Human leukocyte antigen (HLA)-B*57:01-restricted activation of drug-specific T cells provides the immunological basis for flucloxacillin-induced liver injury



The role of the adaptive immune system in adverse drug reactions that target the liver has not been defined. For flucloxacillin, a delay in the reaction onset and identification of human leukocyte antigen (HLA)-B*57:01 as a susceptibility factor are indicative of an immune pathogenesis. Thus, we characterize flucloxacillin-responsive CD4+ and CD8+ T cells from patients with liver injury and show that naive CD45RA+CD8+ T cells from volunteers expressing HLA-B*57:01 are activated with flucloxacillin when dendritic cells present the drug antigen. T-cell clones expressing CCR4 and CCR9 migrated toward CCL17 and CCL 25, and secreted interferon-gamma (IFN-γ), T helper (Th)2 cytokines, perforin, granzyme B, and FasL following drug stimulation. Flucloxacillin bound covalently to selective lysine residues on albumin in a time-dependent manner and the level of binding correlated directly with the stimulation of clones. Activation of CD8+ clones with flucloxacillin was processing-dependent and restricted by HLA-B*57:01 and the closely related HLA-B*58:01. Clones displayed additional reactivity against β-lactam antibiotics including oxacillin, cloxacillin, and dicloxacillin, but not abacavir or nitroso sulfamethoxazole. Conclusion: This work defines the immune basis for flucloxacillin-induced liver injury and links the genetic association to the iatrogenic disease. (HEPATOLOGY 2013;)


DILI, drug-induced liver injury; HLA, human leukocyte antigen; PBMC, peripheral blood mononuclear cells.

Adverse drug reactions are a major complication of drug therapy and an impediment to drug development. Immunological reactions are extremely important because of their severity and they account for many cases of drug withdrawal. They cannot easily be predicted and no simple dose-response is discernible. Skin is the tissue most commonly targeted by immune cells; however, other organs, including the liver, can be damaged either in isolation or as part of a generalized hypersensitivity syndrome.1

T-lymphocytes are believed to cause drug-induced skin injury through the action of cytokines and cytolytic molecules. To stimulate a T-cell response the drug must bind to human leukocyte antigen (HLA) and in some way crosslink specific T-cell receptors. Recently, a number of cutaneous drug reactions have been strongly associated with expression of HLA alleles (e.g., abacavir hypersensitivity [HLA-B*57:01],2 carbamazepine hypersensitivity in Caucasians, and Japanese [HLA-A*31:01],3,4 carbamazepine-induced Stevens Johnson syndrome in Han Chinese [HLA-B*15:02]),5 which implies a direct/indirect effect of the gene product on the disease. For abacavir and carbamazepine, it has been possible to relate the genetic association to the mechanism of disease by characterizing drug-specific CD8+ T-cell responses in volunteers expressing HLA-B*57:01 and B*15:02, respectively.6,7

The role of T cells in drug reactions targeting the liver is less well defined. In 1997, Maria and Victorino8 described lymphocyte proliferative responses to drugs in over 50% of patients with drug-induced liver injury (DILI). More recently, histological examination of an inflamed liver from a single patient exposed to sulfasalazine revealed an infiltration of granzyme B secreting T-lymphocytes.9 The discovery of HLA alleles as risk factors for DILI (e.g., flucloxacillin [B*57:01],10 ximelagatran [DRB1*07:01],11 lumiracoxib [DRB1*15:01]12) is supportive of an immune mechanism; however, biological data showing HLA restriction of drug-responsive cytotoxic T cells is lacking. The strength of the association described for flucloxacillin—approximately 85% of cases carry at least one copy of HLA-B*57:01—prompted us to investigate (1) the cellular response in patients with flucloxacillin-induced liver injury, and (2) whether flucloxacillin activates naive CD8+ T cells from HLA-B*57:01-positive volunteers.

Patients and Methods

Human Subjects.

Six patients with flucloxacillin-induced liver injury and six flucloxacillin-exposed tolerant controls were recruited. Table 1 lists the clinical features of the reactions and concomitant medications. Three HLA-B*57:01-positive flucloxacillin-naive individuals were also selected from our frozen cell bank containing peripheral blood mononuclear cells (PBMCs) from 400 healthy volunteers recruited from northwest England. A total of 100 mL of blood was collected for both DNA and PBMC isolation. Genomic DNA was extracted using Chemagic magnetic separation (Chemagen, Baesweiler, Germany) and high-resolution sequence-based HLA typing was performed by the Histogenetics laboratory (Histogenetics, Ossining, NY) at the following loci: HLA-A, -B, -C, -DRB1, -DQB1, and DQA1. Approval for the study was acquired from the Liverpool local Research Ethics Committee and informed written consent was obtained from each donor.

