The developmental transcription factor p63 is redeployed to drive allergic skin inflammation through phosphorylation by p38α

Abstract

Keratinocytes, the epithelial cells of the skin, reprogram their gene expression and produce immune effector molecules when exposed to environmental and endogenous triggers of inflammation. It remains unclear how keratinocytes process physiologic signals generated during skin irritation and switch from a homeostatic to an inflammatory state. Here, we show that the stress-activated protein kinase p38α is crucial for keratinocytes to prompt changes in their transcriptome upon cytokine stimulation and drive inflammation in allergen-exposed skin. p38α serves this function by phosphorylating p63, a transcription factor essential for the lineage identity and stemness of the skin epithelium. Phosphorylation by p38α alters the activity of p63 and redeploys this developmental transcription factor to a gene expression program linked to inflammation. Genetic ablation and pharmacologic inhibition of p38α or the p38α-p63 target gene product MMP13 attenuate atopic dermatitis-like disease in mice. Our study reveals an epithelial molecular pathway promoting skin inflammation and actionable through treatment with topical small-molecule therapeutics.

Introduction

The epithelium of the skin, or epidermis, not only serves as a physical barrier against environmental threats but also generates instructive signals for stromal and hematopoietic-derived cells with roles in immune homeostasis and inflammation (1–3). Unlike single-layered and pseudostratified epithelia, such as those forming the mucosal barrier in the intestine and airways, the epidermis consists of multiple layers of keratinocytes stacked in a gradient of differentiation. The innermost, basal keratinocyte layer harbors stem-like cells that self-renew and give rise to differentiated progenies in the suprabasal layers during epidermal turnover and post-injury regeneration (4). Constantly exposed to injurious stress and immune insults from the environment, the epidermis is encumbered with the need to achieve a balance between immunity and tolerance as well as tailor its immune response to the context of skin tissue. These requirements could be fulfilled by coupling the molecular pathways governing immune function with those directing epidermal lineage commitment and maintenance, yet it is unclear whether such developmental-immune interconnection is indeed a built-in feature of the epidermal molecular circuitry and enables keratinocytes to mount a context-appropriate immune response.

The protein kinase p38α (also known as MAPK14) is activated by environmental and endogenous physiologic stimuli associated with tissue damage and infection. The signals relayed by p38α prompt the adaptive transition of cell state and function for protection and resilience. When excessive or uncontrolled, however, p38α-driven cell responses become maladaptive, debilitating, and self-destructive. p38α was discovered in studies that tracked down proteins triggering a protein phosphorylation cascade in response to physicochemical stress, lipopolysaccharide, and interleukin (IL)-1 (5–7) or being targeted by a group of small-molecule anti-inflammatory compounds (8). Additional studies have since placed the p38α signaling module downstream of a growing number of receptors for microbial products, cytokines, and antigens, substantiating its extensive role in immunity and inflammation (9). As with epithelial cells of other barrier tissues, keratinocytes reside in a milieu inherently teeming with p38α-activating stimuli. Ultraviolet radiation-B (UVB) is a potent activator of p38α; loss of p38α signaling in keratinocytes has been shown to prevent skin irritation and injury in UVB-exposed mice (10, 11). Little is known, however, about the contribution of epithelial p38α signaling to the pathophysiology of skin disease resulting from chronic inflammation and immune dysregulation.

The immune effector mechanisms brought into action during allergic inflammation have evolved to protect against ectoparasites and environmental toxins and mainly operate in the skin and other barrier tissues, which are in dire need of such defense and where such defense can be most effective (12). When misdirected towards harmless environmental substances and unrestrained, however, allergies cause detrimental conditions of varying severity and scales, ranging from temporary annoyances to chronic maladies to fatal anaphylactic attacks. Atopic dermatitis and other allergic skin diseases are major health threats worldwide with their enormous and escalating socioeconomic burdens. Atopic dermatitis is initiated upon immune sensitization to offending allergens and manifests as full-blown disease during the mobilization of allergen-primed lymphocytes and other immune effectors in response to epicutaneous allergen re-encounter (13). Epidermal barrier disruption, skin dysbiosis, and other compounding events can further this process and aggravate the disease (13, 14). Crucial to developing broadly effective and safe therapies for allergic skin diseases are an advanced understanding of their pathogenesis and the discovery of actionable molecular targets tied to their core pathogenic mechanisms.

Here we discover that p63, a transcription factor essential for epidermal development and maintenance, is redeployed to drive skin inflammation. p38α signaling mediates this redeployment and shifts keratinocyte gene expression from a homeostatic to an inflammatory program. We assess the role of this molecular pathway in mouse models of skin disease and find that p38α signaling in keratinocytes drives atopic dermatitis-like disease through the induction of p63 target genes linked to allergic inflammation.

Materials and Methods

Primary cells and cell lines

Primary human keratinocytes (American Type Culture Collection) in a proliferative state were cultured in keratinocyte-SFM medium (Thermo Fisher Scientific). To prepare differentiated keratinocytes, proliferative keratinocyte populations were cultured in the presence of calcium chloride (1.2 mM) for five days. HaCaT cells (American Type Culture Collection) were cultured in Dulbecco’s modified Eagle’s medium containing supplemented with 10% fetal bovine serum and penicillin-streptomycin (50 unit/ml and 50 μg/ml, respectively; all from Thermo Fisher Scientific). The following cytokines were used in cell stimulation: IL-1α (200–01A), IL-1β (200–01B), IL-4 (200–04), IL-13 (200–13), IL-17A (200–17), IL-22 (200–22), IL-24 (200–35), IL-33 (200–33), IFN-β (300–02BC), and IFN-γ (300–02; all from PeproTech).

