Research ArticleCancer

Epidermal EGFR Controls Cutaneous Host Defense and Prevents Inflammation

See allHide authors and affiliations

Science Translational Medicine  21 Aug 2013:
Vol. 5, Issue 199, pp. 199ra111
DOI: 10.1126/scitranslmed.3005886

Abstract

The epidermal growth factor receptor (EGFR) plays an important role in tissue homeostasis and tumor progression. However, cancer patients treated with EGFR inhibitors (EGFRIs) frequently develop acneiform skin toxicities, which are a strong predictor of a patient’s treatment response. We show that the early inflammatory infiltrate of the skin rash induced by EGFRI is dominated by dendritic cells, macrophages, granulocytes, mast cells, and T cells. EGFRIs induce the expression of chemokines (CCL2, CCL5, CCL27, and CXCL14) in epidermal keratinocytes and impair the production of antimicrobial peptides and skin barrier proteins. Correspondingly, EGFRI-treated keratinocytes facilitate lymphocyte recruitment but show a considerably reduced cytotoxic activity against Staphylococcus aureus. Mice lacking epidermal EGFR (EGFRΔep) show a similar phenotype, which is accompanied by chemokine-driven skin inflammation, hair follicle degeneration, decreased host defense, and deficient skin barrier function, as well as early lethality. Skin toxicities were not ameliorated in a Rag2-, MyD88-, and CCL2-deficient background or in mice lacking epidermal Langerhans cells. The skin phenotype was also not rescued in a hairless (hr/hr) background, demonstrating that skin inflammation is not induced by hair follicle degeneration. Treatment with mast cell inhibitors reduced the immigration of T cells, suggesting that mast cells play a role in the EGFRI-mediated skin pathology. Our findings demonstrate that EGFR signaling in keratinocytes regulates key factors involved in skin inflammation, barrier function, and innate host defense, providing insights into the mechanisms underlying EGFRI-induced skin pathologies.

Introduction

The epidermal growth factor receptor (EGFR) belongs to the ErbB family of receptor tyrosine kinases and is activated by ligand-dependent homo- or heterodimerization (1). Genetic ablation experiments in mice revealed the importance of EGFR for the development of different organs like brain, bone, and heart and several epithelial tissues (2). Depending on the genetic background, EGFR deletion is lethal during embryonic or early postnatal development (3).

Aberrant EGFR signaling is implicated in the development of cancer in various organs (2, 4), and many studies in mice and humans revealed that EGFR overexpression and activating mutations in the EGFR signaling pathway lead to epithelial neoplasms (2, 46). Therefore, the EGFR is an attractive target for anticancer therapy. Notably, cancer patients treated with EGFR inhibitors (EGFRIs) frequently suffer from cutaneous toxicities including papulopustular (acneiform) rash (45 to 100%), dry and itchy skin (12 to 16%), and microbial infections (38 to 70%) (79). Moreover, several studies indicated that development and severity of this rash are the most important clinical predictors for efficacy of EGFRI treatment (7, 10). However, these skin toxicities are frequently also the reason for premature termination of anti-EGFR therapy. It is therefore important to understand the mechanisms underlying the skin toxicities caused by EGFR inhibition to improve anti-EGFR–based cancer therapies and minimize debilitating side effects for patients.

EGFRI-treated patients show infiltration of inflammatory cells in the skin (9). Proinflammatory mediators such as CCL2, CCL5, and CXCL10, which may recruit dendritic cells (DCs), T cells, and neutrophils, are down-regulated in keratinocytes in response to EGFR stimulation. The production of these chemokines induced by cytokines including tumor necrosis factor–α (TNF-α) is enhanced when EGFR signaling is inhibited (11), providing a possible explanation for the infiltrate into the skin of EGFRI-treated patients. Additionally, although inflammatory adverse events are prominent, there is now accumulating evidence on cutaneous and systemic bacterial infections in EGFRI-treated patients (8, 12).

Here, we used both specimens from cancer patients treated with EGFRI and conditional EGFR knockout mice lacking the receptor specifically in the epidermis (EGFRΔep mice) to unravel the role of EGFR signaling in skin homeostasis and to identify the cellular players contributing to the EGFRI-induced inflammatory phenotype. Our study demonstrates that EGFR expression in keratinocytes is essential for maintaining skin homeostasis because, in its absence, the genes controlling inflammation, barrier function, and host defense are deregulated, leading to a characteristic cutaneous inflammatory phenotype in mice and man.

Results

Systemic pharmacological EGFR inhibition results in severe impairment of skin homeostasis

EGFRI therapy in cancer patients is associated with the development of a characteristic, inflammatory papulopustular exanthema (rash), occurring within the first days to weeks of therapy (Fig. 1A and table S1). Additionally, in the course of therapy, many patients display severe skin dryness (xerosis cutis) or cutaneous bacterial superinfections (table S1). Immunohistochemical analysis of early stages of EGFRI-mediated rash revealed a dense, periadnexial inflammatory infiltrate with a marked clustering of CD68+ macrophages, intraepidermal CD1a+ Langerhans cells, and abundant CD4+ and CD8+ T cells (Fig. 1, B to G), which is not found in healthy skin (fig. S1A). Quantitative analysis demonstrated a significant increase of CD4+ T cells, as well as CD8+ T cells and neutrophil elastase–positive (NE+) cells (P = 0.0001, 0.0495, and 0.0500), in lesional skin compared to healthy controls (Fig. 1H and table S2). Giemsa staining showed increased numbers of total as well as highly degranulated (eight or more granuli) mast cells in lesional skin of EGFRI-treated patients (Fig. 1I and table S2). Reduced expression of phospho-EGFR and phospho-ERK (extracellular signal–regulated kinase) was detected in skin samples obtained from patients treated with EGFRI compared to healthy controls, consistent with the idea that cutaneous adverse effects are caused by a direct inhibition of EGFR signaling in the skin, rather than by indirect metabolic effects (fig. S1B).

Fig. 1 EGFRI therapy induces proinflammatory chemokines in epidermal cells.

