Research ArticleCancer

Genetic Ablation of Epidermal EGFR Reveals the Dynamic Origin of Adverse Effects of Anti-EGFR Therapy

See allHide authors and affiliations

Science Translational Medicine  21 Aug 2013:
Vol. 5, Issue 199, pp. 199ra110
DOI: 10.1126/scitranslmed.3005773


Cancer patients treated with anti-EGFR (epidermal growth factor receptor) drugs often develop a dose-limiting pruritic rash of unknown etiology. The aims of our study were to define causal associations from a clinical study of cutaneous and systemic changes in patients treated with gefitinib and use these to develop and characterize a mouse model that recapitulates the human skin rash syndrome caused by anti-EGFR therapy. We examined the patients’ plasma before and after treatment with gefitinib and documented changes in chemokines and leukocyte counts associated with the extent of rash or the presence of pruritus. We established a parallel mouse model by ablating EGFR in the epidermis. These mice developed skin lesions similar to the human rash. Before lesion development, we detected increased mRNA expression of chemokines in the skin associated with early infiltration of macrophages and mast cells and later infiltration of eosinophils, T cells, and neutrophils. As the skin phenotype evolved, changes in blood counts and circulating chemokines reproduced those seen in the gefitinib-treated patients. Crossing the mutant mice with mice deficient for tumor necrosis factor–α (TNF-α) receptors, MyD88, NOS2, CCR2, T cells, or B cells failed to reverse the skin phenotype. However, local depletion of macrophages provided partial resolution, suggesting that this model can identify targets that may be effective in preventing the troublesome and dose-limiting skin response to anti-EGFR drugs. These results highlight the importance of EGFR signaling in maintaining skin immune homeostasis and identify a macrophage contribution to a serious adverse consequence of cancer chemotherapy.


Epidermal growth factor receptor (EGFR) activation through overexpression or mutation contributes to tumor development and progression in multiple tissues. Both small molecules and monoclonal antibodies designed to block EGFR activation have proven effective in cancers dependent on EGFR activity (1). However, all EGFR inhibitors produce skin- and skin adnexa–specific toxicity in patients (2). Starting as early as the first 2 weeks after therapy initiation, many patients develop the typical papulopustular follicular rash and frequently suffer from pruritus, skin xerosis, scaling of the epidermis, and nonscarring alopecia. Depending on severity, the skin response can negatively affect the quality of life, necessitating discontinuation of therapy or dose reduction that may interfere with treatment efficacy. An interesting feature of the rash is that its presence and severity often coincide with a positive tumor response, suggesting a shared mechanism of action of the drug within the skin and the tumor compartments (35). For this reason, there is considerable interest in characterizing the pathogenesis of the rash beside the obvious necessity to minimize the negative effects on patients’ quality of life.

Altered epithelial differentiation is characteristic of the skin of patients undergoing anti-EGFR treatment, and this has been interpreted as the main trigger of the inflammatory process (6). It is worth noting that in biopsies from lesional skin, the inflammatory response and the aberrant epithelial differentiation often coincide (7). Thus, it is not possible to identify the initial causative event solely from the examination of human lesional skin. Transgenic models with reduced or absent EGFR expression also display alterations in the differentiation of the follicular and interfollicular keratinocytes, but the relationship between these alterations and the inflammatory process is unknown (810).

In normal mice, acute pharmacological inhibition of EGFR or one of its downstream effector kinases [extracellular signal–regulated kinase 1/2 (ERK1/2)] is associated with exacerbation of skin inflammatory responses and enhanced expression of chemokines in keratinocytes (11, 12). Interfering with EGFR signaling through blockade of growth factor shedding, preventing ligand-receptor interaction through antibody administration, or blocking tyrosine kinase activity with small inhibitors all enhanced inflammation outcome in human keratinocytes in vitro and in vivo in patients (1, 11, 13). These data indicate that the blockade of EGFR is linked to an inflammation-prone behavior of keratinocytes. Clinical evidence suggests that what may start as a local cutaneous response has systemic repercussions, because biomarker analyses have documented changes in circulating cytokines and chemokines in treated patients (1417). However, a temporal analysis documenting the dynamics of these local cutaneous and systemic changes in patients and a model system that can link the systemic effects of the anti-EGFR treatment to the skin response is lacking. Such an analysis is important to determine the primacy of the response, because clinical samples reveal only the fully formed characteristics of the skin lesions. Using data from a clinical trial of gefitinib treatment for ovarian cancer, we have been able to monitor the pattern of systemic changes in circulating chemokines, cytokines, and blood cells after initiation of treatment and relate this to onset of pruritus and rash. We constructed a mouse model of epidermal targeted ablation of EGFR and observed local inflammation, pruritus, altered keratinocyte differentiation, alopecia, and systemic inflammation. We used this model to document the temporal changes involved in the developing phenotype and applied genetic and pharmacological methods for prevention.


