Hypoxia-inducible factors regulate filaggrin expression and epidermal barrier function.

A functional epidermal skin barrier requires the formation of a cornified envelope from terminally differentiating keratinocytes. During this process, multiple genetic and environmental signals coordinately regulate protein expression and tissue differentiation. Here we describe a critical role for hypoxia-inducible factors (HIFs) in the regulation of filaggrin expression and skin barrier formation. Similar to other mammalian tissues, fetal epidermis in mice is normally O2 deprived. Simultaneous deletion of Hif1a and Hif2a in murine epidermis revealed defects in keratinocyte terminal differentiation and epidermal barrier formation. Mice lacking Hif1a and Hif2a in the epidermis exhibited dry flaky skin, impaired permeability barrier, and enhanced sensitivity to cutaneous allergens. These defects were correlated with stratum granulosum attenuation and reduced filaggrin expression. Hypoxic treatment of primary keratinocytes induced filaggrin (Flg) gene expression in a HIF1α- and HIF2α-dependent manner, suggesting that one mechanism by which Hif1a and Hif2a loss causes epidermal barrier defects in mice lies in Flg dysregulation. Therefore, low O2 tension is an essential component of the epidermal environment that contributes to skin development and function.

Introduction

The epidermis, together with hair follicles, sebaceous glands, and dermal connective tissue, forms the largest organ in the body. Skin performs many important functions, including thermoregulation, sensory perception, immunity, and protection from physical trauma. The protective function of the epidermis derives mainly from its most superficial epithelial layer, the cornified envelope. This barrier is constantly regenerated from differentiating keratinocytes, and abnormalities in this process have been associated with a variety of skin diseases such as ichthyosis, psoriasis, and atopic dermatitis (Irvine et al. 2011).

Skin homeostasis consists of a coordinated process whereby dividing basal keratinocytes detach from the basement membrane, commit to terminal differentiation, and eventually slough off the body surface (Simpson et al. 2011). The course of epidermal development can be delineated spatially and morphologically, as well as by the expression of specific keratin intermediate filaments at distinct differentiation stages. For example, basal keratinocytes express keratin 5 (KRT5) and keratin 14 (KRT14), whereas keratin 1 (KRT1) and keratin 10 (KRT10) are expressed in the spinous and lower granular layers in newly differentiating keratinocytes (Blanpain and Fuchs 2009). Terminally differentiated keratinocytes in the upper granular layer and cornified envelope express cornification proteins such as involucrin (IVL), loricrin (LOR), and filaggrin (FLG). Notably, filaggrin binds intermediate filaments in the upper granular layer, thereby condensing the keratinocyte cytoskeleton into a strong, flattened matrix (Irvine et al. 2011). Other cornified envelope proteins bind this matrix and become crosslinked to epidermal sphingolipids. These changes confer structural integrity and barrier properties on the epidermis.

Numerous regulatory and signaling pathways govern epidermal specification, differentiation, and cornification. For example, Wnt and BMP signaling maintain epidermal stem cell self-renewal (Chen et al. 2012; Lim et al. 2013). Notch and p63 mediated transcription programs control the transition from basal to suprabasal keratinocyte cell fate (Nguyen et al. 2006; Williams et al. 2011), while formation of the cornified envelope is regulated by transcription factor pathways involving KLF4 and IKKα (Gareus et al. 2007; Sen et al. 2012). The epidermal microenvironment is also an important determinant of keratinocyte differentiation: cornified envelope formation is regulated by extracellular calcium gradients, as well as steroid hormone levels (Kömüves et al. 2000; Tu et al. 2012).

