive T cells prior to serial dilution. Peptide concentrations (10–50 μM) were selected based on a lack of intrinsic toxicity and no inhibition of PHA-treated healthy donor PBMC proliferation. Pep2 was reserved for cross-reactivity studies, as it was synthesized in low yield. T cell clones were generated by serial dilution and repetitive mitogen-driven expansion. Briefly, irradiated allogenic PBMCs (5 × 10⁴ cells/well) in medium containing PHA and IL-2 were added to 96-well U-bottom plates. T cells were then diluted and added to the PBMC mixture at 0.3, 1, and 3 cells/well. Cultures were incubated for 14 d (37°C/5% CO₂), and medium was supplemented with IL-2 every 2 d. On day 14, the growing clones were restimulated with PHA and irradiated allogenic PBMCs (5 × 10⁴ cells/well) in IL-2–containing medium and expanded for a further 14 d prior to testing for peptide specificity.

PBMCs from the hypersensitive patients were cultured with supernatant from EBV-producing B-958 cells in the presence cyclosporine A to generate immortalized autologous B cell lines using established methods (14).

To generate a HMy2.C1R-HLA-B*13:01-P2A-B2M cell line, the pLJM1-EGFP plasmid (Addgene plasmid #19319; a gift from David Sabatini) was modified by replacing the EGFP with a Multiple Cloning Site, adding an Eμ enhancer (pLJM1- Eμ-SFFV-NewMCS) and a P2A sequence. β₂-Microglobulin was PCR amplified from cDNA from a volunteer and ligated into the distal end of the P2A sequence using restriction sites Age1/SnaB1. HLA-B*13:01 was PCR amplified from the HLA-B*13:01 pcDNA3.1(⁺) plasmid and ligated into the proximal end of the P2A sequence using restriction sites PME1 and EcoR1 to create the pJLM1-HLA-B*13:01-P2A-B2M plasmid. The pJLM1-HLA-B*13:01-P2A-B2M plasmid was transfected into competent Escherichia coli and colonies grown on agarose plates under carbenicillin selection overnight at 37°C. Positive colonies were identified by colony PCR and grown up overnight in a shaking incubator at 37°C. Plasmids were purified using the QIAprep Spin Miniprep Kit (Qiagen). Correct integration was confirmed by Sanger sequencing. The HEK293 cell line was used as the machinery to generate second-generation lentivirus. HLA-B*13:01-P2A-B2M lentivirus was collected from the supernatant 96 h posttransfection. HMy2.C1R cells were transduced with the HLA-B*13:01-P2A-B2M lentivirus supernatant and selected for positive transduction with puromycin. HMy2.C1R cells (European Collection of Authenticated Cell Cultures 94050320), an EBV-transformed B cell line lacking HLA-B expression, was purchased from the UK Health Security Agency culture collections (https://www.culturecollections.org.uk).

Expanded T cell clones (5 × 10⁴ cells/well) were incubated with irradiated autologous APCs (EBV-transformed B cells; 1 × 10⁴ cells/well) and modified peptides (10 μM) for 48 h in duplicate cultures, and proliferation was measured through addition of [³H]thymidine (0.5 µCi/well) for the final 16 h of the culture period. Peptide-free medium was used as negative control.

T cell clones were phenotyped using flow cytometry for the CD4⁺ (CD4-APC [clone RPA T4]) and CD8⁺ surface receptors (CD8-PE [clone HIT8a]). T cell clones (5 × 10⁴ cells/well) were tested in dose-response studies for cross-reactivity with irradiated APCs (1 × 10⁴ cells/well) and nitroso dapsone–modified peptides (Pep1, 2, and 3; 1–100 μM) in triplicate cultures. Unmodified peptides subjected to the same culture conditions and purification steps were used as a negative control (1–100 μM). Clones were also cultured with soluble nitroso dapsone (1–40 μM; higher concentrations induced toxicity). Proliferation of the clones was measured by addition of [³H]thymidine for the final 16 h of the experiment.

