T-cell mediated adverse drug reactions can be life-threatening, with mortality rates up to 50% in certain severe cutaneous adverse drug reactions (SCAR) (Peter et al., 2017). Antimicrobial drugs are responsible for a significant burden of these reactions; more so in high burden tuberculosis (TB) and human immunodeficiency virus (HIV) settings (Blumenthal et al., 2019). Delayed hypersensitivity reactions to antimicrobials frequently occur in the setting of multidrug regimens e.g., TB infection; and rapid, accurate, and safe diagnostic tools to aid the identification and exclusion of the offending drug are critical (Lehloenya et al., 2020). Drug provocation testing is often considered contraindicated in most settings of severe, life-threatening adverse drug reactions; while systemic reactions to in vivo diagnostics such as patch testing has been reported, especially in the context of HIV infection (Lehloenya et al., 2020). Thus, in vitro and ex vivo diagnostics are appealing and have been increasingly employed in the evaluation of presumed T-cell mediated hypersensitivity reactions (Lehloenya et al., 2020).

The Enzyme-Linked ImmunoSpot (ELISpot) assay, an ex vivo diagnostic, was initially introduced in 1983 (Sedgwick and Holt, 1983) and detects locally secreted cytokine molecules, such as IFN-γ, using antibody-coated plates offering both a qualitative but also a quantitative specific protein measurement, measuring cytokine production on a per-cell basis (Cox et al., 2006). Successful IFN-γ production requires adequate, live T lymphocytes. Despite an increasing use of IFN-γ release ELISpot for drug causality assessment and utilization of a range of antimicrobial concentrations ex vivo, data regarding antimicrobial-associated cellular cytotoxicity and implications for assay performance remain scarcely described in the literature.

We used the peripheral blood monocytes (PBMCs) from three control participants and two patients with IFN-γ ELISpot confirmed rifampicin-associated drug reaction with eosinophilia and systemic symptoms (DRESS). To understand the required blood quantity for each assay, one needs to keep in mind that, for a healthy control, 9 ml of blood gives us approximately 1 × 10⁷ PBMC. About 3 × 10⁷ (or 27 ml of blood) is required for each executed assay (e.g., ELISpot, LDH, or flow cytometry).

For these patients, the blood was collected less than 2 weeks after their acute DRESS reaction and they were not known to have any other drug hypersensitivities. PBMCs were isolated from heparinized peripheral blood [ethylenediaminetetraacetic acid (EDTA) tubes] by density gradient centrifugation (using polysaccharide Ficoll Paque®) as described previously (Trubiano et al., 2018) and preserved in liquid nitrogen or a −80° freezer until assessment. After thawing, prior to sample use, cell viability was measured using 0.4% trypan blue staining solution (Thermo Fisher Scientific®) and the Countess® automated cell counter (Fisher Scientific, 2010).

As IFN-γ release ELISpot is increasingly used for drug causality assessment in patients with life-threatening, severe immune-mediated adverse drug reactions (IM-ADRs), ELISpot has been shown to be an efficient ex vivo tool in the diagnosis of antibiotic-associated SCAR patients (Trubiano et al., 2018) by offering the benefit of observing the patient’s reaction to a drug without re-exposing the patient. To maximise assay sensitivity, it is advantageous to use the highest possible drug concentrations, especially in situations where the mechanism of drug-immune interaction is poorly defined. In this project, we examine antimicrobial-associated cellular cytotoxicity for common offending antimicrobials and its impact on assay performance. We describe two patients with rifampicin-associated DRESS (one HIV negative and one positive); as well as three control participants where the cytotoxicity of different antimicrobials used in clinical diagnostic was evaluated. Our main findings were: i) that in both the HIV negative and positive rifampicin-associated DRESS patients, there was a dose-dependent reduction in SFU/million cells in the ELISpots corresponding to increasing rifampicin-induced drug cytotoxicity; ii) for all the antimicrobials evaluated, the use of Cmax and 10-fold Cmax concentrations did not reach cytotoxicity seen with rifampicin and are therefore unlikely to affect ELISpot performance; and iii) for drugs such as piperacillin/tazobactam and isoniazid even concentrations as high as 100-fold Cmax did not result in cytotoxicity likely to reduce ELISpot performance.