Inflammatory mediators and blood cell count changes in plasma of gefitinib-treated patients are associated with rash and pruritus

We examined changes in inflammatory mediators in plasma collected from 10 patients participating in a gefitinib trial for epithelial ovarian cancer (NCT00049556) (18) to investigate a relationship between rash severity and pruritus. Because of the small number of patients, we looked for trends and significant differences. Although we saw increases in most circulating factors after treatment, the patients whose pretreatment levels of these inflammatory factors were the lowest tended to have greater rash and pruritus after treatment (Fig. 1, A and B). Lower interleukin-18 (IL-18) concentrations before and after the first month of treatment with gefitinib were associated with higher-grade rash (each P = 0.038; Fig. 1A); lower CCL11 showed a similar trend at both times but did not reach statistical significance (each P = 0.067). Lower CCL11 concentrations both before and after 1 month of treatment with gefitinib, and lower IL-18 before treatment also showed a trend to associate with pruritus, which was not significant (each P = 0.095; Fig. 1B). Lower concentrations of IL-1ra and CCL22 after the treatment were associated with higher-grade rash (P = 0.0095 and 0.019, respectively; Fig. 1A), but only lower concentrations of IL-1ra showed a trend to associate with pruritus (P = 0.056; Fig. 1B). In contrast, higher concentrations of CCL5 before the treatment were associated with trends toward pruritus (P = 0.064; Fig. 1B). We analyzed the available blood cell counts for 8 of the 10 patients, looking for trends indicating association with the rash severity or presence of pruritus (Fig. 1, C and D). The magnitude of increase in the percentage of granulocytes and decrease in the percentage of lymphocytes in the blood cell counts were associated with rash severity (P = 0.029) and pruritus (P = 0.036; Fig. 1, C and D). Larger differential increases in the absolute number of platelets were also associated with pruritus (P = 0.036; Fig. 1D), but no association of platelets was found with the rash. These data may indicate that the changes in circulating granulocytes, lymphocytes, and platelets of individual patients can be related to the anti-EGFR treatment but do not discriminate between the effect of the drug on the tumor and a possible role of EGFR blockade in other target organs.

Fig. 1 Inflammatory mediators and blood cell count changes in gefitinib-treated patients.

Patients with ovarian cancer were treated with gefitinib (500 mg daily for 28 days), and matched plasma samples were studied before treatment (T = 0) and after one cycle (T = 1 month). Patients were characterized as having rash grade 1 (limited acneiform or maculopapular rash and limited or no symptoms; blue line) or grade 2 (more extensive rash and/or plaques and symptoms and/or requiring medical intervention; red line) and complaining of pruritus (red) or not (blue). Cytokine and chemokine concentrations were measured by enzyme-linked immunosorbent assay–based multiplex assay. (A and B) Associations of cytokine and chemokine concentrations (pg/ml) with rash (A) and pruritus (B) were analyzed. (C and D) Blood cell count data were available for 8 of the 10 patients and were analyzed by calculating the percentage of granulocytes and lymphocytes present in the blood after 1 month of treatment minus the percentage before the treatment (difference %) and correlating these differences with the extent of rash (C) and presence of pruritus (D). Changes in platelet count (absolute numbers, K/μl) were reported as the number of platelets after treatment minus the number before treatment (D). P values described in the text were calculated with a Wilcoxon rank-sum test. Given the small number of patients and the large number of exploratory tests performed, P < 0.01 was considered statistically significant, whereas 0.01 < P < 0.10 indicated trends.

Epidermal ablation of EGFR in mice recapitulates the skin phenotype of anti-EGFR–treated patients

To gain insight into the role of EGFR inhibition in the skin, we developed an epidermally targeted mouse model of EGFR ablation by crossing Keratin5/promoter-driven Cre recombinase transgenics (19) with mice containing loxP sites flanking exon 3 of the EGFR gene (20). Double transgenics or epidermal EGFR–deleted mice (K5CreWT/+ EGFR flx+/+, indicated as KO) appeared phenotypically and histologically normal at birth and up to 5 to 6 days of age (Fig. 2, A to C). After the first week of age, the skin of EGFR-ablated mice progressively developed the hallmarks of the skin lesions of patients treated with anti-EGFR agents (Fig. 2, D to L): infiltration of leukocytes and pruritus, scaly skin, neutrophilic pustules (red arrows), keratin plugs (yellow arrows), altered sebaceous glands (green arrows), pigment masses (black arrows), and hair follicle destruction. Thus, the local consequences of the genetic ablation of EGFR in the epidermis mimicked the adverse effects observed with anti-EGFR drugs in patients (6, 7).

Fig. 2 Consequences of epidermal ablation of EGFR in mice.

(A to L) Hematoxylin and eosin (H&E) staining in skin sections and corresponding macroscopic phenotype of mice at different ages are shown at 3 days (A to C), 7 days (D to F), 21 days (G to I), and 4 months (J to L). Neutrophilic pustules (red arrows), keratin plugs (yellow arrows), sebaceous glands (green arrows), and pigment masses (black arrows) are highlighted in KO mouse skin sections and photographs (H, K, and L). Objective, 10×. Scale bars, 100 μm. Images were captured by staining skin sections from n > 3 independent litters.