Table 1. Clinical Features of the Patients
Subject IDAgeSexPeak Liver Function Tests at Time of Liver Injury*Time to OnsetSince ReactionInternational Consensus Criterion RUCAM ScoreConcomitant Medication and Comments
  • *

    ULN, upper limit of normal; ALT, alanine aminotransferase; ALP, alkaline phosphatase; GGT, gamma-glutamyl transferase

  • The cases were evaluated by application of the Council for International Organizations of Medical Science scale, also called the Roussel Uclaf Causality Assessment Method (RUCAM). The pattern of liver injury was classified according to the International Consensus Meeting Criteria. Only cases having at least possible causality (score 3+) were included in the study. Diagnosis of DILI was done by expert hepatologists.

   ALTBilirubinALPGGT(weeks)(years)3–5 Possible, 6–8 Probable, >8 Highly Probable 
P173F18x ULN7x ULN6x ULN22x ULN41110Reaction included maculopapular exanthem. Other causes of cholestatic hepatitis excluded
P261M23x ULN3x ULN9x ULNNot measured3117Eosinophilia and arthralgia; Other causes of cholestatic hepatitis excluded
P373F13x ULN12x ULN2x ULN12x ULN423Also taking co-amoxiclav at time of the reaction. Other causes of cholestatic hepatitis excluded
P490F11x ULN13x ULN4.5x ULN14x ULN336Other causes of cholestatic hepatitis excluded
P565F4x ULN36x ULN2x ULN1.5x ULN335Also taking co-amoxiclav at time of the reaction. Reaction included maculopapular rash. Other causes of cholestatic hepatitis excluded
P678F36x ULN17x ULN5x ULN16x ULN3610Reaction included maculopapular exanthem. Other causes of cholestatic hepatitis excluded

Detection of Flucloxacillin-Specific PBMC Responses.

Proliferation of patient PBMC (0.15 × 106/well) against flucloxacillin (0.1-2 mM) and tetanus toxoid (5 μg/mL) was measured using the lymphocyte transformation test.13 interferon-gamma (IFN-γ) and granzyme B-secreting PBMCs were visualized using ELISpot (MabTech, Nacka Strand, Sweden) by culturing PBMC (0.5 × 106/well; 200 μL) with flucloxacillin (1–2 mM) or PHA (5 μg/mL) for 48 hours.

Generation of T-Cell Clones from Patients With Flucloxacillin-Induced Liver Injury.

PBMC (1 × 106/well; 0.5 mL) from patients with DILI were cultured with flucloxacillin (1-2 mM) in RPMI 1640 supplemented with 10% human AB serum (Innovative Research, Class A), 25 mM HEPES, 10 mM L-glutamine, and 25 μg/mL transferrin (Sigma-Aldrich, Gillingham, UK). Cultures were supplemented with 200 IU/mL rhIL-2 (PeproTech, London, UK) on days 6 and 9. On day 14, CD8+ cells were isolated by positive selection using CD14 microbeads (Miltenyi Biotec, Bisley, UK) and the remaining cells designated as CD4+. The separated cells were then cloned by serial dilution.14

Epstein-Barr virus (EBV) transformed B-cell lines were created from PBMC by transformation with supernatant from the virus-producing cell line B9.58. Lines were maintained in RPMI 1640 supplemented with 10% fetal bovine serum (Invitrogen, Paisley, UK), 100 mM L-glutamine, 100 μg/mL penicillin, 100 U/mL streptomycin, and used as a source of autologous antigen-presenting cells.

Protocol for Priming Naive T Cells From Healthy Volunteers With Flucloxacillin.

T-cell priming was performed with naive CD3+ T cells using our recently established protocol.15 CD14+ cells were cultured in medium containing GM-CSF and IL-4 (PeproTech) for 8 days to generate dendritic cells. Twenty-five ng/mL tumor necrosis factor alpha (TNF-α) (PeproTech) and 1 mg/mL lipopolysaccharide (LPS) (E. coli strain 0111:B4, Sigma-Aldrich) were added for the last 16 hours of culture. T cells were isolated from PBMC by negative selection and CD25+ and CD45RO+ cells removed by positive selection (Miltenyi Biotec). The purity of the naive T cells exceeded 97%. The dendritic cells (8 × 104 cells/well) were cultured with naive CD3+ T cells (2 × 106T cells/well; 2 mL) in the presence of flucloxacillin (1-2 mM) for 8 days (37°C; 5% CO2).

After the coculture period, primed T cells (1 × 105/well; 200 μL) were restimulated with flucloxacillin (0.5-2 mM) and fresh dendritic cells (4 × 103/well) for 48 hours. IFN-γ release was assessed by ELISpot. The remaining cells were cloned as described above.