The secretion of IFN-γ from the clones was assessed using ELISPOT. T cell clones were incubated with APCs in the presence and absence of nitroso dapsone–modified peptides and other study compounds in IFN-γ Ab-coated ELISPOT plates (37°C, 5% CO₂) for 48 h. Plates were then developed according to the manufacturer’s instructions (Mabtech, Stockholm, Sweden), and spots were counted using an AID ELISPOT reader. To confirm levels of cytokine produced, the supernatant of four clones (two drug-specific and two not drug-specific) was analyzed using a cytokine bead array according to the manufacturer’s instructions (LEGENDplex, Custom Human 11-plex panel; BioLegend). Briefly, supernatant pooled from triplicate wells for each clone was added to specific Ab-coated beads, forming an analyte–Ab complex. After washing, a biotinylated detection Ab mixture was added that bound to the specific analyte–Ab complexes. Streptavidin-PE was subsequently added, which bound to the biotinylated detection Abs, providing fluorescent signal intensities in proportion to the bound analyte amount. Fluorescent signals were measured using BD FACSCanto II and analyzed using LEGENDplex data analysis software, where concentrations of each analyte are determined using a standard curve generated in the same assay.

To determine whether the detected activation of clones with nitroso dapsone–modified peptides was due to residual dapsone or nitroso dapsone or degradation of the peptides and liberation of free dapsone, clones (5 × 10⁴ cells/well) were cultured with irradiated APCs (1 × 10⁴ cells/well): 1) in the presence of soluble dapsone (125–500 μM), and proliferation was measured by addition of [³H]thymidine; and 2) in the presence of nitroso dapsone–modified Pep1 or soluble nitroso dapsone and glutathione (1 mM), which binds covalently to the nitroso metabolite, preventing protein binding (11), and proliferation was measured by addition of [³H]thymidine. The irreversible binding of nitroso dapsone to APCs was measured in the presence and absence of glutathione (1 mM) using immunofluorescence staining with an anti-dapsone Ab.

To explore the importance of HLA proteins in T cell activation, T cell clones (5 × 10⁴ cells/well) were cultured with Pep1 or Pep3: 1) in the absence of APCs; 2) in the presence of C1R-B*13:01 or C1R-parental APCs (1 × 10⁴ cells/well); and 3) in the presence of C1R-B*13:01 APCs pretreated with either isotype (IgG1) or HLA class I (DX17) or HLA class II (Tu39) blocking Abs for 30 min. T cell proliferation or IFN-γ release was measured using [³H]thymidine or ELISPOT, respectively.

HLA-B*13:01 designer peptides were created by incorporating HLA-B*13:01 anchor residues and a cysteine residue (Fig. 1A). These peptides are predicted as strong binders using the MHC binding prediction tool (pep1, 0.862 µM; pep2, 1.155 µM; and pep3, 1.175 µM) (Fig. 1B). Nitroso dapsone–modified peptides were prepared from conjugation of N-terminal Fmoc-protected peptides with the nitroso metabolite. This procedure was followed by Fmoc deprotection and HPLC purification. The final products were essentially free of soluble dapsone, nitroso dapsone, or unmodified peptide based on HPLC analysis (Fig. 2A). Liquid chromatography-MS/MS analysis of the final purified nitroso dapsone–modified peptides revealed doubly charged ions at m/z 585.235, corresponding to the peptides with a sulfonamide adduct (Δm = 278). The peptide sequence was confirmed by the presence of a series of y and b ions. The modification site was confirmed by a series of adducted B ions, all with a mass addition of 278 atomic mass units (amu), giving evidence of modification at cysteine residue (Fig. 2B, 2C). As expected, further oxidation of the sulfonamide adduct could result in an N-hydroxyl sulfonamide adduct with a mass addition of 294 amu. This adduct was confirmed by the MS/MS spectrum of Pep3 AQDAEACAL, as shown in Fig. 2D. To explore how the nitroso dapsone modification affects the binding affinities of the peptides, two approaches were adopted. First, the cysteine residue of all three peptides was replaced with bulky aromatic amino acids (phenylalanine [F] and tryptophan [W]) to mimic the nitroso dapsone modification in the cysteine-containing peptide, and B*13:01 binding affinity was assessed using NetMHCpan BA 4.1. It was not possible to assess drug peptide modifications directly. Second, in silico models of C- and W-containing peptides binding to HLA-B*13:01 were generated and compared with the predicted binding of the equivalent nitroso dapsone–modified peptide. Interestingly, when the cysteine residues of the designer peptides were replaced with the bulky aromatic amino acids, the binding of these peptides to HLA-B*13:01 was stronger than the cysteine-containing peptides (Fig. 3A), suggesting that nitroso dapsone–modified peptides would be good binders to HLA-B*13:01. Molecular docking of AQDCEAAAL, AQDWEAAAL, and AQDC(DDS)EAAAL to a homology model of HLA-B*13:01 demonstrated that all of these peptides could bind to HLA-B*13:01 similarly within the binding groove (Fig. 3B). Of particular interest, the predicted conformation of dapsone-modified peptide AQDC(DDS)EAAAL is similar to the peptide AQDWEAAAL (Fig. 3C, 3D), with the bulky aromatic groups pointing out the binding groove to ensue favorable interaction of P2 and P9 anchor residues with HLA-B*13:01.