Mice

All mice used in this study were in a C57BL/6J background. Keratinocyte-specific p38α-knockout (KO) mice were generated by crossing mice harboring floxed p38α alleles (Mapk14^tm1.2Otsu; 15) with mice carrying a Cre recombinase transgene expressed under the control of the human KRT14 promoter (Tg[KRT14-cre]8Brn; 16). MMP13-KO (Mmp13^−/−) mice (Mmp13^tm1Smk; 17) had a systemic deficiency of MMP13. Keratinocyte-specific MMP13-KO mice were generated by crossing mice with floxed MMP13 alleles (Mmp13^tm1Werb; 18) with mice carrying another human KRT14-Cre recombinase transgene (Tg[KRT14-cre]1Amc/J; 19). These mice did not display any gross physical or behavioral abnormalities when maintained in specific pathogen-free conditions. Mice with p38α deficiency restricted to myeloid cells, dendritic cells, and T cells were generated in previous studies (10, 20, 21). All animal experiments were conducted under an Institutional Animal Care and Use Committee–approved protocol.

Small-molecule compounds

SCIO469 (Sigma) was dissolved in water. DB04760 (CAS544678-85-5; EMD Millipore) was dissolved in 50% dimethyl sulfoxide.

OVA-induced allergic dermatitis

For allergen sensitization, ovalbumin (OVA; Sigma) and alum adjuvant (Imject; Thermo Fisher Scientific) were combined in saline (0.9% sodium chloride; SAL) to constitute an inoculum (1 mg/ml and 100 mg/ml, respectively) and injected intraperitoneally into mice (200 μl per animal) on day 0 and 7. To elicit dermatitis in OVA-sensitized mice, shaved and depilated dorsal skin was subjected to 15 strokes of tape stripping (Tegaderm; 3M) and then challenged with OVA on day 14, 16 and 19. For this epicutaneous challenge, 100 μl of OVA (1 mg/ml in SAL) was applied to cotton gauze (1 cm × 1 cm) placed on tape-stripped skin; 100 μl of SAL was applied as a control. For topical administration of p38α and MMP13 inhibitors, 125 μl of inhibitor solution (50 mM) was applied to skin prior to OVA challenge; OVA was applied to the inhibitor-administered skin site when it was no longer visibly moist. The relatively high dose was chosen to ensure keratinocyte exposure to topically delivered compounds. Challenged skin was sampled on day 21 for analysis.

Imiquimod-induced psoriasiform dermatitis

Shaved and depilated dorsal skin and pinnae of mice were treated with imiquimod cream (40 mg for dorsal skin and 10 mg for pinnae per animal) daily for four consecutive days. Inflamed skin was sampled 24 h later for analysis. Erythema and desquamation scores were determined daily by a combined assessment of lesion severity and affected area. Pinna thickness was measured daily using a caliper (Mitutoyo).

TPA-induced irritant contact dermatitis

Shaved and depilated dorsal skin of mice was treated with 200 μl of 12-O-tetradecanoylphorbol-13 acetate (TPA; 50 μg/ml in acetone) daily for two consecutive days. Inflamed skin was sampled 48 h later for analysis.

Histology and immunofluorescence

Mouse skin tissue was fixed with formalin and embedded in paraffin. Tissue sections were analyzed by hematoxylin and eosin (H&E) staining or by immunofluorescence analysis using primary antibodies in conjunction with Alexa Fluor 594- or fluorescein isothiocyanate-conjugated secondary antibodies (Thermo Fisher Scientific) and Hoechst 33342 (Sigma). Antibodies specific to the following proteins were used in immunofluorescence straining after 1:100 to 1:200 dilution: CD3 (SP7) and MMP13 (ab39012; both from Abcam); and MBP (2000–124; Mayo Clinic). Epidermal areas in immunofluorescence images were defined by detecting their boundaries with Fiji software in high color saturation conditions.

CRISPR genome editing

Single guide RNA (sgRNA) target sites within or near MAPK14 exon-2 (#1-#3; Supplemental Fig. 1A) were selected using the CRISPOR tool (22). Double-stranded oligonucleotides containing sgRNA target sequences were inserted into pX458-DsRed2 (sgRNA #1 and #2) and pX458-ECFP (sgRNA #3) to generate three CRISPR/Cas9 plasmid constructs for genome editing (23). HaCaT cells were transfected with a mixture of all three plasmids by electroporation and flow sorted into DsRed2⁺ECFP⁺ single cells for cloning (24). HaCaT clones with deletions in MAPK14 were identified by PCR using primers specific to a genomic region predicted to harbor deletions (Supplemental Fig. 1A).

siRNA transfection

Human p63 small interfering RNA (siRNA) was selected from the pre-designed Silencer Select collection (#1-#3) or custom designed (#4) and formed as duplexes of the sense (s) and antisense (as) oligonucleotides: #1s, 5’-gaaccgccguccaauuuuatt-3’; #1as, 5’-uaaaauuggacggcgguucat-3’; #2s, 5’-ugaugaacuguuauacuuatt-3’; #2as, 5’-uaaguauaacaguucaucatc-3’; #3s, 5’-ggaugaagauagcaucagatt-3’; #3as, 5’-ucugaugcuaucuucauccgc-3’; #4s, 5’-gaguuucaguuuguuggaatt-3’; and #4as, 5’-uuccaacaaacugaaacuctt-3’ (all from Thermo Fisher Scientific). Control siRNA was a negative control duplex with medium GC content (Thermo Fisher Scientific). siRNA was transfected with siRNA using the Lipofectamine RNAiMAX Reagent (Thermo Fisher Scientific) as described (25). Cells were used for subsequent analysis 48 h after transfection.

Lentiviral transduction

Lentiviral vectors expressing ΔNp63α and its variants (26) were constructed using the backbone transfer plasmid pCDH-CMV-MCS-EF1-copGFP (System Biosciences). Lentiviral particles were generated by transfecting 293T cells with a transfer vector in conjunction with pRSV-Rev, pMDLg/pRRE, and pMD2.G (all from Addgene) using the Lipofectamine LTX with Plus Reagent (Thermo Fisher Scientific). Cells were infected with lentiviruses in the presence of polybrene (5 μg/ml) and used for subsequent analysis 72 h after infection.

