Keratin 1 maintains skin integrity and participates in an inflammatory network in skin through interleukin-18

The epidermis protects an organism against mechanical injury and dehydration, and regulates immune homeostasis by virtue of epidermal keratinocytes (Simpson et al., 2011). Many of its functions depend on structural proteins including keratins (Kim and Coulombe, 2007; Magin et al., 2007). Keratins constitute the intermediate filament cytoskeleton in all epithelia (Schweizer et al., 2006). The type II keratin KRT1 and its heterodimer type I partner KRT10 form the major cytoskeleton in suprabasal keratinocytes. Their upregulation precedes expression of filaggrin and cornified envelope proteins. Covalent crosslinking of Krt1 but not Krt10 to these proteins, in addition to a complex series of lipids, ultimately forms the cornified/lipid envelope, which, together with Langerhans cells, constitutes the epidermal barrier (Candi et al., 2005; Simpson et al., 2011). This led to the hypothesis that a subset of keratins and the associated protein filaggrin are crucial for barrier function. In fact, KRT1 and KRT10 mutations lead to congenital epidermolytic ichthyosis (EI, MIM 113800) characterized by skin erosions, hyperkeratosis and barrier defects (Arin et al., 2011; Lane and McLean, 2004; Schmuth et al., 2001; Segre, 2006), whereas mutations in filaggrin can cause the inflammatory disorder atopic eczema (AE) (Brown and McLean, 2012; Oji et al., 2010). EI is characterized by skin erosions, hyperkeratosis and barrier defects (Lane and McLean, 2004; Schmuth et al., 2001; Segre, 2006). Although these disorders have established a primary role of keratinocyte-resident proteins in acute and chronic skin diseases (Brown and McLean, 2012; Nestle et al., 2009; Quigley et al., 2009), the underlying pathomechanisms remain poorly understood.

Besides immune cells, keratinocytes regulate skin inflammatory and immune responses by secretion of cytokines, antimicrobial peptides and by expression of MHCII proteins (Nestle et al., 2009). S100A8 and A9 proteins belong to the S100 family of calcium-binding proteins that can act as antimicrobial peptides and are released as heterodimers upon barrier defects by unconventional secretion. Once released, they act as autocrine activators of Toll-like receptor (TLR) 4 on keratinocytes (Eckert et al., 2004; Ehrchen et al., 2009; Nestle et al., 2009; Vogl et al., 2007). They are elevated in psoriatic skin and therefore are linked to inflammatory skin diseases (Nestle et al., 2009). Interleukins (IL) IL-18 and IL-33 are pro-inflammatory cytokines of the IL-1 family produced by suprabasal keratinocytes (Dinarello, 2009). Skin-specific overexpression of IL-18 in mice results in AE-like inflammatory lesions (Konishi et al., 2002), and IL-18 levels are elevated in humans suffering from AE (Kou et al., 2012). IL-18 is synthesized with an N-terminal propeptide requiring cleavage by caspase-1 or other proteases before unconventional secretion from a variety of cells, including keratinocytes (Dinarello, 2009). Activation of caspase-1 is tightly regulated by the inflammasome, a large cytoplasmic multiprotein complex consisting of scaffold proteins such as NLRP3 in keratinocytes, the adaptor protein ASC (apoptosis-associated speck-like protein containing a caspase recruitment domain) and procaspase-1/-5 (Davis et al., 2011). The mechanisms controlling inflammasome activation are partially understood and include TLRs, sensing either pathogen-associated molecular patterns (PAMPs) or NOD and RIG receptors, recognizing self-derived danger signals, including nucleic acids, ATP, cholesterol crystals and amyloid beta (Davis et al., 2011). An intriguing question is whether the keratinocyte cytoskeleton is involved in intracellular damage control.

To test this hypothesis, we have generated Krt1^−/− mice. In contrast to Krt10 deficiency, which led to viable mice with a mild phenotype (Reichelt et al., 2001), loss of Krt1 caused perinatal lethality and a barrier defect. Transcriptional profiling revealed a gene expression signature in Krt1^−/− skin similar to the human inflammatory skin diseases AE and psoriasis. Depletion of IL-18 partially rescued Krt1^−/− mice. Our data identify keratin proteins as gatekeepers of immune responses in skin and link cytokine release to the loss of KRT1.

