Activation of β-catenin signaling programs embryonic epidermis to hair follicle fate

The embryonic surface ectoderm stratifies to form the epidermis, providing an essential protective barrier. Subsets of surface ectodermal cells, rather than stratifying, develop into hair follicles and other ectodermal appendages(Mikkola and Millar, 2006). Hair follicle development is initiated by signaling from the dermis, which induces the formation of local thickenings, or placodes, in the overlying epithelium (Hardy, 1992). Although it is currently unclear whether the initiating dermal signal is uniform or whether it is generated by labile local dermal condensates, the final location of placodes appears to be influenced by reaction-diffusion mechanisms involving competing epithelial signals(Jiang et al., 2004). These interactions result in the formation of a hexagonal array of primary placodes that appears at approximately embryonic day (E) 14.5 in mouse embryos, and gives rise to the large guard over-hairs of the coat. Secondary placodes are initiated in several waves between E15.5 and birth, and develop into awl,auchene and zigzag hair follicles.

Signaling from placode epithelial cells causes condensation of mesenchymal cells that in turn stimulates proliferation and downgrowth of the placode(Hardy, 1992). Rapidly proliferating hair matrix cells surround the dermal condensate, hereafter referred to as a dermal papilla, and differentiate to form the hair shaft and inner root sheath (Fuchs,2007; Millar,2002). An outer root sheath surrounds the follicle and is contiguous with the basal epidermis. Incompletely characterized signals direct melanoblasts and nerve fibers to developing hair follicles(Jordan and Jackson, 2000; Peters et al., 2002).

Although the molecular nature of the primary dermal signal has not been established, several families of secreted signaling molecules participate in hair placode induction. These factors include the TNF family member ectodysplasin-A (EDA), which signals via its receptor EDAR and NFκB to regulate primary placode formation; BMP family members, which negatively regulate placode formation and compete with EDA in establishing a regular patterned placodal array; and sonic hedgehog (SHH), which controls proliferation and downgrowth of hair follicle epithelium (reviewed by Fuchs, 2007; Millar, 2002; Schmidt-Ullrich and Paus,2005). Stable primary placodes fail to form in EDA pathway mutants, but transient pre-placode structures can be detected, suggesting the existence of earlier patterned signals(Schmidt-Ullrich et al.,2006).

Wnt/β-catenin signaling regulates many aspects of development and disease. Signaling is activated by the binding of a secreted Wnt ligand to a frizzled (FZ) receptor and an LDL-related protein (LRP) 5/6 co-receptor,leading to the inactivation of a complex of proteins that phosphorylates cytoplasmic β-catenin and targets it for degradation. Cytoplasmicβ-catenin accumulates, translocates to the nucleus, and complexes with LEF/TCF transcription factor family members to activate target gene transcription (Gordon and Nusse,2006). Wnt/β-catenin signaling is required for hair follicle placode induction and the patterned expression of Edar(Andl et al., 2002). Wnt reporter transgene expression is elevated in epithelial and dermal condensate cells at early stages of follicular development, and is then downregulated in the epithelium until the onset of hair shaft differentiation when it appears in hair shaft precursor cells (DasGupta and Fuchs, 1999; Maretto et al., 2003). β-catenin/LEF complexes can directly regulate the expression of hair shaft keratin genes in vitro(Merrill et al., 2001),suggesting a key role for Wnt/β-catenin signaling in hair shaft differentiation.

Deletion of the intracellular Wnt/β-catenin pathway inhibitor Apc in skin epithelial cells results in the formation an increased number of embryonic hair follicles, and hair follicle abnormalities, but does not affect the development and stratification of interfollicular epidermis(Kuraguchi et al., 2006). Transgenic epidermal expression of stabilized β-catenin causes ectopic hair follicle induction in adult skin; however, alterations in embryonic skin or hair follicle development were not observed(Gat et al., 1998), or not reported (Lo Celso et al.,2004; Silva-Vargas et al.,2005). Thus, Wnt/β-catenin signaling is required for hair follicle placode induction, but it is not clear whether activation of this pathway in mammalian embryonic surface ectoderm is sufficient to globally program ectodermal cells to hair follicle fate. It is also not known whether Wnt/β-catenin signaling activates the factors that attract nerve fibers and melanoblasts to hair follicles, or whether it suppresses epidermal stratification to allow placode development. To address these questions, we mutated the endogenous β-catenin gene to a constitutively active form in vivo using an efficient KRT14-Cre transgenic line.

