Avidity determines T-cell reactivity in abacavir hypersensitivity


Full correspondence: Prof. Werner J. Pichler, Clinic for Rheumatology and Clinical Immunology/Allergology, PKT2 D572, University Hospital of Bern, 3010 Bern, Switzerland

Fax: +41-31-63-24208

e-mail: wernerjoseph.pichler@insel.ch


The antiretroviral drug abacavir (abc) elicits severe drug hypersensitivity reactions in HLA-B*5701+ individuals. To understand the abc-specific activation of CD8+ T cells, we generated abc-specific T-cell clones (abc-TCCs). Abc reactivity could not be linked to the metabolism and/or processing of the drug, since abc metabolizing enzymes were not expressed in immune cells and inhibition of the proteasome in APCs did not affect TCC reactivity. Ca2+ influx assays revealed different reactivity patterns of abc-TCCs. While all TCCs reacted to abc presented on HLA-B*5701 molecules, a minority also reacted immediately to abc in solution. Titration experiments showed that the ability to react immediately to abc correlated significantly with the TCR avidity of the T cells. Modifications of soluble abc concentrations revealed that the reactivity patterns of abc-TCCs were not fixed but dynamic. When TCCs with an intermediate TCR avidity were stimulated with increasing abc concentrations, they showed an accelerated activation kinetic. Thus, they reacted immediately to the drug, similar to the reaction of TCCs of high avidity. The observed immediate activation and the noninvolvement of the proteasome suggest that, in contrast to haptens, abc-specific T-cell stimulation does not require the formation of covalent bonds to produce a neo-antigenic determinant.


Hypersensitivity reactions to drugs can lead to a variety of clinical symptoms and these involve different immune mechanisms [1]. Some of these reactions depend on genetic factors, among which HLA molecules play a particularly important role [2-4]. A striking example of such a genetic association is found in hyper-sensitivity reactions to the antiretroviral drug abacavir (abc), whereby it is strongly associated with the HLA-B*5701 allele [5]. Similar to other severe drug reactions [1], abc hypersensitivity reactions involve drug-reacting T cells. This was illustrated by the presence of CD8+ T cells in skin biopsies of hypersensitive patients [6]. Moreover, a population of CD8+ T cells from HLA-B*5701+ individuals secrets IFN-γ in response to abc in vitro, irrespective of previous exposure to abc [7]. However, how abc is presented and subsequently stimulates T cells is still unclear.

Currently, two models account for the stimulation of T cells by drugs. According to the hapten model, compounds bind to certain amino acids via covalent bonds, with or without previous metabolism of the drug. These hapten-modified proteins are then processed into antigenic peptides and are loaded onto MHC molecules of APCs. In this instance, the haptenization of molecules is important for the activation of the innate immune system [8]. If this was the case in abc hypersensitivity, these hapten complexes would be exclusively presented by the HLA-B*5701 molecules. Although Chessman et al. [7] identified two components of the MHC class I processing machinery involved in the stimulation of abc-reacting T-cell lines (TCLs), TAP and tapasin, abc-haptenized peptides, or potential metabolites have not been detected so far.

An alternative model for drug interactions with the immune system has been proposed whereby a noncovalent and processing independent binding of the drug to HLA molecules or TCR stimulates an immune response [4]. This model is named the ``p-i concept'' that stands for pharmacological interactions with immune receptors [9]. In this model, these drug–receptor interactions result in T-cell activation and subsequent delayed drug hypersensitivity reactions [1]. The p-i concept has been developed mainly based on in vitro studies using drug-specific T-cell clones (TCCs) or TCR transfected cell lines [10-12]. An important argument for the p-i concept was provided by Ca2+ influx measurements that revealed that TCR mediated signaling took less than 200 s, whereby uptake, metabolism, processing, and presentation of the reactive drug could not have taken place [11, 13]. This is further supported by a computer model that suggested a noncovalent binding of abc to the HLA-B*5701 protein [14]. In addition to abc hypersensitivity, other drug hypersensitivity reactions also show striking HLA-associations [2, 15]. In the case of carbamazepine hypersensitivity, associated with HLA-B*1502, elution of peptides after the addition of carbamazepine did not reveal any bound carbamazepine or its metabolite [16]. This supports a drug interaction with the HLA-peptide complex via van-der-Waals bonds rather than via covalent interactions [14, 16].

The involvement of the p-i mechanism in drug hypersensitivity reactions has major implications in understanding drug-specific stimulations of the human immune system. As the p-i concept does not require generation of hapten-modified peptide antigens, a structure of a small molecular compound and its ability to interact with proteins via noncovalent interactions may also determine the immunogenic potential of a drug.

