Negative regulation by PD-L1 during drug-specific priming of IL-22 secreting T-cells and the influence of PD-1 on effector T-cell function

Abstract

Activation of PD-1 on T-cells is thought to inhibit antigen-specific T-cell priming and regulate T-cell differentiation. Thus, we sought to measure the drug-specific activation of naïve T-cells after perturbation of PD-L1/2/PD-1 binding and investigate whether PD-1 signaling influences the differentiation of T-cells. Priming of naïve CD4+ and CD8+ T-cells against drug antigens was found to be more effective when PD-L1 signaling was blocked. Upon restimulation, T-cells proliferated more vigorously and secreted increased levels of IFN-γ, IL-13 and IL-22, but not IL-17. Naïve T-cells expressed low levels of PD-1; however, a transient increase in PD-1 expression was observed during drug-specific T-cell priming. Next, drug-specific responses from in vitro primed T-cell clones and clones from hypersensitive patients were measured and correlated with PD-1 expression. All clones were found to secrete IFN-γ, IL-5 and IL-13. More detailed analysis revealed two different cytokine signatures. Clones secreted either FasL/IL-22 or granzyme B. The FasL/IL22 secreting clones expressed the skin homing receptors CCR4, CCR10 and CLA and migrated in response to CCL17/CCL27. PD-1 was stably expressed at different levels on clones; however, PD-1 expression did not correlate with the strength of the antigen-specific proliferative response or the secretion of cytokines/cytolytic molecules. This study shows that PD-L1/PD-1 binding negatively regulates the priming of drug-specific T-cells. ELIspot analysis uncovered an antigen-specific FasL/IL-22 secreting T-cell subset with skin homing properties.

INTRODUCTION

Immunological drug reactions represent a major clinical problem because of their severity and unpredictable nature. In recent years, genome-wide association studies have identified specific HLA alleles as important susceptibility factors for certain reactions (1,2). Drug antigen-specific CD4+ and/or CD8+ T-cell responses are detectable in blood/tissue of patients presenting with mild and severe forms of skin (3-5) and liver injury (6,7) and are therefore believed to participate in the disease pathogenesis. For a limited number of drugs, the drug-derived antigen has been shown to interact specifically with the protein encoded by the HLA risk allele to activate T-cells. However, one must emphasize that, with the exception of abacavir hypersensitivity, the majority of individuals who carry known HLA risk alleles do not develop clinically relevant immunological reactions when exposed to a culprit drug. Thus, there is a need to characterize the immunological parameters that are superimposed on HLA-restricted T-cell activation to determine why particular individuals develop drug hypersensitivity. Infection, especially reactivation of the herpes virus family (8,9), has been put forward as an additional risk factor. Virus infection alone however does not fully explain the unpredictable nature of drug hypersensitivity.

Thus, our current study focuses on two model drug haptens, nitroso sulfamethoxazole (SMX-NO) and flucloxacillin, to investigate whether the programmed death (PD) pathway regulates the drug-specific priming of T-cells from healthy, drug-naïve blood donors. Both compounds have been shown previously to activate CD4+ and CD8+ T-cells isolated from patients presenting with drug-induced tissue injury (SMX-NO, skin injury; flucloxacillin, liver injury) (4,6,10-14). SMX-NO is a cysteine reactive drug metabolite that binds extensively to cellular protein, while flucloxacillin binds directly to lysine residues of serum proteins. This very different chemistry of antigen formation obviates compound-specific effects; as such, any regulation of T-cell priming must involve signaling pathways downstream of the drug interaction with protein. Activation of the PD-1 receptor, which is transiently expressed on activated T-cells (15,16), leads to clustering between T-cell receptors and the phosphatase SHP2, dephosphorylation of T-cell receptor signaling and suppression of antigen-specific T-cell responses (17). PD-1 has two ligands; PD-L1 (CD274) and PD-L2 (CD273); PD-L1 is expressed on a variety of immune cells, while PD-L2 expression is limited to dendritic cells, bone-marrow-derived mast cells and activated macrophages. The PD-1 pathway has already been shown to regulate autoimmunity in several experimental models. Furthermore, genome-wide association studies have identified single nucleotide polymorphisms in the PD-1 gene in humans that are associated with a higher risk of developing autoimmune disease (18).