(A) Characteristic papulopustular rash in a 52-year-old patient treated with erlotinib. (B) Hematoxylin and eosin (H&E) staining of lesional skin of a patient treated with erlotinib. (C to G) Immunohistochemical analysis of CD1α, CD68, CD4, CD8, and NE in lesional skin from patients treated with erlotinib. (H) Quantitative analysis of immunohistochemistry-based staining. Values indicate relative units (RU) of positive cells per area and represent means ± SEM. P = 0.0001, 0.0495, and 0.0500. (I) Quantitative analysis of Giemsa staining on sections from healthy skin or lesional skin from patients treated with erlotinib. Activated mast cells with more than eight granuli were indicated as degranulated. Values indicate numbers of mast cells per microscopic field (MF) and represent means ± SEM. n = 3. P = 0.0217 and 0.0334. (J) Expression analysis of the indicated genes in keratinocytes treated with erlotinib alone or in the presence of proinflammatory cytokines (TNF-α and IL-1β). Values were normalized to the mean of untreated controls and represent means ± SEM. n = 2 to 5. P = 0.0392 and 0.0144 (CCL27); 0.0001, 0.0037, 0.0038, and 0.0046 (CXCL14); 0.0001, 0.0034, 0.0001, and 0.0001 (CCL5); and 0.0017 and 0.0001 (CCL2). (K) Immunohistochemical analysis of CCL27, CXCL14, CCL5, and CCL2 expression in the skin of healthy individuals (left) and lesional skin of patients treated with erlotinib (right; n = 3). Representative sections are shown. (L) Transwell chemotaxis assay of T cells (total) and CD4+ or CD8+ T cells subjected to conditioned medium from keratinocytes treated with erlotinib plus TNF-α and IL-1β, supplemented with either isotype control or anti-CCL27 antibody. Migration is expressed as a percentage of supplied cells migrating toward the conditioned medium and shown as the mean ± SEM of three independent experiments. P = 0.0034, 0.0062, and 0.0119. (M and N) CCL27 protein expression in the sera of healthy individuals (n = 12), patients receiving erlotinib therapy (n = 17), atopic dermatitis patients (AD; n = 7), and psoriasis vulgaris patients (PsV; n = 5) (M). P = 0.0001 (healthy/erlotinib), 0.0105 (erlotinib/atopic dermatitis), and 0.0061 (erlotinib/psoriasis vulgaris). Differences in serum concentrations of CCL27 between erlotinib-treated patients with rash severity grade 2 and 3 were not significant (N). Data represent means ± SEM. n = 8 to 12. P = 0.0001 (healthy/grade 2) and 0.0002 (healthy/grade 3). Scale bars, 200 μm (B to G) and 100 μm (K).

To analyze the molecular and cellular mechanisms that direct leukocyte subsets to the sites of EGFRI-induced cutaneous inflammation, we performed comprehensive analyses of the chemokine expression profile of primary human keratinocytes treated with the EGFRI erlotinib in vitro. Chemokines represent a superfamily of cytokine-like proteins that mediate the directional migration of leukocytes in vitro and regulate their trafficking in vivo (1315). Erlotinib induced the expression and/or secretion of proinflammatory and skin-associated chemokines such as CCL27, CXCL14, CCL2, and CCL5 in primary human keratinocytes (Fig. 1J, fig. S2A, and table S2). Inflammatory conditions mimicked by the addition of TNF-α and interleukin-1β (IL-1β) further enhanced the expression of CCL2 and CCL5, but not of CCL27 or CXCL14 (Fig. 1J and table S2), and induced the expression of CXCL2, CXCL9, and CXCL10 (fig. S2B and table S2). In contrast, erlotinib led to the down-regulation of CCL20, CXCL1, CXCL2, and CXCL8 (IL-8), as well as IL-1α and IL-1β (fig. S2B and table S2). TNF-α levels were not changed by erlotinib alone but in combination with TNF-α and IL-1β (fig. S2B and table S2).

Furthermore, we found a strong up-regulation of CCL27, CXCL14, CCL5, and CCL2 within cutaneous lesions (rash) of patients treated with erlotinib, as shown by immunohistochemical staining and quantitative reverse transcription polymerase chain reaction (qRT-PCR) (Fig. 1K, fig. S2C, and table S2). Transwell chemotaxis assays with conditioned medium of erlotinib-treated primary human keratinocytes demonstrated a marked chemotactic effect on total, CD4+, and CD8+ T cells. Neutralizing anti-CCL27 antibodies impaired this migratory response, suggesting an important role for erlotinib-induced CCL27 production during the recruitment of effector T cells to sites of EGFRI-induced skin inflammation (Fig. 1L and table S2). Furthermore, serum samples obtained from rash patients under erlotinib therapy for about 3 weeks (19.5 ± 5.8 days) contained significantly more CCL27 than sera obtained from healthy volunteers (Fig. 1M and table S2; P = 0.0001) and even exceeded the levels observed in sera of patients suffering from chronic inflammatory skin diseases such as psoriasis vulgaris or atopic eczema (Fig. 1M and table S2) (16, 17). Rash severity in erlotinib-treated patients did not correlate with serum concentration of CCL27 (Fig. 1N and table S2). The other chemokines of particular interest (CCL5 and CXCL14) were not detectable in patient sera. In summary, these results show that the expression of several chemokines is altered in keratinocytes upon EGFR inhibition and in the epidermis of EGFRI-treated patients.

EGFRI therapy impairs host defense

Microbiological analyses of pustular eruptions (Fig. 2B) or lesions suspicious for infection in a set of patients treated with the EGFRI erlotinib revealed a cutaneous bacterial colonization in 70% of the cases. Staphylococcus aureus was detected in the vast majority (53.6%) of these analyses and was followed by Enterobacteriaceae (17.9%; table S1). To elucidate the molecular and cellular mechanisms of the increased susceptibility to bacterial cutaneous infections in patients treated with EGFRI, we investigated the expression of antimicrobial peptides (AMPs), which are key players in the innate defense against microbial agents at the host-environment barrier (1820). In particular, we tested for the expression of human β-defensin-3 (HBD3), cathelicidin LL37, and ribonuclease 7 (RNase 7). Erlotinib impaired the expression of HBD3, LL37, and RNase 7 in primary human keratinocytes in vitro (Fig. 2A and table S2). Moreover, erlotinib significantly suppressed the production of RNase 7 protein (fig. S2D and table S2; P = 0.0500). To investigate the biological relevance of these findings, we performed colony-forming unit (CFU) assays. Conditioned medium of cultured primary human keratinocytes treated with erlotinib reduced the bactericidal activity against clinical isolates of S. aureus in a dose-dependent manner (Fig. 2B and table S2).

Fig. 2 EGFR inhibition impairs the cutaneous immune defense.

(A) Relative expression of the indicated genes in keratinocytes treated with erlotinib alone or in the presence of proinflammatory cytokines (TNF-α and IL-1β). n = 3 to 4. P = 0.0001, 0.0001, 0.0022, and 0.0003 (RNase 7); 0.0001, 0.0001, 0.0033, and 0.0002 (LL37); and 0.0130, 0.0112, 0.0086, and 0.0081 (HBD3). (B) CFU analyses demonstrating a concentration-dependent reduction of the antimicrobial activity of conditioned medium of primary human keratinocytes cultured in the presence or absence of erlotinib against clinical isolates of S. aureus. Shown are a clinical presentation of a S. aureus superinfection of the rash in an erlotinib-treated patient, a representative CFU analysis (at the indicated dilutions of bacteria), and statistical analysis of the antimicrobial activity. n = 6. P = 0.0014 and 0.0001. (C) Relative expression of claudin-1, claudin-3, and occludin in healthy skin or erlotinib-induced lesions. n = 7 to 9. P = 0.0005.

Next, we set out to investigate additional factors that might affect cutaneous antimicrobial defense as well as the molecular causes leading to xerosis cutis in EGFRI-treated patients. Skin barrier function depends on the proper formation of tight junctions in the outer epidermal layers. Hence, we analyzed the expression of tight junction genes in lesional skin of EGFRI-treated patients and healthy controls. Whereas expression of claudin-3 and occludin was comparable in patients and controls, claudin-1 expression was significantly reduced in lesional skin (Fig. 2C and table S2; P = 0.0005). These results indicate that EGFRI treatment of patients might either directly or indirectly (because of skin inflammation) affect epidermal differentiation, with the subsequent barrier defect resulting in a progressively increased permeability and skin infections.