Phenotype and Specificity of T-Cell Clones.

Drug specificity was assessed by culturing autologous irradiated EBV-transformed B cells (1 × 104/well) and flucloxacillin (1-2 mM) with T-cell clones (5 × 104/well; 200 μL) for 48 hours. Proliferation was measured by [3H]thymidine incorporation (0.5 μCi/well, 5 Ci/mmol, Morovek Biochemicals, Brea, CA) for the last 16 hours of culture followed by scintillation counting. Clones with a stimulation index of greater than 2 were expanded by repetitive stimulation with irradiated allogeneic PBMC (5 × 104/well; 200 mL) and 5 mg/mL phytohemagglutinin (PHA) in interleukin (IL)-2 containing medium (250 IU/mL). Drug-specificity of the selected T-cell clones was assessed by proliferation and ELISpot for IL13, IFN-γ, FAS ligand, perforin, and granzyme B (Mabtech). Autologous irradiated EBV-transformed B cells (1 × 104/well), flucloxacillin (1-2 mM), and T-cell clones (5 × 104/well; 200 μL) were incubated for 48 hours. T-cell clones were also tested for reactivity against piperacillin, penicillin G, amoxicillin, oxacillin, cloxacillin, and dicloxacillin (all 0.1-2 mM), nitroso sulfamethoxazole (10-100 μM), and abacavir (10-100 μM). Dose ranges have been shown to be optimal for the activation of T cells.6,13-15 Cell phenotyping was performed by flow cytometry on a BD FACSCanto II using CD4, CD8, CDR1, CCR2, CCR3, CCR4, CCR5, CCR8, CCR9, CCR10, CXCR3, CXCR6, and CLA antibodies (BD Biosciences) and the IOtest Beta Mark TCR Vβ repertoire kit.

Migration Assay.

The 24-well transwell chambers were used with 5-μm pores. A total of 0.1 × 106T cells (n = 4 CD8+ clones) in 100 μL chemotaxis buffer (RPMI 1640 + 0.5% bovine serum albumin [BSA]) were placed in the upper chambers. CCL17 (CCR4 ligand) and CCL25 (CCR9 ligand; 100 ng/mL) in 600 μL chemotaxis buffer were placed in the lower wells and the cells were incubated for 2 hours. Cells migrating to the lower chamber were collected and counted using a hemocytometer.

Mechanistic Studies of Antigen Presentation.

A step-wise approach was adopted to study pathways of flucloxacillin presentation to T-cell clones. First, clones (5 × 104/well) were stimulated with autologous EBV-transformed B cells (1 × 104/well) in the presence of anti-HLA class I and class II blocking antibodies (5 μL; BD Biosciences, Oxford, UK); second, clones were stimulated with autologous and allogeneic EBV-transformed B cells expressing different HLA-B allotypes; and third, autologous EBV-transformed B cells were subjected to glutaraldehyde fixation (0.05%; Sigma-Aldrich) to terminate metabolic processes and/or were incubated for 1, 4, 16, or 48 hours with flucloxacillin (2 mM) followed by three washes to remove soluble drug.

Characterization of β-Lactam Albumin Binding in Culture.

To identify the key drug-modified lysine residues in albumin, we utilized our recently described mass spectrometry methods.13,14,16 Flucloxacillin was incubated with EBV-transformed B cells in RPMI 1640 medium containing 10% human AB serum for 1-48 hours. At each timepoint the cells were removed by centrifugation at 450g. Serum proteins were precipitated from culture supernatant by the addition of nine volumes of ice-cold methanol followed by centrifugation at 14,000g and 4°C for 15 minutes. Flucloxacillin, piperacillin, penicillin G, amoxicillin oxacillin, cloxacillin, and dicloxacillin were also incubated with human serum albumin at a molar ratio of drug to protein of 10:1 for 16 hours and methanol-precipitated.

Prior to mass spectrometry, all samples were reduced and alkylated before again being subjected to methanol precipitation. They were reconstituted in ammonium bicarbonate buffer (50 mM), digested with trypsin overnight at 37°C, and then desalted using C18 Zip-Tips (Millipore). Samples (2.4-5 pmole) were delivered into a QTRAP 5500 hybrid quadrupole-linear ion trap mass spectrometer (ABSciex) by automated in-line liquid chromatography (U3000 HPLC System, 5 mm C18 nano-precolumn, and 75 μm × 15 cm C18 PepMap column [Dionex, Sunnyvale, CA]) by way of a 10-μm inner diameter PicoTip (New Objective, Woburn, MA). A gradient from 2% ACN/0.1% FA (v/v) to 50% ACN/0.1% FA (v/v) in 70 minutes was applied at a flow rate of 280 nL/min. The ionspray potential was set to 2,200–3,500V, the nebulizer gas to 18, and the interface heater to 150°C. Established multiple reaction monitoring (MRM) transitions specific for drug-modified peptides were employed. MRM survey scans were used to trigger enhanced product ion dual mass spectrometry (MS/MS) scans of drug-modified peptides. Total ion counts were determined from a second aliquot of each sample analyzed by conventional liquid chromatography (LC)-MS/MS on the same instrument and were used to normalize sample loading on column. MRM peak areas were determined by MultiQuant 1.2 software (ABSciex). The relative intensity of MRM peaks for each of the modified lysine residues within a sample were compared and were normalized across samples.

