In vitro detection and characterization of drug hypersensitivity using flow cytometry


  • Edited by: Hans-Uwe Simon

M. Martin, Federal Institute for Drugs and Medical Devices, 53175 Bonn, Germany.


Background:  The lymphocyte transformation test (LTT) is the only in vitro test for detecting drug sensitization at the cellular level irrespective of the reaction’s phenotype. However, the LTT includes working with radioactive substances and is considered impracticable for routine laboratory investigation.

Objective:  The aim of this study was to assess drug-specific cytokine production by means of flow cytometry as an alternative nonradioactive approach which may be more appropriate for routine testing and may provide in addition more information about the pathophysiology of the reaction than proliferation-based assays, like the LTT.

Method:  Peripheral blood mononuclear cells of 19 patients were incubated with culprit drugs (n = 28) or irrelevant antigens (n = 10). Ten healthy persons served as controls for all different drugs (n = 15). Intracellular interleukin (IL)-5, interferon (IFN)-γ and IL-10 production was investigated using flow cytometry. Accuracy of the flow cytometry test system was confirmed using different statistical tests, i.e. receiver operating characteristic curve and Mann–Whitney rank test. In addition, drug-specific secretion of IL-5, IL-2 and IFN-γ were analysed using enzyme-linked immunosorbent assay (ELISA).

Results:  Drug-specific cytokine production could be demonstrated in 75% of the patients using flow cytometry and in 79% using ELISA respectively. Combining ELISA and flow cytometry increased the sensitivity to 100%. Analysis of involved T-cell subsets [e.g. CD4+ or CD8+; T helper (TH) 1 or TH 2] allowed characterization of the in vitro lymphocyte reactivity pattern.

Conclusions:  Analysis of drug-specific cytokine production by means of flow cytometry proved a useful and reliable approach for the in vitro detection and characterization of drug hypersensitivities.


enzyme-linked immunosorbent assay






lymphocyte transformation test


maculopapular exanthema


normalized mean fluorescence intensity


optical density


peripheral blood mononuclear cells


phosphate buffered saline






receiver operating characteristics


stimulation index


T helper

Drug hypersensitivity reactions account for only about one-seventh of all adverse drug reactions (1), but they represent a major problem in drug safety as result of their severity and their unpredictability (2, 3).

However, conclusive diagnosis of drug allergy still remains a major problem in daily clinical practice (4, 5). In vivo tests, such as patch, prick and intracutaneous tests often do not yield positive reactions, even in patients with well-documented histories of drug-allergic reactions (6, 7). Moreover, anaphylactic reactions associated with skin testing have been reported (8, 9). Challenge tests, which are considered to be the gold standard, are frequently not tolerated by the patients, bear the risk of severe reactions and do not differentiate between allergic and pseudoallergic reactions (10, 11).

As with in vivo tests, the usefulness of in vitro tests is also limited. Certain methods in use, e.g. the basophil activation test are linked to the effector phase of the reaction. Hence, many drug-allergic reactions like the common maculopapular exanthema (MPE), which usually do not involve formation of drug-specific IgE, cannot be detected by means of these diagnostic methods (1).

Currently, the lymphocyte transformation test (LTT) is the only available in vitro tool for detecting drug sensitization at the cellular level, irrespective of the effector mechanism and the clinical phenotype of the reaction (1).

However, the LTT imposes limitations in terms of practicability and sensitivity (1). In this respect, others and we have demonstrated that the in vitro detection of drug-specific cytokine production by peripheral blood mononuclear cells (PBMC) appears to be an adequate alternative for detecting drug hypersensitivities (2, 3).

In this study, this approach was further explored by focusing on flow cytometry as a test system and by including a panel of different cytokines. Cytokine secretion using ELISA-techniques as described elsewhere was assessed in parallel for validation.