The epidermal ablation of EGFR has systemic consequences

We measured the plasma levels of cytokines and chemokines in WT and KO littermates at 2, 3, and 12 weeks of life to verify that our model could also reproduce the systemic imbalance of inflammatory mediators observed in patients after 1 month of gefitinib treatment. In mice with EGFR-ablated epidermis, we observed increases in plasma concentration of CCL11, CCL22, and IL-1ra as in human patients, as well as CCL2, CCL17, IL-6, and IL-17 (Fig. 3A and table S4). In contrast, plasma concentrations of the cytokine IL-18 were not significantly different between WT and KO mice. Blood cell counts reflected the pattern of gefitinib-treated patients: increased circulating neutrophils and platelets and decreased lymphocytes in KO mice versus normal littermates (Fig. 3B). With aging, systemic symptoms of myeloproliferative disease became evident with extramedullary hematopoiesis and enlarged spleen and lymph nodes (Fig. 3C).

Fig. 3 Systemic consequences of EGFR ablation in the epidermis.

(A) Multiplex assay of plasma concentration of chemokines and cytokines expressed in pg/ml for 2-, 3-, and 12-week-old WT (white bars) and KO (black bars) mice. Graphs show mean pg/ml ± SD (n = 3 individual readings, each from a pool of plasma from six to seven mice per genotype per time point; *P ≤ 0.05 by t test for KO compared to WT). Individual P values are listed in table S3. (B) Neutrophil (P = 0.0001), platelet (P = 0.0007), and lymphocyte (P = 0.0008) mean counts ± SD in 3-week-old mice (blue, WT; red, KO; n = 4 mice, by t test). (C) Spleen and lymph nodes from 2-month-old WT and KO littermates.

Immune cell infiltration precedes changes in expression of epidermal differentiation markers

Skin barrier alterations are often associated with the establishment of local and systemic inflammatory conditions (21). Changes in epidermal differentiation markers and cornified envelope proteins are detected in the skin of patients with anti-EGFR rash, but there is not a common consensus if they are primary or secondary events (7). Immunoblot analysis of total protein from skin of EGFR-ablated mice compared to WT littermates (fig. S1A) showed an early increase in the levels of CD45 antigen (total leukocytes) and keratin 6 (7 days of age) that anticipated the up-regulation of keratin 1, keratin 10, and loricrin (2 weeks of age). Filaggrin total protein levels remained unchanged. In the absence of epidermal EGFR, the skin barrier formed normally, as shown by the dye permeation assay in newborn mice (fig. S1B). Moreover, follicular and interfollicular keratinocytes maintained a normal expression of differentiation markers at day 7 when higher numbers of leukocytes (CD45-positive cells) were already invading the tissue as shown by immunohistochemistry (fig. S2). At the same time, keratin 6 expression was slightly altered in the hair follicles of EGFR-ablated mice. At 21 days of age, keratin 6, keratin 1, loricrin, and filaggrin were aberrantly distributed and overexpressed throughout the skin of KO mice (fig. S2). Thus, changes in leukocyte population recruited in the EGFR-ablated skin anticipated changes in differentiation marker expression.

We further characterized which subsets of leukocytes contributed to the increase of the total CD45 population, starting when the KO mice had no apparent histological skin phenotype (3 days) to the later stages of life (20 weeks). Immunohistochemical analysis of skin tissue (Fig. 4, A and B, and table S4) indicated early infiltration of macrophages (detected with F4/80 antibody) and mast cells (toluidine blue staining), followed by dermal T cells (CD3) and eosinophils [major basic protein (MBP)] in KO skin. B cells (CD45R) were not differentially recruited or retained in the skin of the two genotypes. Infiltration of neutrophils occurred later, and detectable myeloperoxidase (MPO) levels were observed after 2 weeks (Fig. 4, C and D). Neutrophils accumulated in numerous pustules were often associated with keratin plug formation and hair follicle destruction (Fig. 4D). Cultures of skin scrapings for bacteria and fungi were performed on multiple young and old WT and KO mice from two background strains (table S1). Several bacterial species were detected in scrapings from the skin of young and old mice of both genotypes of both backgrounds without evidence of a pattern or predominant pathogen associated with skin phenotype. No fungi were isolated.

Fig. 4 Time course analysis of inflammatory infiltrate in EGFR-ablated mouse skin.

(A and B) Immunohistochemical analysis of skin-infiltrating cells was performed: F4/80 (macrophages), toluidine blue (mast cells), CD3 (T cells in the dermis and γδ T cells in the epidermis), CD45R (B cells), and MBP (eosinophils). The cells were counted in ten 200× fields at the indicated ages (A) for each mouse and expressed as the mean cell counts ± SD (n = 3 mice; *P ≤ 0.05 by t test; individual P values listed in table S3). White bars, WT; black bars, KO. (B) Representative 100× fields at day 7 in WT and KO mouse skin. Scale bars, 100 μm. (C) MPO assay of skin biopsies isolated from the back of WT (blue) and KO (red) mice. The enzyme activity is monitored as changes in OD (optical density) over time. MPO activity is not detectable earlier than 2 weeks in both WT and KO mice. P = 0.011 at 2 weeks, P = 0.039 at 3 weeks. n = 4 mice per time point per genotype; P calculated by t test. (D) MPO staining in sections from back skin and bone marrow of 3-month-old WT and KO mice. Red arrows are pointing to microabscesses. Scale bars, 50 μm.