So far immunopathogenesis of drug hypersensitivity studies has been limited by the paucity of animal models and most in vitro studies have been limited to immune cells isolated from drug-allergic patients. However, the strong association of HLA-B*5701 with abc hypersensitivity provided a new opportunity to study its immune mechanism. To this aim, we generated abc-reactive TCCs (abc-TCCs) from naïve individuals, allowing the analysis of their abc-reactivity patterns by means of Ca2+ influx measurements. Our results provide evidence for a metabolism-independent, direct, and noncovalent abc interaction with the HLA-B*5701 molecule, which forms an antigenic structure that is able to induce a T-cell response. Further characterization revealed that the differences in velocity and intensity of the ensuing reaction are based on different TCR avidities for the generated abc-HLA complex implying a potential role of drug concentrations in vivo for drug hyper-sensitivity.


Generation of abc-reacting TCCs from HLA-B*5701+ individuals

After 12 days of in vitro stimulation, abc-reacting CD8+ T cells were observed in all five HLA-B*5701+ healthy donors (HD) included in the study. The reactivity of TCLs was assessed by intracellular IFN-γ expression on FACS. Abc-specific TCLs could not be generated in HLA-B*5701 individuals (Supporting Information Fig. 1). From abc-reacting TCLs, TCCs were generated by limiting dilution. Identification of abc-specific TCCs was performed by IFN-γ secretion on ELISpot and by 51Cr-release assay. TCCs were considered abc specific if at least one of the two tested parameters gave a positive result as defined in the methods. Two hundred and forty-nine TCCs, which were exclusively CD8-positive, could be produced from five individuals (Table 1 and Supporting Information Table 1). FACS analysis of several TCCs of HD207 and HD530 revealed different TCR Vβ-segments, demonstrating the polyclonality of abc-TCCs (Supporting Information Table 1).

Table 1. Phenotypes of abc-TCCs
Individual IDabc-TCCs (n)Cytotoxic (n)IFN-γ releasing (n)Cytotoxic and IFN-γ releasing (n)
HD 20716106
HD 530191414
HD 5434781128
HD 588106181969
HD 589615947

Absence of alcoholdehydrogenase (ADH) in immune cells

Metabolism is thought to play an important role in the generation of haptens and neo-antigens [17]. Abc was shown to be mainly metabolized in the liver through ADH to its 5′-carboxylic and its 5′-glucuroconjugated form, with the potential formation of reactive aldehyde intermediates [18]. However, it is unclear, whether this enzyme is expressed in peripheral immune cells or not. Furthermore, which isoform of ADH is relevant is also unknown. To clarify the presence of ADH, we tested the expression of six different ADH isoenzymes in various subsets of immune cells and keratinocytes by RT-PCR. In contrast to a control liver cell line, relevant ADH expression could not be detected in PBMCs, EBV-B-lymphocyte cell line (BLCL), monocytes, DCs, and T cells (Fig. 1A) or in keratinocytes (Supporting Information Fig. 2). Only a formaldehyde-dehydrogenase (ADH5), an enzyme not described to be involved in abc metabolism, was ubiquitously expressed.

Figure 1.

Abc metabolizing enzymes are not expressed in immune cells of peripheral blood. (A) mRNA from PBMCs, EBV-BLCLs, monocytes, DCs, abc-TCC-A1/HD543 (TCC A1), and abc-TCC-D1/HD543 (TCC D1) was reverse transcribed and analyzed for the expression of six different ADH isotypes. Amplification of the housekeeping-gene actin served as an internal positive control. The data are representative of three independent experiments. (B) TCC-10B/HD589 was stimulated with autologous PBMCs, pulsed with abc for 14 h in the presence or absence of 4-MP. TCC activation was analyzed by annV positivity by flow cytometry after 4 h of stimulation. TCC incubated with unpulsed PBMCs served as a negative control. (C) TCC-1A/HD589 was stimulated with autologous PBMCs, pulsed with abc for 14 h, in the presence or absence of either Lactacystin (top) or Bortezomib (below). TCC activation was analyzed by annV positivity by flow cytometry after 4 h of stimulation. TCC incubated with unpulsed PBMCs served as a negative control. (B and C) Data are shown as mean ± SD across three independent experiments performed in duplicates.

Since an involvement of other ADH could not be excluded, the competitive inhibitor of ADH, 4-methylpyrazole (4-MP), was used to block abc-TCC activation (Fig. 1B). Previously, it was shown that 4-MP blocks abc metabolism in vitro [18]. However, our experiment showed that the incubation of APCs with 4-MP before and during TCC stimulation did not show a reduction in TCC reactivity.