Materials and methods

Patients’ characteristics

Nineteen patients (average age: 53 years, range 29–84 years) with a history suggestive of drug hypersensitivity were included in this investigation serving for 38 antigen testings (n = 28 culprit drugs, n = 10 irrelevant antigens) (Table 1). In all individuals, drug allergy was confirmed by stringent drug allergy history category A according to Nyfeler and Pichler (1).

Table 1.   Clinical phenotypes of the 19 patients investigated
ReactionPatient Sex/ageTime spanDrug of main interestAdditional drug of interest*Irrelevant antigen
  1. TEN, toxic epidermal necrolysis; SJS, Stevens-Johnson syndrome; Misc., miscellaneous.

  2. *History of drug hypersensitivity to the substance, although not further evaluated.

  3. †Exanthema not further specified.

Bullous reactions n = 3
 TEN1F/528 weeksLamotrigine  
 SJS2F/548 weeksAmoxicillineAmpicilline
Penicillin G
 SJS3M/604 weeksProguanil  
Urticaria n = 24F/444 weeksDoxycycline Penicillin G
5F/448 weeksAmoxicilline Cotrimoxazole
Angioedema n = 4
 6F/5820 yearsPenicillin G  
 7M/842 weeksAmpicillineCotrimoxazole
 8M/528 weeksOmeprazoleIbuprofen 
 9F/592 yearsPenicillin G Metamizole
Maculopapular exanthema n = 3
 10F/584 weeksAmoxicillinePenicillin G 
 11F/624 weeksClindamycine Penicillin G
 12M/4112 weeksAmoxicilline  
Exanthema†n = 4
 13F/4120 weeksCarbamazepine Valproic acid
 14F/4012 weeksClarithromycineAmoxicillinePantoprazole
 15F/294 weeksClindamycine Penicillin G
 16F/675 yearsNa-PerchlorateIodixanol 
Misc. n = 3
 Dyspnoea17F/448 weeksDoxycycline Cotrimoxazole
 Nausea18F/525 yearsNadroparine  
 Unknown19F/6020 yearsPenicilllin GOmeprazole 

A suspected atopy was not defined as a criterion for exclusion. The following classification is based on analysis of our own questionnaires and patients’ files: (i) atopy excluded (7/19); (ii) atopy confirmed (4/19) and (iii) atopy neither excluded nor confirmed (8/19).

Two control groups containing nonatopic individuals were included. In the first control group (n = 6), a healthy control person, if possible age- and sex-matched without allergy to the tested drug and without intake of the drug in the last 12 months was matched to a respective patient. PBMC of these control persons were studied to rule out unspecific stimulating effects of the tested drugs. In the second control group, PBMC of four individuals were investigated who were currently or in the last 3 months exposed to the respective drug under investigation without clinical signs of a drug-allergic reaction.

Altogether, these control persons (n = 10) were tested with 15 different antigens used in this study. It was ensured that every substance tested in a patient was also tested at least once in a control person, although the total number of testings in patients exceeded those in controls because of several-fold testings.

This study was approved by the ethics committee of the Medical faculty of the RWTH Aachen (EK 100/06) and all patients gave written informed consent prior to participation.

Drugs and antigens

Drugs were used at nontoxic concentrations as described elsewhere (12). Directly before use, all drugs were dissolved in phosphate buffered saline (Dulbecco’s PBS; PAA Laboratories GmbH, Pasching, Austria) and diluted in up to three different concentrations which were analysed in the ELISA. The medial concentration was analysed in the flow cytometry system, which had been proved to be optimal for the stimulation and detection of PBMC in our own preceding experiments. All drugs except for the penicillins (1.0, 0.5, 0.1 mg/ml) and omeprazole (50, 10, 1 μg/ml) were included in concentrations of 100, 10, 1 μg/ml. The following antigens were obtained as pure substances: phytohemagglutinin (PHA), amoxicillin, carbamazepine, ibuprofen, penicillin G, valproic acid (all from Sigma-Aldrich, St Louis, MO, USA). Solutions for infusion were used for ampicilline, clindamycine, cotrimoxazole, doxycycline, metamizole (all from Ratiopharm, Ulm, Germany), cefuroxime (Fresenius, Bad Homburg, Germany), clarithromycine (Abbott Laboratories, North Chicago, IL, USA), nadroparine-Ca (Glaxo Smith Kline, London, UK), metronidazole (Infectopharm, Heppenheim, Germany), omeprazole (AstraZeneca, Wedel, Germany), pantoprazole (Nycomed, Konstanz, Germany) and tetanus toxoid (Sanofi Pasteur MSD, Paris, France). Additional solutions were used for Na-perchlorate (Bayer Vital, Leverkusen, Germany) and iodixanol (GE Healthcare Buchler, Braunschweig, Germany). Lamotrigine (1A Pharma, Oberhachingen, Germany) and proguanil (AstraZeneca) solutions were prepared based on crushed tablets.

