Cell type-specific targeting dissociates the therapeutic from the adverse effects of protein kinase inhibition in allergic skin disease

Abstract

The kinase p38α, originally identified because of its endotoxin- and cytokine-inducible activity and affinity for antiinflammatory compounds, has been posited as a promising therapeutic target for various immune-mediated disorders. In clinical trials, however, p38α inhibitors produced adverse skin reactions and other toxic effects that often outweighed their benefits. Such toxicity may arise from a perturbation of physiological functions unrelated to or even protective against the disease being treated. Here, we show that the effect of interfering with p38α signaling can be therapeutic or adverse depending on the targeted cell type. Using a panel of mutant mice devoid of p38α in distinct cell types and an experimental model of allergic skin disease, we find that dendritic cell (DC)-intrinsic p38α function is crucial for both antigen-specific T-cell priming and T-cell–mediated skin inflammation, two independent processes essential for the immunopathogenesis. By contrast, p38α in other cell types serves to prevent excessive inflammation or maintain naïve T-cell pools in the peripheral lymphoid tissues. These findings highlight a dilemma in the clinical use of p38α inhibitors, yet also suggest cell-selective targeting as a potential solution for improving their therapeutic index.

The kinase p38α, the most abundant and ubiquitously expressed p38 MAP kinase isoform in mammals, was discovered based on its binding affinity for antiinflammatory compounds (1). The nature of the stimuli that elicited p38α activation and enabled its identification—proinflammatory cytokines, microbial products, and injurious environmental insults—also hinted at a role for p38α in the immune and stress response (2–4). Pharmacological inhibition of p38α has since held promise for the treatment of allergic, autoimmune, and other diseases of inflammatory etiology. A series of recent clinical studies, however, revealed the frequent occurrence of adverse events, ranging from skin rashes to liver damage, after the use of p38α inhibitors (5–7). These toxicities limited the dose and frequency of p38α inhibitor treatment and have become a liability to fulfilling its promise as an effective therapeutic strategy.

The therapeutic index is a relative measure of the efficacy versus toxicity of a treatment regimen. Toxic side effects have often been the cause of the failure of an otherwise effective therapeutic agent such as p38α inhibitors. We questioned whether the adverse effects of p38α inhibitors arose from interference with a physiological function of the protein kinase and, if so, whether the therapeutic and the adverse effects of p38α inhibition were based on distinct cell type-specific mechanisms. In this study, we used a mouse model of allergic contact dermatitis to examine the disease responses of conditional p38α knockout (KO) mice. In these mice, ablation of p38α expression was targeted to keratinocytes, myeloid cells, dendritic cells, or T cells, which represented a simulation of p38α inhibition in one cell type at a time. We observe cell type-specific effects of p38α loss on allergic skin inflammation and propose that cell-selective targeting may help increase the therapeutic index of p38α inhibition.

Results and Discussion

Keratinocytes, the epithelial cells of the skin, play an active role in inflammatory responses, not only serving the epidermal barrier function, but also producing various inflammatory mediators (8). Best characterized as sentinels and effectors of innate immunity, myeloid cells such as macrophages and neutrophils are central to the triggering and regulation of inflammation (9, 10). Our previous study showed that p38α protein kinase signaling was pivotal in eliciting inflammatory responses to acute tissue injury, yet the contribution of epithelial and myeloid p38α differed depending on the mode of injury (11). Meanwhile, the precise roles of epithelial and myeloid cells, let alone those of p38α in these cell types, in antigen-specific T-cell–mediated inflammation remain ill defined. To address this knowledge gap, we sought to determine the effects of cell type-specific p38α deficiency on contact hypersensitivity (CH). In this skin immune reaction, topical contact with a small-molecule allergen (hapten) leads to a priming of specific T cells (sensitization phase); the resulting effector T cells, particularly CD8⁺ effectors (12–15), are recruited to and activated in the skin upon reencounter with the hapten, a process associated with actual skin disease (challenge phase).