Local overexpression of cytokines and chemokines is an early event in the EGFR-ablated skin

We examined early changes in chemokines and their receptors to provide insights into the character of the skin infiltrates. Real-time polymerase chain reaction (PCR) of skin mRNA showed up-regulation of a subset of inflammatory mediators. In particular, Ccl5, Ccl11, Ccl2, Ccl22, Il1β, Tnfα, and Nos2 were expressed at higher levels in EGFR-ablated skin within the first week of life (Fig. 5 and table S4). Ccl5 was one of the earlier mediators to be up-regulated in KO skin. The mRNA for Ccl5 receptors Ccr5, Ccr3, and Ccr1 (Fig. 5 and table S4) was also elevated in KO skin starting at 3 days (Ccr5) and further increased at 1 week (Ccr3 and Ccr5) and 2 weeks (Ccr1, Ccr3, and Ccr5). Chemokine receptors Ccr1 and Ccr5 are expressed on macrophages, and we observed those in higher numbers in EGFR-ablated skin at 3 days (Fig. 4A and table S4). Supernatants from keratinocytes isolated from EGFR-null mice (22) showed enhanced basal and cytokine [cocktail of IL-1β, IL-6, IL-17, and tumor necrosis factor–α (TNF-α)]–induced CCL5 protein expression when compared with supernatants from cells isolated from WT littermates (fig. S3A and table S4). We tested these supernatants as chemotactic stimuli on the macrophage cell line RAW264.7. Supernatants from EGFR-null keratinocytes, both untreated and after stimulation with a cocktail of cytokines, showed a stronger chemotactic activity toward macrophages (fig. S3, B and C). In the presence of CCL5-neutralizing antibodies or a combination of CCL5 receptor inhibitors (DAPTA and UCB35625), the chemotactic activity induced by EGFR-null supernatants was significantly reduced (P < 0.0001; fig. S3D). Besides Ccl5, another ligand promoting the infiltration of Ccr3-positive cells (mast cells and eosinophils) is Ccl11, which is expressed at high levels in 3-day-old KO skin (Fig. 5 and table S4). Isolated keratinocytes do not release detectable levels of CCL11 in supernatants in vitro (fig. S3A and table S4). We performed CCL11 immunohistochemistry combined with toluidine blue staining and observed high expression of this chemokine in dermal mast cells (fig. S3E). Although levels of both Ccl2 and its cognate receptor Ccr2 were highly expressed in KO skin starting from the first week of life, abundant levels of the chemokine Ccl22 were not associated with enhanced expression of the corresponding receptor Ccr4 (Fig. 5 and table S4). Many of the inflammatory mediators described as elevated in the skin in the early phase were persistently elevated in the circulation of older KO mice (Fig. 3A, fig. S4, and table S4).

Fig. 5 Time course of cytokine and chemokine changes in EGFR-ablated mouse skin.

SYBR Green–based real-time PCR analysis of cytokines, chemokines, and chemokine receptors expressed over time after birth in mouse skin. Values are expressed as fold induction over WT values at 24 hours, normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Mean expression ± SD; n = 4 mice per genotype per time point; *P ≤ 0.05 by t test of KO (black) versus corresponding WT (white); individual P values listed in table S3.

Macrophage recruitment and hair phenotype in KO mice are not altered by crosses with mice deficient for CCR2, MyD88, TNFR1/2, B and T cells, and NOS2

In the attempt to neutralize or ameliorate the skin inflammatory phenotype as a potential therapeutic strategy, we crossed the epidermal targeted EGFR KO mice with several strains of mice defective in major pathways regulating immune signaling: TNFR1/R2−/− (23), MyD88−/− (24), NOS2−/− (25), CCR2−/− (26), and Rag1−/−. None of these regulatory mutants prevented or reversed the induction or maintenance of the skin phenotype in the EGFR KO mice (Fig. 6). The persistent infiltration of F4/80 macrophages recruited in the skin of the multiple mutants in the absence of EGFR (Fig. 6) prompted us to determine the contribution of this cell type to the skin phenotype.

Fig. 6 Macrophage recruitment and hair phenotype in EGFR KO mice crossed with mice deficient for CCR2, B and T cells, MyD88, TNFR1/2, and NOS2.

Representative F4/80 (left panels) and H&E staining (right panels) in skin sections of TgK5Cre(WT/+)/Egfr(f/f)/CCR2−/−, TgK5Cre(WT/+)/Egfr(f/f)/Rag1−/−, TgK5Cre(WT/+)/Egfr(f/f)/MyD88−/−, and TgK5Cre(WT/+)/Egfr(f/f)/TNFR1/2 −/− (all on a C57BL/6 background) and TgK5Cre(WT/+)/Egfr(f/f)/NOS2−/− (on an FVB/N background). All mice were 21 days old. Objective, 10×. Scale bars, 100 μm. Images were captured by staining skin sections from n ≥ 3 independent litters per genotype.