KRT14-Cre Ctnnb1^(Ex3)fl/+ and KRT5-rtTA tetO-Dkk1 mice were generated as described previously(Chu et al., 2004; Liu et al., 2007). For tetracycline-inducible β-catenin mutation, mice carrying tetO-Cre (Mucenski et al.,2003) and KRT5-rtTA(Diamond et al., 2000)transgenes and Ctnnb1^(Ex3)fl(Harada et al., 1999) were placed on doxycycline chow (1 mg/kg, Bio-serv, Laurel, MD). To generate 4-hydroxytamoxifen-inducible β-catenin mutants, 50 μl of 4-hydroxytamoxifen in 95% ethanol (10 mg/ml, Sigma-Aldrich, St Louis, MO) or vehicle control was applied to dorsal back skin after hair clipping. Mice were genotyped by PCR analysis of tail biopsy DNA. All experiments were performed with approved animal protocols according to University of Pennsylvania and University of Michigan institutional guidelines.

Histological analysis, whole-mount and section in situ hybridization and immunostaining, TUNEL, X-gal staining, Oil Red O staining and transmission electron microscopy (TEM) were performed as described previously(Allen et al., 2003; Andl et al., 2004; Andl et al., 2006; Braun et al., 2003; Chu et al., 2004; Hutchin et al., 2005; Langbein et al., 2007; Liu et al., 2007). For epidermal barrier assays, methanol-fixed embryos were stained in 1% Toluidine Blue O solution (Sigma-Aldrich).

Dissected dorsal skin was dispase treated (BD Bioscience, Sparks, MD) to separate epidermis and dermis. Wild-type, floxed and deleted-floxed Ctnnb1 alleles were detected by PCR of extracted genomic DNA. RNA was extracted using RNeasy Mini Kit (Qiagen, Valencia, CA). Primer pairs for qRT-PCR were purchased from Superarray Bioscience. Reactions were performed in triplicate using SYBR green on an MJ Opticon II thermocycler (Bio-Rad,Hercules, CA). Relative expression levels were standardized using β-actin as an internal control. Data were analyzed using the Opticon III program.

ChIP assays were performed using the Chromatin Immunoprecipitation Kit(Upstate, Charlottesville, VA). Epidermal cells were dissociated, fixed in 1%formaldehyde for 15 minutes at room temperature, sonicated and incubated with anti-β-catenin antibody (14, BD Bioscience, San Jose, CA) or control IgG overnight at 4°C, after which Protein A agarose beads were added. Purified DNA was subjected to semi-quantitative PCR with primers that amplify the Sp5 promoter, or Gapdh sequences.

Total RNA was subjected to one round of amplification/labeling to obtain biotinylated cRNA for hybridization to Affymetrix GeneChip Mouse Genome MOE430 2.0 oligonucleotide microarrays at the Microarray Core, University of Pennsylvania (Philadelphia, PA). Two entirely independent samples of each tested condition were used for data analyses. Scanned microarray images were imported into Gene Chip Operating Software (GCOS, Affymetrix, Santa Clara, CA)to generate signal values and present/absent calls for each probe set using the MAS 5.0 statistical expression algorithm. Data files were imported into GeneSpring 7.3.1 (Agilent, Santa Clara, CA), and replicate microarrays grouped and compared. Transcripts that were increased at least 2-fold in one population over the other and called present in both replicates were considered significant for further analysis.

Full-length mouse Sp5 cDNA cloned into pCMV-3Tag (Stratagene, La Jolla, CA) or a pCMV-3Tag control plasmid were transfected into HaCAT cells using the Cell Line Nucleofector Kit V (Amaxz GmbH, Gaithersburg, MD). mRNA was extracted after 3 weeks of culture in 500 μg/ml G-418 and subjected to quantitative RT-PCR for Sp5, KRT10, involucrin and DLX3.