The epidermal microenvironment is further characterized by low oxygen (O2) availability. Studies in humans and rodents have demonstrated that O2 saturation in adult epidermis ranges from 0.5% to 5% (Evans et al. 2006). The transcriptional response to low O2 is mediated primarily by hypoxia inducible factors (HIFs) (Keith et al. 2012). HIFs are heterodimeric proteins comprised of an O2-labile subunit (HIF1α or HIF2α) and constitutively-expressed HIF-β subunit, also known as aryl hydrocarbon receptor nuclear translocator (ARNT). HIF1α activity in the epidermis is important in cutaneous O2 sensing, skin innate immunity, wound healing, and melanoma transformation (Elson et al. 2000; Bedogni et al. 2005; Boutin et al. 2008; Peyssonnaux et al. 2008). In comparison, little is known about the function of HIF2α in the skin. However, both HIF1α and HIF2α have well-characterized roles in the determination and differentiation of other O2-deprived tissues such as the placenta, hippocampal neurons, skeletal muscle, and bone (Dahl et al. 2005; Amarilio et al. 2007; Mazumdar et al. 2010; Majmundar et al. 2012; Rankin et al. 2012).

In this study, we investigated the effect of epidermal O2 levels on keratinocyte differentiation. O2-deprived keratinocytes specifically stimulated Flg expression in a HIF1α- and HIF2α-dependent manner. Consistent with other mammalian tissues, low-O2 conditions occurred naturally in the skin during murine development. This ‘hypoxic’ state correlated with epidermal expression of HIF1α and HIF2α. Epidermal deletion of Hif1a and Hif2a led to reduced Flg expression, decreased corneocyte integrity, and impaired epidermal barrier function. Similar to Flg mutant animals, loss of Hif1a and Hif2a in the epidermis resulted in ichthyosiform skin and percutaneous sensitization to a common skin allergen. Consequently, HIF is an essential regulator of keratinocyte differentiation and epidermal barrier function.

Results and Discussion

Developing murine epidermis is naturally low in O2

Murine epidermis begins to differentiate by day 14 of gestation (E14) from a single layer of epidermal keratinocytes, such that maturing epidermal layers expressing late differentiation-specific keratins are present by E16 (Byrne et al. 1994). The cornified envelope, which is the most superficial layer, undergoes further maturation between E18 and birth (P0) (Candi et al. 2005). To characterize the epidermal O2 environment during this developmental process, we injected animals with the O2-sensitive dye pimonidazole. Positive pimonidazole staining indicated that the epidermis encounters low O2 levels (<2%) between E16 and P0 (Figure 1a). This finding is consistent with O2 electrode measurements in sheep embryos, which showed that ovine epidermis is hypoxic between days 60 and 120 of gestation (Scheid et al. 2002). Low O2 levels in the newborn epidermis correlated with HIF1α and HIF2α protein expression (Figure 1b). HIF1α and HIF2α were also stabilized in primary keratinocytes when cultured at 0.5% O2 (Figure 1c). These results suggested that murine epidermal development occurs in an O2-deprived setting.

Figure 1

Murine epidermis develops under naturally low O2 tension

(a) Hypoxyprobe was injected into pregnant mice at E14, E16, E18, and P0 pups. Frozen skin sections were stained using hematoxylin and eosin, anti-hypoxyprobe-FITC (green), and DAPI (blue). Uninjected mice of the same age were used as negative controls. (b) Skin from wild type P2 pups were immunostained for HIF1α and HIF2α. Inserts (400×) highlight nuclear immunostaining. (c) Primary murine keratinocytes were grown at 21% or 0.5% O2 for 48 hours and cell lysates immunoblotted for HIF1α and HIF2α. Scale bar, 50 μm.

Interestingly, pimonidazole staining appears stronger in more superficial epidermal layers (Fig. 1a). This disparity is most apparent at E16, but diminishes by the time of birth. In contrast, HIF immunostaining is comparable between basal and suprabasal nuclei (Fig. 1b). This finding indicates that HIF1α and HIF2α are stabilized in the epidermis below a certain threshold of hypoxia. However, the mechanisms by which low O2 tension is maintained in the epidermis are poorly understood. Possible causes of O2 deprivation in the superficial epidermis include the absence of vascularization as well as low O2 saturation in amniotic fluid (Bejar et al. 1971). Additionally, non-O2-dependent processes such as PI3K/AKT signaling or the generation of reactive oxygen species in proliferating keratinocytes may contribute to HIF stabilization in the basal epidermis.