Although PD-1 has been classified as a marker of cell exhaustion (19,20), recent studies from independent laboratories describe an alternative perspective. Duraiswamy et al. showed that most PD-1^high human CD8+ T-cells are effector memory cells rather than exhausted cells (21). Zelinskyy et al showed that although virus-specific CD8+ T-cells upregulate PD-1 expression during acute infection, the majority of PD-1^high cells were highly cytotoxic and controlled virus replication (22). Finally, Reiley et al. showed that PD-1^high CD4+ T-cells were highly proliferative and appeared to maintain effector T-cell responses during chronic infection (23). Consequently, in the present study T-cell clones were isolated directly from SMX hypersensitive patient PBMC and from healthy drug naïve donors following in vitro priming to characterize the cytokine signatures(s) of antigen specific T-cells and to study whether PD-1 expression/signaling governs the differentiation of T-cells into effector/helper subsets.

MATERIALS AND METHODS

Human subjects

120ml of venous blood was collected from drug naïve donors for T-cell priming. Blood (60ml) was also collected from four SMX hypersensitive patients for cloning. Table I lists the clinical features of the adverse reactions. Approval for the study was acquired from the Liverpool local research ethics committee and informed written consent was obtained.

Table I. Clinical details of the hypersensitive patients and the origin, phenotype and specificity of the T-cell clones.

Patient ID	Clinical details	Clones tested (n)	SMX-NO specific clones	Phenotype (%)		Proliferation (cpm)
				CD4	CD8	Control	SMX-NO
1: Female, Age 30	Maculopapular rash day 2 of treatment; 7 years since reaction	336	21	100	-	6840 ± 7406	14837 ± 13224
2: Female, Age 25	Maculopapular rash day 4 of treatment; 5 years since reaction	394	29	100	-	3899 ± 5522	23970 ± 18651
3: Female, Age 34	Maculopapular rash day 6 of treatment; 8 years since reaction	216	6	100	-	1058 ± 203	2740 ± 469
4: Male, Age 23	Maculopapular rash day 10 of treatment; 20 years since reaction	152	12	100	-	1849 ± 1659	5086 ± 3649

Cell separation

PBMC were isolated using lymphoprep (Axis-shield, Dundee) by density gradient separation. CD14+ monocytes and different T-cell populations were separated using magnetic beads and columns according to the manufacturer’s instructions (Miltenyi Biotech; Bisley, UK). CD14⁺ cells were positively selected from total PBMC. For isolation of naïve T-cells, pan negative T-cell separation was performed using an anti-T-cell antibody cocktail. CD3⁺ cells were then subject to positive selection to remove the unwanted CD25+ T_reg and memory cells (CD45RO⁺). Monocytes and naïve T-cells were frozen and stored at −150°C prior to use.

T-cell priming assay

CD14⁺ cells were cultured in medium (RPMI-1640, 100μg/ml penicillin, 100U/ml streptomycin, 25μg/ml transferrin, 10% human AB serum [Innovative Research; MI, USA], 25mM HEPES buffer, and 2mM L-glutamine) supplemented with GM-CSF and IL-4 (800U/ml; 37°C/5% CO₂) for 7 days to generate dendritic cells. On the penultimate day, 25ng/ml TNFα and 1μg/ml LPS were added as maturation factors. Dendritic cell phenotype (CD1a,CD11a,CD11c,CD14,CD40,CD80,CD83,CD86,CD274,MHC class II) was measured by flow cytometry.

Mature dendritic cells were plated (0.8×10⁵ per well) and cultured with naive CD3+ T cells (2.5×10⁶ per well; 24 well plated total volume 1.5ml) and SMX-NO (50μM) or flucloxacillin (2mM) for 8 days_. Anti-PD-L1 and/or PD-L2 antibodies (Biolegend UK Ltd, London, UK; 10μg/ml) were added to certain wells. Both antibodies have previously been shown to block the receptors (24). Antibody concentrations were optimized in dose-ranging studies around the concentration suggested by the supplier. Where indicated, TGF-β (5ng/ml), IL-1β (10ng/ml) and IL-23 (20ng/ml) or TNFα (50ng/ml) and IL-6 (20ng/ml) were added to the cultures to induce the differentiation of Th17 and Th22 cells, respectively. All experiments were performed at least three times using cells from different blood donors with no previous history of sulfonamide/flucloxacillin exposure.