Recently, data of Chessman et al. [7] suggested that inhibition of TAP and tapasin resulted in diminished IFN-γ production in abc-reacting TCLs. However, the role of proteasomes in the abc processing pathway remains unclear. To investigate their potential role in the presentation of abc, we performed proteasome inhibition assay with a TCC-1A/HD589 using two different inhibitors, Lactacystin and Bortezomib (Fig. 1C). Neither of the two compounds led to a significant reduction of abc-TCC reactivity. These results suggest that abc-TCC reactivity is independent of abc metabolism and proteasome processing. Validity of the inhibition assay was tested with a TCL specific for the Hepatitis C Virus-related protein NS3/4A. Presentation of the NS3/4-derived peptide was inhibited by proteasome blockade (Supporting Information Fig. 3).

Three different kinetic patterns of TCC reactivity

To understand how TCCs recognize abc, kinetic analysis of abc-specific TCC reactivity was performed by monitoring intracellular Ca2+ concentration over an hour after TCR stimulation (Fig. 2). After 5 min of baseline measurements, EBV-BLCL previously pulsed with abc for 14 h were applied to abc-TCCs (Fig. 2A). All tested TCCs reacted immediately to abc presented on HLA-B*5701, confirming the specificity of the clones. To verify that the reactivity to abc is specific, Ca2+ influx measurements of a sulfamethoxazole-specific TCC were assessed using abc and sulfamethoxazole, showing that sulfamethoxazole-specific TCC reacted only to sulfamethoxazole and not to abc (Supporting Information Fig. 4) [19].

Figure 2.

Abc-specific TCCs exhibit different abc recognition patterns. Forty-seven TCCs were analyzed in a Ca2+ influx assay for their abc recognition patterns. (A) TCC-2D/HD589, -6B/HD543, and -11C/HD588 were either stimulated with APCs pulsed with abc for 14 h or were incubated with unpulsed APCs (negative controls). Addition of drug-pulsed APCs is indicated with an arrow. Representative data of three independent experiments are shown. (B) TCCs were stimulated with abc in solution in the presence of APCs (addition is indicated with a black arrow) or with PHA (addition is indicated with a gray arrow). TCCs incubated with APCs in CM served as a negative control. Representative data of three independent experiments are shown. (C) Abc-TCCs (n = 47) were pooled into ``immediate,'' ``fast,'' and ``delayed'' responders, according to their reactivity patterns as defined in Fig. 2B. **p < 0.01, ***p < 0.001, one-way ANOVA.

TCCs were then stimulated with abc in solution in the presence of unpulsed EBV-BLCL. This experiment revealed three different TCC activation patterns (Fig. 2B). Some clones, for example, TCC-2D/HD589, reacted immediately and very strongly to abc in solution. The peak of activation was reached within 200 s and the activation curve was very steep. These clones were named ``immediate responders.'' Other TCCs, for example, TCC-6B/HD543, showed a slower and less intense response to abc in solution, reaching the peak of activation at approximately 15 min. As the reaction in response to the stimulus was still fast, such TCCs were named ``fast responders.'' Finally, some TCCs, for example, TCC-11C/HD588, did not respond to abc in solution within 1 h. Because such clones consistently reacted to abc-pulsed APCs but not to freshly added abc in solution, they were named ``delayed responders.'' Interestingly, the pool of immediate responders consistently showed the highest signal intensities (F/F0 ratio) compared with those of fast or delayed responders (Fig. 2C). Most of the tested 47 TCCs belonged to the pool of delayed responders (n = 28), whereas 14 TCCs accounted for fast reacting clones and a minority (n = 5) for immediate responders. Therefore, the reactivity pattern of abc-reacting TCCs was shown to be nonhomogenous as clear differences were observed in signal intensities and reactivity kinetics.

Abc-TCC reactivity patterns correlate with abc avidity

Since we observed notable differences in the maximum Ca2+ intensity signals among the three reactivity patterns, we hypothesized that this feature could be linked to different TCR avidities. Therefore, avidity titrations were performed by stimulating TCCs in the presence of APCs with increasing concentrations of abc (1 ng/mL to 50 μg/mL). TCC activation was measured by annexinV (annV) positivity on FACS (Supporting Information Fig. 5), revealing a correlation between TCC activation and antigen concentration (Fig. 3A). For every analyzed clone, EC50 was calculated and used as a descriptive parameter of abc avidity (Fig. 3B). There was a significant correlation between the different reactivity patterns amongst TCCs (``delayed,'' ``fast,'' and ``immediate'' responders) and abc avidity of these TCCs (p < 0.01) (Fig. 3C). All clones with an immediate reactivity showed the strongest avidity for abc. On the other hand, TCCs with a delayed reactivity consistently revealed the weakest avidity for the drug and the fast responders exhibited an intermediate abc avidity. Consequently, the immediate responder TCCs were named ``high avidity'' (HA)-TCCs, the fast responder TCCs ``intermediate avidity'' (IA)-TCCs, and the delayed responder TCCs ``low avidity'' (LA)-TCCs.