Cell preparation and culture

The LTT served as a methodological platform in the experiments and hence, the first steps were performed in accordance with the LTT protocol as described elsewhere (1). Briefly, PBMC from patients and controls were isolated from heparinized whole blood by gradient centrifugation over Ficoll-Hypaque solution (Sigma-Aldrich, St Lois, MO, USA) and cultured for up to 6 days in 96-well round-bottom plates as 10-fold culture in RPMI 1640 supplemented with 5% autologous heat inactivated plasma, sodium pyruvate (1 mM) and l-glutamine (2 mM) at a concentration of 1 × 106 cells/ml under humidified conditions and 5% CO2 (all reagents PAA Laboratories GmbH).

At the beginning of the incubation (=day 0) the indicated drugs dissolved in PBS or with PBS alone (control) were added to the PBMC cultures in different concentrations. Incubations of the PBMC with PHA and tetanus toxoid served as positive controls for mitogen-induced and memory T-cell antigen-specific lymphocyte reactivity respectively.

From the 10 wells assigned to each drug concentration, four wells were dedicated to flow cytometry and six wells for the determination of interleukin (IL)-2, IL-5 and interferon (IFN)-γ in the supernatants by means of ELISA (two wells for each cytokine). Cell culture supernatants for analysis in the ELISA were taken at different time points from separate wells reflecting optimal conditions for the individual cytokine (see Results).

Flow cytometry test system

In contrast to the ELISA, only the wells incubated with the middle drug concentration were selected for flow cytometry analysis. Cells were harvested after 5 days of incubation, for the last 16 h in the presence of brefeldin A (6 μg/ml) and then stained with anti-CD3 PerCP-Cy5.5, anti-CD4 FITC, anti-CD8 APC (BD Biosciences, Milpitas, CA, USA) for T-cell phenotyping. Fixed and permeabilized cells (Cytofix/Cytoperm Plus Kit with BD GolgiPlug containing brefeldin A, BD Bioscience) were subsequently stained with phycoerythrin (PE)-labelled anti-IFN-γ, anti-IL-5 and anti-IL-10 specific monoclonal antibodies (BD Biosciences) for intracellular cytokine detection. Stained cells were analysed in a FACSCalibur (BD Biosciences) and data were processed with CellQuest software (BD Biosciences). 10 000 CD3+ cells were assessed in each sample. The recommended PE-conjugated IgG1κ and IgG2aκ were used (BD Biosciences) as isotype controls.

ELISA test system

The sandwich enzyme-linked immunosorbent assays (ELISA) for human IL-5 (range: 7.8–500 pg/ml), IL-2 (range: 7.8–500 pg/ml) and IFN-γ (range: 4.7–300 pg/ml) were obtained from BD Biosciences. The minimum detectable dose was defined by us as 2 SD above the mean optical density (OD) of 20 replicates of the zero standards. In preliminary testings minimum detectable doses between 0.5 and 1.5 pg/ml were obtained.

According to similar instruction manuals for precoated plates, we defined a mean value of 1.0 pg/ml for all cytokines (IL-2, IL-5 and IFN-γ) as the cut-off. A Dynex MRX micro plate reader (Dynex Technologies, Berlin, Germany) was used for the assays. The results of the highest value (based on three drug concentrations) are expressed as cytokine concentration (pg/ml).