Initial testing of ∼400 T cell clones derived from nitroso dapsone–modified Pep1- or Pep3-treated PBMCs involved culture of T cells with APCs and the peptides or medium (as a negative control) in duplicate culture and comparison of proliferation. Almost 50% of the clones generated displayed reactivity against either nitroso dapsone–modified Pep1 (n = 124) or Pep3 (n = 48) and the strength of the proliferative response varied from a stimulation index (proliferation in test incubations with Ag/proliferation in control incubations with medium) of 2 to >30 (Fig. 4). These T cell clones were expanded and analyzed for CD phenotype. Clones expressing the CD8⁺ receptor with no CD4 expression were used in the experiments described below.

A panel of up to 30 T cell clones from both patients were used to assess cross-reactivity. All clones were stimulated to proliferate with nitroso dapsone–modified Pep1 or Pep3 with no discernible difference between the strength of the induced response observed (Fig. 5A); however, proliferative responses were not detected when the clones were cultured with unmodified peptides (Fig. 5B, shows Pep1 data). All clones were also activated with soluble nitroso dapsone, which forms adducts with protein in the cell culture assay, and the strength of the maximal response induced was similar with nitroso dapsone and nitroso dapsone–modified peptides (Fig. 5C).

Three clones were used in an ELISPOT assay to study IFN-γ with Pep2. Clones secreted IFN-γ in the presence of all three nitroso dapsone–modified peptides (Pep1, Pep2, and Pep3) (Fig. 6A; the number of experiments with Pep2 was limited, as the peptide was synthesized in small quantities). To confirm levels of cytokine produced, the supernatant from four clones (two drug-specific, two not drug specific) was analyzed using a cytokine bead array. Clones that were stimulated to proliferate also secreted IFN-γ, IL-5, IL-13, perforin, and granzyme B (Fig. 6B–E).

Given the complete cross-reactivity profile of the clones with different peptides, well-growing clones from either patient were selected and tested with Pep1 or Pep3 for the HLA restriction studies detailed below.

To confirm that T cell activation with the nitroso dapsone–modified peptides was not due to residual dapsone, clones were cultured with an optimal concentration of the parent drug. Clones were stimulated to proliferate with nitroso dapsone–modified peptide and soluble nitroso dapsone; however, proliferative responses were not detected with dapsone itself (Fig. 7A). Glutathione was used to differentiate between nitroso dapsone–modified peptide and soluble nitroso dapsone T cell proliferative responses. The addition of glutathione to cell culture medium inhibits the covalent binding of nitroso dapsone to EBV-transformed B cells (Fig. 7B) and the activation of clones with soluble nitroso dapsone. In contrast, glutathione did not alter the activation of clones with nitroso dapsone–modified peptides, in which the nitroso moiety is already bound covalently to the cysteine residue in the peptide sequence (Fig. 7C).

In in vitro culture conditions, soluble nitroso dapsone interacts with multiple HLA proteins to activate CD4⁺ and CD8⁺ clones from hypersensitive patients (14). The optimized culture conditions with extensive covalent modification of cellular protein potentially overrides the exquisite HLA restriction observed in patients. Thus, a stepwise approach was used to explore the restriction of the nitroso dapsone–modified peptide-specific T cell response. First, with the exception of a small number of self-presenting clones, T cells were not stimulated to proliferate with nitroso dapsone–modified peptides when APCs (EBV-transformed B cells) were excluded from the assays (Fig. 8A). Second, C1R-B13:01 APCs, expressing HLA-B*13:01, but not the other HLA class I alleles expressed by the patients were used as APCs in the place of autologous EBV-transformed B cells. Clones were activated and secreted IFN-γ when cultured with either nitroso dapsone–modified peptides or soluble nitroso dapsone and C1R-B*13:01 cells (Fig. 8B). Third, IFN-γ secretion above control levels was not detected when the experiment was repeated with nitroso dapsone–modified Pep1 or Pep3 and C1R-parental cells (Fig. 9A). Finally, pretreatment of C1R-B*13:01 APCs with an anti–HLA class I blocking Ab inhibited peptide-induced IFN-γ secretion, whereas an anti–HLA class II blocking Ab had no effect (Fig. 9B).