Figure 3.

Specific TCCs exhibit different TCR avidities for abc. (A) TCC-2D/HD589 was stimulated with the indicated concentrations of abc in solution in the presence of APCs for 4 h. TCC activation was monitored by annV positivity by flow cytometry and shown are representative data of three independent experiments (left). Data pooled from n = 6 samples from three independent experiments are shown as MFI ± SD (right). (B) Representative example of the responsiveness of an ``immediate'' (2D/HD589), a ``fast'' (14D/HD589), and a ``delayed'' (20A/HD589) TCC after stimulation with increasing concentrations of abc in solution. Data are shown as mean ± SD across two independent experiments performed in duplicates. (C) Cumulative data on the TCR avidity of abc-TCCs (n = 15). These clones belonged to the TCC pool (n = 47) presented in Fig. 2C. The EC50 corresponds to the abc concentration needed to achieve a half-maximal response. **p < 0.01, one-way ANOVA.

Reactivity patterns are dependent on HLA-B*5701 surface densities of APCs

Since the amount of antigen seemed to be an important factor in determining the strength of abc-TCC activation, we investigated the influence of HLA-B*5701 density of APCs on TCC reactivity. For that purpose, the human lymphoid cell line 721.221, lacking HLA I, was transduced with the HLA-B*5701 molecule with a retroviral vector. Three lines were generated expressing different HLA-B*5701 surface densities (B*5701high, B*5701medium, and B*5701low, Fig. 4A). These cells were pulsed with abc for 14 h and were subsequently used as APCs. HLA-expression levels had a strong effect on the measured Ca2+ influx in all analyzed TCC types. A total of 721.221 cells with a high HLA-B*5701 surface expression activated the TCCs strongly (Fig. 4B) and vice versa, implying that the intensity of TCC stimulation was proportional to the HLA-B*5701 density on the cell surface.

Figure 4.

Antigen density on APCs influences abc-specific TCC reactivity patterns. (A) HLA-B*5701-transduced 721.221 cells were quantified for their HLA-B*5701 and β2-microglobulin surface expression. Mock-transduced 721.221 cells served as a negative control. Data are shown as mean absolute numbers ± SD (n = 4 per group pooled from two independent experiments). (B) In a Ca2+ assay, HA-TCC-2D/HD589, IA-TCC-18A/HD589, and LA-TCC-21C/HD589 were stimulated with abc-pulsed 721.221 cells, expressing different levels of HLA-B*5701. TCCs incubated with unpulsed 721.221 cells served as a negative control (only B*5701high shown). Addition of the stimulus is indicated with an arrow. Representative data of two independent experiments are shown.

Accumulation of abc on APCs surface enhances stimulation of IA-TCCs

In the next step, we addressed the effect of time on the presentation of abc on HLA-B*5701 molecules. EBV-BLCL were abc-pulsed for 2 h, 1 h, or 15 min and were subsequently washed three times to remove unbound abc molecules. These were then used to stimulate TCCs in a Ca2+ assay (Fig. 5). On the one hand, HA-TCCs reacted equally strongly and rapidly to the three different stimuli. On the other hand, IA-TCCs exhibited different activation patterns in response to the duration of pulsing. Two-hour pulsing phases of EBV-BLCL elicited the strongest reaction and a reduction of abc-pulsing time resulted in a lower response. This correlation between pulsing time and activation intensity of the TCCs was consistently found in all analyzed IA-TCCs (n = 4, data not shown). Finally, LA-TCCs reacted to all stimuli with an equally low activation signal without an ability to discriminate between the longer or shorter APCs-pulsing time. This observation was monitored for all analyzed LA-TCCs (n = 4, data not shown). However, APCs that were pulsed with abc for 14 h elicited the strongest response in all analyzed IA- and LA-TCCs but did not result in a higher activation of HA-TCCs (data not shown). This implies that HA-TCCs were already able to reach the peak intensity with short duration of pulsing.

Figure 5.