Mice lacking EGFR in the basal layer of the epidermis develop a lethal, inflammatory skin disease

Systemic treatment with EGFRI affects EGFR signaling not only in keratinocytes but also theoretically in all cells expressing the EGFR. To address whether the skin phenotype observed in EGFRI-treated patients is caused by a direct effect on keratinocytes, we generated mice lacking the EGFR specifically in epidermal cells (EGFRΔep) by crossing mice carrying conditional EGFR alleles EGFRf/f (21) with K5-Cre transgenic mice (5). Southern blot analysis revealed that EGFR was deleted by more than 90% in total skin and tail biopsies, whereas no deletion could be detected in other tissues (fig. S3A). EGFRΔep mice showed reduced EGFR protein in the skin compared to littermate controls (fig. S3B). EGFRΔep mutants were born at almost Mendelian ratio (21.5%) but developed a phenotype similar to EGFR−/− mice (Fig. 3, A to E) (3). Their eyes were open at birth (Fig. 3A), and they displayed growth retardation (Fig. 3, B and F, and table S2) and hair growth defects (Fig. 3, C and D). Most EGFRΔep mutant mice died shortly after birth, and only less than 10% lived longer than 3 weeks (Fig. 3G). These survivors displayed severe inflammation and necrotic lesions and eventually lost all hair (Fig. 3, D and E). In contrast, inducible EGFR deletion in adult epidermis with the tamoxifen-inducible K5-CreERT transgenic line (EGFRΔepER) (5) did not lead to these severe phenotypes, although EGFR protein was absent in the epidermis (fig. S4, A to D). Only mild derangements of hair follicles could be detected, and the life span of the mice was comparable to controls (fig. S4, A to E). Notably, long-term oral erlotinib treatment in wild-type mice also did not lead to any skin toxicities, and qRT-PCR on skin samples from these mice did not reveal changes in various cytokine and chemokine levels as observed in human patients (fig. S5, A to D, and table S2). From these results, it appears that in mice, the timing of EGFR deletion and/or inhibition affects the severity of the skin phenotype. We therefore concentrated our investigations on the skin phenotype of EGFRΔep mice, which closely resembled EGFRI-treated human skin. Starting within the first days after birth, mast cells not only were increased in numbers in the skin of EGFRΔep mice but were also highly degranulated (eight or more granuli), indicating an activated state of these cells (Fig. 3H), thus confirming the observations made in lesional skin of EGFRI-treated patients (Fig. 1I and table S2) (22). As additional indicators of inflammation in adult EGFRΔep mice, we detected major histocompatibility complex (MHC) II expression on keratinocytes, fibroblasts, and endothelial cells and an increased number of leukocytes in both the epidermis and dermis (Fig. 3, I and J, fig. S6, and table S2). Similar to EGFRI-treated human skin, the mouse skins also had increased numbers of DCs (Fig. 3, K and L, and table S2) as well as CD11b+GR1high cells, which most likely represent neutrophil granulocytes (Fig. 3M, figs. S6D and S7, A to F, and table S2) (14). Resident Langerhans cells (LCs) in epidermal ear sheets decreased with disease progression (fig. S7G and table S2), but we observed a significant increase in Langerin, MHC II+ CD45+ cells (P = 0.0146), suggesting that there was active recruitment of inflammatory DCs into the epidermis in the absence of epidermal EGFR (Fig. 3, K and L, and table S2). Skin-resident T cells were activated, and the number of T cell receptor (TCR)low cells was increased in EGFRΔep skin compared to normal skin (Fig. 3N, fig. S7, J to K, and table S2). Although control mice showed a network of dendritic CD3ε+ cells in epidermal ear sheets, cells found in EGFRΔep mice were round and irregularly distributed (fig. S7, H and I). Staining with antibodies detecting αβ TCR and γδ TCR demonstrated that wild-type epidermis harbored almost exclusively γδ T cells, whereas in EGFRΔep epidermal ear sheets, most of the T cells were αβ T cells (fig. S7, J and K). Similarly to EGFRI-treated human skin, significantly increased numbers of CD8α+ cells could be detected in the mouse dermal compartment (fig. S6F and table S2; P = 0.0008). These results show that EGFRΔep mice develop a strong skin inflammation characterized by the increased presence of DCs, granulocytes, monocytes, degranulated mast cells, and αβ T cells.

Fig. 3 Epidermis-specific deletion of EGFR leads to early lethality and severe skin inflammation.

(A to D) Phenotype of EGFRΔep mutant mice (indicated by an asterisk) and control littermates at postnatal day 2 (P2; A), P10 (B), P20 (C), and >3 months of age (D). (E) Magnification of lesions on the back skin of a 3-month-old EGFRΔep mutant. (F) Body weight of EGFRΔep mice is significantly reduced. n = 4. P = 0.0001. (G) Kaplan-Meier curve showing reduced survival in EGFR mutants (n = 110). (H) Quantitative analysis of Giemsa staining of skin samples from EGFRΔep and wild-type mice isolated at the indicated time points after birth reveals an increased number of (activated) mast cells in EGFR mutant skin (degranulated defined as eight or more granuli). n = 35 to 73. P = 0.0001, 0.0024, 0.0002, and 0.0344. (I to N) Flow cytometric analysis of percentages of the indicated cell types in epidermal cell suspensions isolated from wild-type and EGFRΔep mice. Data represent means ± SEM. n = 6 to 12. P = 0.0001 (I), 0.0450 (J), 0.0049 (K), 0.0146 (L), 0.0006 (M), and 0.0001 (N).

EGFR deletion in the epidermis leads to skin barrier and differentiation defects

To understand the reason for the high postnatal lethality, we investigated whether EGFR mutant mice suffered from a defect in the barrier function of the skin, which would eventually lead to water loss or sepsis by invading microbes. A qualitative in situ assay for skin permeability revealed a delay in the skin barrier formation in EGFRΔep embryos, as evidenced by their increased uptake of toluidine blue dye (Fig. 4A). At birth, the skin of EGFRΔep mutants appeared similar to the skin of EGFRf/f mice (Fig. 4B). Functional analysis of the skin barrier with a Tewameter revealed that transepidermal water loss (TEWL) was comparable between EGFRΔep and wild-type skin for about 2 weeks after birth (Fig. 4C and table S2). However, TEWL progressively increased in EGFRΔep mice older than 3 weeks, which also displayed dry and scaly skin, demonstrating a close resemblance to the progressive skin dryness observed in EGFRI patients (Fig. 4C and table S2) (9).

Fig. 4 Lack of epidermal EGFR affects the skin barrier and induces proinflammatory mediators.

(A and B) Toluidine blue staining of EGFRΔep mutants (indicated by an asterisk) and control littermates at embryonic day 17 (E17) (A) and P1 (B). (C) TEWL measured at the indicated time points. n = 4 to 15. P = 0.0001 and 0.0001. (D to F) Expression of tight junction genes is reduced in EGFR mutant epidermal cells as shown by qPCR (D; n = 4; P = 0.0207 and 0309) or staining of epidermal ear sheets from 3-month-old mice with an anti–claudin-1 antibody (F) compared to wild-type controls (E). (G) Relative expression of AMPs in cultured keratinocytes of the indicated genotypes. n = 4 to 6. P = 0.0253, 0.0053, and 0.0264. (H and I) Relative expression of proinflammatory cytokines and chemokines in total epidermis (H; n = 4 to 5; P = 0.0001, 0.0165, 0.0171, 0.0001, 0.0366, and 0.0230) or cultured keratinocytes of 3-week-old mice (I; n = 5 to 7; P = 0.0253, 0.0031, 0.0423, 0.0001, 0.0447, 0.0112, and 0.0386). (J and K) Relative expression of CCL2 (J) and CCL5 (K) in EGFR mutant and wild-type epidermal cells at the indicated time points. n = 3 to 5. P = 0.0492 (CCL2) and 0.0117 and 0.0001 (CCL5). (L) Fold induction of CCL2 and CCL5 expression in wild-type keratinocytes treated with the EGFRI BIBW2992. n = 4. P = 0.0264. Data represent means ± SEM. Scale bars, 50 μm (E and F).