Immune Epitope Database Analysis (IEDB) Software.

The MHC class I T cell epitope prediction tool on the IEDB website ( was used to predict which peptides from human serum albumin were likely to be high affinity binders to HLA-B*57:01. The IEDB recommended prediction method was used, the relevant allele was selected, and all lengths of peptides were allowed. Peptides with median inhibitory concentrations (IC50s) of <50 nM (high affinity binders) or <500 nM (intermediate affinity binders) containing lysine residues modified by all of the β-lactams studied were listed.

Statistical Analysis.

Student t test was used to analyze the proliferation and ELISpot data.


Flucloxacillin-Specific PBMC Responses in Patients With DILI.

Five of the six patients with DILI were positive for the risk allele HLA-B*57:01 (Supporting Table 1). The remaining patient, patient 6, expressed HLA-B*44:02 and 55:01. PBMCs were not stimulated to proliferate with flucloxacillin; however, flucloxacillin-specific PBMC responses were detected using an IFN-γ ELIspot. PBMCs from five out of the six patients, including the patient who was positive for HLA-B alleles other than B*57:01 (patient 6), were activated with the drug (Fig. 1A). The ELIspot assay was repeated with PBMC from three patients at least 1 month after the initial test and the flucloxacillin-specific response remained the same. Figure 1B, C shows the dose-dependent secretion of IFN-γ and granzyme B from PBMC and a T-cell line generated from patient 6. PBMC from flucloxacillin-tolerant patients and drug-naive volunteers were not stimulated with flucloxacillin to proliferate or secrete cytokines (data not shown).

Figure 1.

Stimulation of DILI patient lymphocytes and T-cell clones with flucloxacillin. (A) Flucloxacillin-induced IFN-γ release in PBMC from patients (P1-6). (B) Dose response to flucloxacillin using PBMC from patient 6 in an IFN-γ and Granzyme B ELISpot. (C) IFN-γ and Granzyme B ELISpot for a T-cell line from patient 6. (D) Table summarizing the phenotype of T-cell clones isolated from patients 1-6. (E) Proliferative response (black lines) and the secretion of cytokines and cytolytic molecules (colored lines and pictures) from T-cell clones following flucloxacillin stimulation. Five representative clones isolated from patients 2-6 are shown. The data show the mean of replicate wells.

Characterization of Flucloxacillin-Responsive CD4+ and CD8+ T-Cell Clones From Patients With DILI.

A total of 38 flucloxacillin-responsive T-cell clones expressing different Vβ receptors were isolated from PBMC of the four IFN-γ ELIspot-positive patients expressing HLA-B*57:01. Of these, 35 were identified as CD8+ by flow cytometry (Fig. 1D; Supporting Table 2). Flucloxacillin-specific proliferation was dose-dependent, with clones displaying different response profiles up to a concentration of 2 mM. The proliferative response of CD4+ and CD8+ clones was associated with the secretion of T helper (Th)1 and Th2 cytokines and the cytolytic molecules perforin, granzyme B, and FasL (Fig. 1E). Interestingly, the levels of cytolytic molecules secreted from the CD4+ clones was lower when CD4+ and CD8+ clones were compared.

Seven flucloxacillin-responsive CD4+ clones were isolated from patient 6, who expressed HLA-B*44:02/55:01. Over 100 other CD8+ clones were also isolated, but flucloxacillin responses were not detected.

Level of Flucloxacillin-Albumin Binding at Specific Lysine Residues Correlates With the Processing-Dependent Activation of T-Cell Clones.

To investigate the role of drug protein binding in the generation of flucloxacillin-derived antigens for T cells, antigen-presenting cells were pulsed with the drug for 1, 4, 16, and 48 hours, prior to washing and exposure to clones. A 16 to 48-hour incubation period was required to stimulate a reproducible proliferative response with all clones (Fig. 2A) and the strength of the response was similar to that seen with the soluble drug.

Figure 2.