To address isotype-specific keratin functions during epidermal differentiation and barrier formation (Kim and Coulombe, 2007; Magin et al., 2007) we generated Krt1-deficient mice (Fig. 1B,D; supplementary material Fig. S1B). Surprisingly, the absence of Krt1 caused neonatal lethality in a mixed 129/Ola×C57BL/6 genetic background, unlike Krt10 depletion (Reichelt et al., 2001). Histology of neonatal Krt1^−/− skin revealed a largely intact stratified epidermis, without local inflammation and localized keratinocyte lesions, possibly from handling-induced trauma. Further, the granular layer was reduced (Fig. 1A). Expression of Krt10, the obligate heterodimer protein partner of Krt1, was significantly reduced in Krt1^−/− skin extracts, but its intracellular distribution remained unaltered (Fig. 1C,D). Electron microscopy revealed sparse keratin aggregates in Krt1^−/− skin due to loss of Krt1, indicating lack of compensatory keratins, and normal filaments in wild-type epidermis (supplementary material Fig. S1A). Krt5 and Krt14, typically restricted to basal keratinocytes, were present in suprabasal Krt1^−/− skin, and their protein level was slightly elevated, similar to other epidermis-specific knockouts (supplementary material Fig. S1C–E). In Krt10^−/− mice, no Krt1 aggregates were detectable, but atypical intermediate filament between Krt1 and Krt14 were present and contributed to skin integrity (Reichelt et al., 2001). To address whether Krt10 assembled with type II keratins Krt5 or Krt6 in Krt1^−/− epidermis, high resolution confocal imaging was performed. This revealed the presence of numerous Krt10 aggregates without Krt5 or Krt6 in spinous layer keratinocytes, indicating that upon expression of Krt10, no type II partner is available to form intermediate filaments (Fig. 2A,B). In the absence of Krt1, the distribution and staining intensity of major desmosomal proteins desmoplakin, desmoglein1 and 2 and plakoglobin were unaffected (Fig. 2C–E). This suggested that both presence and intracellular organization of Krt5/14 filaments in suprabasal epidermis maintained desmosome functionality. Distribution and expression levels of cornified envelope (CE) proteins loricrin, involucrin and filaggrin (supplementary material Fig. S1F; Fig. S2B, and data not shown) appeared unaltered, supporting occurrence of terminal differentiation. A strong hint on disturbed skin homeostasis came from the appearance of stress-associated Krt6 and Krt16, induced upon hyperproliferation, barrier defects and wound healing (Kim and Coulombe, 2007) (Fig. 1E; supplementary material Fig. S1C). The mild skin defects of Krt1^−/− mice and the notion that Krt1^−/−, in contrast to Krt10^−/− mice, are not viable (Reichelt et al., 2001; Reichelt and Magin, 2002), point to a novel and unique function of Krt1 in vivo. This is also in agreement with the recently described phenotype of Krt1/10 doubly deficient mice (Wallace et al., 2012).

The known participation of KRT1, but not KRT10 to cornified envelope formation (Candi et al., 2005; Candi et al., 1998) and a mild weight loss of Krt1^−/− mice that manifested postnatally (supplementary material Fig. S2A) prompted us to examine epidermal barrier integrity employing an established dye-penetration assay (Segre et al., 1999). This established an intact outside-in epidermal barrier in Krt1^−/− newborns, as no dye penetrated the skin (Fig. 3A). In contrast, trans-epidermal water loss (TEWL) was increased twofold in neonatal Krt1^−/− mice, revealing a defective inside-out barrier (Fig. 3D). Possibly, the biotin assay performed in Krt1/10-doubly-deficient mice may not fully reveal the barrier state of these mice (Wallace et al., 2012). Furthermore, the number of intact cornified envelopes was significantly reduced to ∼17% in Krt1^−/− compared to ∼84% in Krt1^+/+ and ∼55% in Krt10^+/+mouse skins (Fig. 3B,C,E; supplementary material Fig. S2C,D). Together with the notion that in Krt1/10 doubly deficient mice 60% of cornified envelopes remained intact (Wallace et al., 2012), our data substantiated a major role of the type II Krt1 in barrier integrity. To address the known role of tight junctions which contribute to barrier functionality by sealing the paracellular spaces between granular layer keratinocytes (Kubo et al., 2012), tight junction protein expression and distribution were examined. The unaltered expression of claudin-1 (Cldn), Cldn4 and Cldn5 (Fig. 3F) and of ZO-1 and occludin (data not shown) in Krt1^−/− mice further supported a primary defect directly related to Krt1.