Keratinocyte- and myeloid cell-specific p38α KO mice (11), designated K-KO and M-KO, respectively, were examined for the severity of CH reaction to the hapten 2,4-dinitrofluorobenzene (DNFB). Expression of p38α was efficiently ablated and activation of its downstream kinases (16), such as MAP kinase-activated protein kinase 2 (MK2) and mitogen- and stress-activated kinase 1 (MSK1), was attenuated in keratinocytes and macrophages from K-KO and M-KO mice, respectively (Fig. 1 A and B). Both mutant mice exhibited more severe DNFB-induced edema and rash compared with wild-type (WT) mice, as indicated by greater swelling of ear skin as early as 24 h after hapten challenge (Fig. 1 C–F). We performed adoptive T-cell transfer experiments to assess the role of p38α during the sensitization and the challenge phase. Hapten-sensitized K-KO and M-KO mice could generate lymph node (LN) T-cell populations that transfer hapten sensitivity to naïve WT recipient animals. The severities of inflammatory reactions in the recipients that received T cells from WT, K-KO, and M-KO donors were comparable (Fig. 1 G and H). However, naïve K-KO and M-KO mice developed more severe inflammation compared with WT after passive acquisition of hapten sensitivity from donor-derived T cells (Fig. 1 G and H). These results show that p38α in keratinocytes and myeloid cells is dispensable for hapten-specific T-cell priming; rather, the role of p38α in these cells appears to be antiinflammatory and confined to moderating inflammatory responses during the challenge phase. Pharmacological interference with antiinflammatory mechanisms in epithelial and myeloid cells may account for the adverse effects of p38α inhibitors seen in clinical settings (5–7).

Fig. 1.

Keratinocytes and myeloid cell-specific p38α ablation exacerbates skin inflammatory responses during the challenge phase of CH. (A and B) Keratinocytes and bone marrow-derived macrophages from the indicated mice were treated with 12-O-tetradecanoylphorbol-13-acetate (TPA; 100 nM) and lipopolysaccharide (LPS; 100 ng/mL). Whole-cell lysates were prepared after the indicated durations of stimulation and analyzed by immunoblotting with antibodies against the proteins indicated on the left. p-, phosphorylated. (C–F) WT, K-KO, and M-KO mice were sensitized on two consecutive days by topical application of DNFB on the shaved abdomen and chest; 4 d later, the left ear was challenged with DNFB and the right ear with vehicle. Skin tissue sections from the auricles of the indicated mice were prepared 48 h after hapten treatment and analyzed by hematoxylin and eosin staining (C and E). (Scale bars: 100 μm.) Hapten-specific ear swelling was determined at the indicated time points (D and F). Data represent mean ± SE (n = 5–7). (G and H) LN T cells were prepared from donor (D) mice 4 d after hapten sensitization and transferred to naïve recipient (R) mice as indicated. One day later, the left and right ears of the R mice were challenged as in C–F, and ear swelling was determined 24 h after hapten challenge. **P < 0.01; *P < 0.05.

We extended our investigation of p38α function to other cell types that participate in hapten-specific T-cell priming and hapten-induced inflammatory responses. We generated mice in which Mapk14, the p38α gene, was specifically deleted in DCs, the most proficient antigen-presenting cells and a proximal regulator of antigen-induced T-cell responses. Expression of p38α in LN DCs from these mutant mice, D-KO, was almost abolished (Fig. 2A) and, consequently, MK2 and MSK1 activation after treatment with CD40 ligand (CD40L), a potent DC stimulant, was markedly reduced (Fig. 2B). D-KO mice did not display any apparent phenotypic changes or spontaneously develop immune-mediated disorders in specific pathogen-free conditions. The numbers (Fig. 2C) and relative fractions (Fig. S1) of specific immune cell subpopulations in the lymphoid tissues of naïve D-KO mice were normal, indicating no major baseline abnormalities in the immune system. In marked contrast to the CH response in K-KO and M-KO mice, DNFB-induced skin swelling and rash in D-KO mice were greatly diminished compared with WT (Fig. 2 D and E). T cells from hapten-sensitized D-KO mice could not transfer hapten sensitivity to naïve recipient mice as efficiently as WT donor T cells, nor were D-KO recipient mice competent to develop hapten-induced skin reactions after receiving T cells from sensitized WT mice (Fig. 2F). Therefore, DC p38α was critically required not only for hapten sensitization, but also for hapten-induced skin inflammation.