Flucloxacillin binds irreversibly to protein to stimulate T-cell clones. (A) Stimulation of a panel of 10 clones with flucloxacillin-pulsed antigen-presenting cells. Antigen-presenting cells were washed repeatedly after pulsing to remove unbound drug. The data shows the mean of replicate wells. Results were analyzed by Student's t test. (B) Flucloxacillin binds to multiple lysine residues on serum albumin in a time-dependent manner in cell culture, as detected by mass spectrometry. (C) Correlation between the time-dependent relative level of flucloxacillin covalent binding to albumin in serum-supplemented culture medium and the strength of the mean drug-specific T-cell response (n = 10 clones; cpm in control wells subtracted). (D) Stimulation of a panel of 10 clones with flucloxacillin and irradiated or glutaraldehyde-fixed antigen-presenting cells. Fixation blocks protein processing. The data show the mean of replicate wells. Results were analyzed by Student's t test. (E) Variable crossreactivity of flucloxacillin-specific clones with related β-lactam antibiotics and mass spectrometric analysis of the profile of lysine modification with each drug in vitro (table shows the Lys residues modified after 16 hours [10:1; drug:albumin]). Four representative clones that illustrate the different response profiles are shown (see Supporting Fig. 1 for dose-response curves). The data show the mean of replicate wells. (F) Identification of peptides from serum albumin with high binding affinity to HLA-B*57:01, which contain lysine residues consistently modified with the crossreactive β-lactam antibiotics. IEDB analysis software was used to identify the peptides.

Mass spectrometric analysis of albumin in flucloxacillin-treated cell cultures revealed an irreversibly bound hapten of the predicted mass of 453 amu, which was formed from direct adduction of flucloxacillin. After 48 hours, 12/59 lysine residues were modified, including Lys190 and Lys212, which are modified on albumin isolated from plasma of all flucloxacillin-exposed patients.16 Adduct formation on albumin at each modified Lys residue was dependent on incubation time (Fig. 2B) and a strong positive correlation between the level of flucloxacillin binding and the strength of the proliferative response stimulated by flucloxacillin-pulsed antigen-presenting cells was observed (Fig. 2C). Glutaraldehyde fixation of antigen-presenting cells, which inactivates protease enzyme activity and antigen processing, inhibited the proliferation of clones against flucloxacillin-pulsed antigen-presenting cells (Fig. 2D).

Flucloxacillin-Responsive Clones Display Additional Reactivity Against Alternative β-Lactam Antibiotics That Form Similar Haptenic Determinants on Albumin.

To study chemical restriction of the flucloxacillin-specific T-cell response, clones were cultured with antigen-presenting cells and 4 β-lactam antibiotics (flucloxacillin, piperacillin, amoxicillin, and penicillin G) and hapten albumin binding profiles and proliferation were measured. Greater than 80% of the clones displayed additional reactivity against at least one β-lactam, which form similar haptenic determinants with Lys residues on albumin. In fact, drug modifications were detectable at eight Lys residues (Lys132, 190, Lys199, Lys212, Lys351, Lys432, Lys 525, and Lys541) with all four drugs. Figure 2E shows the drug structures, the proliferative response of four representative clones that show the different crossreactivity profiles observed, and the sites of lysine modification on albumin. Supporting Fig. 1 shows the dose-dependent proliferative response of clones with optimal stimulatory concentrations of the different drugs.

In subsequent experiments, hapten albumin binding profiles and T-cell responses were studied with oxacillin, cloxacillin, and dicloxacillin (Fig. 3A), which are used in certain countries as an alternative to flucloxacillin. Hapten modifications were detected on the same 12 lysine residues with all four drugs and the level of modification at each site was comparable (Fig. 3B). Furthermore, all of the flucloxacillin-responsive clones tested displayed reactivity against the structurally related drugs (Fig. 3C).

Figure 3.

Flucloxacillin-responsive CD8+ clones are stimulated with oxacillin, cloxacillin, and dicloxacillin, which, bind irreversibly to similar lysine residues on albumin. (A) Structures of the drugs. (B) Mass spectrometric analysis of the profile of lysine modification on albumin with each drug in vitro after 16 hours (10:1; drug:albumin). (C) Stimulation of six flucloxacillin-responsive T-cell clones with oxacillin, cloxacillin, and dicloxacillin. Proliferation was measured by incorporation of [3H]thymidine. Results show mean cpm in the presence and absence of drug.

The drug metabolite nitroso sulfamethoxazole, which binds irreversibly to cysteine residues on multiple proteins generating antigenic determinants for T cells, was used as a specificity control. Flucloxacillin-responsive clones were not activated with the nitroso metabolite (Supporting Fig. 2a).