To unravel previously unnoted and essential functions of Krt1 at the molecular level, we performed global gene expression analysis on back skin from Krt1^−/− mice and Krt1^+/+ littermates at the onset of Krt1 expression [embryonic day (E)15.5], at E18.5 and at postnatal (P) day 0. At E15.5, when Krt1 expression commences during mouse development, Krt1 itself was the only gene differentially expressed in the two genotypes (data not shown). At E18.5, we found 163 genes upregulated and only two genes downregulated in Krt1^−/− skin. In newborn skin, 627 genes were upregulated and 284 were downregulated in Krt1^−/− mice. Gene set enrichment analysis, followed by gene ontology classification and grouping according to function (Keller et al., 2008) of genes upregulated in Krt1^−/− skin revealed a similar distribution of function categories in E18.5 and newborn samples (supplementary material Fig. S3A,B). Strikingly, cluster analysis and heat maps identified a significant over-representation of genes linked to CE formation/epidermal differentiation (Fig. 4A,C) and inflammatory/immune defense pathways (Fig. 4B,D) in E18.5 and newborn Krt1^−/− skin, demonstrating that both gene sets were altered prenatally. To validate array data, selected candidate gene expression was confirmed by quantitative RT-PCR (supplementary material Fig. S3C–E). Prenatal upregulation of pro-inflammatory genes strongly supported a direct involvement of Krt1 in the regulation of the above pathways. The pattern of genes upregulated in Krt1^−/− skin, e.g. Il-1β, IL-18, defensins and S100 proteins, bore resemblance to the human inflammatory skin diseases atopic eczemia (AE) and psoriasis, characterized by barrier defects, inflammation and immune dysregulation (Barnes, 2010; Gudjonsson et al., 2010; Gudjonsson et al., 2009; Sääf et al., 2008; Suárez-Fariñas et al., 2010; Suárez-Fariñas et al., 2011). We then focused on IL-18, as cytokine analysis of E18.5 serum samples confirmed its elevation (see below, Fig. 5A). To substantiate this further, global comparisons between the Krt1 data set (our study) and genome-wide transcriptional profiling data sets derived from AE and psoriasis patients were performed (Gudjonsson et al., 2010; Sääf et al., 2008). The intersection of similarly regulated genes in Krt1^−/− skin and AE or psoriatic skin was remarkably high: 55 (E18.5) and 154 (P0) genes similar to AE skin, and 86 (E18.5) and 166 genes (P0) similar to psoriatic skin (Fig. 4E; supplementary material Tables S1, S2). Genes similarly regulated in Krt1^−/− and AE skin are IL-18, S100A8/A9, Il6/Il6R, SOCS3, KLK, TGM and SPRR (supplementary material Tables S1, S2), genes that have been described to be associated with AE (Barnes, 2010; Sääf et al., 2008; Suárez-Fariñas et al., 2011). In analogy, genes upregulated in Krt1^−/− and psoriatic skin were compared. This uncovered a group of genes previously ascribed to psoriasis, including epidermal barrier-coding genes, e.g. SPRRs, S100, KLKs, IL-1 family members and genes of lipid metabolism (ELOV and fatty acid binding protein, FABP) (Gudjonsson et al., 2010; Gudjonsson et al., 2009; Suárez-Fariñas et al., 2010) (supplementary material Tables S1, S2). In striking contrast, the comparison of genes upregulated in Krt1^−/− versus Krt5^−/− mouse skin, which serves as a model for the keratin-associated disease epidermolysis bullosa simplex (EBS) was very low, with 14 genes at E18.5 and 23 genes at P0 (Lu et al., 2007) (Fig. 4E). Furthermore, only 11 upregulated genes at E18.5 and 18 genes at P0 in Krt1^−/− skin were common to human EBS patient skin (Bchetnia et al., 2012). Based on these data we conclude that the gene expression signature from Krt1^−/− mice shows more similarities to human inflammatory skin diseases than to keratin-associated defects in mice and humans (supplementary material Table S3).