Fig. 2.

DC-specific p38α ablation diminishes CH reactions, attenuating both the sensitization and the challenge phase. (A) Whole-cell lysates from WT and D-KO LN DCs were analyzed by immunoblotting with antibodies against the proteins indicated on the left. The numbers indicate individual animals. (B) WT and D-KO LN DCs were treated with CD40 ligand (CD40L; 1 μg/ml). Whole-cell lysates were prepared after the indicated durations of stimulation and analyzed by immunoblotting with antibodies against the proteins indicated on the left. p-, phosphorylated. (C) The number of cells expressing the indicated markers in the lymph nodes (LN) and spleen (Sp) of the indicated mice was determined by flow cytometry. (D and E) WT and D-KO mice (n = 5) were sensitized and challenged with DNFB, and hapten-specific ear swelling was determined as in Fig. 1. (F) Recipient (R) mice received LN T cells from donor (D) mice and were challenged with hapten and analyzed as in Fig. 1. (G) CD69⁺ cells among the indicated T-cell subpopulations in the LNs of WT and D-KO mice were analyzed by flow cytometry 1 d after hapten sensitization. (H) LN CD8⁺ T cells were prepared from WT or D-KO mice 4 d after hapten sensitization, mixed with hapten-pulsed LN DCs from C57BL/6 mice as antigen-presenting cells (APC) at the indicated ratios. Seventy-two hours later, amounts of IFN-γ and IL-13 in the culture supernatant were determined by ELISA (n = 3). **P < 0.01; *P < 0.05.

DNFB-sensitized D-KO mice showed very weak T-cell expansion in the LNs draining the sensitization site, mainly owing to the lack of increase in CD8⁺ T-cell number (Fig. S2). Furthermore, LN CD8⁺ T cells isolated from sensitized D-KO mice contained much lower fractions of activated CD69⁺ cells (Fig. 2G) and produced substantially reduced amounts of cytokines upon hapten restimulation in vitro (Fig. 2H). These results showed that D-KO mice failed to generate a population of activated hapten-responsive T cells. Nevertheless, the ability of p38α-KO DCs to uptake antigens and present them to T cells in general seemed intact, given the undiminished capacity of p38α-KO DCs to mediate in vitro T-cell responses to DNFB (Fig. S3A). Also, in vitro migration of WT and p38α-KO DCs in gradients of the chemokines CCL19 and CCL21 was comparable (Fig. S3B). Hence, ablation of p38α in DCs likely preserved most of the fundamental cell-autonomous properties required for antigen-presenting cell function.