Statistical analysis

The Mann–Whitney rank test, boxplots and receiver operating characteristic (ROC) curve were employed using spss software version 17.0 (SPSS Inc., Chicago, IL, USA).


Preceding experiments

Preceding experiments were carried out to determine the conditions (including time courses) for the analysis of in vitro drug-specific activated PBMC in primary cultures in the following two systems: (i) flow cytometry (detection of intracellular cytokines) and (ii) ELISA (secreted cytokines). Concerning the flow cytometry test system, a maximum yield of drug-specific IL-5, IFN-γ and IL-10 producing T cells was obtained after 5 days of incubation with the causative drug and with tetanus toxoid. In the ELISA, maximal concentrations of IL-5 and IFN-γ were detected after 6 days of incubation and maximal secretion of IL-2 was observed after 4 days of incubation (data not shown).

Establishing flow cytometry as an in vitro tool

Table 2 provides an overview of the results obtained in the flow cytometry test system following stimulation of PBMC from 19 patients with drug hypersensitivity. In this system, intracellular detection of selected cytokines in T cells and their subsets was used as read-out parameter in contrast to the detection of T-cell proliferation in the conventional LTT (Fig. 1).

Table 2.   Flow cytometry and ELISA results of 28 culprit drugs
Culprit drugPatients*T-cell subset detected using flow cytometrySecreted cytokine detected using ELISA (pg/ml)
  1. *Patient number (Table 1).

Amoxicilline (n = 5)5++5
Ampicilline (n = 2)2++
Clindamycine (n = 2)11++
Doxycycline (n = 3)43
Penicillin G (n = 5)2+
Omeprazole (n = 2)8242
Sensitivity (%) 504643434357
Figure 1.

 Gating strategy for the detection of intracellular cytokine production. In this example, upon exposure of PBMC from a patient with carbamazepine-induced hypersensitivity to the drug of interest carbamazepine, a substantial increase for the respective cytokine was noted. The gating strategy is shown exemplarily (A) for the following IL-5 analysis. 10 000 CD3+ T cells were selected (gate R1) and further differentiated into CD4+ or CD8+ T cells (B). Data were expressed as delta values (difference of calculated absolute cell count of cytokine positive T cells with and without stimulation).

Employing an ROC curve, the accuracy of this assay was analysed. The coordinates of the curve provided guidance to determine what delta values (difference of cytokine positive T cells in the cultures with antigen and without antigen) should serve as the cut-off to distinguish positive from negative results (Fig. 2). Based on the analysis of 54 results considered positive and 300 negative results (results considered negative n = 174 plus controls n = 126), a cut-off for the delta value of 100 cells was applied.

Figure 2.

 Characterization of the accuracy of both test systems: flow cytometry and ELISA ROC curve for differentiating between positive and negative results. The relationship between sensitivity and specificity is plotted. Flow cytometry (A): the optimum cut-off for the identification of positive results at a delta value of 100 cells provided a sensitivity of 100% and a specificity of 97%. The area under the curve (AUC) was 0.986 ± 0.006. ELISA (B): the delta value of an optical density difference of 0.035 provided a sensitivity of 100% and a specificity of 99%. AUC = 0.996 ± 0.002.

Determination of cytokine secretion by means of ELISA

In parallel to the analysis of intracellular cytokine production using flow cytometry, secretion of cytokines by the stimulated T cells was analysed by means of ELISA. In Table 2, the highest absolute cytokine concentration obtained at three different concentrations is shown.

Based on the analysis of 64 results considered positive and 509 results considered negative, a cut-off for an OD difference between unstimulated and stimulated sample of 0.035 was applied. Statistics were performed with the OD and not the cytokine concentration, due to the fact that the OD is the parameter that is primarily detected.