RNA analysis

For transcriptome analysis, total RNA was isolated using the Trizol Reagent (Thermo Fisher Scientific), further purified using the RNeasy Mini Kit (Qiagen), and subjected to sequencing library construction (TruSeq Stranded mRNA Prep, Illumina) and 75-cycle single-end sequencing (NextSeq 550, Illumina) at the Cutaneous Biology Research Center of MGH. Sequencing reads were mapped to a Homo sapiens reference genome (assembly hg19) using the STAR genome alignment algorithm. Normalized transcript reads (reads per kilobase per million mapped reads) were calculated using HOMER software. All RNA sequencing (RNA-seq) data are available in the NCBI GEO database (GSE180870). For qPCR analysis, total RNA isolated using the Trizol Reagent were subjected to cDNA synthesis using the SuperScript IV VILO Master Mix (Thermo Fisher Scientific) and PCR using the SYBR Green PCR Master Mix (Applied Biosystems) and gene-specific primers. Heat maps of gene transcript abundances were generated using Morpheus software (Broad Institute).

Protein analysis

Proteins in whole-cell lysates were prepared and analyzed by immunoblotting as described (27). Antibodies specific to the following proteins were used in immunoblotting after 1:1000 dilution: p-ERK (9101), p-JNK (9251), p-MK2 (3007), p-MSK1 (9595), p-p38α (9211), p63 (13109), p-Ser^66/68–p63 (4981), p-STAT1 (5806), and YAP1 (4912; all from Cell Signaling Technology); IκBα (sc-371), IRF4 (sc-6059), NF-κB1 (sc-7178), NF-κB2 (sc-298), Notch1 (sc-6014), p38α (sc-535), RelA (sc-372), RelB (sc-226), STAT1 (sc-346), STAT3 (sc-482), p-STAT3 (sc-8059), and vinculin (sc-73614; all from Santa Cruz Biotechnology); MMP13 (ab39012) and p-STAT6 (ab54461; both from Abcam); p-Ser³⁰¹–p63 (PA5–39827) and p-Ser³⁶¹–p63 (PA5–38380; both from Thermo Fisher Scientific); β-catenin (610153; BD Biosciences); and IRF6 (IMG-3484; IMGENEX). Phosphorylated p63 was enriched from whole-cell lysates by immunoprecipitation with the p63-specific monoclonal antibody 4A4 (ab735; Abcam) and protein G-agarose beads (Sigma) before detection by immunoblotting. Proteins in culture medium and serum were analyzed by enzyme-linked immunosorbent assay.

Statistical analysis

Data values are expressed as means ± SEM. P values were obtained by the unpaired two-tailed Student t test with Welch’s correction or the two-way repeated measures ANOVA with the Geisser-Greenhouse correction and calculated with Prism software (GraphPad).

Results

p38α-dependent gene expression program in keratinocytes

To examine the contribution of p38α signaling to the inflammatory response of the skin epithelium, we first sought to establish a model system that would allow us to capture the transcriptomic output of p38α signaling in keratinocytes. To this end, we tested a multitude of cytokines for their ability to induce the phosphorylation (hence activation) of p38α and its downstream kinases, MSK1 and MK2, in HaCaT keratinocytes. This screen and subsequent validation in primary human keratinocytes identified IL-1α and IL-1β, which bind to and signal through the same receptor, as potent activators of the p38α signaling module (Fig. 1, A and B). Of note, IL-1-responsive p38α activation in primary human keratinocytes was markedly augmented after calcium-induced differentiation in vitro (Fig. 1B). We next generated HaCaT cells harboring a large deletion in exon-2 of the p38α gene (MAPK14) by CRISPR genome editing (Supplemental Fig. 1) and used them to determine the effects of genetic ablation of p38α. RNA-seq analysis of the transcriptome of wild-type (WT) and p38α-KO HaCaT cells (Fig. 1C) and confirmation by qPCR (Fig. 1D) revealed distinct sets of genes whose steady-state and IL-1-induced expression was dependent on p38α.

Figure 1. Loss of p38α signaling prevents the induction of specific genes during the keratinocyte response to an inflammatory stimulus.

(A and B) HaCaT cells (A) and human primary keratinocytes (B) were left unstimulated or stimulated with the indicated cytokines (20 ng/ml). Whole-cell lysates were prepared after the indicated durations of stimulation and analyzed by immunoblotting. The primary keratinocytes were proliferative or had been induced to differentiate by calcium when stimulated. The panel showing data from HaCaT cell analysis (A) consists of image strips derived from two gels processed in parallel and collated from different parts of the resultant immunoblots.

(C and D) Parental WT and CRISPR-edited p38α-KO HaCaT cells were left unstimulated or stimulated with IL-1α (20 ng/ml, 4 h). RNA prepared from the cells was analyzed by RNA-seq (C) and qPCR (D). Relative mRNA amounts for individual genes are presented on a color-coded scale shown at the bottom. #1 and #2, independent CRISPR-edited clones.

The IL-1-inducible p38α-dependent genes identified in the transcriptome analysis included those encoding CXCL chemokines specific to the CXCR2 receptor, the IL-13 receptor chain IL13RA2, the matrix metalloproteinase MMP13, and several other inflammatory mediators (Fig. 1C). CXCR2 and its ligands play a role in neutrophil recruitment in various pathophysiological settings (28). IL13RA2 was originally regarded as a decoy receptor interfering with IL-13 signaling (29) but has recently been shown to mediate IL-13-dependent immune responses in both allergic and non-allergic inflammation (30–33). Increased MMP13 expression has been detected in the skin and nasal mucosa of atopic dermatitis and allergic rhinitis patients, respectively (34, 35). Deficiency of p38α did not result in complete refractoriness to IL-1 stimulation in keratinocytes, with the induction of many genes remaining intact in p38α-KO cells (Fig. 1, C and D). This led us to postulate a rather specialized contribution of p38α signaling in vivo and explore the specific disease contexts in which the p38α-driven changes in the keratinocyte transcriptome might translate substantially to inflammation and tissue disruption.