To further dissect the role of Krt1 in inflammation, gene set enrichment analysis (GSEA) was performed to identify commonly regulated pathways in Krt1^−/− versus AE and Krt1^−/− versus psoriatic skin. Remarkably, the pathway network obtained for the Krt1^−/− versus AE intersection (Fig. 4F; supplementary material Tables S4, S5) showed a striking similarity to the AE pathway. Resulting data were compared with an ingenuity pathway analysis based on 81 validated AE genes (Barnes, 2010), which detects expression patterns of genes whose expression is linked to distinct signaling pathways. This highlighted the functional significance of IL-1 family members such as IL-18, TLR/MYD88 signaling, and NOD-like receptor family members (NLR) in Krt1^−/− mice and in AE. In contrast, the pathway network obtained for the Krt1^−/− versus psoriasis intersection was less extended and showed a limited interconnection of functional clusters (data not shown). These data suggested a link between Krt1 and IL-18.

To further substantiate Krt1 involvement in an inflammatory network in keratinocytes, we focused on IL-18 and related pro-inflammatory cytokines. Quantitative RT-PCR showed a significant upregulation of IL-18 in E18.5 Krt1^−/− skin (supplementary material Fig. S3D), consistent with its marked elevation at the protein level in suprabasal, neonatal Krt1^−/− skin and in the dermis (Fig. 5A). Moreover, IL-18 levels were significantly elevated in the serum of newborn Krt1^−/− mice (Fig. 5D). In line with the AE-like phenotype resulting from ectopic expression of IL-18 in mouse epidermis and the recent identification of high serum IL-18 in AE patients (Kambe et al., 2010; Konishi et al., 2002; Kou et al., 2012), our data support a role of IL-18 in the pathology of Krt1^−/− mice. Furthermore, the levels of IL-33, another member of the IL-1 family, were markedly increased in epidermal extracts of Krt1^−/− mice, but only mildly increased in newborn Krt1^−/− serum (Fig. 5D). IL-33 is known to be expressed in the suprabasal epidermis and the lung where it mediates a pro-inflammatory response by activating mast cells and Th2 lymphocytes. Its upregulation has been associated with allergic airway diseases and asthma (Haraldsen et al., 2009). The antimicrobial peptides S100A8 and S100A9, two damage-associated molecular pattern (DAMP) molecules linked to compromised skin barrier function and inflammatory skin diseases (Ehrchen et al., 2009), were strongly induced in suprabasal Krt1^−/− epidermis, where Krt1 is normally expressed, but also in the dermis, supporting an inflammatory response (Fig. 5B,D). Moreover, thymic stromal lymphopoietin (TSLP), a cytokine produced by differentiated keratinocytes (Ziegler and Artis, 2010), was increased in Krt1^−/− epidermal extracts (supplementary material Fig. S4C). TSLP has been implicated in atopic dermatitis and is a master switch for allergic inflammatory airway diseases in mice (Ziegler and Artis, 2010). The elevated expression of these cytokines in Krt1^−/− mice and in AE (Ehrchen et al., 2009; Suárez-Fariñas et al., 2011) lends further support to a role of Krt1 in controlling skin inflammation.

Next, we examined whether elevated cytokine levels might be connected to erythema and to additional systemic defects noted in epidermolytic ichthyosis (Arin et al., 2011). Staining for platelet endothelial cell adhesion molecule (PECAM) was increased, indicating increased vascularization at the dermal/epidermal interface (Fig. 5C; supplementary material Fig. S4A,B). In line with these findings, lungs from P0 Krt1^−/− pups had higher blood cell numbers (supplementary material Fig. S4D,E) and displayed capillary leakage (supplementary material Fig. S4F). Finally, the expression of the prominent pro-inflammatory cytokines IL-1α and TNFα remained unaltered in Krt1^−/− mice (data not shown), underscoring the selectivity of Krt1 function.