To determine whether forced activation of Wnt/β-catenin signaling alters the fate of embryonic surface ectoderm, we stimulated β-catenin signaling in vivo by mutating the endogenous β-catenin gene(Ctnnb1) to a dominant-active form in epithelial cells. Ctnnb1^(Ex3)fl/fl mice, in which exon 3 of Ctnnb1,encoding all phosphorylation target serine/threonine residues, is flanked by loxP sequences that are targets for Cre-mediated recombination, were mated to KRT14-Cre line 43 transgenic mice expressing Cre recombinase under the control of a keratin 14 promoter(Liu et al., 2007). KRT14-Cre-mediated recombination of a ROSA26R Cre reporter gene (Soriano, 1999) was observed in surface epithelium by E11.5 (see Fig. S1 in the supplementary material).

KRT14-Cre Ctnnb1^(Ex3)fl/+ embryos died perinatally, and displayed limb and craniofacial defects. Their skin appeared grossly abnormal by E14.5, with multiple raised papules, and by birth was almost completely covered in large keratinized plaques (Fig. 1A-J). Precocious pigmentation was observed in embryonic ectoderm associated with keratinized plaques (Fig. 1J). Control mice displayed hair shaft pigmentation several days after birth, but control embryo skin was never pigmented(Fig. 1I).

Embryonic mutant skin showed elevated expression of β-catenin protein in the epithelium from E11.5 onwards, with many ectodermal cells exhibiting both cytoplasmic and nuclear localization of β-catenin(Fig. 1K,L). To determine the pattern of Wnt/β-catenin pathway activation in mutant skin, we generated KRT14-Cre Ctnnb1^(Ex3)fl/+ mice carrying two independent Wnt reporter genes, BATgal(Maretto et al., 2003) or Topgal (DasGupta and Fuchs,1999). Expression of either reporter was significantly elevated and more widespread in mutant compared with littermate control skin, and the normal regular pattern of hair follicle placode formation was disturbed in the mutant (Fig. 1M-P).

Thickened epithelial structures expressing the placode marker Shh(St-Jacques et al., 1998)appeared in stabilized β-catenin mutant embryonic dorsal skin by E12.5,whereas hair follicle placodes were not apparent in control littermate embryos until E14.5 (Fig. 2A-F).

Interestingly, immunofluorescence staining for the pan-neural marker PGP9.5 revealed that E14.5 mutant skin was more extensively innervated than control littermate skin, with larger, more developed nerve fibers(Fig. 2G,H). These data indicate that secreted factors activated by cell autonomous mutantβ-catenin accelerate the attraction of nerve fibers to the skin.

To visualize the global pattern of placode development, we carried out whole-mount in situ hybridization for the placode marker Wnt10b(St-Jacques et al., 1998). At E12.5, elevated Wnt10b expression localized specifically to mammary and vibrissa placodes in control embryos(Fig. 2I,K). Stabilizedβ-catenin mutant embryos displayed numerous placodes covering the trunk,head and limbs (Fig. 2J,L). At E13.5, mutant placodes were enlarged (Fig. 2M,N). Counting of the placodes in a defined area of head skin at E13.5 revealed a mean of 7±3 placodes in control embryos (n=3)and 33±9 placodes in mutant embryos (n=3; P=0.03). Wnt10b-positive placodes were observed in the footpads of E12.0 mutant embryos but not control embryos (see Fig. S2 in the supplementary material). Thus activated β-catenin can promote placode development even in normally hairless skin regions. Accelerated development of sweat glands was not observed in mutant footpads at later embryonic stages, suggesting that footpad placodes did not represent precocious sweat gland anlagen. Mammary glands developed in mutant female embryos, but exhibited histological abnormalities that will be described elsewhere.

The expression of Edar, assayed at E14.5, was elevated and more widespread in mutant than in littermate control skin(Fig. 2O,P). To determine whether EDAR signaling is activated in mutant placodes, we generated KRT14-Cre Ctnnb1^(Ex3)fl/+ mice carrying the NFκB reporter gene(Igκ)_3×cona-lacZ(κGal) (Schmidt-Ullrich et al., 1996). NF-κB activity was detected in KRT14-Cre Ctnnb1^(Ex3)fl/+ κGal skin at E14.5 and E15.5,but failed to show its normal regular pattern(Fig. 2Q-T). NFκB activity was not detected in mutant skin at earlier stages despite placodal expression of Wnt10b and Shh. Precociously initiated placodes may thus be more similar to secondary awl than primary hair follicle placodes, as development of the former, but not the latter, is independent of NFκB signaling.