Role of epithelial p38α signaling in allergic skin inflammation

We sought to determine the role of p38α signaling in distinct mouse models of skin disease where varying immunopathogenic mechanisms drove skin inflammation. To examine the contribution of epithelial p38α signaling and assess its impact relative to that of p38α signaling in other cell types, we established a panel of mice with p38α gene ablation restricted to keratinocytes (Mapk14^fl/fl-K14Cre; p38α-EKO), myeloid cells (Mapk14^fl/fl-Lyz2Cre; p38α-MKO), dendritic cells (Mapk14^fl/fl-ItgaxCre; p38α-DKO), or T cells (Mapk14^fl/fl-LckCre; p38α-TKO). Discovery cohorts of these mice were provoked to develop skin disease, and their degree of dermatitis was evaluated based on macroscopic and histological features of inflammation (erythema, desquamation, acanthosis, and edema formation). In this in vivo screen, p38α-EKO mice, but not other cell type-specific p38α-KO mice, exhibited markedly attenuated inflammation compared to WT mice when subjected to a protocol inducing atopic dermatitis-like disease through sensitization to the model allergen OVA and subsequent epicutaneous OVA exposure (Supplemental Fig. 2, A–C). The degree of psoriasiform dermatitis (Supplemental Fig. 2, D–F) and irritant contact dermatitis (Supplemental Fig. 2, G–I), induced by epicutaneous exposure to imiquimod and TPA, respectively, was comparable without significant differences across all groups of mice tested.

The mouse atopic dermatitis model used for the phenotypic screen involved intraperitoneal OVA and alum injection for sensitization and OVA administration to tape-stripped skin for the elicitation of dermatitis (36). Sensitization to and skin challenge with OVA resulted in epidermal thickening and dermal enlargement, indicative of acanthosis and edema formation, respectively, in WT mice (Fig. 2, A–C). Mock challenge of tape-stripped skin with SAL did not bring about these changes in OVA-sensitized mice (Fig. 2, B and C). The lesional skin of OVA-challenged WT mice also displayed infiltration of T cells and eosinophils, detected by CD3 and major basic protein (MBP) immunostaining, respectively (Fig. 2, D and F). All of these changes associated with inflammation were absent or blunted in p38α-EKO mice challenged with OVA (Fig. 2, A–G). We examined whether reduced inflammation in p38α-EKO mice was a result of inadequate immune sensitization to OVA and a consequent unavailability of molecular and cellular effectors of OVA-specific immunity. Total immunoglobulin E (IgE) concentrations and OVA-specific IgE titers in serum were comparable in WT and p38α-EKO mice (Fig. 2, H and I). Allergen-specific T cells primed in vivo can be induced to produce effector cytokines ex vivo by exposing them to the allergen in the presence of antigen-presenting cells. For such restimulation of OVA-primed T cells in WT and p38α-EKO mice, their skin-draining lymph node cells were incubated with OVA in culture medium; lymph node cells from OVA-challenged WT and p38α-EKO mice produced similar amounts of IL-4, IL-13, and IL-5 upon ex vivo restimulation (Fig. 2, J–L), indicating that OVA-specific T cell priming occurred in both groups of mice and led to the generation of T cell pools of comparable size that had acquired the potential to produce effector cytokines in response to allergen re-encounter. In our experimental setting, T cells in the skin-draining lymph nodes were primed to become potent producers of IL-13 and IL-5 but not IL-4 (Fig. 2, J–L), suggesting their acquisition of a specialized effector phenotype. Taken together, these findings showed that p38α-EKO mice possessed intact allergen-specific immune effector functions but failed to harness them to drive skin inflammation.

Figure 2. Epithelial-restricted p38α gene ablation suppresses allergic skin inflammation in mice.

(A-L) WT and p38α-EKO mice were sensitized to the experimental protein allergen ovalbumin (OVA; 200 μg) by intraperitoneal injection in conjunction with alum (20 mg) on day 0 and 7. Shaved and depilated back skin of the sensitized mice was tape-stripped and challenged with control saline (SAL) or OVA (100 μg) on day 14, 16 and 19 (n = 2 [SAL] and 7 [OVA] per group). Skin tissue, serum, and lymph node cells were collected on day 21 for analysis. *, P < 0.05; **, P < 0.01 (two-tailed unpaired Student’s t test). Scale bar, 100 μm.

(A-C) Skin sections were analyzed by H&E staining (A). Epidermal and dermal thicknesses (B and C, respectively) were determined from H&E-stained skin images and are shown as means ± SEM.

(D-G) Skin sections were analyzed by immunostaining for the indicated protein markers and DNA counterstaining (D and F). Solid and dotted line, epidermal margins and the epidermal-dermal boundary, respectively. CD3⁺ T cell and MBP⁺ eosinophil densities (E and G, respectively) were determined based on fluorescence signal-bearing cell numbers per image field and are shown as means ± SEM.

(H and I) Serum was analyzed by ELISA. Total and OVA-specific IgE concentrations in serum (H and I, respectively) are shown as means ± SEM.