To validate our hypothesis that KRT1 restrains inflammation and an innate response in the epidermis by promoting barrier acquisition and restricting IL-18 release, we decided on a rescue of Krt1^−/− mice following antibody-mediated depletion and loss-of-function approaches. Surprisingly, systemic administration of IL-18-blocking antibodies to pregnant females extended the lifespan of Krt1^−/− pups for at least 2 days. The antibody-mediated IL-18 blockade (Fig. 5E–G; supplementary material Fig. S4G) increased the percentage of intact CEs (∼32% versus ∼15% in isotype-treated controls), indicating a crucial role of IL-18. To gather additional evidence, a genetic rescue experiment was conducted, crossing Krt1^−/− to Il18^−/−-knockout mice (Hochholzer et al., 2000). Double heterozygous Krt1^+/−Il18^+/− mice were interbred and offspring that survived until 4- to 6-weeks of age were assessed for homozygosity. Loss of IL-18 resulted in a partial rescue of Krt1^−/− mice, since 12 Krt1^−/−Il18^−/− survivors were obtained from 16 expected survivors (75%), compared with only six Krt1^−/−Il18^+/+ expected from 16 survivors (37.5%) in a Il18 wild-type genetic background (Fig. 6A). In contrast, the equivalent rescue experiment using a S100A9-deficient background (Manitz et al., 2003) generated no survivors (data not shown). Rescued Krt1^−/−Il18^−/− double-deficient mice exhibited an apparently normal epidermal morphology similar to Krt1^−/− mice (Fig. 6B,C; supplementary material Fig. S5D,E). Analysis of cornified envelopes revealed a twofold increase in intact CEs in newborn Krt1^−/−Il18^−/− skin compared with corresponding Krt1^−/− skin at P0, in line with the results obtained by the antibody blocking experiment (Fig. 6D,E; supplementary material Fig. S5A–C). Thus, we conclude that Krt1 acts upstream of IL18.

Having established a major role of IL-18 in Krt1 pathology, we sought to identify the underlying mechanism responsible for its release from keratinocytes. Engagement of intracellular NOD-like receptors (NLRs) by pathogen- or danger-associated molecular pattern molecules activates inflammasomes and caspase-1, which cleaves IL-18 into its active form (Davis et al., 2011; Schroder and Tschopp, 2010; Strowig et al., 2012). Remarkably, NLRP3 (nucleotide-binding-domain, leucine-rich repeat-containing protein) was localized in a keratin-dependent manner in mouse keratinocytes. In contrast to control keratinocytes, NLPR3 distribution appeared more diffuse in keratinocytes in which we had depleted the entire keratin family (Seltmann et al., 2012; Vijayaraj et al., 2009), suggesting a contribution of keratins to the regulation of inflammasome activity (Fig. 7C). We analyzed whether IL-18 processing occurs in a KRT1-dependent manner in human keratinocytes that are known to express higher amounts of KRT1 compared to murine cells upon differentiation (Feldmeyer et al., 2010). Unstimulated human keratinocytes produce significant amounts of proIL-18 but do not release mature IL-18 (Fig. 7A). Upon UVB irradiation, known to increase inflammasome activity and IL-1β release from keratinocytes (Feldmeyer et al., 2007), mature IL-18 was released into the supernatant. Using two caspase-1 inhibitors, IL-18 release could be blocked (Fig. 7A). This demonstrated functionality of the cell system. A knockdown of KRT1 in differentiated human primary keratinocytes resulted in secretion of cleaved IL-18, whereas a simultaneous knockdown of caspase-1 blocked the release of this cytokine (Fig. 7B). This highlighted an important role of KRT1 in inflammasome activation. Thus, Krt1, in contrast to its heterodimer partner Krt10, has unique functions in the maintenance of the cornified envelope, barrier formation and in restraining IL-18 release from keratinocytes. Both Krt1 and Krt10 share an important function in the maintenance of epidermal integrity (Table 1) (Wallace et al., 2012).