To further substantiate the functional defects of p38α-KO DCs and identify the associated gene expression changes, we compared gene expression in WT and p38α-KO DCs by DNA microarray analysis. As an in vitro model for inducing p38α activity and p38α-dependent gene expression, we used Pam₃CSK₄, an agonist of Toll-like receptor 2 (TLR2), as a stimulant because it strongly induced p38α phosphorylation in DCs, and TLR2 signaling has been shown to be relevant to and crucial for CH (17, 18). TLR2-activated p38α-KO DCs underwent functional maturation, gaining the ability to present a peptide antigen to and induce the proliferation of CD4⁺ and CD8⁺ T cells (Fig. 3A). Correspondingly, the WT and p38α-KO DCs exhibited gene expression profiles indicative of maturation, inducing cytokine and other immune regulatory gene expression at similar intensities and with similar kinetics (Fig. 3B). There was, however, a small group of genes whose expression was substantially decreased in p38α-KO compared with WT DCs. Among the p38α-dependent genes identified was Ccl17, which encodes the chemokine CCL17, also known as thymus- and activation-regulated chemokine (TARC). The induction of CCL17 mRNA and protein was much lower in Pam₃CSK₄- and CD40L-stimulated p38α-KO DCs (Fig. 3 C and D), yet the production of cytokines such as tumor necrosis factor, interleukin (IL)-6, and IL-12 (p70) was not impaired in the same stimulation conditions (Fig. S4). It has been demonstrated that CCL17 expression is restricted to DCs and that CCL17 deficiency attenuates CH reactions (19). DC-derived CCL17 was found to promote the interaction of DCs with CD8⁺ T cells and, hence, the activation of CD8⁺ T cells (20, 21) and the migration of skin-resident dendritic cells (22). Hapten-induced emigration of epidermal dendritic cells was impaired in D-KO mice (Fig. S5). Of note, however, CCL17-KO DCs (22), but not p38α-KO DCs (Fig. S3B), displayed cell-autonomous defects in CCL19- and CCL21-directed migration in vitro. This discrepancy is presumably due to incomplete loss of CCL17 expression in p38α-KO DCs. Overall, p38α-dependent CCL17 expression in DCs accounts, at least in part, for the failure of D-KO mice to support hapten-specific CD8⁺ T-cell priming and CD8⁺ T-cell-mediated skin inflammation.

Fig. 3.

Transcriptional induction of specific genes in DCs requires p38α. (A) CFSE-labeled naïve CD8⁺ T cells from OT-I mice and CD4⁺ T cells from OT-II mice were mixed with ovalbumin peptide-pulsed WT and D-KO LN DCs as indicated, cultured for 3 d, and analyzed by flow cytometry. Cell proliferation was determined by CFSE dilution. (B) WT and D-KO LN DCs were treated with Pam₃CSK₄ (1 μg/mL). Total RNA was isolated at the indicated time points and subjected to DNA microarray analysis. Relative RNA amounts of the genes indicated on the left are presented in color-coded arbitrary units. Arrows on the right indicate genes showing lower expression in D-KO relative to WT. Genes with maximum expression in WT cells at 2 h and 4 h posttreatment are arranged in upper (Tnf to Edil3) and lower (Il1a to Serpinb9b) rows, respectively. (C and D) WT and d-KO LN DCs were treated with Pam₃CSK₄ (1 μg/mL) and CD40 ligand (CD40L; 1 μg/mL). Total RNA was isolated at the indicated time points and analyzed by quantitative PCR (C). Culture supernatants were collected 24 h after stimulation, and CCL17 amounts were determined by ELISA (D).

Having identified the suppressive effect of DC-specific p38α ablation on T-cell–mediated allergic skin disease, we turned our investigation to the T cell itself and generated T-cell–specific p38α-KO mice. These mice, T-KO, showed highly efficient p38α ablation in T cells from the lymphoid tissues (Fig. 4A). Naïve adult T-KO mice presented no signs of spontaneous disease, but examination of their lymphoid tissues revealed moderate atrophy with the size and cellularity of the thymus, LNs, and spleen decreased to varying extents ranging between 60% and 80% of those of WT organs (Fig. S6A). The CD4⁺CD8⁺, CD4⁺CD8⁻, and CD4⁻CD8⁺ T-cell subsets developed normally in the thymus of T-KO mice (Fig. S6B). In contrast, the fractions of CD8⁺ T cells, but not CD4⁺ T cells, among total LN cells and splenocytes were markedly reduced in T-KO mice (Fig. S6C). Consequently, T-KO mice had significantly lower absolute numbers of T cells, particularly CD8⁺ T cells, in the peripheral lymphoid tissues (Fig. 4B).

Fig. 4.