To give an estimate for comparison, an OD difference of 0.035 corresponded with concentration differences of about 2–3 pg/ml depending on individual test conditions (e.g. calibration line).

Statistical analysis of investigated parameters

In Fig. 3, the obtained data for reactions considered positive or negative and controls are depicted by means of boxplots containing the respective five-number summaries: the smallest observation, lower quartile, median, upper quartile and largest observation. The differences between the positive results and the corresponding controls were tested using the Mann–Whitney test and P-values were calculated for each parameter and method. Concerning flow cytometry, statistical significant differences between positive delta values from drug-allergic patients in comparison with delta values from controls were observed for all parameters. The corresponding statistical P-values for significance ranged between 0.000 and 0.008. With regard to the ELISA findings, Mann–Whitney testing revealed statistical significant differences between stimulation index (SI) considered as positive or negative and controls, respectively, for all cytokines investigated (P-values of 0.000 for IL-2, IL-5 and IFN-γ, data not shown).

Figure 3.

 Comparison of different parameters. The boxplots depict the delta values (△) of patients tested positive or negative and respective controls, through their five-number summaries: the smallest observation, lower quartile, median, upper quartile and largest observation. Statistically confirmed outliers (bsl00001, •) are shown separately. Differences between the study groups were tested using the Mann–Whitney test and are described as P-value.


Based on the testing of 28 drugs of interest, the sensitivities of both in vitro approaches were analysed separately according to the different single parameters and the two different methods (Table 2). Single cytokine analysis led to sensitivities ranging from 43% up to 57%, whatever method used. By testing a panel of relevant cytokines a sensitivity of 75% in the flow cytometry and 79% in the ELISA was observed. Combining both methods increased sensitivity to 100%.


No significant cytokine concentrations or productions were detected when PBMC of control persons were incubated with the various drugs investigated. Substantial production and secretion of cytokines were detectable following stimulation with tetanus toxoid or the mitogen PHA, serving as positive control for PBMC of patients and controls. Furthermore, in sensitized patients no T-cell activation could be detected following stimulation with nonsuspected tolerated co-medication or other irrelevant antigens (n = 10) (Table 1) resulting in a specificity of 100%.

In vitro lymphocyte reactivity pattern

Table 2 provides an overview of the results of all parameters. Flow cytometry offered information about the in vitro phenotype of the reactions and allowed to differentiate between the involvement of: (i) CD4+ or CD8+ T cells; (ii) T helper (TH) 1 (IFN-γ) or TH 2 (IL-5) T cells. The ELISA provided additional information about the cytokines secreted. No consistency between the involved T-cell subset and either the clinical phenotypes or the class of drugs could be observed in both methods.


In this study, we could demonstrate that determination of drug-induced IL-5, IFN-γ, IL-10 production by means of flow cytometry proved a nonradioactive, alternative approach for the in vitro detection of drug sensitization. To the best of our knowledge, the employment of flow cytometry in this specific context has not been described before (13, 14).

As a precondition for our statistical model, we hypothesized that in vitro lymphocyte reactivity should be demonstrable in PBMC of all 19 selected patients (category A according to Nyfeler and Pichler; 1). To establish a test system that proves itself on ‘real-life conditions’, no restriction concerning culprit drugs or clinical phenotypes was made.

The TH 2-like cytokine IL-5 and the TH 1-like IFN-γ were selected as in vitro read-out parameters for both methods based on studies suggesting their inclusion in this context (2, 15, 16).

Likewise, secretion of the T-cell growth factor IL-2 was analysed in the ELISA system with reference to its inclusion in the previous studies. In the flow cytometry test system, we additionally investigated IL-10, because this TH 2-like cytokine is assumed to skew TH 1 into a TH 2 immune response (17, 18). The usefulness of such a cytokine panel containing IL-5 and in addition IL-2 or IFN-γ has also been described for extracellular cytokine detection (19).