The epidermal barrier is formed by keratinocytes contributing tight junctions and the cornified envelope, and by Langerhans cells providing immune functions (Simpson et al., 2011). Barrier dysfunction and cutaneous sensitization can give rise to chronic inflammatory disorders including AE and psoriasis. Although both cell types are crucial for barrier functionality, the significance of epithelial keratinocytes in acute and chronic immune disorders has long been disputed. The recent discovery of filaggrin mutations as an underlying cause of AE has highlighted a major role of keratinocytes in shaping immune responses following barrier disruption and secretion of cytokines including IL-18, IL-33 and TSLP (Haraldsen et al., 2009; Wittmann et al., 2009; Ziegler and Artis, 2010). These seminal findings raise the issue of whether additional keratinocyte-resident proteins contribute to barrier defects and immune disorders through analogous mechanisms.

Until recently, keratins were predominantly regarded as intracellular scaffolds protecting epithelia against mechanical insults, in apparent agreement with phenotypes observed in epidermal keratinopathies (Arin et al., 2010; Arin et al., 2011; Chamcheu et al., 2011; Kim and Coulombe, 2007; Magin et al., 2007). Here, we show for the first time that Krt1, a member of the keratin family of cytoskeleton proteins, plays a major role in epidermal barrier formation and in restricting IL-18 release from keratinocytes, in addition to its role in preserving cell integrity upon elevated mechanical stress. Most importantly, cornified envelope defects, transcriptional upregulation and release of distinct cytokines manifest prenatally upon loss of Krt1, underlining its primary involvement. The specificity of Krt1 is further supported by the very mild phenotype of mice lacking Krt10, its heterodimer partner (Reichelt et al., 2001). As one might predict, combined deletion of Krt1 and Krt10 was accompanied by skin fragility more severe than with single gene knockouts (Wallace et al., 2012). In the absence of a side-by side analysis of both mice we speculate that the apparently more severe phenotype of Krt1/10^−/− compared with Krt1^−/− is connected to the presence of cytoplasmic Krt10 aggregates in the latter. The limited extent of cell fragility in all three mouse models and the notion that Krt1, but not Krt10, is covalently cross-linked to the CE (Candi et al., 1998) together with our data underscore unique Krt1 functions (see also Table 1).

The prenatal upregulation of IL-18 that occurred at transcriptional and posttranscriptional levels, together with data showing KRT1- and caspase-1-dependent IL-18 release from cultured keratinocytes place Krt1 upstream of IL-18. The importance of IL-18 in this setting is underscored by the partial pharmacological and genetic rescue of the Krt1 phenotype. Collectively, these data suggest a link between components of the epithelial cytoskeleton and inflammasome activation which might contribute to immune responses shifted from a Th1 to a Th2 profile (Kou et al., 2012; Wittmann et al., 2009). In support of this, absence of Krt17 was reported to attenuate inflammation in models of acute dermatitis, characterized by a change in inflammatory cytokines from a Th1- and Th17- to a Th2-dominated inflammatory response (Depianto et al., 2010). More recently, antimicrobial activity of fragments of Krt6A in corneal keratinocyte extracts was reported (Tam et al., 2012). Given the slightly elevated fragility of Krt1/IL-18^−/− epidermis, such an activity, if conserved in mouse epidermal keratinocytes, might contribute to the absence of strong inflammation in surviving animals.