Albumin Peptide Sequence Incorporating the Flucloxacillin Binding Sites Lys190 and Lys212 Interact With HLA-B*57:01.

The Immune Epitope Database Analysis Resource was utilized to explore potential high affinity HLA-B*57:01 binding peptides derived from human serum albumin. Several peptide sequences that contain flucloxacillin-modifiable Lys residues were identified as high affinity HLA-B*57:01 binders (Fig. 2F), including Lys 190 and Lys212, which are targets for flucloxacillin-derived haptens in drug-exposed patients.

Flucloxacillin Activates Naive CD45RA+CD8+ T Cells From Volunteers Expressing HLA-B*57:01.

In an attempt to prime drug-specific T-cell responses, naive CD3+ T cells from HLA-B*5701-positive volunteers (n = 3) were cocultured with dendritic cells in the presence of flucloxacillin (HLA type of the volunteers shown in Supporting Table 1). After 8 days the primed T cells were restimulated with fresh dendritic cells and the drug and antigen-specificity were assessed using an IFN-γ ELIspot. IFN-γ release with flucloxacillin was found to be antigen-specific (sulfonamides did not activate the cells) and dose-dependent (Fig. 4A). IFN-γ release was not detectable when the volunteer PBMCs were cultured with flucloxacillin (data not shown).

Figure 4.

Priming of naive T cells from HLA-B*57:01 volunteers with dendritic cells and flucloxacillin. (A) IFN-γ ELISpot after T-cell priming with flucloxacillin. (B) Phenotype of T-cell clones isolated from three volunteers. (C) Proliferative response (black lines) and the secretion of cytokines and cytolytic molecules (colored lines and pictures) following flucloxacillin stimulation in five clones. The data show the mean of replicate wells.

A total of 600 CD4+ and CD8+ T-cell clones were generated from the flucloxacillin-primed PBMC. Of these, 35 CD8+ clones were identified as flucloxacillin-responsive by analysis of proliferation in the presence and absence of the drug (Fig. 4B). The flucloxacillin-specific dose-dependent proliferative response and the profile of cytokines and cytolytic molecules released were similar when CD8+ clones from patients with DILI and volunteers were compared (Fig. 4C). Flucloxacillin-responsive CD4+ clones were not detected in volunteers expressing HLA-B*57:01.

Flucloxacillin-Responsive CD8+ T Cells Are Restricted by HLA-B*57:01 and HLA-B*58:01.

Activation of CD8+ T-cell clones from patients with DILI and volunteers expressing HLA-B*57:01 by flucloxacillin was inhibited with an anti-HLA class I, but not a class II, blocking antibody (Fig. 5A). Thus, flucloxacillin responses are dependent on the drug-derived antigen interacting with MHC class I molecules. HLA-B*57:01 restriction was studied using 16 flucloxacillin-responsive clones from patients and volunteers and antigen-presenting cells from 10 donors expressing different HLA-B molecules (Fig. 5C). Six of the donors were selected based on expression of either HLA-B*57:01 or the structurally related HLA-B*58:01, which have an overlap in the peptides they display.17 Importantly, abacavir-specific T-cell responses are not detectable using antigen-presenting cells expressing HLA-B*58:01.6 Flucloxacillin-responsive clones were stimulated to proliferate with flucloxacillin-pulsed autologous antigen-presenting cells and antigen-presenting cells from the three donors expressing HLA-B*57:01. Furthermore, several clones were stimulated by flucloxacillin-derived antigens presented on antigen-presenting cells expressing HLA-B*58:01. Antigen-presenting cells expressing other B-alleles did not stimulate the clones (Fig. 5B; Supporting Fig. 3).

Figure 5.

Flucloxacillin-specific stimulation of CD8+ clones is restricted by HLA-B*57:01 and B*58:01. (A) Inhibition of flucloxacillin-specific proliferation with HLA-class I blocking antibodies in six clones using autologous antigen-presenting cells. (B) Flucloxacillin-specific activation of six clones with flucloxacillin-pulsed (16 hours) antigen-presenting cells expressing HLA-B*57:01 and B*58:01, but not other B alleles (open bars: medium only, filled bars: 2 mM flucloxacillin). Soluble drug was not used to prevent self-presentation by MHC class I molecules expressed on the clones. Parts (A) and (B) show data from different clones. (C) HLA-B alleles expressed by antigen-presenting cells used in (B).

Expression of Distinct Patterns of Homing Receptors on Flucloxacillin-Specific T-Cell Clones.