Hypoxic keratinocytes stimulate filaggrin expression

To determine the effect of low O2 on epidermal development, we examined the expression of early and late differentiation genes in keratinocytes cultured under hypoxia. Keratin 1 (Krt1) and keratin 10 (Krt10) are expressed in early differentiating cells, whereas loricrin (Lor), involucrin (Ivl), and filaggrin (Flg) expression are induced at the terminal stage of epidermal differentiation. Transcriptional upregulation of Flg and the related filaggrin-2 (Flg2) gene was observed in hypoxic primary keratinocytes at the mRNA and protein level (Figure 2a-b). Although mRNA levels of early epidermal markers Krt1 and Krt10 were somewhat reduced after hypoxic culture, KRT10 protein levels and that of another early epidermal marker, KRT5, were unchanged (Figure 2a-b). Treatment with 2-oxoglutarate-dependent dioxygenase inhibitors, including dimethyloxalylglycine (DMOG) and deferoxamine (DFO), which inhibit HIFα degradation post-translationally, also induced filaggrin expression (Figure 2c). Similar upregulation of filaggrin mRNA and protein was observed in hypoxic primary human keratinocytes (Figure 2d), indicating that this response to low O2 tension is conserved.

Figure 2

Hypoxia induces filaggrin expression

(a, b) Primary murine keratinocytes were grown at 21% or 0.5% O2 for 48 hours. (a) Epidermal gene expression was assessed by qRT-PCR and (b) cell lysates immunoblotted for epidermal proteins as indicated. (c) Primary murine keratinocytes were grown in 1 mM DMOG or 100 μM DFO for 48 hours and analyzed for filaggrin transcripts (upper panel) and protein (lower panel). (d) Primary human keratinocytes were grown as above and FLG expression evaluated. (e) Primary murine keratinocytes were grown at 21% or 0.5% O2 for 48 hours to 70% or 100% confluency and lysates immunoblotted for filaggrin. (f) Primary murine keratinocytes were grown at 1.4 mM Ca2+ at 21% or 0.5% O2 for 48 hours and lysates immunoblotted for filaggrin.

Although filaggrin expression in cultured keratinocytes is known to be induced by cell confluency and extracellular calcium (Resing et al. 1993), hypoxic treatment of primary keratinocytes led to elevated filaggrin accumulation in both subconfluent and confluent cultures (Figure 2e). Similarly, hypoxia further increased filaggrin expression even at high calcium concentrations (Figure 2f). We concluded that O2 deprivation induces filaggrin expression; furthermore, this effect is independent of known environmental regulators of epidermal development.

Filaggrin expression is HIF-dependent

To characterize the requirement for HIF in hypoxic Flg induction, we generated mice lacking HIF1α and HIF2α in the epidermis using the Cre transgene driven by the keratin 14 (Krt14) promoter, expressed in the basal epidermis (Dassule et al. 2000). Mice lacking either HIF1α or HIF2α in the epidermis have been described previously and do not develop epidermal abnormalities (Boutin et al. 2008; Rezvani et al. 2011). Krt14-Cre+Hif1afl/wtHif2afl/fl mice were crossed to produce control (Krt14-Cre+Hif1afl/wtHif2afl/fl) or epidermis-specific double knockout mice (DKO; Krt14-Cre+Hif1afl/flHif2afl/fl) (Supplementary Figure S1a). Control mice, which lack epidermal HIF2α, are phenotypically indistinguishable from wild type animals (data not shown).

In the absence of Hif1a and Hif2a, epidermal filaggrin expression in ear, trunk, and tail skin was significantly reduced (Figure 3a and data not shown). Consistent with a role for HIF in transcriptional regulation of Flg, DKO epidermis exhibited lower levels of Flg transcripts, profilaggrin, and filaggrin monomers (Figure 3b). Flg induction was similarly abrogated in hypoxic primary keratinocytes lacking Hif1a and Hif2a (Figure 3c and Supplementary Figure S1b). Primary keratinocytes lacking Arnt also did not express filaggrin under hypoxic conditions (Figure 3d). These results indicated that HIF is required for filaggrin expression both in vitro and in vivo.