Despite the formation of enlarged placodes in stabilized β-catenin mutant embryos at E14.5, hair follicles failed to extend into the dermis or to form multilayered structures (Fig. 3A-J). In some regions of the mutant embryos, large evaginations were noted that accumulated keratinized material and corresponded to the keratinized papules seen in whole mounts(Fig. 3B,D,J). Stratification of the epidermis appeared impaired, and fibroblasts in the upper dermis were more densely packed than were those in littermate controls(Fig. 3G,H).

To determine the molecular basis for these phenotypes, we examined the expression of hair follicle markers at E17.5. Shh is a specific marker for hair follicles within the skin(Chiang et al., 1999; St-Jacques et al., 1998), and,at this stage, its expression was confined to control hair follicle bulbs(Fig. 3K). In mutant embryos,expression of Shh was observed throughout the surface ectoderm,suggesting a widespread acquisition of hair follicle fate(Fig. 3L). Alkaline phosphatase activity marks the hair follicle dermal condensate and dermal papilla(Fig. 3M). Strikingly, in mutant embryos, alkaline phosphatase activity was detected broadly throughout most of the upper dermis (Fig. 3N). Together with the increased density of upper dermal fibroblasts, this abnormal staining indicated that upper dermal cells had broadly acquired at least some characteristics of the hair follicle dermal condensate. These data suggest that a secreted factor(s) expressed in response to cell-autonomous, stabilized epithelial β-catenin induced dermal condensation and alkaline phosphatase expression.

To determine whether markers for hair follicle differentiation were induced in mutant surface ectoderm, we analyzed the expression of GATA3, which is expressed specifically in inner root sheath precursors within the hair follicle and is required for their differentiation(Kaufman et al., 2003); Sox9, a marker for the hair follicle outer root sheath(Vidal et al., 2005); and hair shaft keratins that are specifically recognized by the AE13 antibody(Lynch et al., 1986). Expression of GATA3 and Sox9 was detected in the inner and outer root sheaths of control hair follicles, respectively, between E17.5 and birth, but the expression of either marker was barely detectable in mutant ectoderm(Fig. 3O-R). Immunostaining for AE13 was confined to hair shaft precursor cells in controls, but, remarkably,in the mutant was broadly and strongly present throughout the surface ectoderm(Fig. 3S,T). Thus, among the various hair follicle epithelial lineages, mutant ectodermal cells appeared to differentiate predominantly towards hair shaft rather than towards inner or outer root sheath.

Consistent with hair shaft-like differentiation of mutant ectodermal cells,at E17.5 mutant dorsal skin displayed a mean of 10±3 Ki67-positive cells per field at 20× magnification compared with 29±5 in the equivalent region of control skin (n=3 embryos and two fields counted per embryo for each genotype; P=0.0024; Fig. 3U,V). The decreased proliferation observed in mutant ectoderm was largely due to the absence of defined hair follicle matrix cell populations. TUNEL-positive cells were detected broadly in mutant surface ectoderm, but in controls were mostly detected in terminally differentiating upper layers of the epidermis(Fig. 3W,X).

Abnormal hair follicle development could be due to global alterations in the surface ectoderm prior to the stage when placodes are normally initiated,rather than to specific effects within developing follicles. To test this, we generated KRT5-rtTA tetO-Cre Ctnnb1^(Ex3)fl/+embryos, in which the mutation of β-catenin to a stabilized form can be induced in epithelial cells by dosage of the pregnant mother with oral doxycycline, and KRT5-rtTA tetO-Cre Ctnnb1^+/+ littermate controls. Untreated KRT5-rtTA tetO-Cre Ctnnb1^(Ex3)fl/+ embryos and adult mice lacked overt phenotypes. The induction of β-catenin mutation from E13.5, the stage at which primary placode initiation is just beginning,resulted in the formation of enlarged, abnormal placodes, and a failure of hair follicle downgrowth (Fig. 4A,B). Similarly, induction from E15.5, a stage after the formation of primary hair follicle placodes, blocked hair germ extension and resulted in defects in epidermal stratification (defects were assayed at birth; Fig. 4C-F). Pigment accumulation was noted in mutant but not control littermate skin at birth(Fig. 4G,H). Thus, the activated β-catenin mutation causes defective hair follicle development and abnormal pigmentation when induced after the onset of hair follicle placode formation.