Interestingly, we found drug-specific cytokine detection to be time dependent in our experiments. Concerning flow cytometry, preceding investigations indicated that for all cytokines studied, analysis is best performed after 5 days of incubation, as no positive reactions were observed with a shorter incubation time. Conversely, the vulnerability of the PBMC, e.g. during the staining process, increased substantially beginning with the 6th day of incubation.

In the ELISA, maximal concentrations of IL-5 and IFN-γ were observed on the 6th day of incubation as reported elsewhere (2, 14). Interleukin-2 was detectable as early as at day 2 with a maximum yield obtained on the 4th day. Subsequently IL-2 concentrations decreased probably because of consumption by T cells.

It is important to point out that in the flow cytometry system the delta values were calculated with absolute cell numbers, irrespective of the cell volume. By contrast, cell volume impacts on the normalized mean fluorescence intensity (NMFI), which is calculated by multiplying the percentage of cells considered positive with the mean fluorescence intensity of the cell population of interest (20). In this study, the calculation of a NMFI was not useful, as the sensitivity of flow cytometry test system would have been decreased as a result of the impact of large cells in the stage of proliferation. However, the aim of this study was to create a test system which allows detecting cells that were stimulated, although neither proliferation nor cytokine secretion maybe observable.

The sensitivity was calculated for all parameters investigated and both methods applied. As one result, analysis of a single cytokine only – even if assumed to be involved in the pathophysiology of the particular clinical phenotype – was not helpful to detect lymphocyte reactivity in all different clinical phenotypes of drug sensitization (see Results).

Based on a panel of cytokines investigated, we calculated a sensitivity of 75% for the flow cytometry and 79% for the ELISA respectively. Although the ELISA test system seemed to be superior in relation to flow cytometry, it has to be noted that the statistics of the ELISA were based on the results of three different concentrations in contrast to flow cytometry (result based on a single concentration). Concerning the ELISA test system, other groups reported similar or even higher sensitivities for distinct clinical reactions. For instance, Livni et al. (21) calculated a sensitivity of 50% for urticaria and angioedema based on the IFN-γ release of drug-specific activated PBMC. In a study with test conditions similar to ours, a high sensitivity for in vitro detection of IL-5 (92%), but not IFN-γ (36%) and IL-10 (50%) was observed, however, including patients with MPE only (2).

Concerning in vivo tests, Barbaud et al. (22) investigated patients with cutaneous adverse drug reaction and described a sensitivity of 50% and 65% for patch and intradermal tests respectively. These sensitivity results are comparable to the sensitivity of single cytokine analysis in the present study, whatever method was used. However, if a panel of cytokines was investigated, sensitivity increased to 75% and 79% by flow cytometry and ELISA, respectively, suggesting at least noninferiority compared with the in vivo tests. As a broad spectrum of different antigens and clinical phenotypes was investigated, the cytokine panel included seemed to be suitable for the detection of drug hypersensitivity.

Concerning the TH 1/TH 2 dichotomy, Fernandez et al. (23) described that the expression of cytokines and cytokine receptors in both, skin biopsies and peripheral CD4+ T cells, are in accordance with a TH 1 profile in patients with MPE. Our in vitro data concerning a preferential involvement of CD4+ and/or CD8+ T cells and the TH 1/TH 2 dichotomy appear inconsistent and in view of the clinical phenotype not following theoretical considerations or dogmas (Table 2). These inconsistencies, however, may be explained by the differences in the individual test systems applied. In addition, quantitative differences of cytokines may not necessarily translate into qualitative differences as the biological activity of equal amounts may be different. Further on, it is unclear, if the intracellular cytokine detection pattern is representative for the overall (production/secretion) cytokine patter.

Overall, this investigation suggests that the flow cytometry test system is a helpful nonradioactive approach for the detection and characterization of drug hypersensitivities. It may provide a viable diagnostic alternative, for instance, in patients with a history of severe cutaneous drug reactions or life-threatening anaphylaxis in whom drug challenge testing is not recommendable. However, further studies comprising larger patient populations and a broader range of investigated substances are warranted to corroborate the findings of our study.