Figure 3

Filaggrin expression is HIF-dependent

(a) Ear skin from control and DKO pups were immunostained for filaggrin. Scale bar, 50 μm. (b) Tail skin from control and DKO pups were analyzed for filaggrin transcripts (left) and protein (right). (c, d) Primary murine keratinocytes from control (CONT) and DKO pups (c) or KO pups lacking Arnt (d) were grown at 21% or 0.5% O2 for 48 hours and analyzed for filaggrin transcripts (left) and protein (right). (e) Alignment of the hypoxia response element (HRE) at position -193 and flanking sequences indicates high mammalian conservation. (f) Fragments of the human FLG promoter were cloned into pGL3 luciferase reporter. Rectangles indicate putative HREs. (g) Primary human keratinocytes transfected with promoter construct and Renilla luciferase were exposed to 21% or 0.5% O2 for 30 hours. Luciferase activity under hypoxia was normalized to normoxic values. Data are presented as the mean of three independent experiments. *, p<0.05

Filaggrin accumulation is regulated by the epidermal serine proteases caspase 14 (CASP14) and serine protease inhibitor matriptase (PRSS8), which cleave profilaggrin to monomeric filaggrin (List et al. 2003; Denecker et al. 2007). Casp14 and Prss8 expression was unchanged in DKO epidermis (Supplementary Figure S1c). However, the expression of Spink5, an inhibitor of epidermal serine proteases, was increased (Supplementary Figure S1c). Of note, SPINK5 deficiency is correlated with increased monomeric filaggrin in the epidermis (Chavanas et al. 2000). Increased Spink5 expression in DKO mice may therefore inhibit the conversion of profilaggrin to monomeric filaggrin.

FLG promoter contains hypoxia response elements

Nucleotide analysis of the human FLG promoter for HIF binding sites matching the consensus sequence (A/G)CGTG identified three putative hypoxia response elements (HREs) at 193, 1399, and 2008 bases upstream of the FLG transcriptional start site. The proximal consensus site at -193 is highly conserved between mammals (Figure 3e). Fragments of the FLG promoter containing these HRE sequences were cloned into a luciferase expression vector and transfected into primary keratinocytes (Figure 3f). In cells transfected with wild type promoter sequences (WT4 and WT5), luciferase activity was increased under hypoxic conditions relative to normoxic control (Figure 3g). Sequential mutagenesis of the two proximal HREs (Δ4 and Δ42) reduced hypoxic reporter activity progressively (Figure 3g). In contrast, mutation of the most distal HRE (Δ5) had no effect on hypoxic stimulation of reporter activity (Figure 3g). Therefore, the FLG promoter contains two functional binding sites for HIF1α and/or HIF2α.

Surprisingly, chromatin immunoprecipitation did not reveal HIF1α or HIF2α occupancy at the human FLG promoter (data not shown). While we cannot exclude the possibility that hypoxic stimulation of FLG is an indirect consequence of HIF activity, promoter mutagenesis experiments clearly demonstrated that functional binding sites exist for HIF1α and HIF2α. Therefore, it is likely that the interaction between HIF and the FLG promoter lies beneath the threshold of detection by ChIP, implying that the proximal HREs represent consensus sequences possessing relatively weak HIF/DNA interactions.

HIF-deficient mice develop skin abnormalities

DKO mice were indistinguishable from control littermates at birth. However, 4 to 5 days later (P4-P5), DKO mice exhibited dry, flaky skin on their extremities (Figure 4a). By P8, DKO mice had developed thickened and diminutive external ears, constricted hind digits, and hyperkeratotic tail constrictions (Figure 4b). This set of phenotypes is associated with epidermal barrier defects in other mouse models and phenocopies the flaky tail (ft) mice, which possess a spontaneous mutation in the Flg gene (Lane 1972; Presland et al. 2000; Fallon et al. 2009). Adult DKO mice retained underdeveloped external ears and dry, flaky tail skin (data not shown).