The inflammatory signature in Krt1^−/− mice bears similarity to AE, linked to mutations in the keratin-associated protein filaggrin (Brown and McLean, 2012). The similarity of cytokine profiles and of commonly regulated pathways in Krt1^−/− mice and AE patients raise the issue of whether Krt1 acts through filaggrin-dependent mechanisms. We find that the absence of Krt1 does not affect filaggrin processing (supplementary material Fig. S2B). Furthermore, elevation of pro-inflammatory cytokines in filaggrin-related AE represents a late event, possibly resulting from a defective barrier and the responding immune system (Brown and McLean, 2012; Kezic et al., 2012; Kou et al., 2012). In contrast to prediction, deletion of filaggrin in mice does not affect survival or cause an elevated trans-epidermal water loss, but results in enhanced responses in contact hypersensitivity (Kawasaki et al., 2012). The unexpectedly mild phenotype arising from complete loss of filaggrin leads to questions about the involvement of KRT1. In a wider context, our data raise the issue of whether conditions in which KRT1 is mutated (e.g. epidermolytic ichthyosis) or downregulated (e.g. squamous cell carcinoma, cutaneous wound healing and HPV infections) (Arin et al., 2011; Depianto et al., 2010; Longworth and Laimins, 2004) carry an inflammatory signature similar to that seen in Krt1^−/− mice. Our data connect the presence of KRT1 to a keratinocyte-based network that restrains inflammation and, in conjunction with Krt17 (Depianto et al., 2010), provide strong evidence for context-dependent functions of keratin isotypes that extend beyond formation of cytoskeletal scaffolds in epithelia.

Krt1^−/− mice were generated by targeted disruption of the Krt1 gene by homologous recombination in embryonic stem cells (supplementary material Fig. S1B). Hence, exon1, intron 1 and ∼415 bp from the 5′ UTR were deleted and replaced with an Hprt-minigene cassette (Reichelt et al., 2001). All mice were from a mixed 129/SvJ×C57BL/6 background. Experiments were performed according to institutional regulations. Offspring from heterozygous intercrosses were genotyped by PCR analysis of tail DNA (supplementary material Table S6 for details). Genotypes of Krt1^−/− offspring were confirmed by western blot analysis of skin protein extracts with a polyclonal anti-Krt1 rabbit antiserum.

See supplementary material Table S7.

Preparation of tissue samples and immunofluorescence analysis were carried out as described previously (Roth et al., 2009). Horseradish-peroxidase-conjugated secondary antibodies were detected using the Super Sensitive Link-Label IHC Detection System (BioGenex Laboratories, Fremont, CA, USA) with diaminobenzidine (Dako, Hamburg, Germany). Images were acquired at ambient temperature with an Axioplan 2 fluorescence microscope (Carl Zeiss, Jena, Germany) equipped with Zeiss Plan-Apochromat 63×/1.4 and Plan-Neofluar 40×/1.3 oil immersion objectives, and recorded with an AxiocamHR camera (Carl Zeiss, Jena, Germany). Image analysis and processing were performed using the AxioVision 4.6 software (Carl Zeiss, Jena, Germany). Confocal images were recorded on an LSM710 microscope equipped with a Zeiss63×LCI Plan Neofluar objective (Carl Zeiss, Jena Germany). Fluorochromes were sequentially scanned to avoid cross-talk between channels using Pinhole ‘airy 1’ Zeiss standard settings. Z-stacks of 8–12 consecutive sections were recorded and processed as maximum intensity projections. Image analysis and processing were performed with the Zen software (Carl Zeiss, Jena, Germany). Images were cropped and analyzed in Adobe Photoshop CS4; Adobe Illustrator CS4 was used for figure design.

Full-thickness skin samples were lysed in 5× Laemmli buffer and homogenized (Ultra Turrax). Immunoblot analysis was performed as described previously (Löffek et al., 2010).

Full-thickness mouse skin was floated on 0.5 M ammonium thiocyanate (Sigma-Aldrich, St Louis, MO, USA) as described previously (Reichelt and Magin, 2002) to isolate epidermal sheets.

Epidermal sheets or full-thickness skin samples were homogenized (Ultra-Turrax) in TRIzol (Life Technologies, Darmstadt, Germany) supplemented with ribonucleoside–vanadyl complexes (New England BioLabs, Ipswich, MA, USA). Total RNA was phenol/chloroform extracted, precipitated, treated with DNaseI (Fermentas Life Science, Leon-Rot, Germany) and purified with the RNeasy MinElute Cleanup kit (Qiagen, Hilden, Germany).