Duration of abc-pulsing affects IA-TCC activation patterns. HA-TCC-17D/HD589, IA-TCC-15A/HD589, and LA-TCC-21C/HD589 were stimulated with APCs, pulsed with abc for 2 h, 1 h, or 15 min and specific TCC activation was assessed in a Ca2+ influx assay. Addition of the stimulus is indicated with an arrow. TCCs incubated with unpulsed APCs served as a negative control. Representative data of two independent experiments are shown.

Abc concentration determines the activation kinetic of specific TCCs

Antigen density on the surface of APCs and the pulsing time of APCs are important factors in determining the strength of the elicited TCC responses (Figs. 4 and 5). Moreover, titration experiments have shown a correlation between avidity and activation kinetics. These findings suggest that the distinct activation patterns of abc-TCCs might be shaped by the amount of abc molecules presented on APCs. To directly investigate the role of antigen concentration on the abc-TCC reactivity, we stimulated the clones with varying concentrations of abc in solution in the presence of unpulsed APCs (Fig. 6). Interestingly, TCC activation kinetics could be modified by altering the abc concentrations. HA-TCC-2D/HD589, for example, exhibited an immediate activation kinetic upon stimulation with a concentration of 10 μg/mL abc. However, decreasing the drug concentrations consistently lowered the strength and the velocity of the response. At lower drug concentrations, the reactivity pattern of HA-TCCs could be altered to match those of IA- or LA-TCCs (Fig. 6A). Even though abc in solution at 10 ng/mL did not result in detectable Ca2+ influx, APCs pulsed with the same abc concentration for 14 h could still elicit an immediate and clear activation of TCCs (Fig. 6A). Similar observations were made with IA-TCCs, whose abc-reactivity pattern resembled those of LA-TCCs upon stimulation with lowered drug concentrations (<10 μg/mL abc, data not shown). On one hand, the activation patterns of HA- and IA-TCCs can be changed to those of IA- and LA-TCCs, respectively, by reducing the concentration of abc. On the other hand, increasing the abc concentrations to 66 μg/mL raised the activation kinetic curve of IA-TCCs, for example, TCC-6B/HD543, to match those of HA-TCCs (Fig. 6B). Although it is possible that LA-TCCs could behave in the similar manner in the presence of even higher abc concentrations, the reactivity pattern of LA-TCCs could not be assessed since the experiments were limited by the maximum abc concentration that could be used to reliably perform Ca2+ influx measurements. Altogether, our data indicate that abc-TCC reactivity patterns are not static but are dynamic; the available antigens for TCR recognitions determine the activation kinetics of T cells.

Figure 6.

TCC activation patterns are defined by abc concentration. (A) TCC-2D/HD589 was stimulated with decreasing concentrations of abc in solution (10–0.01 μg/mL) in the presence of APCs (left). TCCs incubated with APCs in CM served as a negative control. Addition of the stimuli is indicated with an arrow. TCC-2D/HD589 was stimulated with APCs, pulsed with 0.01 μg/mL abc for 14 h (right). Addition of the stimulus is indicated with an arrow. Representative data of two independent experiments are shown. (B) TCC-6B/HD543 was stimulated with increasing abc concentrations (10–66 μg/mL) in the presence of APCs, as in (A) (left). TCC-6B/HD543 was stimulated with APCs and pulsed with 10 μg/mL abc for 14 h, as in (A) (right). Representative data of two independent experiments are shown.


The association between abc hypersensitivity and the HLA-B*5701 provided a fresh opportunity to study the involvement of T cells in delayed-typed drug hypersensitivity and their perplexing interaction with drug and MHC molecules. Taking advantage of this, we generated abc-TCCs from HLA-B*5701+ donors and analyzed their specific reactivity patterns. TCCs stimulation with a constant abc concentration of 10 μg/mL resulted in three different TCC reactivity patterns, misleadingly suggesting that there are different presentation mechanisms. However, our study clearly showed that the observed differences in the activation kinetics originated from different TCR avidities for abc. The experiments with varying abc concentrations revealed that the reactivity patterns were not solely dependent on the TCC itself, but were also shaped by antigen availability. The fact that some TCCs were activated within minutes of abc exposure strongly argues for a direct interaction of a drug with HLA molecules, bypassing the need for metabolism and processing. The fact that these same TCCs showed slower reactivity pattern in response to lower drug concentrations further challenges the previously held assumption that different presentation mechanisms account for diverse T-cell activation patterns. Finally, this fluidity in T-cell reactivity may provide an essential piece of information in understanding the perplexing immunology of drug hypersensitivity, in particular, in reference to the role of drug levels in vivo.