Distinct chemokine receptor expression profiles control, at least in part, the migration of immune cells. A panel of antibodies were used to demonstrate that flucloxacillin-responsive CD8+ clones from patients with DILI and HLA-B*57:01-positive volunteers express high levels of the receptors CCR2, CCR4, CCR9, and CXCR3, but only low levels of CCR10 and CLA. Other chemokine receptors including CCR1, CCR3, CCR5, and CXCR6 were expressed at low levels on a limited number of clones (Table 2). To show that the chemokine expression was functionally relevant, migration assays using transwells and the CCR4 and CCR9 ligands, CCL17 and CCL25, respectively, were established. Both chemokines induced the migration of CD8+ clones from patients with DILI and drug-naive HLA-B*57:01+ subjects (Supporting Fig. 4).

Table 2. Tissue Homing Receptors Expressed on T-Cell Clones
 Patient ClonesVolunteer Clones
Clone ID123456123456
  • *

    (—) indicates a value of 1.5 or less.

  • Data presented as mean fluorescent index (fluorescence with antibody/fluorescence with isotype).

  • Clones also used in chemotaxis assay (see Supporting Fig. 4).


Flucloxacillin-Responsive CD8+ T Cells Are Not Activated With Abacavir.

The drug-specific T-cell response in abacavir hypersensitive patients is exclusively HLA-B*5701 restricted; thus, we conducted experiments to determine whether flucloxacillin-responsive clones are activated with abacavir. All 14 clones tested proliferated in the presence of flucloxacillin, but not abacavir (Supporting Fig. 2b).


PBMCs were isolated from six patients who developed cholestatic liver injury following flucloxacillin therapy to characterize for the first time the drug-specific T-cell response. PBMC activation with flucloxacillin was detected with five patients, including a HLA-B*57:01-negative individual, using ELIspot to detect antigen-driven cytokine release. Subsequently, CD4+ and CD8+ T cells were isolated from flucloxacillin-treated PBMC and cloned to characterize the cellular pathophysiology of the reaction in each patient. Flucloxacillin-responsive CD8+ clones expressing a range of different Vβ receptors were successfully isolated from the four HLA-B*57:01 ELIspot-positive patients. Contrary to the findings of Chessman et al.,6 showing that abacavir-activated T cells were exclusively CD8+, we were also successful in isolating flucloxacillin-responsive CD4+ clones, albeit in low numbers. Activation of clones with flucloxacillin was concentration-dependent and provoked the secretion of IFN-γ and cytolytic molecules (granzyme B, FasL, and perforin). Individuals given flucloxacillin achieve peak serum levels of 60 μM. However, as there is significant biliary excretion, it is likely that a local concentration in the liver of 100 μM, which activates clones, could be achieved. All of the clones expressed the chemokine receptors CCR2, CCR4, and CCR9. CCR2 and CCR9 are thought to be involved in the migration and accumulation of immune cells in the liver.18,19 Migration of the clones in response to CCR4 and CCR9 ligands demonstrated that the receptor expression was functionally relevant. In contrast to our previous studies with clones from patients with anticonvulsant-induced cutaneous eruptions,20–22 the skin homing lymphocyte receptors CCR10 and CLA were detected at low levels. Flucloxacillin-responsive clones were also isolated from the HLA-B*44:02/55:01-positive patient. However, the clones were all CD4+. Collectively, these data argue that immune phenomena contribute to the development of flucloxacillin-induced liver injury.

It is well established that an obligatory step in β-lactam allergy is the formation of covalent bonds between drug and lysine residues on protein. For protein binding, the β-lactam ring is targeted directly by nucleophilic lysine residues. Nucleophilic attack leads to ring opening and binding of the penicilloyl group. The protein conjugate can also be formed through binding of the reactive degradation product penicillenic acid. Using mass spectrometric methods, we recently identified albumin as a major circulating protein modified with β-lactam antibiotics including flucloxacillin, defined the profile of drug protein conjugation at specific lysine residues with respect to dose and incubation time, and characterized for the first time the sites of modification associated with the stimulation of a clinically relevant drug-specific T-cell response.13,14,16 Herein, we show that (1) the stimulation of clones with flucloxacillin-pulsed antigen-presenting cells, and (2) the detection of flucloxacillin haptens on albumin in culture are time-dependent. Furthermore, simultaneous measurement of antigenicity and immune responsiveness revealed that the cumulative level of flucloxacillin protein binding at each timepoint studied correlated directly with the strength of the T-cell proliferative response. Aldehyde fixation of antigen-presenting cells, which inhibits processing, blocked the flucloxacillin-specific stimulation of clones (Fig. 2D). Collectively, these data indicate that flucloxacillin-protein binding is critical for the formation of functional T-cell antigens. Interestingly, the β-lactam antibiotics piperacillin, amoxicillin, and penicillin G, which bind to similar albumin lysine residues, also stimulated the flucloxacillin-responsive clones. The detection of broadly crossreactive T cells is in contrast to our recent studies describing highly drug-specific T cells in piperacillin-hypersensitive patients,14 and suggests that MHC binding peptides and the core penicilloyl structure provide the binding energy to drive T-cell responses in patients with flucloxacillin-induced liver injury.