T-cell readouts

Primed T cells (1×10⁵; 200μl) were harvested and re-stimulated with autologous dendritic cells (4×10³) and drug (SMX-NO [5-50μM], flucloxacillin [0.5-2mM]) and assessed for cytokine secretion (in duplicate cultures per condition) as well as proliferation (in triplicate cultures per condition). After 48h, [³H]thymidine (0.5μCi/well) was added to the proliferation plate. Incorporated radioactivity was counted after a further 16h incubation using a MicroBeta TriLux 1450 LSC β-counter (Perkin Elmer, Cambridge, UK). Proliferation was also assessed using CSFE-labelled cells according to our recently published protocol (11). ELISpot was used, according to the manufacturer’s instructions (Mabtech, Nacka Strand, Sweden) to visualize secreted cytokines (IFN-γ, IL-13, granzyme B, IL-17 and IL-22). Cell phenotype during priming and following restimulation was assessed by staining with CD3-APC, CD4-APC, CD8-PE, CD45RA-FITC, CD45RO-PerCP-Cy5.5 and/or PD-1-PE (CD279) antibodies.

Generation and characterization of T-cell clones

T-cells from the priming assay (i.e., from drug-naïve donors) were cloned directly by serial dilution and repetitive mitogen-driven expansion using previously described methods (21). PBMC (1×10⁶/well; 0.5ml) from hypersensitive patients were cultured with SMX-NO (25μM). Hence, the clones from hypersensitive patients reacting to SMX-NO almost certainly derive from the memory compartment. On days 6 and 9 culture medium was supplemented with IL-2 (200IU/ml) to expand the number of antigen specific T-cells prior to cloning on day 14. Autologous Epstein-Barr virus (EBV)-transformed B-cell lines were used as antigen presenting cells in assays with clones.

Antigen-specificity was assessed by culturing irradiated EBV-transformed B-cells (1×10⁴/well) and SMX-NO with clones (5×10⁴/well; 200μl) for 48h in duplicate cultures per experimental condition. Proliferation was measured by the addition of [³H]thymidine followed by scintillation counting. Clones with a stimulation index (mean cpm drug-treated wells/mean cpm in control wells) of greater than 2 were expanded by repetitive stimulation with irradiated allogeneic PBMC (5×10⁴/well; 200μl) and 5μg/ml PHA in IL-2 containing medium. Dose-dependent proliferative responses (± PD-L1 block; SMX-NO [5-50μM]) and the profile of secreted cytokines (IFN-γ, IL-5, IL-13, granzyme B, FasL, perforin, IL-17 and IL22 ELISpot) were then measured. Cell phenotyping was performed by flow cytometry using CD4-FITC, CD8-PE and PD-1-PE, CCR4-PE, CCR10-PE and CLA-FITC antibodies.

Twenty-four well transwell chambers with 5-μm pores were used to measure chemotaxis. T-cells (0.1×10⁶; n=4 clones) were placed in the upper chambers. CCL17 (100ng/ml)/CCL27 (400ng/ml) were placed in the lower wells and the cells were incubated for 0.5-24h. Cells migrating to the lower chamber were counted using a hemocytometer.

Flow cytometry

Cells were acquired using a FACSCanto II (BD Biosciences) and data analyzed by Cyflogic. For CFSE analysis, a minimum of 50,000 lymphocytes were acquired using forward scatter/side scatter characteristics.

Statistics

For in vitro priming of naïve T-cells, all experimental data show the mean of three triplicate incubations. All priming experiments were conducted on at least three separate occasions. To characterize cytokine signatures with clones, multiple clones (up to 17 per experiment) from different donors were analysed. Experiments were conducted in duplicate or triplicate, depending on the availability of cells. Mean values and standard deviations were calculated, and statistical analysis was performed using paired T test (Sigmaplot 12 software).