Macrophage depletion improves KO skin phenotype

Reducing macrophages by subcutaneous injection of liposomes containing clodronate (a bisphosphonate) (27) caused changes in the architecture of KO skin with more parallel growing hair follicles, less dermal cellularity, and fewer keratin plugs (Fig. 7A and fig. S5, A and B). Those effects are visible as a gradient radiating from the skin immediately adjacent to the injection site (red arrow, fig. S5B) to more distal areas. The reduction of F4/80-positive cells (Fig. 7B) was associated with a decrease of keratin 1 levels in skin of clodronate-treated KO mice. Clodronate also influenced WT skin by stimulating an anagen-like phenotype (elongated hair follicles and expanded dermal compartment), suggesting a role for macrophages in maintaining telogen (Fig. 7B).

Fig. 7 Macrophage depletion and changes in KO skin phenotype and differentiation marker expression.

(A) H&E staining in skin sections from WT and KO littermates treated with PBS liposomes or clodronate liposomes. Objective, 20×. Scale bars, 100 μm. (B) Immunohistochemistry for F4/80 macrophages and keratin 1 in mice treated as in (A). Objective, 20×. Scale bars, 100 μm. Images were captured by staining skin sections from n = 3 independent litters treated with clodronate injections.

Macrophage depletion alters the expression of inflammatory genes and reduces epithelial differentiation markers in KO mouse skin

RNA from clodronate-treated KO skin was compared to RNA from KO skin treated with phosphate-buffered saline (PBS) liposomes. Changes in RNA expression of inflammatory and differentiation markers were expressed as percentage of change (clodronate/PBS) to globally visualize factors that may contribute to the amelioration of the phenotype (Fig. 8, A and B) and factors that may prevent further normalization (Fig. 8C). The levels of chemokine receptors (Ccr1, Ccr3, and Ccr5), chemokines (Ccl11), and cytokines (Il17), highly expressed in 3-week-old KO skin, were normalized or reduced in clodronate-treated animals (Fig. 8A). A similar trend was followed by epithelial differentiation markers (Fig. 8B). Clodronate-induced macrophage apoptosis also stimulated some unwanted adverse effects like enhanced presence of neutrophils (Cxcr2 and its ligands Cxcl1 and Cxcl2); higher expression of Il6, Icam1, and Nos2; and superinduction of Gcsf expression (Fig. 8C) (28). Clodronate treatment also reduced infiltrating mast cells (fig. S5C). Although there were some quantitative differences, the responses to clodronate/liposomes were qualitatively similar in mice of both FVB/N and C57BL/6 backgrounds.

Fig. 8 Differential effect of macrophage depletion on mRNA in EGFR KO mouse skin.

(A to C) SYBR Green–based real-time PCR analysis of chemokines, chemokine receptors, inflammatory mediators, and epithelial differentiation markers. Values are expressed as percent change in clodronate-treated KO compared to PBS-treated KO. (A and B) Values below zero on the y axis represent a decrease in KO mRNA levels after treatment with clodronate, with −100% corresponding to WT levels. (C) Changes with positive values (above zero on the y axis) represent a further increase in the mRNA expression in KO skin upon clodronate administration. No effect of clodronate is represented in values around zero. Experiments were performed on littermates from two litters of C57BL/6 (black circles) and one litter of FVB/N mice (empty circles).


Although the clinical data reported here are limited by the small number of patients, and thus the lack of statistical power for some of the endpoints, the study has the advantages of monotherapy with gefitinib and both baseline and defined after-treatment measurements to correlate with the more dynamic mouse studies. Similarly, the advantage of the mouse model is its ability to use epidermal targeting to distinguish primary skin events from events caused by systemic anti-EGFR therapy. The disadvantage of the mouse model is that it lacks the setting of neoplastic disease characteristic of chemotherapy. Mouse studies have shown that cutaneous EGFR activity is important for the proper development and cycling of the hair follicles (810) and for maintaining cutaneous immune homeostasis (11, 12, 29). Thus, the adverse skin phenotype characteristic of cancer patients on EGFR inhibitory drugs is consistent with previous mouse studies. However, data illuminating the underlying pathways involved in either species are sparse. The human studies presented here suggest that individual variation in the baseline concentration of chemokines or cytokines, because of disease, therapy, or individual genetic variation, could potentially influence the severity of adverse skin reactions. Confirmation of this finding in a larger group of patients could be helpful in predicting those patients who are most likely to experience cutaneous reactions. However, the similarity in skin phenotype, hematological changes, and common chemokine and cytokine alterations in these patients and KO mice suggests that the mouse model will be useful to further dissect the mechanism that causes the skin adverse effect of anti-EGFR drugs. Our genetic approaches to inhibit the skin phenotype with complementary knockout of a proinflammatory pathway demonstrated the complexity of the inflammatory response that follows EGFR blockade. The neutralization of a single pathway was not sufficient to ameliorate the phenotype of KO mice. It is likely that a subset of inflammatory mediators acts in concert during lesion formation and that infiltrating cells respond to multiple stimuli at the same time. The antagonists of both TNF-α and IL-1 pathways have been suggested to alleviate the skin lesions induced by an anti-EGFR antibody in mice (30). Crossing our KO mice with mice lacking MyD88 or TNFR1/2 indicated that these pathways are not essential for the development and exacerbation of the lesions.