In parallel, we could show that the expression of tight junction genes such as claudin-1, claudin-3, and occludin was reduced both in cultured keratinocytes (Fig. 4D) and in the epidermis of adult mice (Fig. 4, E and F, fig. S8, A and B, and table S2). These data indicate that the EGFR is required for the efficient establishment of the skin barrier during embryonic development as well as for the maintenance of a functional barrier in adults, which is consistent with the reduced expression of claudin-1 in lesional skin from EGFRI-treated patients. Furthermore, the epidermis of adult EGFRΔep mice displayed a differentiation defect. Immunofluorescence staining revealed epidermal hyperplasia as shown by increased numbers of keratin 1–positive, suprabasal cell layers in EGFRΔep mice (fig. S8, C and F). Involucrin expression was not altered in these animals, but loricrin, a marker for terminal differentiation, not only was detectable in the cornified layer but was also weakly expressed in the basal layer of EGFRΔep epidermis, indicating that keratinocyte differentiation was incomplete (fig. S8, D, E, G, and H).

Analogous to observations made in humans, we also found impaired innate defense mechanisms in murine EGFRΔep keratinocytes, which displayed decreased expression of β-defensin-14 (mDefb14), the mouse homolog of HBD3 (Fig. 4G and table S2). However, other defensins like mDefb1 and mDefb2 were up-regulated in EGFRΔep skin. These progressive defects in epidermal differentiation and innate defense mechanisms in EGFR-deficient epidermis might contribute to the dysfunctional skin barrier and host defense.

Lack of EGFR signaling in epidermal cells induces proinflammatory chemokines and cytokines

We have shown that the skin of cancer patients treated with EGFRI harbors enhanced levels of proinflammatory cytokines and chemokines such as CCL2, CCL5, CCL27, or CXCL14 (Fig. 1 and table S2). EGFR-deficient murine keratinocytes and epidermis also showed increased expression of CCL2, CCL5, and CCL27, whereas CXCL14 was unaffected in keratinocytes and down-regulated in total epidermis (Fig. 4, H and I, and table S2). TNF-α and IL-6 also appeared to be increased, but the difference was not significant. Other cytokines, such as CXCL1, the mouse homolog of CXCL8, CXCL2, IL-1α, IL-1β, and GM-CSF (granulocyte-macrophage colony-stimulating factor) (not significantly), were down-regulated in keratinocytes (Fig. 4I and table S2). However, substantial amounts of IL-1α, IL-1β, CXCL1, and CXCL2 appeared to be produced by inflammatory cells because their expression was increased in EGFR-deficient epidermis but not in isolated primary keratinocytes (Fig. 4, H and I, and table S2). To determine whether up-regulation of proinflammatory cytokines in keratinocytes upon EGFR deletion is a cell-autonomous effect or a consequence of defective skin barrier function, we analyzed their expression in fetuses. CCL2 was already expressed at higher concentrations in the epidermis of EGFRΔep fetuses (E18.5) and neonates born by Caesarian section (P0) before being exposed to pathogens (Fig. 4J and table S2). The expression of most cytokines such as CCL2 and CCL5 was highest in 3-week-old mice, at an age when most of the mice died, and was decreased again in the epidermis of surviving EGFRΔep mice (Fig. 4, J and K, and table S2). Similar to human keratinocytes, treatment of murine wild-type keratinocytes with the EGFRI BIBW2992 (5) resulted in up-regulation of CCL2 and CCL5 (Fig. 4L and table S2), suggesting that the EGFR negatively regulates the expression of these cytokines in a cell-autonomous manner.

Skin inflammation is not induced by hair follicle degeneration

Human patients undergoing EGFRI therapy frequently display hair growth defects and hair follicle degeneration similar to mice with defects in the EGFR signaling pathway (9, 23, 24). Hair follicles have an immune privilege (25), which might be perturbed by EGFR inhibition and thus contribute to the inflammatory phenotype. To clarify this issue, we crossed EGFRΔep mice with immunologically competent hairless (hr/hr) mice, which carry a mutation in the hairless (hr) gene and progressively lose hair, becoming completely bald 3 to 4 weeks after birth (26). At the age of 3 weeks, EGFRΔep hr/hr double-mutant mice still harbored some hair follicles in the skin, even though their EGFRf/f hr/hr littermate controls were already bald. At the age of 4 months, the very few surviving EGFRΔep and EGFRΔep hr/hr mice had almost completely lost their hair (Fig. 5, A and B). Histological and flow cytometric analysis of skin sections of 3-week-old or 4-month-old double-mutant EGFRΔep hr/hr mice revealed that the quality and quantity of the inflammatory infiltrate were comparable to EGFRΔep mice (Fig. 5, C to H, fig. S6, and table S2). Analysis of epidermal and dermal cell suspensions revealed high expression of MHC II on keratinocytes, fibroblasts, and endothelial cells and increased numbers of hematopoietic cells in both EGFRΔep and EGFRΔep hr/hr mice compared to the respective littermate controls (Fig. 5, C to H, fig. S6, and table S2). From these results, we conclude that the strong inflammation observed in EGFRΔep mice does not result from degenerating hair follicles.

Fig. 5 Crossing EGFRΔep with hairless mice does not ameliorate the inflammatory phenotype.

(A and B) Phenotype of 4-month-old EGFRΔep, EGFRΔep hr/hr, and control mice. (C to H) Flow cytometric analysis of epidermal cells isolated from mice with the indicated genotypes. Data represent means ± SEM. n = 4 to 7. P = 0.0001 and 0.0001 (C); 0.0012, 0.0106, and 0.0053 (D); 0.0114, 0.0308, and 0.0098 (E); 0.0001 and 0.0050 (G); and 0.0427 (H).

Lymphocytes, Langerhans cells, and IL-1 signaling are not responsible for the EGFRΔep phenotype

Immune cells like DCs, macrophages, mast cells, granulocytes, or T cells recruited to the skin may play important roles in the phenotype of EGFRΔep mice and patients treated with EGFRI. To investigate which immune cell contributes to the skin inflammation, we crossed EGFRΔep mice to Rag2−/− mice, which lack mature T and B cells (27). In the skin of EGFRΔep mice, γδ T cells were activated and partially replaced by αβ T cells (fig. S7, G, J, and K, and table S2). However, the lack of lymphocytes in EGFRΔep Rag2−/− double-mutant mice neither rescued the growth retardation nor ameliorated the skin inflammation (Fig. 6, A and B, fig. S9A, and table S2). The epidermal immune cell infiltrate was mainly composed of DCs, macrophages, and neutrophils/granulocytes (Fig. 6, C and D, and table S2), indicating that recruitment of these cells into the lesions is not mediated via a T cell–driven axis but rather directly by keratinocyte-derived factors.