RESULTS

PD-L1 block enhances the priming of naïve T-cells against drug derived antigens

For in vitro priming, naïve CD3+ T cells from drug-naive donors were co-cultured with autologous mature dendritic cells in the presence of SMX-NO (± PD-L1/PD-L2 block). Dendritic cells were routinely stained for co-stimulatory receptors and characterized as CD1a^negativeCD11a^highCD11c^highCD14^negativeCD40^highCD80^highCD83^lowCD86^highMHC class II^high and PD-L1^high (data not shown). As described previously, SMX-NO does not significantly alter the maturation status of mature dendritic cells. Upon re-stimulation, dose-dependent antigen-specific proliferation was clearly detectable (Figure 1A; P<0.05 SMX-NO 12.5-50 μM). Inclusion of PD-L1 blockade markedly enhanced the proliferative response (P<0.05; at each SMX-NO concentration). PD-L2 block however gave proliferative responses comparable to those without PD-ligand block (Figure 1A). Blockade of PD-L1 and PD-L2 together produced enhanced proliferation compared to medium alone (P<0.05 SMX-NO 12.5-50 μM), but less so than for PD-L1 block.

Figure 1. SMX-NO specific priming of naïve T-cells with and without PD-L1/PD-L2 block.

(A) Antigen-specific T-cell responses to SMX-NO measured by [³H]thymidine incorporation. (B, C) Antigen-specific T-cell responses measured by IFN-γ (B) and granzyme B (C) ELIspot. Line graphs show mean±SD of 3 experiments (*P<0.05). (D) Flucloxacillin-specific priming of naïve T-cells with and without PD-L1. Antigen-specific T-cells responses measured by IFN-γ and granzyme B ELIspot. Line graphs show mean±SD of 2 experiments conducted in triplicate (*P<0.05).

Additional SMX-NO priming experiments were performed using IFN-γ and granzyme B secretion as readouts. IFN-γ (Figure 1B) and granzyme-B (Figure 1C) were released from SMX-NO primed cells following restimulation and the response was enhanced when the anti-PD-L1 antibody was included in the co-culture.

To explore whether PD-L1 regulates priming against other drug antigens, naïve T-cells from HLA-B*57:01+ donors were co-cultured with autologous dendritic cells and flucloxacillin (2mM; ± PD-L1/PD-L2 block). In initial experiments, activation of naïve T-cells with flucloxacillin was only detected with PD-L1 block (Figure 1D; left hand side); IFN-γ and granzyme B release above that of background was discernible with flucloxacillin. The granzyme B ELIspot was then repeated on 2 further occasions with naïve T-cells from different donors (using triplicate samples per culture condition). Low levels of granzyme B secretion were detectable when T-cells primed in the absence of PD-L1 block were stimulated with flucloxacillin and this was significantly enhanced using T-cells primed in the presence of PD-L1 block (Figure 1D right hand side).

To confirm antigen specificity, SMX-NO- and flucloxacillin-primed T-cells were cultured with the alternative drug prior to analysis of proliferation and/or cytokine secretion. SMX-NO- and flucloxacillin-primed cells were not activated with flucloxacillin and SMX-NO, respectively (results not shown).

PD-1 expression is enhanced on drug-primed dividing T cells

CFSE staining allows a more in depth analysis of proliferation by quantifying the number of dividing cells and distinguishing between different cell populations. SMX-NO stimulated the activation of naïve T-cells in a concentration-dependent fashion; similar to the [³H]thymidine data (Supplementary Figure 1). CD4+ cells seemed to be more responsive than CD8+ cells to enhanced proliferation in the presence of PD-L1 block. It should be noted that PD-L1 signaling might also regulate the quantity of cytokines/cytolytic molecules secreted by individual cells; hence, these data do not exclude PD-L1/PD-1 mediated regulation of the CD8+ T-cells. The expression of CD45RO on dividing and non-dividing cells was measured to characterize cell phenotype. As expected, the non-dividing cells were CD45RO negative, while the dividing cells stained CD45RO positive (results not shown)

CFSE staining was also used to measure PD-1 on dividing and non-dividing T-cells. In initial experiments, PD-1 expression was found to be significantly upregulated on dividing CD3+, CD4+ and CD8+ cells 48 h after SMX-NO restimulation (Figure 2A). In subsequent experiments, PD-1 expression was measured during SMX-NO priming and for 72h after restimulation. Little or no expression of PD-1 was detected on naïve CD4+ and CD8+ cells. After priming, a small population of PD-1 positive cells was seen on day 7, both in the presence and absence of PD-L1 block. After restimulation with SMX-NO (day 9), PD-1 expression was rapidly upregulated on 20-40% of CD4+ and CD8+ cells in a transient fashion (Figure 2B; columns 1 and 3). PD-1 reverted back to pre-restimulation levels within 48-72h. Based on the intensity of PD-1 staining detected by flow cytometry, it would seem that activated CD4+ T-cells expressed higher levels of PD-1, when CD4+ and CD8+ T-cells were compared. Collectively, these data show that CD4+ and CD8+ T-cells are activated during drug-specific priming and the activated cells express high levels of PD-1 in a transient fashion.