From the mouse data, it appears that macrophages and mast cells are early responders and actively contribute to the lesion formation. This conclusion was confirmed in a companion paper using similar approaches (31). Because these cells are part of the infiltrate documented in the skin of treated patients (6, 32), they, or their attractants, could be clinical targets for prevention or resolution of the rash. Both bone osteoclasts and macrophages are targeted by bisphosphonates, a class of drugs approved for clinical use to prevent bone fractures in cancer patients (33). Future studies in patients treated with EGFR inhibitors could assess a possible improvement of the skin lesions upon bisphosphonate administration. Patients’ lesions are currently managed with anti-inflammatory and immunomodulatory agents. Tetracyclines, administered to patients to prevent superinfection of the lesions, might function in a dual mode by modulating macrophage differentiation as shown in other systems (34). Another aspect illuminated by our model is the early induction of the inflammatory response followed by the development of the aberrant epithelial differentiation program. In patients, protein alterations involved in cornified envelope formation and the development of skin xerosis are not an immediate consequence of EGFR blockade but develop later in the areas of the papulopustular rash (7, 35). Thus, EGFR blockade impairs the proper differentiation of epithelial cells and skin appendages both in a cell-autonomous way and via the modulation of the immune cell recruitment and activity. Interfering with this mechanism improves the adverse effects caused by EGFR ablation, showing the cross talk of the two mechanisms.

Recent observations highlight how the EGFR-driven control of inflammatory homeostasis is restricted not only to the skin compartment but also to colon epithelium. Diarrhea is another common adverse reaction to anti-EGFR therapy. EGFR expression is necessary to prevent excessive local inflammation in experimental colitis (36, 37). Moreover, a syndrome of inflammatory skin and bowel disease in children (13) has been linked to a loss-of-function mutation in ADAM 17, an EGFR ligand sheddase. Thus, the EGFR pathway may control inflammatory homeostasis in multiple lining epithelia, requiring intense effort to determine how to counteract dysregulation of the pathway in nontarget tissues with expanding use of EGFR inhibitors in the clinical setting.


Study design

Patient sample study. Full details of the phase 2 gefitinib monotherapy trial (NCT00049556) in ovarian cancer patients have been reported previously (18). The protocol mandated monthly blood collection and punch biopsy of tumor tissue before the trial and at 1 month. Human samples were all blinded for assays reported here. For this analysis, patients had to have remaining matched aliquots of pretreatment and 1-month plasma. These aliquots were provided with unique patient identifiers after Institutional Review Board (IRB) approval for their use; cytokine results were unblinded and matched to clinical findings (toxicity and clinical laboratory results) by the clinical team, and the reblinded data set was provided to the statistician. Samples had been isolated and frozen at −80°C within 4 hours of receipt from the patient, and the aliquots provided had never been thawed. Patients are presented according to rash National Cancer Institute (NCI) Common Toxicity Criteria v2, with grade 1 signifying limited acneiform or maculopapular rash and limited or no symptoms, and grade 2 representing more extensive rash and/or plaques and symptoms and/or requiring medical intervention. Table S2 provides limited demographic characteristics of the patients. All clinical investigation has been conducted according to Declaration of Helsinki principles. Human studies have been approved by the IRB of the Clinical Center for Cancer Research. Written informed consent was received from participants before inclusion in the study.

Mouse study. Mouse studies were performed under a protocol approved by the NCI and the National Institutes of Health (NIH) Animal Care and Use Committee. The epidermal deletion of EGFR was achieved by crossing Keratin5/promoter-driven Cre recombinase (19) male transgenics with female mice in which loxP sites were inserted flanking exon 3 of the EGFR gene (20). Double-transgenic K5CreWT/+/EGFR flx+/+, indicated in the text and figures as KO, were compared with K5CreWT/WT/EGFR flx+/+ or K5CreWT/+/EGFR flx+/− littermates, indicated in the text and figures as WT. Mice were observed on a C57BL/6 and FVB/N background. In the text, all the data are from C57BL/6 mice unless otherwise stated. No blinding was performed because of the visibly different phenotype of EGFR-deficient animals and controls. For rescue experiments, we backcrossed our mice on a TNFR1/R2−/− (23), MyD88−/− (24), NOS2−/− (25), CCR2−/− (26), or Rag1−/− (Jackson Lab) background. All of these strains were on C57BL/6 background, except the NOS2 mice, which are FVB/N. Mice were genotyped with small distal tail samples. Taq polymerase mixes were used according to the manufacturer’s instructions, with Platinum PCR SuperMix (Invitrogen) used for EGFRlox3, NOS2, MyD88, and TNFR primers sets and Taq Master Mix (Qiagen) used for Cre, CCR2, and RAG1 sets. GAPDH was used as a housekeeping gene with Cre genotyping to ensure that adequate genomic amplification occurred. PCR products were run with standard electrophoresis methods. The list of primers used is in the Supplementary Materials. Design of individual experiments is reported below.

Clodronate treatment

Clodronate liposomes and control PBS liposomes were purchased from Encapsula NanoSciences. Clodronate (50 μl) or PBS liposomes were injected subcutaneously in the upper and lower dorsum of mice every other day from 6 to 18 days. At 21 days, dorsal skin sections spanning from the neck to the tail were taken for histological analysis as well as for skin RNA preparation.