ELISpot is an assay that requires optimization according to the population and drug of interest as well as the specific cytokine measured (Ji and Forsthuber, 2016). At high concentrations, drugs become toxic for cells via different mechanisms like overproduction of nitric oxide; generation of reactive oxygen species, and subsequent oxidative stress; mitochondrial dysfunction; and DNA damage. Not all drugs have equivalent cytotoxicity at the same concentrations and thus these experiments inform the optimization of diagnostic ELISpot for delayed IM-ADRs. Indeed, rifampicin appeared the most toxic of all tested drugs as illustrated with over 90% cytotoxicity at 100-fold Cmax concentrations and complete abrogation of SFUs in the ELISpot assay. Mechanisms of increased cytotoxicity with rifampicin are uncertain but the formation of drug-antibody complexes, which binds to platelet membrane activating complement and increasing cell death has been proposed (Combrink et al., 2020).

Various drug concentrations are described in the literature for ex vivo assays with varying units of measure including, millimoles (isoniazid, 1–4 mM; rifampicin, 0.1 mM) for tuberculosis related drugs (Usui et al., 2017) or micromoles (El-Ghaiesh et al., 2012) and mg/ml (Rozieres et al., 2009; Tanvarasethee et al., 2013; Sedlackova et al., 2018) for beta-lactam drugs. Thus, a strength of this project is the use of the measured Cmax as a reference to ensure that these units are comparable across drugs and concentrations are optimised as being the point of maximal bacterial cytotoxicity with minimum tolerable cellular injury (Supplementary Table S1).

We noted higher percentage of cytotoxicity with the LDH assay compared with the 7-AAD staining method. The 7-AAD staining involves a nuclear dye, while the LDH assay measures the oxidation of the NADH⁺ after cell membrane damage and LDH release into the extracellular medium. This distinction in assays could explain the difference observed in Figure 1. For the 7-AAD staining, all the drugs tested at the three concentrations, except rifampicin, showed relatively low cytotoxicity (<30%) while for the LDH assay, most drug concentrations 10-fold higher than Cmax resulted notable increases in cytotoxicity (>40%). The localisation of the damage—cytosolic (LDH assay) versus nuclear (7-AAD) may explain these differences, but the correlation of both assay with decreased SFUs in the rifampicin ELISpot provides clear data linking the differing degrees of cytotoxicity with decreasing assay performance.

Using flow cytometry for evaluating cytotoxicity can be performed with the uptake of labeled annexin V, propidium iodide or the uptake of fluorescent DNA probes 7-amino-actinomycin (7-AAD), as in our study (Saade et al., 2012). In apoptotic cells, membrane phosphatidylserine, which is located in the inner part of the membrane in healthy cells, translocates to the outer leaflet and is exposed to the external environment (Saade et al., 2012), where it binds with annexin V staining providing evidence of early cell death. Further, using 7-AAD dye gives information on loss of membrane integrity and later stages of cell death.

The study has limitations that include the inability to evaluate all the different drug concentrations for all antimicrobials in patients where the drug is known to be offending, and to directly match cytotoxicity data with ELISpot performance. However, the rifampicin data shows proof of concept and provides guidelines on when cytotoxicity levels are likely to effect assay performance. In addition, as with any laboratory assay, technical limits particularly for long-lasting cell cultures and complex manipulations as well as operator-dependent variability could influence results. Other potential considerations are sample irregularities in terms of time and condition of sample storage as well as drug formulation and preparation.

Our results show that the known Cmax for each antimicrobial and the 10-fold increased drug concentrations showed a mean percentage of cytotoxicity below 30%, allowing a significant live population to evaluate in ex vivo experiments when these concentrations are employed. An increase in cytotoxicity was translated by fewer SFU in the ELISpot illustrating the impact of cell death on reduced ELISpot performance. These exploratory findings will inform approaches for ex vivo diagnostics such as IFN-γ release ELISpot increasingly utilized in drug causality assessments of rare severe T-cell mediated hypersensitivity.