(J-L) The culture supernatants of lymph node cells incubated with SAL or OVA for 24 h were analyzed by ELISA. Cytokine concentrations in culture supernatants are shown as means ± SEM.

p38α-EKO also displayed reduced severity of atopic dermatitis-like disease induced by other experimental protocols. They developed milder dermatitis than WT mice when sensitized to and challenged with OVA entirely through epicutaneous exposure and also when subjected to topical treatment with MC903 (data not shown), a vitamin D3 analog eliciting atopic dermatitis-characteristic cytokine production in the skin (37). Taken together, our findings from the investigation of p38α-EKO mice highlight the key contribution of epithelial p38α signaling to allergic skin inflammation.

MMP13 as a mediator of p38α-driven allergic skin inflammation

Our analysis of the HaCaT keratinocyte transcriptome showed that cytokine-responsive MMP13 expression was dependent on p38α signaling (Fig. 1C). The mRNA amounts of MMP13 and other p38α-dependent genes identified in this analysis were sharply elevated in primary human keratinocytes stimulated with IL-1α and IL-1β (Fig. 3A). MMP13 and another p38α-dependent gene, IL13RA2, but not the others examined, were also induced in IL-4-stimulated keratinocytes (Fig. 3A). Stimulation with IL-1α and IL-4 led to massive production and accumulation of MMP13 protein in HaCaT keratinocytes (Fig. 3B). Given the reported association of MMP13 expression with allergic inflammatory conditions in barrier tissues (21, 22), we chose to explore whether MMP13 acted downstream of p38α signaling in atopic dermatitis-like skin disease and promoted allergic skin inflammation.

Figure 3. Systemic deficiency in MMP13 suppresses allergic skin inflammation in mice.

(A) Primary human keratinocytes were induced to differentiate by calcium and left unstimulated or stimulated with the indicated cytokines (20 ng/ml, 4 h). RNA prepared from the keratinocytes was analyzed by qPCR. Relative mRNA amounts for individual genes are presented on a color-coded scale shown at the bottom.

(B) HaCaT cells were left unstimulated or stimulated with the indicated cytokines (20 ng/ml). Whole-cell lysates were prepared after the indicated durations of stimulation and analyzed by immunoblotting.

(C-N) WT and MMP13-KO mice were sensitized to ovalbumin (OVA) and challenged with control saline (SAL) or OVA as in Fig. 2 (n = 1–2 [SAL] and 6–11 [OVA] per group). Skin tissue, serum, and lymph node cells were collected for analysis as in Fig. 2. *, P < 0.05; ****, P < 0.0001 (two-tailed unpaired Student’s t test). Scale bar, 100 μm.

(C-E) Skin sections were analyzed by H&E staining (C). Epidermal and dermal thicknesses (D and E, respectively) were determined from H&E-stained skin images and are shown as means ± SEM.

(F-I) Skin sections were analyzed by immunostaining for the indicated protein markers and DNA counterstaining (F and H). Solid and dotted line, epidermal margins and the epidermal-dermal boundary, respectively. CD3⁺ T cell and MBP⁺ eosinophil densities (G and I, respectively) were determined based on fluorescence signal-bearing cell numbers per image field and are shown as means ± SEM.

(J and K) Serum was analyzed by ELISA. Total and OVA-specific IgE concentrations in serum (J and K, respectively) are shown as means ± SEM.

(L-N) The culture supernatants of lymph node cells incubated with SAL or OVA for 24 h were analyzed by ELISA. Cytokine concentrations in culture supernatants are shown as means ± SEM.

We examined how MMP13 deficiency affected allergic skin inflammation in mice. Mice with systemic deficiency of MMP13 exhibited a significant reduction or a trend toward reduction in acanthosis, edema formation, and the infiltration of T cells and eosinophils in skin challenged with OVA after sensitization through the intraperitoneal route. (Fig. 3, C–I). This reduction of skin inflammation occurred without a significant impairment in total and OVA-specific IgE production (Fig. 3, J and K) or the formation of OVA-primed T cells that could produce effector cytokines upon ex vivo restimulation (Fig. 3, L–N). These observations indicated that MMP13, while dispensable for the generation of allergen-specific immune effectors, might play a crucial role in allergen-driven inflammation during the phase of active skin disease. MMP13-KO mice also displayed less severe dermatitis than WT counterparts when exposed to OVA only through the epicutaneous route (data not shown).

Immunofluorescence analysis revealed that the expression of MMP13 in OVA-challenged inflamed skin was detected in the epidermis, mainly in the suprabasal layers (Fig. 4A, upper panel). Suprabasal epidermal MMP13 expression was also observed in mouse skin irritated with TPA (Fig. 4A, lower panel). The basal epidermis contains proliferative keratinocytes, whereas the suprabasal epidermal layers comprise differentiated keratinocytes; keratinocytes in these distinct epidermal compartments might differ in their responsiveness to inflammatory stimuli and express MMP13 and other p38α-dependent genes with dissimilar magnitudes during the inflammatory response. Indeed, primary human keratinocytes stimulated with IL-1α or IL-1β while in a proliferative state showed a modest induction of p38α-dependent genes, but the induction of most of these genes, except CXCL2, was substantially stronger in differentiated keratinocytes (Fig. 4B).

Figure 4. Epithelial-restricted MMP13 gene ablation suppresses allergic skin inflammation in mice.

(A) Skin sections from mice with ovalbumin (OVA)-induced allergic dermatitis (upper panel) and TPA-induced acute dermatitis (lower panel) were analyzed by MMP13 immunostaining and DNA counterstaining. Dotted line, the epidermal-dermal boundary. Scale bar, 50 μm.

(B) Primary human keratinocytes were left unstimulated (None) or stimulated with the indicated cytokines (20 ng/ml, 4 h). RNA prepared from the keratinocytes was analyzed by qPCR. Relative mRNA amounts for individual genes are presented on a color-coded scale shown at the bottom. The keratinocytes were proliferative or had been induced to differentiate by calcium when stimulated.