Fig. 6 The inflammatory phenotype is not rescued in the absence of lymphocytes, LCs, and IL-1 signaling.

(A) Phenotype of EGFRΔep, EGFRΔep RAG−/−, and control mice around 2 weeks after birth. (B to D) Flow cytometric analysis of epidermal cells isolated from mice with the indicated genotypes, shown as a percentage of live epidermal cells (B and C; n = 2 to 3) and as a percentage of hematopoietic cells in the epidermis (D; n = 2 to 3). (E) Analysis of the percentage of hematopoietic cells in the epidermis of mice depleted of LCs for at least 3 weeks (n = 2 to 5). (F) Phenotype of EGFRΔep, EGFRΔep MyD88−/−, and control mice 4 weeks after birth. (G) Phenotype of EGFRΔep and EGFRΔep CCL2−/− mice at around 8 months of age. (H) Analysis of the percentage of hematopoietic cells in the epidermis of the indicated mice (n = 2 to 5). (I and J) Analysis of the percentage of hematopoietic cells in the epidermis (I) and TCRlow γδ T cells in the dermis (J) of mice treated with the mast cell stabilizer ketotifen or vehicle [phosphate-buffered saline (PBS)]. Data represent means ± SEM. n = 3 to 4. *P = 0.0203.

LCs, the antigen-presenting cells of the epidermis, were activated in EGFRΔep mice and showed increased migration to skin-draining lymph nodes compared to controls (figs. S7, G to I, and S9B and table S2). We used transgenic mice expressing the diphtheria toxin receptor (DTR) from the Langerin promoter (Lan-DTR) to kill LCs in the skin of EGFRΔep mice by injection of diphtheria toxin (DT) (28). When LCs were depleted in 2- to 3-week-old EGFRΔep Lan-DTR mice for 3 or more weeks, there was no amelioration of the skin phenotype of EGFRΔep mice. We could not detect any difference in the number of inflammatory cells in the skin when compared to DT-treated EGFRΔep littermate control mice (Fig. 6E and table S2). However, long-term depletion of LCs in the inflamed skin of EGFRΔep mice was not as efficient as in control Lan-DTR mice (fig. S9C and table S2).

IL-1 signaling has been suggested to be critically involved in causing the inflammatory infiltrate in EGFRI-treated patient skin (29). We thus crossed EGFRΔep mice to MyD88 knockout mice, which show defective IL-1 signaling (30). Macroscopically, EGFRΔep MyD88−/− double knockout mice could not be distinguished from EGFRΔep littermate controls (Fig. 6F). The composition of the immune cell infiltrate, including up-regulation of DCs and MHC II on keratinocytes, was unchanged and comparable to EGFRΔep mice (fig. S9D).

Chemokines like CCL2 are up-regulated very early in the absence of epidermal EGFR and are further induced by TNF-α upon EGFR inhibition (11) (Figs. 1J and 4J and table S2). Moreover, CCL2 may be responsible for recruiting inflammatory monocytes or mast cells to the skin. However, the growth retardation and inflammatory skin phenotype of EGFRΔep mice were not rescued in a CCL2−/− background (Fig. 6, G and H, and table S2). Surprisingly, the accumulation of DCs in the epidermis was also not influenced by the lack of CCL2 (fig. S9E and table S2), suggesting compensation by other cytokines/chemokines.

EGFRΔep mice display increased numbers of degranulated mast cells in the dermis (Fig. 3H). To analyze the contribution of mast cells to the skin phenotype, we tried to pharmacologically inhibit these cells in 2- to 3-month-old EGFRΔep mice by ketotifen injection. Macroscopically, no difference in skin inflammation could be observed in mutant mice 5 days after treatment. However, the number of infiltrating immune cells was reduced in EGFRΔep mice after ketotifen treatment (Fig. 6I and table S2). The percentage of TCRlow-expressing γδ T cells was strongly reduced in the dermis (Fig. 6J and table S2) and to a lesser extent in the epidermis (fig. S9F and table S2) of ketotifen-injected EGFRΔep mice. Consistent with recently published results (31), only a mild decrease in mast cell degranulation could be observed at this dose of ketotifen, which was not statistically significant (fig. S9G). These data suggest that mast cells might be responsible for the recruitment of γδ T cells into the dermis of EGFRΔep mice.

Together, these results demonstrate that lymphocytes and LCs are not responsible for the skin inflammation observed in EGFRΔep mice, and support the notion that the direct communication of keratinocytes with dermal macrophages, mast cells, and/or granulocytes is the driver of this characteristic inflammatory phenotype.

Discussion

Here, we propose a concept for the control of homeostasis at barrier organs. Our data suggest that epidermal keratinocytes govern cutaneous immunity by controlling (i) innate immune cell–driven inflammation, (ii) barrier function, and (iii) antimicrobial defense through EGFR signaling under physiological conditions. Conversely, impaired EGFR signaling results in an uncontrolled inflammatory cascade, which is aggravated by an impairment of the skin barrier and antimicrobial defense. Findings made in the skin may also serve as a model for EGFRI-induced extracutaneous adverse effects (7, 32) such as diarrhea, which represents the second most frequent complication of EGFRI treatment. The development of diarrhea was also reported to be associated with rash occurrence in EGFRI patients, pointing toward common mechanisms (33).

Recent clinical observations in patients treated with EGFRI support the idea that the EGFR is a key regulator of inflammation (32, 3436). Analyses of EGFRI-induced skin rash lesions revealed marked infiltration of mast cells, macrophages, LCs, dermal DCs, and T cells. EGFRI treatment led to the induction of proinflammatory chemokines such as CCL27, CXCL14, CCL5, and CCL2 in primary human keratinocytes in vitro. Correspondingly, CCL2, CCL5, and CCL27 were up-regulated in EGFR-deficient murine keratinocytes, whereas expression of CXCL14, which may have different functions in human and murine skin (37), was unchanged. In lesional skin of EGFRI-treated patients, the expression of these cytokines was confined not only to epidermal cells expressing EGFR but also to infiltrating immune cells. In humans, CXCL14 has been shown to be responsible for attracting activated monocytes, immature DCs, and their precursors, as well as natural killer cells, into the skin (3740). Moreover, loss of CXCL14 expression in tumor cells has been reported to impair immune surveillance (39).

CCL27 has been shown to regulate effector memory T cell homing into the skin under homeostatic and inflammatory conditions (41). Hence, the induction of CCL27 by EGFRI might systemically increase immune surveillance (13, 41). In the sera of EGFRI-treated patients, we also observed an induction of CCL27, which was stronger than what is observed in patients suffering from diseases such as psoriasis vulgaris or atopic eczema, where serum CCL27 concentration correlates with disease severity (16, 17, 40). Thus, CCL27 could serve as a serological biomarker for monitoring EGFRI-treated cancer patients because, at this time, the severity of the rash remains the most important indicator of EGFRI tumor response (7, 10, 42). Further studies with higher numbers of patients are needed to test this hypothesis, correlating rash severity, tumor response, and CCL27 serum concentrations before and under EGFRI therapy.