To analyze defects in epidermal stratification at the molecular level, we carried out immunofluorescence for specific markers for basal, spinous and granular epidermal layers at E17.5. Stabilized β-catenin mutants exhibited reduced levels of the basal layer marker KRT14 compared with littermate controls (Fig. 5A,B). Immunofluorescence for the spinous layer markers KRT1 and KRT10 and the granular layer markers filaggrin and loricrin, was prominent in control littermate embryos, but virtually undetectable in mutant surface ectoderm (Fig. 5C-H). Expression of the desmosomal components DSC1 and DSG1, which normally localize to suprabasal epidermal layers, was markedly reduced in mutant surface ectoderm (Fig. 5I-L). These data are consistent with a complete failure of epidermal differentiation and a global conversion of ectodermal cells to a hair follicle fate.

The striking defects in epidermal stratification in stabilizedβ-catenin mutants suggested that barrier formation was abnormal. To test this, we subjected mutant and control littermate embryos to dye exclusion assays between E16.5 and E18.5. Control embryos exhibited barrier formation on their dorsal surface at E17.5 and were able to exclude dye completely by E18.5. By contrast, mutant embryos failed to exclude dye at all stages examined (see Fig. S3 in the supplementary material).

To determine the impact of these changes on skin ultrastructure, we subjected control and stabilized mutant skin to TEM at E16.5. Control epidermis displayed the expected stratifying structure, with readily identifiable basal cells, spinous cells characterized by dense networks of keratin filaments, and granular cells containing keratohyalin granules. By contrast, mutant surface ectoderm lacked the characteristics of suprabasal,spinous and granular cells (Fig. 5O,P). Pigment granules were noted in some regions of the mutant ectoderm (Fig. 5Q). Consistent with the unaffected expression of laminin α5 in mutant skin(Fig. 5M,N), the basement membrane was present in the mutant, and lamina lucida and lamina densa structures were similar to controls. However, hemidesmosomes were reduced in number compared with control skin (Fig. 5R,S). Notably, large spaces were apparent between mutant epithelial cells (Fig. 5P),suggesting defective cell-cell adhesion. Consistent with cell-adhesion defects, E-cadherin was downregulated in mutant basal cells; by contrast,P-cadherin localized intensely to control hair germs, but in mutants was strongly expressed throughout the surface epithelium (see Fig. S4 in the supplementary material).

To begin to dissect the molecular mechanisms underlying the dramatic effects of stabilized β-catenin on embryonic surface ectodermal development, innervation and pigmentation, we used microarray analysis to obtain global transcript profiles of mutant and stabilized β-catenin mutant dorsal dermis and epidermis at E15.5, and of unseparated dorsal skin at E14.5 (GEO accession number GSE10733). We analyzed >18,000 transcripts that were flagged as present in at least two of four samples, representing>14,000 genes. Genespring software was used to normalize the raw data, and to identify genes with a Student's t-test P-value of less than 0.05 and a change in relative expression levels of at least 2-fold in KRT14-Cre Ctnnb1^(Ex3)fl/+compared with control samples.

Analysis at E15.5 revealed major changes in the transcriptional profile of mutant compared with control epithelium, with 3124 probe sets identified as>2-fold upregulated and 3173 probe sets >2-fold downregulated in mutant epithelium at E15.5 (see GEO accession number GSE10728). These numbers overestimate the actual number of differentially expressed transcripts, as some transcripts were detected by multiple probe sets. Selected genes differentially expressed in mutant and control epithelium at E15.5 are listed in Table 1. A smaller number of probe sets displayed differential expression in mutant compared with control dermis at E15.5 (97 >2-fold upregulated and 55 >2-fold downregulated;see GEO accession number GSE10727). The altered gene transcript profile in mutant dermis presumably occurs in response to secreted factors expressed downstream of mutant epithelial β-catenin. A subset of the affected epidermal and dermal genes, including the placode markers Shh and Ptch1, have already demonstrated altered expression at E14.5 (121>2-fold upregulated and 49 >2-fold downregulated signals; see GEO accession number GSE10726). The upregulation of additional Shhpathway members, Ptch2, Smo and Gli3, was detected in mutant ectoderm at E15.5. Note that the expression of some epithelial-associated genes appears in the E15.5 dermal transcript data set. In situ hybridization and immunostaining experiments indicate that these aberrations are due to difficulties in completely separating mutant epithelial and dermal tissue before RNA extraction, rather than to misexpression(Fig. 3S,T; data not shown). Wnt/β-catenin signaling pathway genes known to be regulated by Wnt showed increased expression in mutant ectoderm, providing validation of the microarray data (see Table 1).