Figure 2. PD-1 expression on CD3+, CD4+ and CD8+ T-cells.

(A) PD-1 expression on dividing and non-dividing CD3+, CD4+ or CD8+ cells after restimulation with SMX-NO. (B) PD-1 expression on dividing and non-dividing during priming and after SMX-NO restimulation (+/-PD-L1 block). An aliquot of CFSE labelled cells was taken throughout the culture period and PD-1 expression measured.

In the presence of PD-L1 block, the increase in PD-1 expression was sustained (Figure 2B; columns 2 and 4). Greater than 30% of CD4+ cells stained positive for PD-1 72h after restimulation.

Cytokine signatures secreted from drug-primed T-cells

Cutaneous reactions to drugs have been classified previously according to the phenotype and function of antigen-specific T-cells. However, the discovery of new T-cells subsets (e.g., Th17, Th22 cells) may render this classification somewhat obsolete. Thus, to study whether PD-1 expression/signaling governs the function of effector/helper T-cell subsets it was first important to characterize the cytokine signatures(s) of antigen specific T-cells following SMX-NO priming under different polarizing conditions.

In initial experiments, naive CD3+ T-cells were cultured with dendritic cells and SMX-NO in the absence of polarizing cytokines. These cells were then harvested, restimulated with SMX-NO and assayed for IFN-γ, IL-13, IL-17 and IL-22 secretion. SMX-NO-specific secretion of IFN-γ, IL-13 and IL-22 was observed. However, IL-17 release was not detected (Figure 3).

Figure 3. IL-22 secretion by T-cells exposed to SMX-NO.

Naïve CD3+ T-cells were co-cultured with SMX-NO and dendritic cells for 8 days. The cultures were plated and restimulated with fresh dendritic cells and SMX-NO. Antigen-specific T-cell responses were measured by IFN-γ, IL-13, IL-17 and IL-22 ELIspot. Figure show representative data from 1 out of 3 experiments.

When the priming assay was repeated under Th22 polarizing conditions, a substantial number of IL-22 secreting colonies were observed for a second time. In contrast, SMX-NO-specific IL-17 secretion was not detected when naïve T-cells were primed with SMX-NO under Th17 or Th22 polarizing conditions (Supplementary Figure 2).

Generation of CD4+ T-cell clones and characterization of cytokine secretion profiles

To characterize the functionality of drug-responsive T-cells and the way in which PD-1 signaling influences effector T-cell responses, clones isolated following in vitro priming and from hypersensitive patient PBMC were studied.

Two hundred and eighty three T-cell clones were generated from drug-naïve donors following SMX-NO priming of in the presence of PD-L1 block. Nineteen CD4+ clones were found to proliferate in the presence of SMX-NO (no drug, 1651±410cpm; SMX-NO, 4880±913cpm). SMX-NO responsive clones secreted of IFN-γ, IL-5 and IL-13 following activation (Figure 4). For comparison, SMX-NO responsive clones were generated following priming in the absence of PD-L1 block. They were found to secrete a similar panel of cytokines (results not shown).

Figure 4. SMX-NO specific activation of T-cell clones isolated from hypersensitive patients and following in vitro priming.

Clones were incubated with antigen presenting cells +/− SMX-NO and (A) proliferative responses and (B) IFN-γ, IL-5 and IL-13 were measured using [³H]thymidine incorporation and ELIspot, respectively (*P<0.05 when SMX-NO and control wells were compared).

The number of clones generated from hypersensitive patients, their CD phenotype and the SMX-NO-specific proliferative response are summarized in Table I. No significant difference in the secretion IFN-γ, IL-5 or IL-13 was observed when clones from hypersensitive patients and the in vitro priming assay were compared (Figure 4).