Besides the strong mast cell and neutrophil infiltrate, we also observed a switch in the cutaneous T cell repertoire from γδ T cells to αβ T cells in the epidermis of EGFRΔep mice. T cells are recruited very early into the skin of patients treated with EGFRI, providing clinical evidence for an important role of lymphocytes in EGFRI-induced skin rash. However, T cells do not seem to be responsible for the cutaneous symptoms, because crossing EGFRΔep mice into a Rag2−/− background did not alleviate the skin inflammation. Ablation of cutaneous LCs also did not rescue the skin inflammation. Therefore, it is likely that cells of the innate immune system such as macrophages, granulocytes, and mast cells are responsible for the skin rash. Indeed, treatment with the mast cell stabilizer ketotifen affected immigration of γδ T cells into the skin of EGFRΔep mice. Whether this has any impact on the severity of skin inflammation could not be evaluated because long-term treatment with mast cell stabilizers (cromolyn sodium or ketotifen) was not possible in EGFRΔep mice.

Our findings in EGFRΔep hr/hr double-mutant mice, which almost completely lack hair follicles and which develop a very similar skin inflammation to EGFRΔep mice, provide strong evidence that hair follicle degeneration is not responsible for the inflammatory phenotype seen in the absence of EGFR (9). Rather, we propose that interfollicular keratinocytes are sufficient for the development of the inflammatory skin phenotype and demonstrate that EGFR signaling is not only playing a role in maintaining skin homeostasis in follicular spaces.

One major limitation of our EGFR mutant mouse model is its high postnatal lethality, compared to patients undergoing EGFR therapy, who suffer from more benign skin toxicities. This discrepancy raises the concern of whether this mouse model indeed recapitulates the side effects of EGFRI in patients. Another mouse model with epidermis-specific EGFR ablation displayed a milder or similar skin phenotype, depending on the genetic background and Cre line used (43, 44). Inducible EGFR deletion in adult mice did not affect the life span of the mice but led to mild derangements of hair follicles. It therefore seems that the timing of EGFR deletion in mice affects the severity of the skin inflammation. This is likely the case because adult mice treated with EGFRIs such as erlotinib do not display any skin toxicities. Although the EGFRΔep mouse model is not ideal to mimic skin toxicities induced in EGFRI patients, we have found many similarities in the immune infiltrate and also the expression pattern of chemokines and AMPs between EGFRΔep mice and human patients, suggesting that similar mechanisms underlie the skin rash in both systems.

The reason why most of the EGFRΔep mice die within the first weeks after birth remains elusive. Proinflammatory factors such as TNF-α, IL-6, CXCL1, CXCL2, and G-CSF were expressed at elevated levels in the epidermis and serum of EGFRΔep mice, especially at the age of 3 weeks when most mutant animals died. Large amounts of cytokines such as TNF-α or ILs have been implicated in the pathogenesis of several human diseases and may lead to lethality (45, 46). However, crossing of EGFRΔep mice to CCL2 or MyD88 knockout mice did not ameliorate the skin phenotype. In accordance with our data, Mascia et al. (47), who performed a very similar study in a different mouse model lacking epidermal EGFR, also report that crossing EGFR mutants to mice deficient for TNF-α receptors, MyD88, NOS2, CCR2, or T cells failed to rescue the skin toxicities. However, local depletion of macrophages improved the skin phenotype (47).

It has recently been published that lack of ADAM17, the sheddase that cleaves EGFR ligands, results in a similar phenotype as described here. In their study, Franzke et al. (44) confirmed that aberrant EGFR signaling is mainly responsible for the observed skin inflammation and barrier defects. This outside-in barrier defect was thought to contribute to atopic dermatitis–like skin inflammation by exposing the skin to exogenous agents (for example, bacteria) or antigens, thereby attracting immune cells to the site of infection. Moreover, they showed that inflammatory macrophages were not the primary cause for the skin barrier defect. As expected, we found a delay in skin barrier formation in EGFRΔep embryos and defects in its maintenance in adult mice, along with a reduced expression of tight junction proteins like claudin. Similar phenotypes were also seen in EGFRI-treated lesional human skin. Our data suggest that the absence of EGFR signaling in keratinocytes creates a milieu capable of directly recruiting inflammatory cells. This is supported by the finding that in vivo expression of CCL2 and CCL5 is altered even in the absence of bacterial colonization or antigens in embryonic skin. Skin inflammation in EGFRI-treated patients can be provoked or aggravated even by unspecific physical trauma or exposure to intense sunlight (Koebner phenomenon) independently of the presence of bacteria or antigens, indicating that the EGFR controls inflammation independent of the skin barrier status (47). Nevertheless, barrier defects are likely to promote bacterial colonization or infection in EGFR-deficient skin, as indicated by the clinical observations that EGFRI-treated patients show frequent bacterial colonization (table S1) and infections of the skin (8). RNase 7 is constitutively produced by keratinocytes and represents the principal AMP of healthy human skin responsible for killing S. aureus (19). We found that in human keratinocytes treated with EGFRI as well as in EGFR-deficient mouse skin, RNase 7 was decreased and could therefore explain the predominant proliferation of S. aureus on the skin of EGFRI-treated patients. RNase 7, along with HBD3 (mDefb14), has been reported to be induced upon EGFR activation after wounding or inflammation (19, 20, 4850). Moreover, RNase 7 has recently been shown to contribute to the cutaneous defense against Enterococcus faecium (51). This finding is reflected in the fact that Enterobacteriaceae were identified as the second most frequent pathogens in our microbial analyses of EGFRI-treated patients (table S1). Hence, we propose that the lack of EGFR induces deficiency in AMP expression and a barrier defect, resulting in a distinct but potentially temporary impairment.

In summary, our results support the following model of EGFRI-induced cutaneous adverse effects (Fig. 7): Under homeostatic conditions, the EGFR serves as a central regulator of cutaneous inflammation, barrier function, and antimicrobial defense. Conversely, impairment of EGFR signaling enhances the production and release of proinflammatory mediators that facilitate the recruitment of immune cells. Activated infiltrating inflammatory cells produce many different cytokines and chemokines that sustain leukocyte recruitment and activation, finally leading to the development of erythematous papules and pustules. In addition, EGFR inhibition suppresses the production of AMPs and skin barrier proteins, resulting in an impaired host defense and barrier function, finally leading to cutaneous bacterial colonization and frequent infections as well as an increased TEWL and progressive xerosis. Our findings endorse the concept that not the leukocyte but the epithelial keratinocyte is the key regulator of cutaneous immunity. Furthermore, our study may suggest potential surrogate markers for EGFRI therapy, as well as targets for the design of mechanism-based management strategies for EGFRI-associated toxicities.

Fig. 7 EGFRI treatment causes cutaneous rash and infectious complications.

EGFR signaling controls cutaneous immune defense by modulating barrier genes (blue, jagged lines), AMPs (green dots), as well as chemokine/cytokine production (red dots). In vivo, blockade of EGFR signaling by EGFR antagonists (represented as gray circles) results in the down-regulation of epithelial AMPs such as RNase 7, LL37, and HBD3, as well as tight junction genes such as claudin, but up-regulation of skin-associated chemokine production, leading to cutaneous inflammation and decreased host defense, thereby allowing S. aureus infection. MØ, macrophages.