The first important finding of our study is that TCR avidity determines the reactivity pattern of abc-TCCs. Since the amount of drug-MHC complexes required for T-cell activation is defined by avidity, we modified these parameters to observe their effect on T-cell reactivity. TCCs stimulation with APCs expressing different levels of HLA-B*5701 molecules indeed showed that the strength of TCC activation was proportional to the MHC density. This confirmed that the reactivity pattern is MHC dependent. Furthermore, the dose of drug influenced the TCC kinetics, supporting the concept that avidity determines the reactivity. Finally, the similar findings were shown in the experiment where APCs were incubated with drugs for different durations. This is likely explained by the fact that prolonged pulsing results in accumulation of stimulating antigenic complexes on the cell surfaces. Compared with HA- or LA-TCCs, the activation strength of IA-TCCs was more affected by the above parameters, implying that a relatively moderate variation in the available drug-MHC complexes is critical for their reactivity. Clinically, this phenomenon may be important in understanding the profound effect of drug doses on the pathology of drug hypersensitivity. Hypersensitivity reactions develop more often for drugs such as antibiotics or antiepileptics, which are used in high doses or for longer periods [20]. This is consistent with our findings on TCC avidity predicting that more T cells will react with higher antigen concentration, increasing the probability of developing clinical phenotypes.

Unlike peptides, abc is a small molecule. It is unclear how a small compound binds to MHC molecules and activates T cells, especially as it cannot fill the entire MHC groove by itself. The contribution of HLA-B*5701-embedded peptides and their potential interactions with abc remains to be answered. Computer modeling of Yang et al. [14] suggested a direct interaction of abc with HLA-B*5701. In this instance, we postulated that the drug must first bind to the specific HLA molecule, forming a stable complex with the potential to activate specific TCR [4]. Although structural data are unavailable, such abc/HLA complexes must be stable since extensively washed APCs were still able to stimulate effective T-cell responses.

The TCR sequence arrangement is definitive; however, we noted that the activation patterns of TCCs were dynamic. HA-TCCs changed their activation kinetics to match those of IA-TCCs in the presence of low abc concentrations and vice versa was also true for IA-TCCs in the presence of higher doses. Since HA-TCCs and IA-TCCs avidity could not be changed, it implies that the available antigen is the determining factor in shaping the activation kinetics. Using affinity-matured TCR, Thomas et al. [21] made similar observations in peptide-specific primary CD8+ T cells. In the natural affinity range, they noticed that high-affinity TCR–antigen interactions were more rapidly initiated than low-affinity interactions, which our data also verified.

In contrast to noncovalently bound drugs, haptens cannot be removed from fixed APCs by extensive washing steps [11, 22]. Building on this, TCCs that reacted to antigen-pulsed APCs but not to soluble antigen were considered to be TCCs that recognize haptenized-peptide complexes [23, 24]. However, our observation of the fluidity of the activation pattern suggests that this interpretation is flawed. We have shown that different antigen concentrations in same TCCs can lead to distinct recognition patterns. Therefore, one cannot draw a conclusion that metabolism, processing, and haptenization are involved when T cells react to pulsed APCs only as long incubation also leads to antigen accumulation on the surface.

The immediate activation pattern of some TCCs is incompatible with metabolism and processing of abc by the MHC class I machinery. Furthermore, the absence of ADH and the inability of the proteasome inhibitors to suppress abc-specific responses strongly argue against this hapten mechanism. In contrast, Chessman et al. [7] found that reactivity of abc-specific TCLs was affected by the inhibition of TAP and the absence of tapasin, implying that the abc presentation depends on the MHC class I pathway. However, the inhibition of TAP and the absence of tapasin not only affect peptide loading, but also reduce the expression of HLA molecules on the cell surface [25, 26]. It is possible that the effect of TAP and tapasin inhibition is linked to the density of HLA molecules expressed on the cell surface rather than to peptide processing and uptake into the ER. Assuming that the majority of abc-specific T cells within the TCLs had low or intermediate TCR avidities, it is plausible that T-cell activations were affected by TAP or tapasin inhibition due to reduced HLA expression and subsequent insufficient T-cell stimulus. However, since TAP and tapasin participate in peptide loading on MHC class I [27, 28], we cannot exclude that these proteins are also involved in the loading of abc onto HLA-B*5701 molecules as an additional supply.