The MouseWG-6v2.0 Expression BeadChip kit (Illumina, Inc., San Diego, CA, USA) was used to probe triplicate Krt1^+/+ and Krt1^−/− samples. Data analysis was based on the R Statistical language (R Development Core Team 2007, 2.8.0) and Beadstudio 3.1.1.0 software. Data were quantile-normalized. A fold-change/P-value filter was used to select differentially expressed genes; P-values smaller than 0.05, expression changes higher than twofold and a difference between mean intensity signals greater than background were considered statistically significant. The false discovery rate (FDR) of P-values was adjusted by the Benjamini–Hochberg method. Heatmaps were generated based on average linkage and the Pearson correlation coefficient, with blue indicating low level, white intermediate and red high levels of gene expression. The gene expression data sets were deposited in the GEO database (GES32951).

Transcriptome data from AE (E-GEOD-12511), psoriasis (E-GEOD-13355), Krt5 P0 (E-GEOD-7663) and EBS (E-GEOD-28315) were downloaded from the EBI server for the further comparison analysis (Bchetnia et al., 2012; Gudjonsson et al., 2010; Lu et al., 2007; Sääf et al., 2008). For data processing GNU R (www.r-project.org) with the Bioconductor packages (www.bioconductor.org) were used. Data were normalized by RMA algorithms. Krt1 data were individually analyzed or evaluated in conjunction with the additional transcriptome data sets.

Network figures of the enriched pathways were generated using the Cytoscape tool version 2.6.1 (www.cytoscape.org) with the Enrichment Map plug in (Merico et al., 2010). The top 100 enriched pathways with a P-value <0.001 were used.

cDNA synthesis was carried out using a RevertAid H Minus First Strand cDNA Synthesis kit (Fermentas Life Science, Leon-Rot, Germany). Real-time PCR was performed with a Maxima SYBR Green/ROX qPCR Master Mix (Fermentas Life Science, Leon-Rot, Germany) and run on a Mastercycler ep realplex (Eppendorf, Hamburg, Germany). 18S RNA was used as a reference. Mouse qPCR primer sequences are shown in supplementary material Table S8.

IL-18 and IL-33 were measured by ELISA (Bender MedSystems, Vienna, Austria; Quantikine, R&D Systems, Minneapolis, MN, USA). Mrp8 and Mrp14 protein levels were determined by ELISA as described previously (Vogl et al., 2007). TSLP protein levels were determined by mouse cytokine antibody array analysis (C series 1000.1; RayBiotech, Inc., Norcross, GA, USA) as described previously (Roth et al., 2009).

TEWL was assessed in triplicate readings on the dorsal neck skin of newborn mice using a Tewameter (TM 300, Courage and Khazaka Electronic, Cologne, Germany).

Females were injected intraperitoneally at E18.5 with 20 µg anti-IL-18 blocking antibody (clone 93-10C, R&D Systems, Minneapolis, MN, USA) or IgG1 immunoglobulins (R&D Systems, Minneapolis, MN, USA) as control. Newborn pups were treated with 4 µg of antibody or IgG1 at P0 and P1. At P2, mice were killed and analyzed.

In human primary keratinocytes (Feldmeyer et al., 2007), differentiation was induced by culture in KBM-2 (Cambrex, East Rutherford, NJ, USA) for 6 days. Keratinocytes were UV-irradiated with 50 mJ/cm² (Medisun HF-54, Schulze and Boehm, Bruehl, Germany) in the presence of the synthetic tetrapeptide YVAD (Tyr-Val-Ala-Asp; 20 µM), VAD (valyl-alanyl-aspartyl; 10 µM; Enzo Life Sciences, Lorrach, Germany) or DMSO (vehicle) and harvested after 4 h. For knockdown experiments, 10 nM siRNAs and Interferin (Polyplus transfection, Illkirch, France) were used. Cells and supernatants were harvested after 6 days; total protein from the supernatant was acetone precipitated. Sequence information of siRNAs (Sigma-Aldrich, St. Louis, MO, USA) used is provided in supplementary material Table S9.

Statistical significance was determined by unpaired t-tests for two groups and analysis of variance tests for more than two groups. The alpha level was set at <0.05. Data were analyzed with Sigma Plot 11.0 and plotted with Excel. All statistical data are shown in supplementary material Table S10.

We thank Roland Bornheim for the generation of the Krt1^+/− mouse line. We thank B. Fuss and M. Hoch for access to their LSM710.