Materials and Methods

Study design

The rationale for this study was to undertake a comprehensive systematic analysis of EGFRI-associated rashes and identify their underlying molecular and cellular mechanisms by using cultured keratinocytes and skin biopsies isolated from patients treated with EGFRI and from mice genetically lacking EGFR in the epidermis. The study included 107 patients at two centers (Department of Dermatology, University Hospital Duesseldorf, and Department of Dermatology and Allergy, Ludwig-Maximilian University, Munich, Germany) who had been enrolled from March 2007 to October 2009. Inclusion criteria were treatment with an EGFRI, such as erlotinib (Tarceva, Hoffmann–La Roche) or cetuximab (Erbitux, Merck), a chimeric monoclonal antibody (mAb) directed against EGFR, and the development of a papulopustular rash (table S1). The study was approved by the local ethics committees (Ethics Committee, Medical Faculty, University of Duesseldorf, Duesseldorf, and Ethics Committee, Medical Faculty, Ludwig-Maximilian University, Munich).

In general, patients initially developed papular lesions, which gradually progressed to develop the characteristic pustular aspect. Serum samples and lesional skin biopsies were taken after obtaining informed consent from patients treated with erlotinib or from healthy individuals undergoing plastic surgery (negative control). Skin samples were either immediately stored in 4% paraformaldehyde (PFA) or snap-frozen in liquid nitrogen and stored at −80°C. Microbiologic skin swabs were performed in all patients presenting pustular lesions or signs of skin infection (n = 40) and were analyzed by the local institutes for microbiology (Institute of Medical Microbiology, Medical Faculty, University of Duesseldorf, Duesseldorf, and Department of Microbiology, Ludwig-Maximilian University, Munich). All human analyses were performed by investigators blinded to the treatment.

In accordance with the local animal ethics policies (European Commission and Austrian Federal Ministry of Science and Research), the experiments were designed to use the smallest number of mice needed to obtain the requested data. The number of animals ranged from 4 to 10 and is specified for each experiment. This number of animals allowed us to perform adequately powered statistical analyses. Each experiment was independently repeated at least twice. After genotyping, mice were assigned randomly to treatment groups. Skin samples were analyzed by histology, immunohistochemistry, or immunofluorescence or processed further for RNA, DNA, or protein extraction, fluorescence-activated cell sorting analysis, and keratinocyte isolation. Because of the severe skin phenotype, blinded assessment of experimental outcomes was not always applicable, but it was used whenever possible (such as for counting the number of cells in skin biopsies). Mice were also treated with pharmacological inhibitors (see below).

Mice

Mice were kept in the animal facilities of the Medical University of Vienna or the Heinrich-Heine-University Düsseldorf in accordance with institutional policies and federal guidelines. Animal experiments were approved by the Animal Experimental Ethics Committee of the Medical University of Vienna and the Austrian Federal Ministry of Science and Research (animal license numbers: GZ 66.009/124-BrGT/2003; GZ 66.009/109-BrGT/2003; BMWF-66.009/0073-II/10b/2010 BMWF-66.009/0074-II/10b/2010).

Mice of mixed C57BL/6 × 129/Sv background, lacking the EGFR specifically in the epidermis (EGFRΔep), were generated by crossing mice carrying conditional EGFR alleles EGFRf/f (21) with K5-Cre (5) or K5-CreERT (5) transgenic mice. In mice expressing both K5-Cre and the conditional EGFR allele, EGFR is constitutively deleted in the basal layers of the epidermis starting from E14.5 (5). Recombination was confirmed by Southern blot as previously described (5, 21).

Hairless (hr/hr; Skh-1, purchased from Charles River Laboratories), Rag2−/−, CCL2−/−, and MyD88−/− mice (all of C57BL/6 background) have been previously described (26, 27, 30, 52, 53). EGFRΔep mice were bred with Langerin-DTR-EGFP mice (Lan-DTR; C57BL/6 background) (28). LCs were depleted by injection of 100 ng of DT intraperitoneally every other day from week 3 on for at least 3 weeks. To inhibit mast cells, EGFRΔep mice were injected into the tail vein with ketotifen (12.5 mg/kg) on five consecutive days and analyzed 1 hour after the last injection. Male hairless mice (Skh-1:Hr; Charles River Laboratories) were housed in the animal facility of the Heinrich-Heine-University Düsseldorf in accordance with institutional policies and federal guidelines. At the age of 8 weeks, the mice were treated with erlotinib (40 μg/g per day) for up to 18 weeks. Afterward, skin biopsies were taken from the dorsal skin, and analysis of skin parameters was performed with DermaLab skin probes (Cortex Technology).

Immunohistochemistry of human skin biopsies

Skin biopsies of early-stage erlotinib-induced rashes or healthy skin were fixed and stained with antibodies directed against CCL27, CXCL14 (R&D Systems), CCL2 (Abcam), CCL5 (Santa Cruz Biotechnology), CD68, CD8, CD4, NE, and EGFR (DAKO) and against phospho-EGFR and phospho-ERK (Cell Signaling) as described previously (41). Histologic images were acquired with an Axiovert2 MOT microscope and AxioCam MRc with AxioVision 4.6 software (all Zeiss). For the quantification of CD4, CD8, and NE staining, computer-assisted image analyses were performed as described previously (41).

Culture of human epidermal cells

Human primary epidermal keratinocytes were isolated and cultured in keratinocyte SFM cell growth medium (Invitrogen) at 37°C, 5% CO2. Cells were treated with erlotinib (500 and 1000 nM; Roche Pharmaceuticals) or medium containing dimethyl sulfoxide only (vehicle control) in the presence or absence of TNF-α (10 ng/ml; AbD Serotec) and IL-1β (5 ng/ml; R&D Systems). Twenty-four hours later, supernatants and cells were harvested and RNA was extracted.

RT-PCR and ELISA of human samples

qRT-PCR analyses were performed with standard protocols (41). Quantification of CCL27, CXCL14, and RNase 7 protein concentrations in cell culture supernatants was performed by enzyme-linked immunosorbent assay (ELISA) according to the DuoSet ELISA Development kit protocol (R&D Systems).

Measurement of chemoattractive activity

Transwell chemotaxis assays with column-purified human T cells and conditioned medium of erlotinib-treated human primary keratinocytes were performed as previously described (41).

Measurement of antimicrobial activity

Wild-type S. aureus was isolated from patients with skin and soft tissue infections at the Institute for Microbiology, University of Duesseldorf. Bacteria were cultivated and diluted to a final concentration of 105 colonies/ml by serial dilution in culture medium (PAN-medium; Biotech GmbH) plus l-tryptophan (5 μg/ml; Sigma-Aldrich). Human epidermal keratinocytes were incubated as described earlier. Subsequently, bacterial suspensions were added. Twenty-four hours later, cell supernatants were removed, and serial dilutions (10 μl; 10−1 to 10−8) were plated onto agar plates. CFUs were determined by counting bacterial colonies after 24 hours of incubation. Antimicrobial activity [%] = [1 − (CFU treated culture)/(CFU control)] × 100.

Mouse epidermal cell cultures

Mouse epidermal cells were isolated as previously described (5) and cultured on vitrogen/fibronectin-coated dishes in low-calcium minimum essential medium (Sigma) containing 8% chelated fetal calf serum (FCS). For EGFR inhibition, 80% confluent cells were treated with the EGFRI BIBW2992 (10 nM; Selleck) for 24 hours.