Since hypersensitive patient clones and clones isolated from in vitro priming secreted similar Th1 and Th2 cytokines, a panel of 17 clones (from multiple donors) with a strong growth pattern were selected to study IL-17 and IL-22 secretion. Approximately 50% of clones were found to secrete IL-22 following exposure to SMX-NO. In contrast, IL-17 secretion was only detected with 1 clone (Figure 5A). IL-22 secreting clones were isolated from hypersensitive patient PBMC and following in vitro priming (Figure 5B). Importantly, the isolation of SMX-NO responsive, IL-22 secreting clones from the priming assay was not dependent on the presence of Th22 polarizing cytokines and clones were not maintained under Th22 polarizing conditions.

Figure 5. Cytokine secretion by SMX-NO responsive CD4+ clones.

(A) Analysis of SMX-NO-specific cytokine and cytolytic molecule secretion from 17 clones by ELIspot. (B) Cytokine profile of representative SMX-NO-responsive IL-22^high and IL-22^low secreting clones generated from (1) drug naïve donors after in vitro priming and (2) hypersensitive patients.

ELIspot was also used to study secretion of the cytolytic molecules perforin, granzyme B and FasL. Interestingly, the clones were found to secrete either FasL or granzyme B, but not perforin. The IL-22^high clones belonged exclusively to the FasL producing subset (Figure 5A).

To confirm antigen specificity, SMX-NO clones were cultured with flucloxacillin or carbamazepine prior to analysis of proliferation. SMX-NO responsive clones were not activated with either drug (Supplementary Figure 3).

IL-22^highFasL^high clones express skin homing receptors and migrate in response to CCL17 and CCL27

A panel of seven clones showing the different cytokine signatures (2 clones secreting FasL^high IL-22^highgranzyme B^low; 2 clones secreting FasL^high IL-22^lowgranzyme B^low; 3 clones secreting FasL^low IL-22^lowgranzyme B^high) was then selected to explore which clones express CCR4, CCR10 and CLA and hence have the ability to migrate towards skin. All FasL^high clones expressed high levels of CCR4, CCR10 and CLA (Figure 6A) and migrated in the presence of CCL17 and CCL27 (Figure 6B).

Figure 6. Skin-homing receptor expression and the migration of clones towards CCL17 and CCL27.

Flow cytometric analysis of CCR4, CCR10 and CLA expression on SMX-NO responsive clones. Seven clones (2 clones [clone 1 and 2] secreting FasL^high IL-22^highgranzyme B^low; 2 clones [clones 3 and 4] secreting FasL^high IL-22^lowgranzyme B^low; 3 clones [clones 5-7] secreting FasL^low IL-22^lowgranzyme B^high) were selected for the analysis (B) Chemotaxis of FasL/IL-22 and granzyme B secreting clones promoted by CCL17 or CCL27 (cells were only available for clones 1, 2, 6 and 7).

PD-L1 signaling does not regulate the functionality of antigen-specific T-cells

Although PD-1 is most commonly described as a marker of cell exhaustion, it has also been reported that PD-1^high cells are highly cytotoxic and/or proliferative.^{22, 23} Thus, our SMX-NO-specific clones were used to (1) measure PD-1 expression on individual clones, (2) explore the relationship between PD-1 expression and effector function and (3) analyse whether PD-L1 block alters the levels or profile of cytokines secreted following antigen stimulation.

Flow cytometric analysis of PD-1 on 40 clones revealed a 4 fold difference in expression (Figure 7A). PD-1 was stably expressed on the surface of clones; Figure 7B show PD-1 staining on 2 representative clones maintained in culture for 10 days (± SMX-NO stimulation). PD-1 expression did not correlate with the strength of the drug-specific proliferative response or secretion of IFN-γ, IL-5, IL-13, IL-17, IL-22, perforin, granzyme B or FasL (r² less than 0.2 for all parameters tested; results not shown; 17 clones depicted in Figure 5A were used for the comparisons). Despite this, PD-L1 block resulted in a modest increase in IFN-γ, IL-13 and granzyme B secretion when clones were stimulated with SMX-NO (Figure 7C).

Figure 7. PD-1 expression on CD4+ clones and SMX-NO specific cytokine secretion with and without PD-L1 block.