Our study is limited by the fact that we have designed the experiments in vitro as the human physiological condition is far more complex than what can be replicated in TCCs or TCLs. However, as there are no equivalent animal models, our study is uniquely placed to provide an insight into the immunopathology of drug hypersensitivity. Furthermore, preliminary data on the TCR repertoire of abc-TCCs revealed polyclonality of the involved T cells (Supporting Information Table 1). Ko et al. [29] recently linked the reactivity of carbamazepine-specific T cells to specific TCR clonotypes. Due to the lack of a systematic analysis on the TCR repertoire of abc-TCCs, further investigations are required to determine whether such dominant clonotypes are present in the generated clones. In addition, the bulk of our findings are based on Ca2+ influx assay that is limited by the reliability of the measurements at high end drug concentrations. Subsequently, the effect of a broader abc concentration range on TCCs could not be fully investigated. However, we feel that our data based on Ca2+ influx assay are robust enough to reliably conclude that avidity determines T-cell reactivity.

In conclusion, our data provide evidence for a direct and noncovalent association of abc with the HLA-B*5701 complex generating an immunogenic determinant with the potential to stimulate CD8+ T cells. Furthermore, the avidity of drug-specific T cells influences their reactivity patterns and these cells respond uniquely to different drug concentrations. This means that an evaluation of the immunogenic potential of a drug beyond its ability to form covalent bonds must be considered in the drug development process.

Material and methods

Healthy donors

Five HIV-negative, HLA-B*5701+ individuals who were never exposed to abc were enrolled in the study (HD207, HD530, HD543, HD588, HD589). These individuals were selected from Bern's blood donation center according to their HLA-B*5701-positive status. All individuals gave informed consent and the study was approved by the local ethical committee.

Primary stimulation, TCLs, and TCC generation

PBMCs were isolated by Ficoll density gradient centrifugation and cultured in RPMI-1640 (Gibco, Basel, Switzerland) supplemented with 10% of heat-inactivated human AB serum (Swiss Red Cross, Bern, Switzerland), 2 mM L-Glutamine (Biochrom, Berlin, Germany), 25 μg/mL transferrin (Biotest, Dreieich, Germany), 50 U/mL penicillin and 50 μg/mL streptomycin (Bioconcept, Allschwil, Switzerland). Abc-specific T cells were enriched by culturing lymphocytes (4 × 106 cells in 2 mL) in 10 μg/mL abc-sulfate (Reseachem, Burgdorf, Switzerland). Cells were fed with 50 IU/mL IL-2 (Roche, Basel, Switzerland) on days 5, 7, and 9 to maintain antigen-specific proliferation. On day 12, the cells were washed three times and were incubated with autologous PBMCs in the presence or absence of 10 μg/mL abc for 6 h. After 2 h, Brefeldin A (10 μg/mL, Sigma-Chemicals, Buchs, Switzerland) was added. Surface staining was performed with anti-CD3-allophycocyanin-Cy7, anti-CD4-PE-Cy7, and anti-CD8-PerCp (BD Biosciences, Basel, Switzerland). Intracellular staining with anti-IFN-γ-allophycocyanin was performed according to the Cytofix/Cytoperm permeabilization kit (BD Biosciences). FACS analysis was performed on a FACScanto-I cytometer using FACS-Diva software (BD Biosciences). Abc-specific TCLs were further expanded for three rounds of restimulation. Cloning was performed by limiting dilution as described previously [30]. Specificity testing was performed by IFN-γ ELISpot and cytotoxicity assays as described elsewhere [31, 32]. TCCs were considered abc specific if cytotoxicity in the culture exceeded >30% and/or when the difference in the number of spot forming cells between wells with and without abc exceeded 50. TCCs were expanded further as previously described [23]. Clones were selected for the functional studies based on antigen specificity and the availability of cells. Clones that were resistant to expansion were not used in functional studies. TCCs could be kept in culture for approximately 6–12 weeks before they underwent exhaustion.

Abc T-cell stimulation

T cells were either stimulated with abc-pulsed APCs or with freshly added abc in the presence of unpulsed APCs. The later is subsequently referred as ``abc in solution.'' For abc-pulsing, APCs were incubated in abc containing culture medium (CM) at 37°C for the indicated times. To completely remove unbound abc, APCs were washed three times. Unless otherwise stated, autologous EBV-transformed BLCL were used as APCs.