T-cell-specific p38α ablation results in smaller lymphoid tissues and decreased CD8⁺ T-cell pools. (A) Whole-cell lysates were prepared from CD3⁺ cells in the thymus (Th), LNs (LN), and spleen (Sp), and from liver (Lv) tissues of WT and T-KO mice, and analyzed by immunoblotting with antibodies against the proteins indicated on the left. (B) The numbers of cells expressing the indicated markers in the LNs (Left) and spleen (Right) were determined by flow cytometry. (C and D) WT and T-KO mice (n = 5) were sensitized and challenged with DNFB, and hapten-specific ear swelling determined as in Fig. 1. **P < 0.01; *P < 0.05. (E) Recipient (R) mice received LN T cells from donor (D) mice and were challenged with hapten and analyzed as in Fig. 1. (F) CD69⁺ cells among the indicated T-cell subpopulations in the LNs of WT and D-KO mice were analyzed as in Fig. 2. (G) LN CD8⁺ T cells were prepared from hapten-sensitized WT or D-KO mice, restimulated in vitro, and amounts of IFN-γ and IL-13 in the culture supernatant were determined by ELISA (n = 3) as in Fig. 2.

CH reactions in T-KO mice were weaker than in WT mice (Fig. 4 C and D). This result could either indicate a role for T-cell p38α in the immune response to hapten or simply reflect the decreased pool of available CD8⁺ T cells in T-KO mice. When equal numbers of T cells from hapten-sensitized WT and T-KO mice were transferred to naïve WT mice, the two recipient groups developed hapten-induced skin reactions at similar strengths (Fig. 4E), suggesting that p38α-KO T cells were primed in the donor and gained full capacity to induce skin disease. Indeed, CD8⁺ T cells from sensitized WT and T-KO mice contained comparable fractions of activated CD69⁺ cells (Fig. 4F) and produced similar amounts of cytokines upon hapten restimulation (Fig. 4G). T-KO mice on the recipient side developed skin reactions similar in severity to those in WT recipients after transfer of WT donor T cells (Fig. 4E). Therefore, T-KO mice appeared to retain the ability to generate hapten-primed T cells, and p38α-KO T cells remained fully functional throughout the course of CH. The main reason for the impaired immune response in T-KO is likely that they had fewer T cells before and after hapten exposure (Fig. S7), a condition that may be difficult to achieve by a short-term pharmacological approach.

Our findings indicate cell-selective targeting as a solution for bypassing the toxicity while preserving the efficacy of protein kinase inhibitors. In our specific model of allergic skin disease, DC-intrinsic p38α function was specifically linked to the immunopathogenesis. Cell type-specific delivery of p38 inhibitors or targeting of p38α-dependent disease mechanisms uniquely operating in the specific cell type will help improve the clinical outcome, an idea widely applicable to other therapeutics with a low therapeutic index.

Materials and Methods

Animals.

M-KO (Mapk14^fl/fl-LysMCre) and K-KO (Mapk14^fl/fl-K14Cre) mice were described (11). D-KO and T-KO mice were generated by crossing Mapk14^fl/fl mice (23) with CD11cCre (24) and LckCre (25) mice, respectively. These two Cre mice, and OT-I and OT-II T-cell receptor transgenic mice, were obtained from the Jackson Laboratory. All animals were on a C57BL/6J background. All animal studies were conducted under Institutional Animal Care and Use Committee-approved protocols.

Cell Isolation and Culture.

Primary keratinocytes and macrophages were prepared and cultured as described (11). DCs were isolated from cervical, axillary, and inguinal LNs by collagenase digestion and enriched by positive selection using CD11c-specific magnetic microbeads (Miltenyi Biotec). CD3⁺ T cells were isolated from the thymus, LNs, and spleen by negative selection using the Pan T Cell Isolation Kit II (Miltenyi Biotec). CD4⁺ and CD8⁺ T cells were isolated from the LNs and spleen similarly by using biotinylated anti-CD4 (GK 1.5) and anti-CD8 (53-6.70) antibodies (eBioscience). DCs were stimulated with 1 μg/mL Pam₃CSK₄ (InvivoGen) and 1 μg/mL CD40L (R&D Systems).