(C-I) WT and MMP13-EKO mice were sensitized to OVA and challenged with control saline (SAL) or OVA as in Fig. 2 (n = 5–6 [SAL] and 8 [OVA] per group). Skin tissue was collected for analysis as in Fig. 2. *, P < 0.05 (two-tailed unpaired Student’s t test). Scale bar, 100 μm.

(C-E) Skin sections were analyzed by H&E staining (C). Epidermal and dermal thicknesses (D and E, respectively) were determined from H&E-stained skin images and are shown as means ± SEM.

(F-I) Skin sections were analyzed by immunostaining for the indicated protein markers and DNA counterstaining (F and H). Solid and dotted line, epidermal margins and the epidermal-dermal boundary, respectively. CD3⁺ T cell and MBP⁺ eosinophil densities (G and I, respectively) were determined based on fluorescence signal-bearing cell numbers per image field and are shown as means ± SEM.

To determine whether loss of epithelial MMP13 function was sufficient to suppress allergic skin inflammation, we generated mice with keratinocyte-restricted MMP13 gene ablation (Mmp13^fl/fl-K14Cre; MMP13-EKO; Supplemental Fig. 3). Similar to mice with systemic MMP13 deficiency, MMP13-EKO mice displayed attenuated dermatitis when subjected to OVA sensitization and challenge (Fig. 4, C–I), suggesting that MMP13 expressed in the epidermis served as a major driver of allergic skin inflammation.

MMP13 induction in keratinocytes driven by p38α-mediated p63 phosphorylation

Activated p38α modulates the abundance, subcellular location, and functional activity of substrate proteins by direct phosphorylation or indirectly through intermediary kinases such as MK2/3 and MSK1/2 (9). Transcription factors are a principal class of proteins targeted for modulation by p38α signaling. p38α has been shown to phosphorylate transcription factors of particular interest in keratinocyte biology, including p63 (26). p63 plays a pivotal role in epidermal lineage identity and stemness. Epidermal development is completely absent in mice with p63 gene deficiency (38–41). ΔNp63α, a p63 isoform highly expressed in keratinocytes, governs a gene expression program essential for the proliferation and differentiation of stem-like epidermal keratinocytes (40–42). Intriguingly, transgenic mice overexpressing ΔNp63α in the epidermis were found to develop inflammatory skin disease (43, 44); it remains unclear, however, whether this pathology represented an intrinsic ability of ΔNp63α to activate an inflammatory program in keratinocytes or was secondary to skin fragility or a perturbation of epidermal barrier formation caused by dysregulation in developmental gene expression.

We investigated p63 as a possible mediator of p38α-dependent changes in keratinocyte gene expression, in particular MMP13 gene induction during the inflammatory response. Integration and re-analysis of two data sets generated from chromatin immunoprecipitation sequencing (ChIP-seq) of primary human keratinocytes (GSE59827 and GSE139685; 45, 46) revealed p63 occupancy at the enhancers and promoters of many p38α-dependent genes, including MMP13, even prior to immune stimulation (Fig. 5A, and Supplemental Fig. 4A). The p63-occupied genomic regions in steady-state keratinocytes were subdivided into distinct clusters depending on the surrounding chromatin context (Supplemental Fig. 4A). The p63-bound DNA sites within or near genes serving immune-related functions (Supplemental Fig. 4, B and C), which included many p38α-dependent genes, formed a cluster that exhibited weak histone H3 methylation and acetylation at Lys⁴ and Lys²⁷, respectively, and low RNA polymerase II densities (Supplemental Fig. 4A), together indicative of a transcriptionally inactive state. By contrast, the p63-bound enhancers and promoters of genes serving developmental functions (Supplemental Fig. 4B) formed separate clusters associated with a transcriptionally active chromatin landscape (Supplemental Fig. 4A). In unprovoked steady-state keratinocytes, p63 conceivably functions to maintain p38α-dependent genes in an inactive or poised state until signaling events elicited by immune stimuli change this configuration to turn on their transcription.

Figure 5. Knockdown of TP63 or abrogation of p63 phosphorylation by p38α prevents MMP13 induction during the keratinocyte response to an inflammatory stimulus.

(A) Normalized ChIP-seq reads for DNA crosslinked to p63 and H3K27Ac in human primary keratinocytes (retrieved from the data series GSE59827 and GSE29611, respectively) are visualized as custom tracks on the UCSC genome browser. The ChIP-seq data were from proliferative keratinocytes (p63-Day 0 and H3K27Ac) or keratinocytes cultured four days with calcium in the medium (p63-Day 4).

(B and C) HaCaT cells were transfected with negative control (NC) siRNA or siRNA specific to sequences within the p63-coding region (#1-#3) for TP63 knockdown (KD). Whole-cell lysates prepared from the cells were analyzed by immunoblotting (B). RNA was prepared from the cells left unstimulated or stimulated with IL-1α (20 ng/ml, 4 h) and analyzed by qPCR (C). Relative mRNA amounts for individual genes are presented on a color-coded scale shown at the bottom.

(D) Parental WT and CRISPR-edited p38α-KO HaCaT cells were left unstimulated or stimulated with IL-1α (20 ng/ml). Whole-cell lysates were prepared after the indicated durations of stimulation and analyzed by immunoblotting. p-, phosphorylated.

(E and F) HaCaT cells transfected with negative control (NC) siRNA or siRNA specific to a sequence within the 3’-untranslated region (3’-UTR) of TP63 (#4) were infected with lentiviruses expressing the indicated p63 derivatives. The cells were left unstimulated or stimulated with IL-1α (20 ng/ml, 4 h). Whole-cell lysates prepared after the indicated durations of stimulation and analyzed by immunoblotting (E). RNA prepared from the cells was analyzed by qPCR (F). Relative mRNA amounts for individual genes are presented on a color-coded scale shown at the bottom.