RNA was extracted from 105 TCC-A1/HD543 and H4/HD543, EBV-BLCL, PBMCs, monocytes, DCs, and Huh7 cells (hepatocarcinoma cell line) with the RNeasy Protect Mini Kit (Qiagen). Reverse-transcription was performed at 42°C for 60 min with SuperScriptII reverse transcriptase (Invitrogen, Carlsbad, CA, USA) using oligo dT as a primer. PCR primers, specific for the human ADH classes I–V, were used to amplify potential ADH products. ADH1A, 1B, 1C (class I), forward primer: 5′-ATGAGCACAGCAGGAAAAGT. ADH1A reverse primer: 5′-TCAAAACATCAGAATGGTACGG. ADH1B and ADH1C, reverse primer: 5′-TCAAAACGTCAGGACGGT. ADH4 (class II), forward primer: 5′-ATGGGCACCAAGGGCAAA. ADH4 reverse primer: 5′-TCAAAAGATGAGGATTGTTCGG. ADH5 (class III), forward primer: 5′-ATGGCGAACGAGGTTATCAA. ADH5 reverse primer: 5′-TTAAATCTTTACAACAGTTCGAATG. ADH6 (class V), forward primer: 5′-ATGAGTACTACAGGCCAAGTCA. ADH6 reverse primer: 5′-TTAAAGTAACAGGATACAGCGGA. Amplification of the housekeeping-gene actin served as a positive control. Actin forward primer: 5′-GGCATCGTGATGGACTCCG, reverse primer: 5′-GCTGGAAGGTGGACAGCGA. Thirty-five PCR cycles were conducted as follows: denaturation at 95°C for 30 s, annealing at 55°C for 60 s, and elongation at 72°C for 45 s.

ADH inhibition

Autologous PBMCs were pulsed in 10 μg/mL abc for 14 h in the presence or absence of 4-MP (Sigma-Chemicals). After three washing steps, these APCs were incubated with TCC at a ratio of 1:2. TCC reactivity was measured by annV positivity [33] on FACS.

Proteasome inhibition

Autologous PBMCs were incubated in CM containing either Lactacystin or Bortezomib at the indicated concentrations for 12 h at 37°C. Abc-pulsing was initiated by the addition of 10 μg/mL abc. APCs were washed three times and were then incubated with TCC at a ratio of 1:2 for 4 h in the presence of Lactacystin or Bortezomib, respectively. TCC activation was monitored by means of annV positivity on FACS.

Calcium influx assay

TCCs were incubated with 2 μg/mL Fluo-4 AM (Invitrogen) as described previously [23]. Cells were plated in half-area clear bottom 96-well plates (Corning, CA, USA) at 105 cells/well. Measurement was performed on a Synergy-4 instrument (BioTek, Highland Park, VT, USA) with an excitation band of 485/20 nm and fluorescence was measured at 528/20 nm. Baseline signal (F0) was recorded during 5 min before the addition of antigens. Subsequently, continuous fluorescence measurements were performed for 60 min. PHA stimulation (2 μg/mL) served as a positive control. The results are shown as normalized fluorescence (F/F0).

Avidity measurement

TCCs were incubated with EBV-BLCL in a ratio of 2:1 for 4 h in the presence of various abc concentrations ranging from 50 μg/mL to 1 ng/mL. TCC activation was monitored on FACS by means of annV positivity. EC50 was determined as the peptide concentration needed to achieve a half-maximal response.

Viral transduction of 721.221 cells and HLA quantification

RNA was isolated from EBV-BLCL of HD543 and cDNA was generated in the similar manner as the method for ADH reverse-transcription described above. HLA-B*5701 was amplified with sequence-specific primers, flanked by EcoRI restriction sites (for 5′-GAATTCATGCTGGTCATGGCGCCCCGAA and rev 5′-GAATTCAGCTCCGATGACCACAACTGCTAGG) and then inserted into the retroviral vectors pMSCV/BlaR and pMSCV/PuroR (Clontech Laboratories, Mountain View, CA, USA). Recombinant retroviruses were produced as previously described [34]. The HLA-B-negative lymphoblastoid cell line 721.221 [35] was infected with either of the generated constructs or both of these together. Selection was done by adding BlasticidinS (5 μg/mL), Puromycin (3 μg/mL) or both, added to the CM. Quantification of HLA and β2-microglobulin surface expression was performed with QIFIKIT, according to manufacturer's protocol (Dako, Glostrup, Denmark).

Statistical analysis

Statistical analyses were performed using GraphPad Prism5 (GraphPad Software, San Diego, CA, USA). Results are expressed as mean ± SD. Comparisons were drawn using a Kruskal–Wallis test or analysis of variance. Each experiment was repeated at least twice.


This work was supported by the SNF grant Nr. 310030–129828-1, the Ulrich Müller-Gierok Foundation, and the Swiss Center for Applied Human Toxicology. We thank James Yun for critical reading of the manuscript.

Conflict of interest

The authors declare no financial or commercial conflict of interest.




abc-specific T-cell clone


high avidity TCC


healthy donor


intermediated avidity TCC


low avidity TCC


T-cell line