Delayed-type drug hypersensitivity reactions are a serious form of adverse event and represent a challenge to health care professionals attempting to delineate patient susceptibility. Drug-responsive T cells are believed to be the primary effector cells involved in the iatrogenic disease, with drug (metabolite) protein or peptide binding believed to be the molecular initiating event. This interaction may involve the drug molecule binding covalently to cellular or serum proteins, as is the case for β-lactam antibiotics such as flucloxacillin (18–20). The resultant adducts are thought to be processed by APCs into peptide fragments that associate with HLA proteins for presentation to T cells. At the opposite end of the spectrum, drugs such as carbamazepine form labile binding interactions with HLA proteins or peptides within the HLA Ag binding cleft to stimulate a similar effector T cell response (21–23). It should be noted that although the pathways that lead to drug display by HLA proteins differ, the chemical composition and three-dimensional arrangement of molecules at the immunological synapse may be similar, with the only difference being the nature of the drug peptide-binding interaction.

Different forms of drug hypersensitivity reaction are strongly associated with expression of specific HLA class I alleles (24–26). This suggests that a derivative of the drug may interact with exquisite selectivity with the protein encoded by the HLA allele to activate the T cells that instigate the hypersensitivity reaction. Indeed, for the archetypal association between HLA-B*57:01 and abacavir hypersensitivity (27–29), the drug adheres deep within the peptide binding cleft of HLA-B*57:01, altering the structure and the peptides that are displayed by the HLA protein to CD8⁺ T cells (30–32). A similar binding interaction is not observed with closely related HLA proteins. Regrettably, the picture is not so clear for other forms of HLA class I allele–restricted forms of drug hypersensitivity reaction. Even with exemplars such as carbamazepine (HLA-B15:02 ([33] and HLA-A*31:01 [34]) and flucloxacillin (HLA-B*57:01 [35]), the parent drug, drug metabolites, and/or peptide adducts interact with multiple HLA class I and class II proteins to stimulate CD4⁺ and CD8⁺ T cells in hypersensitive patients (19, 36–40). In recent years, we have focused on dapsone hypersensitivity to further define pathways of T cell activation as: 1) the metabolism and protein reactivity of dapsone is well defined (8, 9, 11); and 2) dapsone hypersensitivity is strongly associated with HLA-B*13:01 expression (4). Dapsone and nitroso dapsone activate polyclonal CD4⁺ and CD8⁺ T cells via different pathways, such as pharmacological HLA binding and hapten binding, respectively (13, 14). A number of T cells display dapsone and nitroso dapsone cross-reactivity; however, others are highly selective in that they are stimulated with one molecule and not the other. Thus, exposure of susceptible patients to parent drug and metabolite results in the development of divergent T cell responses that act together produce the adverse event. HLA-B*13:01–restricted dapsone- and nitroso dapsone–responsive CD8⁺ T cells are detectable in assays using APCs from donors expressing matching HLA-B alleles. Nitroso dapsone activates T cells via two pathways: first, through direct covalent modification of peptides embedded within MHC expressed on the surface of APCs; and, second, through formation of protein adducts that undergo Ag processing to generate peptides that associate with MHC before transport to the cell surface for presentation to T cells. The same clone may be activated by both pathways with the surface peptide adduct, presumably mimicking the adduct formed naturally through protein processing. No information is available regarding the nature of protein adducts that activate T cells (including whether they are formed intra- or extracellularly), the different uptake pathways involved in internalizing adducts, and the enzymes involved in breakdown of the adducts. For this reason, we have designed and synthesized nitroso dapsone–modified HLA binding peptides to study the HLA-B*13:01–restricted T cell response. Three nitroso dapsone–modified peptides were synthesized in high purity. Each peptide contained two HLA-B*13:01 anchoring motifs, an alanine backbone (previous studies show that nonanchoring amino acids for the most part do not define the specificity of the T cell response [41, 42]) and a nucleophilic cysteine residue for modification by nitroso dapsone. An Fmoc-protecting group, which was removed before purification, was used to prevent nitroso dapsone N-terminal binding. Reasonable yields of nitroso dapsone–modified Pep1 and Pep3 were obtained after HPLC purification; Pep2 was generated in a lower quantity and, as such, only used in limited T cell cross-reactivity studies. Pep1 contains a cysteine residue in the 4 position, which is an important trinitrophenol hapten binding site for the generation of immunodominant CD8⁺ peptide epitopes (43). In contrast, trinitrophenol modification at more distal positions (e.g., position 7 in Pep3) generates qualitatively different determinants that tend to activate a lower frequency of T cells (43). Modeling revealed that the predicted conformation of dapsone-modified peptides contained the bulky drug aromatic groups, pointing out the binding groove to ensue favorable interaction of P2 and P9 anchor residues with HLA-B*13:01.