We examined the effect of loss of function of p63 on p38α-dependent gene induction in cytokine-stimulated keratinocytes. Given the essentiality of p63 in the identity, proliferation, and survival of keratinocytes, we noted the possibility that a complete and irreversible removal of the p63 gene might reduce their fitness or cause a perturbation in their phenotype, hampering subsequent analysis of specific p63 functions. To bypass this potential problem, we employed an siRNA-mediated gene approach for temporary loss of function. We identified siRNA sequences targeting the human p63 gene TP63 that efficiently eliminated ΔNp63α in HaCaT keratinocytes without perturbing the expression of other transcription factors (Fig. 5, B and C). TP63 KD produced divergent effects on p38α-dependent gene induction by IL-1α, dampening MMP13 induction while enhancing the induction of the other p38α-dependent genes examined (Fig. 5C). This finding points to a role for p63 not only in shaping keratinocyte gene expression during development and homeostasis but also in directing its reshaping during the inflammatory response. p63 appears to function as a transcriptional activator for some p38α-dependent genes and as a repressor for others.

Multiple amino acid residues of p63 have been shown to be phosphorylated by p38α in an in vitro kinase reaction (26). Immunoblotting with phosphoepitope-specific antibodies detected ΔNp63α phosphorylation at Ser^66/68, Ser³⁰¹, and Ser³⁶¹ in IL-1α-stimulated HaCaT cells (Fig. 5D). Phosphorylation at Ser³⁰¹ and Ser³⁶¹, peaking 15 to 30 min after stimulation, was nearly abolished in p38α-KO HaCaT cells; phosphorylation at Ser^66/68, on the other hand, increased gradually and less conspicuously in WT HaCaT cells and was reduced by p38α deficiency only after 60 min (Fig. 5D).

To explore the functional importance of p63 phosphorylation at the individual Ser residues, p63 mutants harboring amino acid substitutions were introduced into HaCaT cells in which endogenous p63 expression was silenced with siRNA (Fig. 5E). For this allele replacement, the siRNA was designed to target a sequence in the 3’-untranslated region of endogenous TP63 mRNA that was absent in the lentiviral p63 gene cassette used for mutant expression. Analysis of these allele-replaced cells revealed that a Ser³⁰¹-to-Ala substitution, but not analogous substitutions at Ser^66/68 and Ser³⁶¹, attenuated IL-1α-induced MMP13 expression (Fig. 5F). The Ser³⁰¹-to-Ala substitution, however, did not affect the ability of p63 to repress the induction of the other p38α-dependent genes examined (Fig. 5F). These results indicate that p63 phosphorylation exerts phosphoacceptor site- and target gene-specific effects, and that phosphorylation at Ser³⁰¹ is a key signaling event for the induction of MMP13, a mediator of p38α-driven allergic skin inflammation.

Therapeutic effects of topical small-molecule inhibitors of p38α and MMP13

Given that both p38α and MMP13 were druggable proteins functioning in the epidermis of allergic-exposed inflamed skin, we assessed the potential of p38α and MMP13 inhibitors as topical therapeutics for allergic skin disease. We first examined the effect of a panel of small-molecule p38α inhibitors on IL-1α-induced gene expression in primary human keratinocytes. All of the six inhibitors tested potently suppressed MMP13 and IL13RA2 induction, whereas they exerted varying degrees of suppression on CXCL5 and CXCL6 induction (Fig. 6A). One of these inhibitors, SCIO469 (47), was selected for in vivo efficacy assessment, as this p38α inhibitor was found to produce pharmacological effects as a topical agent in a preclinical setting (48). We also included DB04760 (49), a small-molecule MMP13 inhibitor, in this assessment.

Figure 6. Topical administration of small-molecule inhibitors of p38α or MMP13 suppresses allergic skin inflammation in mice.

(A) Primary human keratinocytes were induced to differentiate by calcium, preincubated with the indicated small-molecule inhibitors, and left unstimulated or stimulated with IL-1α (20 ng/ml, 4 h). RNA prepared from the keratinocytes was analyzed by qPCR. Relative mRNA amounts for individual genes are presented on a color-coded scale shown at the bottom.

(B-H) C57BL/6 mice were sensitized to ovalbumin (OVA) and challenged with control saline (SAL) or OVA as in Fig. 2 (n = 5 [SAL] and 10 [OVA] per group). Vehicle (water), the p38α inhibitor SCIO469 (p38αi), and the MMP13 inhibitor DB04760 (MMP13i) were administered topically to skin (125 μl of 50 mM solution per cm²) prior to OVA challenge. Skin tissue was collected for analysis as in Fig. 2. *, P < 0.05; **, P < 0.01; ***, P < 0.001 (two-tailed unpaired Student’s t test). Scale bar, 100 μm.

(B-D) Skin sections were analyzed by H&E staining (B). Epidermal and dermal thicknesses (C and D, respectively) were determined from H&E-stained skin images and are shown as means ± SEM.

(E-H) Skin sections were analyzed by immunostaining for the indicated protein markers and DNA counterstaining (E and G). Solid and dotted line, epidermal margins and the epidermal-dermal boundary, respectively. CD3⁺ T cell and MBP⁺ eosinophil densities (F and H, respectively) were determined based on fluorescence signal-bearing cell numbers per image field and is shown as means ± SEM.

SCIO469 and DB04760 were topically applied to the skin of OVA-challenged mice after their immune sensitivity to OVA was fully established. The p38α inhibitor-treated skin displayed a significant reduction in epidermal and dermal thickening as well as T cell and eosinophil infiltration (Fig. 6, B–H). Treatment with the MMP13 inhibitor also significantly reduced OVA-induced epidermal thickening and T cell infiltration (Fig. 6, C and F) and resulted in a trend toward reduction in dermal thickening and eosinophil infiltration (Fig. 6, D and H). These results suggest that blocking p38α signaling or MMP13 action with topical small-molecule inhibitors could effectively treat allergic skin disease even when the inhibitors are administered after allergen sensitization and during active dermatitis.