Figure 4

Impaired epidermal permeability barrier and corneocyte integrity in DKO mice

(a) Flaky skin on hind limbs and tails of DKO P5 pups compared to control litttermates. (b) Thick, shortened pinna of the ear, taut skin of the feet, and constricted tail skin in DKO P8 pups compared to control littermates. (c) DKO epidermis stained with hematoxylin and eosin shows attenuated stratum granulosum compared to control epidermis. Scale bar, 50 μm. (d) Control (CONT) and DKO littermates were stained with toluidine blue dye at E17. (e) Embryos were categorized into groups of low, moderate, or high permeability and results from multiple experiments quantified. (f) Newborn CONT and DKO mice were weighed at regular intervals as indicated. *, p<0.05 compared to CONT. (g) Corneocytes from CONT or DKO P3 epidermis were sonicated and visualized under a microscope. Scale bar, 40 μm. (h) CONT (n=9) and DKO (n=4) littermates were treated topically with ovalbumin for 5 days × 3 cycles and serum IgE levels determined after treatment using ELISA (right). Controls (left) are untreated animals.

Histologically, DKO epidermis resembled ichthyosis vulgaris lesions, which are associated with loss of filaggrin (Compton et al. 2002). DKO epidermis showed attenuation of the stratum granulosum and reduced cornified envelope thickness (Figure 4c). Importantly, changes in the stratum granulosum and cornified envelope did not correlate with changes in keratinocyte proliferation or cell fate. No significant difference in the number of Ki67-positive basal keratinocytes was observed between control and DKO mice (Supplementary Figure S1d-e). KRT10, an early differentiation marker, was appropriately expressed in all suprabasal keratinocytes (Supplementary Figure S1d). These results indicated that HIF deficiency in the epidermis specifically affects keratinocyte terminal differentiation in the granular layer and cornified envelope.

DKO mice exhibit defective permeability barrier and cornified envelope integrity

Formation of the outside-to-in epidermal barrier occurs between E16 and E18 and can be visualized by permeability to toluidine blue dye (Hardman et al. 1998). Most control embryos efficiently excluded toluidine blue at E17; in contrast, many DKO embryos were completely or partially stained (Figure 4d). Because both control and DKO mice displayed some phenotypic variability, embryos were classified into those with high, moderate, or low dye permeability. DKO mice were more likely to exhibit high permeability to toluidine blue, indicative of epidermal barrier dysfunction (Figure 4e).

Inside-to-out barrier function was assessed by measuring the rate of trans-epidermal fluid evaporation. Newborn control pups maintained constant body weight over 3.5 hours, while DKO pups suffered progressive weight loss (Figure 4f). Therefore, the loss of epidermal HIF activity resulted in impaired epidermal barrier development.

We hypothesized that reduced filaggrin expression would render Hif-deficient epidermis more susceptible to mechanical stress. Corneocytes (i.e. terminally differentiated keratinocytes) were isolated from control and DKO epidermis and subjected to ultrasonic disruption. Control corneocytes remained largely intact, whereas DKO corneocytes became fragmented or destroyed (Figure 4g). These experiments indicate that cornified envelope integrity is compromised in DKO mice, resulting in epidermal barrier dysfunction.

Epidermal barrier disruption due to filaggrin deficiency results in percutaneous allergen sensitization (Palmer et al. 2006; Fallon et al. 2009). As a result, FLG mutations and deletions are common causes of atopic dermatitis and ichthyosis vulgaris (Palmer et al. 2006; Smith et al. 2006). We applied the common allergen ovalbumin (OVA) to the skin of control and DKO mice and measured serum IgE levels as a marker of atopic response. Untreated control and DKO mice showed no significant difference in baseline IgE levels (Figure 4h). However, OVA treatment significantly elevated IgE levels in DKO mice (Figure 4h). The loss of HIF activity in the epidermis may therefore create conditions favorable for developing atopic skin disease.