(A) PD-1 expression on SMX-NO specific CD4+ clones. (B) PD-1 expression on dividing and non-dividing CD4+ clones. Clones were cultured with or without SMX-NO and PD-1 expression was measured. (C) Proliferative responses and cytokine secretion from 10 SMX-NO specific clones with and without PD-L1 block. ELIspot images show differences in cytokine secretion from two representative clones ± PD-L1 block.

DISCUSSION

In the present study we have focused on regulation of drug antigen-specific T-cell priming through the PD-1/PD-L1 pathway, the way in which PD-1 signaling influences effector T-cell responses and the functionality of drug-responsive clones generated through in vitro priming and isolated from hypersensitive patient PBMC. The inhibitory function of PD-1 relies on the presence of an immunoreceptor tyrosine based switch motif. On activation, the switch becomes phosphorylated and subsequently recruits the protein tyrosine phosphatase SHP-2. This causes the inhibition of downstream pathways through the dephosphorylation of proteins such as CD3 and ZAP70 (16,25) preventing further T-cell stimulation. To assess the effect of PD-ligand blockade and in particular whether this could be used as an immunogenic boost to enhance drug-specific stimulation of naïve T-cells we utilized an in vitro T-cell priming assay and the model drug haptens SMX-NO and flucloxacillin. In agreement with our previous study, an eight day culture period was sufficient to activate naïve CD3+ T-cells and SMX-NO-specific responses were readily detectable following antigen recall using readouts for proliferation and IFN-γ or granzyme B secretion. CFSE staining revealed that naïve CD4+ and CD8+ T-cells were activated during priming. The dividing cells were CD45RO+, indicating a change in phenotype from naïve to memory. PD-1 expression was induced on dividing T-cells during priming and following antigen recall. An increase in the magnitude of the drug-specific proliferative response and levels of IFN-γ/granzyme B secretion was seen when naïve T-cells were exposed to PD-L1-block. In contrast, PD-L2 block had no effect and even hindered the increased activation of T-cells produced from PD-L1 blockade in 2 out of 3 donors. Previous studies have shown that PD-1/PD-L2 signaling inhibits T-cell receptor-mediated triggering of proliferation and cytokine release. Thus, it is not clear why PD-L2 block did not enhance the priming of naïve T-cells against SMX-NO. One potential explanation is that B7.1 (CD80), a CD28 co-stimulatory ligand, is known to interact with PD-L1, but not PD-L2 (26); this however requires further investigation. Interestingly, CFSE staining suggested the CD4+ cells might be more sensitive to the effects of PD-L1 block than CD8+ cells.

Next, we sought to establish whether PD-L1/PD-1 signaling negatively regulates the priming of T-cells against other drug-derived antigens. To do this, we focused on the β-lactam antibiotic flucloxacillin. In contrast to SMX-NO (27-29), which forms antigenic determinants through the irreversible modification of cysteine, the β-lactam ring of flucloxacillin is targeted by nucleophilic lysine residues (6,30). We have recently characterized drug-responsive CD4+ and CD8+ T-cells that express the gut homing receptors CDR4 and CCR9 from patients with liver injury, but not tolerant controls (6). The HLA-B*57:01 geneotype is a major determinant of flucloxacillin-induced liver injury and through priming of naïve T-cells from blood donors carrying the HLA risk allele it has been possible to link the genetic association to the disease pathogenesis (6,12,31). Importantly, priming naïve T-cells against flucloxacillin generally leads to weak and inconsistent results. Confirmation of priming is often only obtained when flucloxacillin-responsive T-cells are cloned. Herein, we clearly show enhanced priming of flucloxacillin-specific T-cells with PD-L1 block using IFN-γ and granzyme B ELIspot as readouts.