Carey and van Pelt23 raised an antibody specific for flucloxacillin-modified proteins to characterize whether flucloxacillin treatment results in adduct formation in vivo. Interestingly, they found that western blot analysis of liver cytosol from treated rats revealed a single flucloxacillin-modified band with a weight of ∼66 kDa, (i.e., the molecular weight of albumin). Thus, in our next series of experiments the IEDB analysis resource was used to identify HLA-B*5701 binding peptides derived from albumin. Peptide sequences containing Lys190, Lys199, and Lys212 were identified as high affinity binders at HLA-B*57:01. Interestingly, all the penicillins investigated modify albumin at these residues.

Previous studies have reported that it may be feasible to substitute flucloxacillin with an alternative anti-staphylococcal agent such as oxacillin, cloxacillin, or dicloxacillin in HLA-B*57:01-positive individuals.10 Our data shows remarkably consistent albumin binding profiles and stimulation of HLA-B*57:01-restricted CD8+ clones across the different drugs. Thus, loss of fluorine, which has the same volume as hydrogen, and chlorine, does not alter the binding of flucloxacillin to protein or indeed the assembly of haptenic determinants that stimulate T cells. This clearly highlights a potential risk of using alternative oxacillins in HLA-B*57:01-positive patients with liver injury.

If the HLA-B*57:01 genotype is a functional determinant of flucloxacillin-induced liver injury, it should be possible to prime naive T cells from HLA-B*57:01-positive volunteers. To explore whether flucloxacillin activates naive T cells, we employed our recently established T-cell priming assay that recapitulates key elements of events that occur in vivo during elicitation of an immunological drug reaction.15 Flucloxacillin-primed T cells from HLA-B*57:01-positive volunteers were found to secrete IFN-γ following restimulation, whereas the cells that had divided were shown to be CD8+ by T-cell cloning. Over 30 flucloxacillin-specific CD8+ clones generated from three volunteers were found to proliferate and secrete cytokines and cytolytic molecules following drug stimulation. The profile of secretory molecules and chemokine receptor expression was similar to those observed with CD8+ clones from patients with DILI.

Genetic restriction of the flucloxacillin-specific response was studied using a panel of CD8+ clones from patients with DILI and HLA-B*57:01-positive drug-naive volunteers. Activation of CD8+ clones was detected with flucloxacillin-pulsed antigen-presenting cells from volunteers expressing HLA-B*57:01 and B*58:01. Flucloxacillin-pulsed antigen-presenting cells expressing other HLA-B alleles did not activate the clones. Importantly, HLA-B*57:01 and B*58:01 are part of the same HLA-B17 serotype; they differ in structure by only five amino acids and have a significant overlap in their antigenic peptide repertoire.6 Given that the HLA-B*57:01 allele is in the strongest linkage disequilibrium with DRB1*07:01 in worldwide populations,24 it is not surprising that all patients positive for B*57:01 were also carriers of DRB1*07:01. We are actively pursuing whether flucloxacillin-responsive CD4+ T cells are DRB1*07:01-restricted.

Abacavir has recently been shown to interact with amino acid residues located deep within the HLA-B*57:01 (but not HLA-B*58:01) binding groove and alter the repertoire of self peptides that are presented to CD8+ T cells.6,25–28 In contrast, flucloxacillin does not alter peptide binding to HLA-B*57:01,28 which supports our hypothesis that functional flucloxacillin antigens derive from naturally processed drug-protein conjugates. Flucloxacillin-responsive CD8+ clones from patients with liver injury and HLA-B*57:01-positive volunteers were not stimulated with abacavir. Furthermore, abacavir-responsive CD8+ clones generated from the same volunteers were not activated with flucloxacillin (results not shown). Collectively, these data demonstrate that the different chemistries associated with these two drugs result in the presentation of unique HLA-B*57:01-restricted epitopes to T cells.

The discovery of surprisingly strong associations between the expression of HLA alleles and DILI has changed the way in which researchers view this form of iatrogenic disease. Our data characterizing flucloxacillin-responsive T cells in patients represents a fundamental breakthrough in our understanding of the role of the adaptive immune system in liver injury. Moreover, the successful priming of naive CD8 T cells using PBMC from HLA-B*57:01-positive volunteers effectively links the genetic association to the disease pathogenesis. In ongoing studies we are seeking to investigate how and why flucloxacillin-specific T cells kill liver cells in susceptible patients.


The authors thank the patients and volunteers for their generous blood donations.