Comparison with Arnt-deficient mice

In the epidermis, Arnt is required for ceramide synthesis and its absence results in epidermal barrier defects in mice (Takagi et al. 2003; Geng et al. 2006). Additionally, it is required for AHR-dependent keratinocyte differentiation in culture (Takagi et al. 2003; Geng et al. 2006; Sutter et al. 2009). To characterize the role of ARNT in epidermal development in the context of hypoxic gene regulation, Krt14-Cre+ mice were crossed to Arntfl/fl mice to produce control (Krt14-Cre+Arntfl/wt) and KO (Krt14-Cre+Arntfl/fl) mice (Supplementary Figure S2a). Consistent with previous reports, Arnt KO mice were perinatal lethal. Arnt KO epidermis showed reduced FLG expression (Supplementary Figure S2b). Similar to DKO epidermis, stratum granulosum and cornified envelope thickness were also reduced (Supplementary Figure S2b). No difference between KO and control animals was observed in immunohistochemical staining for the proliferation marker Ki67 or early differentiation proteins KRT1 and KRT10 (Supplementary Figure S2c-d). Although previous studies have proposed that epidermal barrier defects in Arnt KO mice are a result of deficiencies in lipid processing, these findings suggested that reduced filaggrin expression in O2-deprived keratinocytes also contributes to this phenotype (Figure 3d).

Conclusion

In this study, we demonstrated the importance of hypoxic signaling in skin development. We showed that a low-O2 environment exists in the murine embryonic epidermis. Low-O2 conditions stabilize HIF1α and HIF2α, the absence of which impaired expression of the HIF target gene filaggrin. As a result, keratinocyte terminal differentiation and epidermal barrier formation are impaired. Interestingly, although HIF1α and HIF2α can sometimes act in distinct and opposing fashion (Keith et al. 2012), they appear to regulate Flg expression and epidermal barrier function in a redundant manner.

The link between epidermal barrier defects and atopic skin diseases became firmly established with the identification of FLG as a susceptibility gene (Palmer et al. 2006). However, only 30% of European patients with atopic dermatitis bear FLG mutations (Bieber 2008). Recent genome-wide association studies have uncovered additional gene loci that confer risk for atopic dermatitis (Esparza-Gordillo et al. 2009; Sun et al. 2011; Paternoster et al. 2012). These investigations have not identified mutations in HIF pathway genes, but it is important to note that no study has controlled for FLG status amongst their cohorts. On the other hand, it is equally possible that HIF-dependent regulation of filaggrin is required for normal development but not disease progression. Further investigation into HIF activity in patients with atopic dermatitis may enhance our understanding of the mechanisms by which the low-O2 environment of the epidermis regulates skin development and function.

Materials and Methods

Mice

Krt14-Cre+ transgenic mice were purchased from JAX. Arntfl/fl, Hif1afl/fl, and Hif2afl/fl mice have been described previously (Ryan et al. 2000; Tomita et al. 2000; Gruber et al. 2007). PCR was performed using allele-specific primers (available on request). All animal experiments were performed in compliance with Institutional Animal Care and Use Committee regulations and approved by the University of Pennsylvania institutional review board.

Cell culture

Primary human epidermal keratinocytes isolated from neonatal foreskin were obtained from the Penn Skin Disease Research Center and maintained in a 1:1 mixture of K-SFM supplemented with human rEGF and BPE (Invitrogen) and Medium 154 supplemented with HKGS (Invitrogen). Primary murine keratinocytes were isolated as described in (Lichti et al. 2008) and maintained at 34°C, 8% CO2 in MCDB 153 media (Sigma) supplemented with 5% chelexed FBS (Hyclone), 100 μM ethanolamine (Sigma), 100 μM phosphorylethanolamine (Sigma), 2 mM L-glutamine (Invitrogen), 10 ng/mL human rEGF (Invitrogen), 1 μM hydrocortisone (Invitrogen), 5 μg/mL insulin (Invitrogen), 0.2 nM choleratoxin (Sigma), 2 nM 3,3′5-triiodo-L-thyronine (Sigma), and 45 μM CaCl2 (Sigma). For hypoxia treatment, cells were cultured in an InVivo2 400 hypoxia workstation (Ruskinn).

Visualization of tissue pO2

1-((2-hydroxy-3-piperdinyl)propyl)-2-nitroimidazole hydrochloride (Hypoxyprobe Inc) was injected intraperitoneally into pregnant mice at E14, E16, E18, and intraperitoneally into P0 pups. After 4 hours, embryos/pups were collected and frozen in OCT (Sakura). Skin cryosections were stained using anti-hypoxyprobe-FITC antibody (Hypoxyprobe Inc).