Whether PD-1 signaling regulates the activation of antigen-specific memory T-cell responses has yet to be fully defined. Previous studies show that PD-1^high cells can be highly cytotoxic and that PD-1 expression might be a marker of effector memory function, which seems counterintuitive (21,22). Thus, using SMX-NO-responsive clones generated from healthy donors through priming and isolated directly from hypersensitive patient PBMC (whose T-cells were primed at the time of the adverse reaction (3,4,13)), we assessed whether PD-1 expression correlated with the strength of the antigen-specific proliferative response and/or secretion of cytokines/cytolytic molecules. Furthermore, PD-L1/2 blocking antibodies were used to assess whether PD-1 signaling regulates the activation of antigen-specific memory T-cells. Detailed analysis of 40 clones revealed (1) a four-fold variation in PD-1 expression on T-cells, (2) PD-1 was stably expressed for up to 10 days after antigen stimulation, and (3) there was no correlation between PD-1 expression and the magnitude of the drug-specific proliferative response or secretion of cytokines. Nevertheless, subtle increases in IFN-γ, IL-13 and granzyme B secretion were observed when clones were stimulated with SMX-NO in the presence of PD-1 block. PD-L2 block had no significant effect. Clearly, our data may relate in part to the manipulations involved in the long-term T-cell culture; however, at present there are no ready alternatives to investigate these effects in humans. Furthermore, fully characterized animal models of drug hypersensitivity are not widely available.

Previous immunohistological studies characterizing the phenotype of T-cells infiltrating inflamed skin of patients with maculopapular skin rashes describe the presence of large numbers of CD4+ T-cells and lower numbers of CD8+ T-cells (32,33). Studies focusing on the SMX-specific T-cell response show that CD4+ and CD8+ T-cells can be activated by the drug to secrete cytolytic molecules; however, keratinocytes are specifically killed by CD4+ T-cells (34). In agreement with the these findings, more recent studies show that most SMX(metabolite)-specific T-cells isolated from hypersensitive patients are CD4+ and secrete a mixed panel of Th1/Th2 cytokines including IFN-γ, IL-5 and IL-13 (4,35,36). However, the discovery of new T-cell populations (e.g., Th9, Th17, Th22) renders this classification out-of-date. For this reason, we conducted a detailed analysis of the cytokines released by SMX-NO specific CD4+ T-cells generated through in vitro priming and from hypersensitive patients. Following antigen recall, the SMX-NO primed T-cells from healthy donors were found to secrete IFN-γ, IL-13 and IL-22, but IL-17 secretion was not detected. CD4+ clones isolated from the priming assay also secreted IFN-γ, IL-5 and IL-13, but no IL-17. IL-22 secretion was detected from approximately 50% of the clones (Figure 5). A similar pattern of cytokine secretion was seen with clones (IFN-γ^high IL-5^high IL-13^high IL-22^low and IFN-γ^high IL-5^high IL-13^high IL-22^high) isolated from SMX hypersensitive patients (Figure 5). IL-22 is a cytokine that modulates tissue responses as expression of the IL-22R1 receptor is restricted to non-haematopoietic cells. In skin, the IL-22 receptor is expressed at high levels on keratinocytes and IL-22 has been found to enhance keratinocyte proliferation and inhibit terminal differentiation (37). Furthermore, IL-22 has been shown to mediate inflammatory responses in patients with psoriasis and IL-22 secreting cells have been identified in patients with allergic contact dermatitis (38-40). Our data is, however, the first to show production of IL-22 alongside IFN-γ by antigen-specific T-cells from drug hypersensitive patients.

Given the heterogeneous secretion of IL-22 by individual CD4+ clones, the release of cytolytic molecules (perforin, granzyme B and FasL) and expression of skin-homing chemokine receptors were also measured using ELIspot and flow cytometry, respectively. These studies clearly show that SMX-NO responsive CD4+ T-cells release cytolytic molecules when activated through their T-cell receptor. Two subsets of drug-specific clone were identified and classified according to the production of either granzyme B or FasL. Importantly, the IL-22 secreting clones produced FasL following antigen stimulation. A preliminary analysis of skin homing receptors on 7 clones revealed they expressed high levels of CCR4, CCR10 and CLA and migrated towards CCL17 and CCL27, indicating that the receptor expression was functionally relevant. Collectively, these studies identify two pathway of killing by drug-specific T-cell clones. The FasL and IL-22 secreting clones may be crucial mediators of the immunological reaction as they are programmed to migrate towards skin.

In conclusion, our in vitro study found that PD-L1/PD-1 signaling negatively regulates the priming of drug antigen-specific T-cells that secrete a heterogeneous pattern of cytokines. These data provide a foundation to explore PD-L1/PD-1 expression and activity in prospective studies of drug immunogenicity.