Discussion

The current understanding of how barrier epithelia contribute to immune homeostasis and inflammation remains rudimentary and has not translated into effective clinical strategies for diagnosis and therapy. We have shown that p63, a master transcription factor essential for epidermal development, is redeployed to orchestrate inflammatory gene expression in keratinocytes. p38α directs this redeployment by phosphorylating p63 and thereby modulating its transcription function. We have demonstrated that this epithelial signaling event serves to promote allergic skin inflammation through the action of MMP13, the product of a gene that depends on p38α and phosphorylated p63 for induction in cytokine-stimulated keratinocytes. p38α and MMP13 have emerged from this study as actionable molecular targets for pharmacological intervention. Corticosteroids, the most common medications prescribed for atopic dermatitis, offer limited efficacy or short-lived symptomatic relief. Many patients initially responsive to corticosteroids often encounter adverse drug effects or develop treatment resistance over time (50). Biologics that block the action of cytokines driving allergic inflammation are more potent and less toxic than corticosteroids and other conventional medications but provide clinical benefit only in subsets of patients and often fail to achieve durable remission (51–53). Inhibition of epidermal p38α and MMP13 activity using topical therapeutics warrants further investigation as a new strategy for the treatment of atopic dermatitis and other allergic skin diseases.

Several p38α inhibitors have been developed and tested in clinical studies for a wide range of indications, including rheumatoid arthritis, chronic obstructive pulmonary disease, neurodegenerative diseases, facioscapulohumeral muscular dystrophy, cancer, and COVID-19 (6). Systemic administration of some p38α inhibitors has been reported to produce hepatic and gastrointestinal toxicity and other detrimental effects (54). These adverse reactions likely resulted from interference with beneficial homeostatic functions of p38α as well as off-target inhibition. For the treatment of allergic skin diseases with new or repurposed p38α inhibitors, it would be desirable to choose compounds highly selective for p38α and limit their action to the skin epithelium by using an optimal formulation and an appropriate route of administration. Another conceivable strategy is to target epithelial molecular events downstream of p38α that are mechanistically linked to allergic skin inflammation but not essential homeostatic processes. The inhibition of MMP13, which we have identified as an epithelial mediator of p38α-driven disease in allergen-exposed skin, is an example that embodies such a strategy.

MMP13 has been implicated in the pathogenesis of osteoarthritis and rheumatoid arthritis (55–57). Many MMP13 inhibitors have been developed for the treatment of these diseases (57–59). Its role and potential as a therapeutic target in allergic diseases, however, have remained unexplored. The expression and action of MMP13 in the skin epithelium make this matrix metalloproteinase a target amenable to inhibition by topically administered therapeutics, as demonstrated in this study. Hydroxamate-based MMP inhibitors (e.g. marimastat), which act by chelating the catalytic zinc ion and generally exhibit low target selectivity, have been shown to produce high rates of adverse effects in clinical settings (59). By contrast, some non-hydroxamate MMP inhibitors interact with a substrate-binding subsite, known as “S1’ pocket,” that is separated from the zinc ion-coordinating histidine residues. X-ray crystallography and nuclear magnetic resonance spectroscopy studies revealed an unusually large S1’ pocket for MMP13 compared with those of other MMPs (60, 61). The MMP13 inhibitor tested in this study, DB04760, fill the S1’ pocket of MMP13 but cannot fit those of other MMPs (49). Hence, this compound is almost exclusively selective for MMP13. Chemical modification of this compound may yield derivatives that preserve this selectivity but feature improved skin penetration. Such derivatives will likely exhibit enhanced pharmacological action for allergic skin diseases when delivered as a topical agent via the epicutaneous route.

Patients with genetic disorders caused by TP63 mutations, such as ankyloblepharon-ectodermal defects-cleft lip/palate syndrome (62), often present with inflammatory skin conditions, which have been considered secondary to impaired epidermal barrier formation or other developmental defects. Our findings suggest that p63 can also contribute directly to skin inflammation by functioning as an activator or repressor of epidermal inflammatory gene transcription. We found that many of the p38α-dependent genes encoding mediators of immune responses possessed enhancer and promoter DNA sites occupied by p63. Furthermore, ablation of p63 expression perturbed cytokine-responsive expression of these genes. Notably, a single nucleotide polymorphism at the TP63 locus (rs28512356) has been associated with susceptibility to psoriasis (63). It is unclear, however, whether this TP63 variation increases or decreases p63 expression and transcription function and how such a change affects skin immunity and psoriasis pathogenesis. Given that imiquimod-induced psoriasiform dermatitis was comparable in WT and p38α-EKO mice, p63 phosphorylation by p38α seems dispensable for this form of inflammation. This, however, does not rule out the possibility that p63 plays a role in psoriasis independently of p38α-mediated phosphorylation. p63 may serve as a point of convergence and integration of diverse physiologic signals—developmental, metabolic, and inflammatory—and coordinate keratinocyte gene expression upon interpretation of these inputs. Phosphorylation by p38α and other signaling events are thought to modify p63 transcription function and represent pharmacologically actionable molecular processes for clinical translation.

Key Points.

• p38α signaling reshapes keratinocyte gene expression during allergic dermatitis.

• p63 phosphorylation by p38α is essential for epidermal MMP13 induction.

• Topical p38α and MMP13 inhibitors suppress atopic dermatitis-like disease in mice.

Abbreviations used in this article:

ChIP-seq

chromatin immunoprecipitation sequencing

H&E

hematoxylin and eosin

IL

interleukin

KD

knockdown

KO

knockout

MBP

major basic protein

OVA

ovalbumin

RNA-seq

RNA sequencing

SAL

saline

sgRNA

single guide RNA

siRNA

small interfering RNA

TPA

12-O-tetradecanoylphorbol-13 acetate

UVB

ultraviolet radiation-B