Two lymphocyte transformation test–positive (dapsone and nitroso dapsone) hypersensitive patients (described in Zhao et al. [14]) expressing HLA-B*13:01 were used to generate nitroso dapsone–modified peptide responsive T cell clones. Almost 50% of clones generated from 14-d nitroso dapsone–modified peptide PBMC cultures were stimulated to proliferate in the presence of either modified Pep1 or Pep3. Clones expressed a CD8⁺ phenotype and were stimulated with the modified peptides in a dose-dependent manner. Clones displayed 100% cross-reactivity between positional derivatives and with soluble nitroso dapsone. Nitroso dapsone extensively modified the surface of APCs, which likely includes binding to cysteine-containing peptides already displayed by HLA class I on the cell surface; hence, the observed cross-reactivity was expected. In contrast to our cross-reactivity data with nitroso dapsone–modified peptides, cross-reactivity between HLA class II binding β-lactam–modified peptide positional derivatives was not observed (16, 44). However, a system used by Honda et al. (45), in which CD8⁺ TCR α and β-chains were alternately fixed prior to assessment of trinitrophenol-modified peptide positional derivative T cell responses, explains how a single TCR is capable of recognizing and responding to hapten structures in different positions. They demonstrated that hapten addition to HLA class I binding peptides (at positions 4 and 6) can cause substantial adjustments to the CD8⁺ TCR structure. Specifically, the β-chain could adjust to interact with the hapten structure irrespective of whether it was at position 4 or 6.

The possibility that the T cell activation with nitroso dapsone–modified peptides may be due to residual soluble nitroso dapsone or degradation of the adduct in culture and liberation of dapsone was excluded through: 1) demonstrating that the clones were not activated with the parent compound; and 2) neutralizing soluble nitroso dapsone-specific, but not nitroso dapsone–modified peptide-specific, T cell responses with glutathione. Glutathione contains a reactive cysteine group and, when in excess, binds to nitroso dapsone, preventing formation of protein adducts.

To explore the importance of HLA proteins in T cell activation, APCs were firstly excluded from T cell proliferation assays. The vast majority of clones were not activated with the nitroso dapsone–modified peptides in the absence of APCs. Next, the HLA-A, B-negative mutant C1R cell line was transduced with HLA-B*13:01 and used as APCs in the place of autologous EBV-transformed B cells. Nitroso dapsone–modified Pep1 and Pep3 and soluble nitroso dapsone stimulated the clones to secrete IFN-γ in the presence of C1R-B*13:01 cells, and the response was inhibited with an HLA class I blocking Ab. Similar activation of the T cell clones was not observed when using the C1R parental cell line as APCs.

Collectively, our study highlights the importance of drug metabolism, drug hapten binding, and, most importantly, the formation of drug metabolite-modified HLA-B*13:01 binding peptides in the activation of CD8⁺ T cells from dapsone-hypersensitive patients. The availability of T cell stimulatory nitroso dapsone–modified HLA-B*13:01 binding peptides and C1R-B*13:01 cells offer the opportunity to study the structural elements of the drug hapten HLA peptide binding interaction.

The authors have no financial conflicts of interest.

We thank the patients and volunteers for agreeing to donate blood samples.