Flow cytometric analysis

Cell suspensions from epidermis and dermis were isolated as described above and then filtered through a 70-μm nylon mesh, counted, and stained with mAbs for 30 min in PBS + 5% FCS at 4°C after blocking with Fc-block (BD Pharmingen). The following mAbs were used: anti-CD3ε–PE (phycoerythrin) (clone145-2C11), anti-CD11c–FITC (fluorescein isothiocyanate) (clone HL3), anti-CD45–APC (allophycocyanin) (clone 30F11), anti–MHC II–FITC (clone 2G9), anti-CD44–FITC (clone IM7), and anti-Vγ3TCR (clone 536), all from BD Pharmingen; anti-CD4–TC (clone RM4-5), anti-CD8α–Alexa Fluor 647 (clone 5H10), anti-CD11b–FITC (clone M1/70.15), and anti-Ly6C/G (clone RB6-8C5), all from Caltag Laboratories; and anti-TCRγδ–APC (clone GL-3) from BioLegend. Dead cells were excluded by adding 7-aminoactinomycin D (Sigma) at a final concentration of 1 μg/ml after the last washing step. Data were acquired on an LSR-II flow cytometer (BD Biosciences) and analyzed with FlowJo software.

Histological analysis of murine skin biopsies

Skin biopsies were fixed in 4% PFA, embedded in paraffin, and cut into 5-μm sections. Before staining, sections were dewaxed in xylene and rehydrated through a series of ethanol dilutions (100, 95, 90, 80, 70, and 30%) and incubated in water for 10 min. Rehydrated sections were stained with Harris H&E (Sigma) according to standard procedures. Giemsa stainings were also performed according to standard procedures. Images were obtained with a Nikon eclipse 80i microscope; histomorphometric analysis was performed with Lucia software. For immunofluorescence staining, mouse tissues were embedded in OCT (Sakura), and 5-μm cryosections were cut and fixed in acetone or 1% PFA before processing. Epidermal ear sheets were prepared by separating epidermis from dermis with 3.5% ammonium thiocyanate and fixed in acetone. For immunohistochemistry and immunofluorescence staining, the following antibodies were used: keratin 1, keratin 5, involucrin, and loricrin (Covance); CD11b, GR1, Langerin, CD3ε, TCRαβ, and TCRγδ (BioLegend); and claudin-1 (Abcam). Secondary antibodies were purchased from Molecular Probes and Vector Laboratories.

Dye exclusion assay

Embryos were incubated for 1 min in 25, 50, and 75% methanol in PBS, followed by a 1-min incubation in 100% methanol and a series of incubations in 75, 50, and 25% methanol in PBS for 1 min. Embryos were then washed in PBS for 1 min and stained with 0.1% toluidine blue O (Sigma) for 10 min.

Total RNA isolation and RT-PCR analysis of murine cells and tissues

Total RNA from epidermis or cultured epidermal cells was isolated with TRIzol Reagent (Invitrogen). Complementary DNA (cDNA) synthesis was performed with SuperScript First-Strand Synthesis System (Invitrogen) according to the manufacturer’s instructions. qRT-PCR was performed with the LightCycler FastStart DNA MasterPLUS SYBR Green I kit together with the LightCycler 2.0 System (Roche).

Western blot analysis

Protein lysates were prepared as previously described (5) and separated by SDS–polyacrylamide gel electrophoresis and transferred onto polyvinylidene difluoride membranes (Millipore). Western blot analysis was performed as described elsewhere (5) with antibodies detecting EGFR (Upstate Biotechnology) and actin (Sigma).

ELISA of murine samples

Mouse CCL2 Immunoassays (R&D Systems) were performed according to the manufacturer’s instructions with 48-hour-old supernatants collected from 80% confluent keratinocyte cultures or with 40 μg of protein from epidermal cell or total skin lysates.

Statistical methods

All experiments except data presented in fig. S5 were repeated at least twice and done in triplicates. Data are represented as means ± SEM. n describes the number of biological replicates. Data were evaluated with a two-tailed, unpaired Student’s t test with 95% confidence intervals. P < 0.05 was taken to be statistically significant, and P values are represented with asterisks in the figures (*P ≤ 0.05; **P ≤ 0.005; ***P ≤ 0.0005). Exact P values are listed in the respective figure legends. Data in Fig. 3F were analyzed by a two-way analysis of variance (ANOVA) test. In Fig. 3G, data were analyzed by a log-rank (Mantel-Cox) test.

Supplementary Materials

www.sciencetranslationalmedicine.org/cgi/content/full/5/199/199ra111/DC1

Fig. S1. EGFRI effectively blocks EGFR signaling in the skin.

Fig. S2. EGFRI differentially regulates the expression of various chemokines.

Fig. S3. EGFR is effectively deleted in epidermal cells.

Fig. S4. Deletion of EGFR in adult mice results in a benign skin phenotype.

Fig. S5. Mice treated with erlotinib do not display the skin inflammation found in EGFRΔep mice.

Fig. S6. Lack of epidermal EGFR results in skin inflammation.

Fig. S7. EGFRΔep mice display strong skin inflammation.

Fig. S8. EGFRΔep mice display an epidermal differentiation defect.

Fig. S9. EGFRΔep mice are not rescued in a Rag2-, CCL2-, MyD88-, or LC-deficient background.

Table S1. Patient characteristics.

Table S2. Raw data of experiments with fewer than 20 biological replicates (Excel file).

References and Notes

  1. Acknowledgments: We are grateful to M. Hammer for maintaining our mouse colonies. We thank A. Bogusch, R. Kubitza, S. Kellermann, U. Wiesner, and A. van Lierop for excellent technical assistance. Funding: This work was supported by the EC programs QLG1-CT-2001-00869 and LSHC-CT-2006-037731 (Growthstop), the Austrian Federal Government’s GEN-AU program “Austromouse” (GZ 200.147/1-VI/1a/2006 and 820966), and Austrian Science Fund grants DK W1212, P18421, P18782, and SFB-23-B13 to M.S. This project was also supported by grants from the German Research Foundation to B.H. (FOR 729; HO2092/5-2; SPP 1190; HO 2092/3-2), a grant from the Research Foundation of the Medical Faculty of the Heinrich-Heine-University to P.A.G., and a research grant from Roche Pharmaceuticals to B.H. and P.A.G. B.M.L. was a recipient of a Boehringer Ingelheim Fonds fellowship. Author contributions: B.M.L., P.A.G., and M.H. designed the experiments, performed most of them, and contributed equally to this work. B.A.B., N.A., and V.S. performed some experiments. P.A.G., B.A.B., H.S., E.B., P.A., R.G., C.M., A.W., J.W.F., A.K., J.H., J.M.S., and B.H. participated in the design, data collection, analysis, and interpretation of the study in human patients. K.R. performed the experiments involving erlotinib treatment in mice. B.H. and M.S. provided funding, supervised the project, and contributed equally to this work. B.M.L., P.A.G., M.H., B.H., and M.S. wrote the manuscript. Competing interests: P.A.G. has received honoraria and speaking fees from Roche Pharmaceuticals, Galderma International SAS, Merck, and Amgen. A.W. has served as a paid advisor for Roche, Merck, and Amgen. B.H. has received consulting and lecture fees from Roche, Leo Pharma, Basilea Pharmaceuticals, Galderma International SAS, and ALK/Abello.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article