Mouse peripheral blood analysis

The mice were anesthetized with Nembutal (Abbott). Blood for differential white blood cell counts was collected by intracardiac puncture with heparinized syringes to prevent coagulation. Total blood cell counts were performed at the Department of Laboratory Medicine at the NIH with an automated hematology analyzer (CELL-DYN 3700, Abbott).

MPO assay

Back whole skin samples were used for MPO assays as previously described (38). In brief, 6-mm punch biopsies taken from the back skin of mice at different ages were homogenized in potassium phosphate buffer (pH 6.0) containing 0.5% hexadecyltrimethylammonium bromide, sonicated, and freeze-thawed three times, after which sonication was repeated. The suspension was centrifuged at 40,000g for 15 min, and 10 μl of supernatant was added to 290 μl of potassium phosphate buffer (pH 6.0) containing o-dianisidine dihydrochloride (0.167 mg/ml) (Sigma-Aldrich) and 0.0005% hydrogen peroxide. Changes in OD were monitored at 460 nm at 25°C over a 4-min period.

Dye permeation assay

For toluidine blue staining, newborn mice were sacrificed and immediately dehydrated by sequential incubation in 25, 50, 75, and 100% methanol. After rehydration in PBS, they were incubated for 10 min in 0.01% toluidine blue and destained with PBS.

Multiplex biomarker analysis

Concentrations of cytokines and chemokines in plasma of patients or mice or cell supernatants were measured with Aushon protein arrays (Aushon Biosystems). Samples were incubated on array plates prespotted with capture antibodies specific for each protein biomarker. Plates were decanted and washed before cocktail of biotinylated detection antibodies was added to each well. After incubating with detection antibodies, plates were washed and incubated with streptavidin–horseradish peroxidase conjugate. Plates were again washed before adding a chemiluminescent substrate. The plates were immediately imaged with the Aushon charge-coupled device imaging system, and data were analyzed with Aushon software. Concentrations were interpolated from a standard curve.

Cell cultures

Mouse keratinocytes were isolated from WT and EGFR-null mice (22) and cultured as extensively described in (39). Twenty-four hours before stimulation, cells were starved with medium containing 0.2% serum and then treated with a cocktail of cytokines (IL-1β, IL-6, IL-17, and TNF-α, all 5 ng/ml) for an additional 24 hours. RAW264.7 cells were obtained from the American Type Culture Collection and maintained in 8% serum/Dulbecco’s modified Eagle’s medium. Subconfluent cultures were starved with 0.2% serum 24 hours before being harvested for the chemotactic assay.

Migration assay

Twenty-six microliters of keratinocyte medium was placed in the bottom of the well of a 48-well microchemotaxis plate (Neuro Probe). A framed filter (5 μm) was then placed over the microplate, and 50 μl of the prepared cells (10,000) was placed on the top of the filter. For experiments involving inhibitors, cells were pretreated for 30 min with the indicated inhibitor before being placed in the chamber wells. The plate was then incubated at 37°C in humidified air with 5% CO2 for 45 min. After incubation, a clean smooth edge wiper was used to quickly wipe the nonmigrated cells from the top of the filter. The filter was then fixed and stained according to Protocol Hema 3 Stain Set (Biochemical) and placed between two glass slides. ImageJ software was used to calculate densitometry on a scanned image of the filter. CCL5-neutralizing antibody and normal goat immunoglobulin G (IgG) (R&D Systems) were used at 200 ng/ml; UCB35625 (or J-113863), an inhibitor of CCR1/3, was used at 5 μM; and DAPTA, an inhibitor of CCR5, was used at 50 nM (Tocris).

Immunohistochemical analysis

Tissues were fixed in 4% neutral buffered formalin and embedded in paraffin. Antigen retrieval was performed with target retrieval solution (citrate, pH 6, DAKO) for CD3, CD45R/B220, keratin 1, keratin 6, keratin 10, filaggrin, loricrin, MPO, and proteinase K (DAKO) for F4/80 and pepsin for MBP. Antigen retrieval was not performed for CD45. Slides were treated with 3% hydrogen peroxide for 15 min and blocked for a minimum of 30 min with serum-free protein block (DAKO). Sections were incubated for 1 hour at room temperature (CD3, CD45R/B220, F4/80, MBP, filaggrin, keratin 1, keratin 10, loricrin, and keratin 6) or overnight at 4°C (CD45 and MPO) with primary antibodies. Slides were treated for 1 hour at room temperature with the corresponding biotinylated secondary antibody. Primary and secondary antibodies were diluted in antibody diluent with background reducing components (DAKO). Slides were then incubated for 20 min at room temperature with the Vector R.T.U. ABC system. PBS washes were performed between each step. ImmPACT DAB Vector was used for staining development. Counterstaining was performed with Hematoxylin QS Vector. All slides were dehydrated and mounted with VectaMount Vector. Primary antibodies and dilutions used were as follows: CD3 (AbD Serotec MCA1477, 1:50), CD45 (R&D Systems AG114, 1:200), F4/80 (eBioscience 14-4801-82, 1:100), keratin 1 (Covance PRB-165P, 1:1000), keratin 6 (Covance PRB-169P, 1:500), keratin 10 (Covance MMS-159S, 1:1000), filaggrin (Covance PRB-417P, 1:500), loricrin (Covance PRB-145P, 1:1000), and MBP antibody (J. Lee, Mayo-Rochester, 1:100). For immunohistochemical analysis, skin sections of three mice for each time point were analyzed. Positive cells were counted in 10 independent fields per slide/mouse and averaged.

