RESULTS/DISCUSSION
In this study, we included PBMC and BFC samples that had been cryogenically stored from four patients with confirmed SCAR (Cases 1-4) including SJS, TEN, DRESS, and generalised bullous fixed drug eruption (GBFDE) identified from previous prospective studies (Supplementary Methods ). All patients had a Naranjo score of 4 or higher[11], a minimum Scorten score of 2 and a minimum Alden score of 4 for SJS/TEN[12, 13]. Cases 2 and 4 had one implicated drug while Cases 1 and 3 had three implicated drugs and all cases were receiving the implicated drug at time of rash onset (Table 2). The latency period for cases (defined as time between drug commencement and rash onset) ranged from 0-38 days with a median value of 18.5 days. PBMC and BFC collection delay for testing had means of 24 and 22 days, respectively. Case 4 had a delayed collection latency for PBMC and BFC of 48 and 49 days, respectively. Baseline demographics, clinical features and biological sampling are shown in Tables 1 and 2 .
IFN-γ ELISpot was performed in matched PBMC and BFC samples from Cases 1-4, as per previously published methods[7] andSupplementary Methods (Figures 1 and 2 ,Supplementary Figures 1 and 2) . Two of these patients displayed positive ELISpot results (defined as SFU ≥ 50U/million cells [7, 8]) upon ex vivo challenge with suspected drugs for both PBMC and BFC (Cases 1-ceftriaxone and 4-oxypurinol), while Cases 2 and 3 only displayed a positive result with BFC (Figure 1).
Case 1 BFC tested positive for both doses of ceftriaxone (200 and 2000µg/mL), whilst matched PBMC only tested positive to the highest dose, with half of the response elicited in BFC. Case 4 BFC tested positive at both concentrations (5 and 50µg/mL) of allopurinol in addition to its metabolite, oxypurinol, while PBMC only showed a positive response to oxypurinol. This suggests that BFC analysis can provide higher sensitivity to drug-allergy testing than PBMC. This is further supported by analysis of Cases 2 and 3 whereby positive IFN-γ release ELISpot responses were detected using BFC but not PBMC samples. Case 2 BFC displayed positivity to all doses of meloxicam (2, 20 and 200 µg/mL) and Case 3 only to the highest dose of sulfamethoxazole (SMX-500µg/mL), its metabolite 4-Nitro-SMX (100µg/mL), as well as to the commercial product (Bactrim©; trimethoprim-sulfamethoxazole) at 50 and 250µg/mL of the sulfamethoxazole component, respectively.
Flow cytometry was used to investigate whether different cellular compositions of matched BFC and PBMC for Cases 1, 3 and 4 could account for differences in ELISpot sensitivities (Figure 2, Supplementary Figure 1-2) . We found that BFC samples were enriched for total T (CD3+) cells and for IFN-γ-secreting cells, relative to matched PBMC (Figure 2A, Supplementary Figure 1 and 2 ). The total proportions of CD4+, CD8+, or double negative (DN) T cells varied across individuals (Figure 2A, Supplementary Figure 1 ), likely reflecting differences in the pathology and/or treatments, with Case 1 displaying a strong bias for CD4+ T cells, which is typical of DRESS[14]. In Contrast, Case 4 BFC were enriched for CD8+ T cells relative to matched PBMC, which may be associated with a delayed BFC sampling, compared to other cases (Table 2 ), possibly reflecting CD8+ T cells egress from peripheral blood[8]. Cases 3 and 4 BFC showed an enrichment for T cell populations with a tissue residency/recruitment (CD69+CD103+) phenotype, which have been implicated in SCAR[8] (Figure 2A, Supplementary Figure 1 ). Case 3 BFC samples further displayed higher proportions of memory (CD45RO) and activated (CD69) T cells, relative to PBMC, whilst remaining similar for Cases 1 and 4, which may partly account for the differences in ELISpot sensitivities between the two samples (Figure 1 ). As unconventional T cells (not HLA-restricted) are also known to produce IFN-γ, and their role in SCAR remains unexplored[15], we assessed the proportions of γδ T cells, mucosal-associated invariant T (MAIT) cells and CD56-expressing T cells (T cells expressing natural killer (NK) markers, likely including natural killer T (NKT) cells)[16]. While MAIT cells and γδ T cells did not show preferential recruitment into BFC (Figure 2A and Supplementary Figure 1 ), they were found among IFN-γ+ populations (Figure 2B, Supplementary Figure 2B ), representing a large proportion of Case 3 PBMC (38.4% and 11.4%, respectively). IFN-γ secreting cells also comprised NK-like T cells (CD56+CD3+), and NK cells (CD56+CD3-) - Case 3 PBMC. Overall, IFN-γ-secreting cells comprised CD4+, CD8+ and DN (CD4-CD8-) T cells, with preferential enrichment for CD8+ T cells in BFC from Cases 3 and 4 and displayed memory and activated phenotypes (CD45RO+/CD69+) (Figure 2,Supplementary Figure 2B ). Overall BFC samples display T lymphocytes that have been recruited from the blood or adjacent tissue with an activated phenotype and cytokine secreting capacity. This leads to higher proportions of cells with an IFN-γ secreting capacity (when compared to blood), which may reflect higher representations of the drug-antigen-specific clones. Collectively, these results suggest a higher sensitivity for BFC samples in ELISpot-testing relative to PBMC, likely reflecting differences in their cellular composition.
Ex vivo drug-allergy diagnostics have an increasing evidence base and clinical demand[3]. By analysing samples from four SCAR patients with distinct drug-allergies and clinical manifestations that are presumed to be T-cell-mediated, this study provides impetus for further work to explore alternative sampling sources for drug-allergy diagnostics. At present there is no gold-standard diagnostic for causality assessment in SCAR, and previous studies, whilst showing promising sensitivity[3], remain limited. Our results suggest higher sensitivity for BFC analysis relative to matched PBMC using ex vivo IFN-γ release ELISpot [3]. Whilst limited by low numbers and cell viability, the rare nature of both blister fluid capture and SCAR cases that have been accurately phenotyped provides a unique insight into the diagnostic potential for this IFN-γ release ELISpot assay.
Our results are consistent with recruitment of known populations involved in the pathology (T cells with a memory/activated phenotype and cytokine-secreting capacity) into blisters. We further reveal that, relative to BFC, PBMC may have lower representation of cells with an IFN-γ-secretion capacity[7, 8]. How much IFN-γ detection by ELISpot is due to direct activation of drug-specific cells or bystander secretion of non-specific cells remains to be understood, and it may vary with the drug causing SCAR. It is possible that some drug-specific cells may produce cytokines other than IFN-γ (such as TNF, IL-4, IL-17) upon activation that have not been tested. Whilst we also assessed IL-17-secretion using flow cytometry, our results do not seem to suggest that this could be a key contributor for the responses studied (Supplementary Figure 3 ). This may require ELISpot assays for other cytokines or markers yet to be identified, or even more generic activation assays using cellular activation markers like CD69, CD107a. Thus, we recommend that clinicians sample BFC, whenever available, for testing with ELISpot assays in drug-allergy diagnostics, whilst retaining correlation with PBMC results. This may prove to be an invaluable resource for future studies aiming at characterising the immunopathogenesis and HLA (or HLA-like) restriction of these drug-induced allergies, including drug-presentation pathways, cell populations involved, and cytokine-outputs. Such knowledge may ultimately lead to improved diagnostics for SCAR patients, improving efforts to lower the significant morbidity and mortality associated with SCAR.