Allergic Skin Disease Model.

CH to DNFB (Sigma-Aldrich) was induced as described (26, 27). For hapten sensitization, 0.5% DNFB in 20 μL of acetone:olive oil (4:1) was applied to the shaved abdominal skin on day 0 and 1. For hapten challenge, 0.35% DNFB in 20 μL of acetone:olive oil (4:1) was applied to the left auricle and 20 μL of acetone:olive oil (4:1) was applied onto the right auricle on day 5. Hapten-specific skin swelling was measured by subtracting the increase in right ear thickness from that in left ear thickness. For adoptive T-cell transfer, CD3⁺ T cells were isolated from the draining LNs of donor mice 4 d after DNFB sensitization. Donor T cells (3 × 10⁷) were injected i.v. to recipient mice. Recipient mice were challenge with DNFB 1 d after T-cell transfer. Mouse ear skin samples were embedded in paraffin, and sections were stained with hematoxylin and eosin.

T-Cell Activation in Vitro.

Hapten-induced T-cell activation was performed as described (27). LN DCs were incubated with 20 mM dinitrobenzene sulfonate (Sigma-Aldrich), a water-soluble analog of DNFB, in serum-free medium for 30 min, and mixed with 3 × 10⁵ LN CD8⁺ T cells isolated from mice 4 d after DNFB sensitization. Culture supernatants were analyzed after 72 h of coculture. For ovalbumin peptide-induced T-cell activation, LN DCs were treated with Pam₃CSK₄ (1 μg/mL) for 24 h and then incubated with ovalbumin peptides, OVA_257–264 and OVA_323–339 (GeneScript), for 4 h. Splenic CD8⁺ and CD4⁺ T cells were isolated from OT-I and OT-II mice, respectively, and incubated with carboxyfluorescein diacetate succinimidyl ester (CFSE; Invitrogen). CFSE-labeled CD8⁺ (0.5 × 10⁶) and CD4⁺ (1 × 10⁶) T cells were mixed with OVA peptide-pulsed DCs at the ratio of 1:3 (DC:T cell). Proliferation of CD8⁺ and CD4⁺ T cells was analyzed by flow cytometry after 48 h and 72 h of coculture, respectively.

Flow Cytometry.

Fluorescent-conjugated antibodies against markers were used as follows: CD3 (2C11), CD4 (RM4-5), CD8 (53-6.7), CD11c (N418), CD45R (RA3-6B2), CD49b (DX5), CD69 (H1.2F3), F4/80 (BM8), and PDCA-1 (eBio927; all from eBioscience). Stained cells were analyzed by flow cytometry using FACSCanto (BD) and the FlowJo software (Tree Star).

Protein and RNA Analysis.

Whole-cell lysates were prepared and analyzed by immunoblot as described (28). Antibodies against the following proteins were used in immunoblotting: p38α (sc-535; Santa Cruz Biotechnology), ERK (9102; Cell Signaling Technology), JNK (554285; BD Pharmingen), phosphorylated MK2 (3007; Cell Signaling Technology), phosphorylated MSK1 (04-384; Millipore), and actin (A4700; Sigma-Aldrich). The following proteins in culture supernatants were measured by ELISA: CCL17 (R&D Systems); IFN-γ, and IL-13 (eBioscience). Total RNA was extracted by using TRIzol (Invitrogen). DNA microarray analysis was performed with GeneChip Mouse Genome 430 2.0 Array (Affymetrix) at the Partners HealthCare Cetner for Personalized Genetic Medicine. RNA analysis by quantitative PCR was performed as described (28) using Ccl17 primers (foward, 5′-CAGGAAGTTGGTGAGCTGGT-3′; reverse, 5′-CATCCCTGGAACACTCCACT-3′).

Statistical Analysis.

Data values are expressed as mean ± SD unless indicated otherwise. P values were obtained with the unpaired, two-tailed Student t test.