Toluidine blue

Slides were stained for mast cells by immersing the sections in toluidine blue working solution (50 mg of toluidine blue O, 5 ml of 70% ethanol, 45 ml of 1% sodium chloride, pH 2.0 to 2.5) for 3 min. Slides were then washed with dH2O.

Skin preparation and protein lysates

Mouse skin was isolated, snap-frozen in liquid nitrogen, and cold-processed at 2000 rpm in a Mikro-Dismembrator S (Sartorius). Protein lysates were obtained from powderized skin with cold radioimmunoprecipitation assay [50 mM tris-HCl (pH 8), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100] lysis buffer supplemented with 1 mM sodium orthovanadate, cocktail of anti-proteases (Roche complete, 1836145), and 1 mM sodium fluoride. Lysates were vortexed and incubated on ice for 10 min and then separated into two fractions. One fraction was centrifuged at 14,000 rpm for 15 min with supernatants quantified with the Bradford method. The second fraction was supplemented with β-mercaptoethanol (1.5 M) and SDS (5%) to optimize keratin and cornified envelope protein extraction.

Immunoblot analysis

Protein samples were analyzed according to standard immunoblotting protocols with 10% tris-HCl gels and nitrocellulose membranes from Bio-Rad. Protein (15 μg) was loaded to detect the differentiation markers, and 70 μg of protein was loaded to detect CD45. All primary antibodies were diluted in 1% bovine serum albumin in tris-buffered saline/Tween 20 and incubated overnight at 4°C. Primary antibodies were the same as those used for immunohistochemistry. Chemiluminescent substrates were purchased from Pierce Biotechnology.

Quantitative reverse transcription PCR

Total RNA was isolated from powderized skin with TRIzol Reagent (Invitrogen) and cleaned of genomic DNA contamination by RNeasy Mini Kit (Qiagen). For complementary DNA (cDNA) synthesis, 1 μg of total RNA was reverse-transcribed with SuperScript III Reverse Transcriptase (Invitrogen). The cDNA was diluted 1:100, and real-time PCR was performed with Bio-Rad iQ SYBR Green SuperMix on myIQ and iQ5 PCR systems (Bio-Rad). Relative expression was calculated with the ΔΔCT method with GAPDH as the housekeeping gene. Melt curves were analyzed from 55 to 95°C at 0.5°C increments to confirm specificity of amplification products. The list of primers used is in the Supplementary Materials.

Statistical analysis

For mRNA, cell counts, migration assays, MPO assays, mouse plasma assays, and supernatant assay, the data were tested for differences with a two-tailed two-sample t test. For these analyses, statistical significance was identified when P ≤ 0.05, and single P values correspondent to * in the figures are reported in table S3. For patient plasma analyses, the difference between groups of patients classified by their rash grade or presence or absence of pruritus was determined by a Wilcoxon rank-sum test. Given the small number of patients and the large number of exploratory tests performed on the parameters, P < 0.01 was considered statistically significant, whereas 0.01 < P < 0.10 reflected trends in the differences. For this group, individual P values are discussed in the Results section and in Fig. 1.


Materials and Methods

Fig. S1. Epidermal differentiation markers and total CD45 protein in WT and KO skin.

Fig. S2. Epidermal differentiation markers and CD45-positive leukocytes in mouse skin lacking EGFR.

Fig. S3. Supernatants of cultured keratinocytes and macrophage chemotaxis.

Fig. S4. Time course analysis of cytokine and chemokine changes in mouse skin after epidermal EGFR ablation.

Fig. S5. Macrophage depletion and KO skin phenotype.

Table S1. Bacteria detected on mouse skin.

Table S2. Demographic characteristics of the patients involved in the study.

Table S3. P values of data highlighted as significant with asterisks.

Table S4. Individual data values (Excel file).


  1. Acknowledgments: We thank J. Lee from Mayo Clinic Arizona for the MBP antibody, M. Custer for her excellent assistance with the mouse colony, and L. Wright for the maintenance of the EGFR-null colony. We are thankful to H. Kong from the Dermatology Branch (NCI) for providing information on patient’s samples and to M. Bryant and F. Benedetti from the Division of Veterinary Resources, Office of Research Services, for their help with the mouse phenotype and bacteria/fungi cultures. The authors are also grateful to J. Udovich from Carl Zeiss Microscopy for his help in scanning and elaborating the images in the extended field of view data. Funding: This work was supported by the Center for Cancer Research Intramural Program of the NIH. Author contributions: F.M. designed and performed the experiments, analyzed the data, and wrote the manuscript; G.L. and C.K. performed the experiments and analyzed the data; C.G. performed the experiments; S.M.S. analyzed the human data; E.K. provided patient samples and assisted with analysis of clinical data; and S.H.Y. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: MBP antibody was obtained with a material transfer agreement from Mayo Clinic, Arizona, and can be requested from J. Lee; mice can be requested from D. Threadgill at North Carolina State University (EGFR flx) and J. Jorcano at Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas in Madrid (K5Cre), or from us with their approval.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article