Histological and immunohistochemical analysis

Paraffin-embedded sections were immunostained with antibodies directed against ARNT (SC-8076; Santa Cruz), filaggrin, HIF1α (Ab2185; Abcam), HIF2α (NB100-132; Novus), IVL, Ki67 (Novocastra), KRT1, KRT6, KRT10, and LOR (Covance unless otherwise noted). Antibody binding was detected using DAB (Vector Labs) or AlexaFluor 488 anti-rabbit immunoglobulin (Invitrogen).

Quantitative RT-PCR

Gene expression was analyzed using specific primers (available on request). Expression levels of 18S rRNA and HPRT1 were used for normalization.

Western blotting

Primary antibodies used were directed against ARNT (NB100-110; Novus), β-actin (Sigma), filaggrin, IVL, HIF1α C-Term (Cayman), HIF2α (NB100-122; Novus), KRT5, KRT10, and LOR (Covance unless otherwise noted).

FLG promoter analysis

Fragments of the human FLG promoter were amplified from keratinocyte genomic DNA using specific primers (available on request) and cloned into pGL3-Basic luciferase expression vector (Promega). Individual constructs were transfected into primary human keratinocytes using FuGENE reagent (Roche). Luciferase activity was measured after normoxic or hypoxic treatment using the Dual-Luciferase Reporter Assay (Promega).

Outside-to-in permeability barrier assay

Epidermal permeability was assessed by toluidine blue staining as described in (Takagi et al. 2003) and categorized as high (dorsal staining), moderate (>50% ventral staining), or low (<50% ventral staining).

Corneocyte sonication

Cornified envelope was purified and sonicated as described in (Koch et al. 2000).

Ovalbumin treatment and serum IgE ELISA

Ovalbumin (OVA) challenge was performed as described in (Palmer et al. 2006). 20 to 50 ug of OVA (fraction V; Sigma) in PBS was applied daily to ventral skin of control or DKO littermates for five consecutive days. OVA challenge was repeated twice with intervening intervals of 7 days. Total serum IgE was measured using PharMingen antibodies (BD Biosciences).

Statistical analysis

Data are presented as mean+s.e.m. with differences between two groups analyzed for significance using Student's t test. A p-value <0.05 was considered to be significant.

Supplementary Figure S1. Epidermal (a) PCR analysis of epidermal and dermal DNA with allele-specific primers indicated site-specific and efficient recombination. (b) Primary murine keratinocytes from control (CONT) or DKO P2 pups were grown at 21% or 0.5% O2 for 48 hours and lysates immunoblotted for HIF1α and HIF2α. (c) Epidermal RNA was extracted from control and DKO P3 pups and subjected to qRT-PCR using specific primers for Casp14, Prss8, and Spink5. (d) Paraffin-embedded ear sections from control (CONT) and DKO P5 pups were immunostained using the indicated antibodies. No observable differences were detected in Ki67, KRT10, and loricrin (LOR) staining intensity between CONT and DKO epidermis. Scale bar, 50 μm. (e) Quantification of Ki67-positive nuclei in the basal epidermis of CONT and DKO animals demonstrated no difference in keratinocyte proliferation.

Supplementary Figure S2. (a) PCR analysis of epidermal and dermal DNA with allele-specific primers indicated site-specific and efficient recombination in the epidermis (upper panel). Cell lysates from primary murine keratinocytes obtained from control (CONT) or KO P0 pups were immunoblotted for ARNT (lower panel). n.s., non-specific band. (b) Paraffin-embedded skin sections from control (CONT) and KO P0 pups were stained with hematoxylin and eosin or with antibodies against filaggrin. (c) Skin sections from control (CONT) and KO pups were immunostained using the indicated antibodies. No observable differences were detected in Ki67, KRT10, and loricrin (LOR) staining intensity between CONT and KO epidermis. Scale bar, 50 μm. (d) Quantification of Ki67-positive nuclei in the basal epidermis of CONT and KO animals demonstrated no difference in keratinocyte proliferation.