Consistent with the results of immunostaining experiments with AE13,transcript levels for multiple hair shaft-specific keratins and keratin-associated proteins were dramatically increased in the mutant. These data were confirmed for selected hair keratins by quantitative RT-PCR(Krt85 and Krt32; Fig. 6A) and immunohistochemistry (KRT82; Fig. 6B,C). Also markedly upregulated was a battery of transcripts expressed in hair shaft precursor cells, and/or required for hair shaft differentiation (see Table 1). Adamts20 and Kitl, which regulate melanoblast migration(Rao et al., 2003), and Foxn1, which controls pigment deposition(Weiner et al., 2007), were overexpressed, and could contribute to the premature pigmentation observed in mutant skin. In line with our finding of increased nerve fiber density in mutant skin, transcripts for several axon guidance molecules, including Sema4c, Sema3c, and the netrin receptor genes Unc5b and Unc5c, were markedly upregulated in the mutant.

The stem cell-associated genes Lrig1(Jensen and Watt, 2006) and Krt15 (Morris et al.,2004) were downregulated in mutant ectoderm, consistent with the lack of expression of Sox9, which is required for establishing the hair follicle stem-cell compartment (Vidal et al., 2005), and with our observation that, in this developmental context, stabilized β-catenin promotes ectodermal differentiation towards hair shaft, rather that maintaining cells in an undifferentiated state.

An unexpected finding was that the transcription factor gene Sp5and its relative Sp8 were substantially upregulated in mutant epidermis. Wnt regulates Sp8 expression in the developing limb(Kawakami et al., 2004). Sp5 is regulated by Wnt in diverse developmental and disease processes, and its proximal promoter region contains five evolutionarily conserved TCF/LEF-binding sites that mediate direct activation by canonical Wnt signaling (Fujimura et al.,2007; Takahashi et al.,2005; Weidinger et al.,2005). Differential expression of Sp5, Sp8 and selected additional genes from the microarray analyses was confirmed by quantitative RT-PCR (Fig. 6A). Sp5expression localized to hair follicle placodes in wild-type embryos in a similar pattern to Topgal Wnt reporter activity(Fig. 6D,E), and was increased in gain of function β-catenin mutant ectoderm(Fig. 6F-I). Conversely, in ectoderm from embryos expressing the Dkk1 Wnt inhibitor under the control of a Krt5 promoter (Chu et al., 2004), Sp5 expression, analyzed by quantitative RT-PCR and in situ hybridization, was downregulated(Fig. 6J-L). We confirmed by ChIP analysis that Sp5 is a direct target of β-catenin in surface ectodermal cells in vivo (Fig. 6M).

Sp5 can function as a transcriptional repressor(Fujimura et al., 2007),suggesting it as a candidate factor mediating the downregulation of epidermal differentiation genes in mutant β-catenin surface ectoderm. To test this,we transfected an expression vector for mouse Sp5 into the HaCAT human keratinocyte cell line under differentiating conditions. Sp5suppressed the expression of mRNAs encoding the squamous differentiation markers KRT10 and involucrin relative to their levels in empty vector-transfected cells (Fig. 6N). Thus, Sp5 may mediate some of the effects ofβ-catenin in downregulating a subset of epidermal differentiation genes. Interestingly, the expression of DLX3, which is predominantly associated with hair shaft differentiation(Morasso et al., 1995), was increased by transfected Sp5.

To test the effects of inducing mutation of β-catenin in adult epidermis, we analyzed hair follicle development in KRT5-rtTA tetO-Cre Ctnnb1^(Ex3)fl/+ experimental and KRT5-rtTA tetO-Cre Ctnnb1^+/+ littermate control mice placed on oral doxycycline from postnatal day (P) 35-P40 or P50-P70, and in adult KRT5-CreERT2 Ctnnb1^(Ex3)fl/+ mice treated from P56-P63 with tamoxifen or vehicle control applied topically to dorsal skin after hair clipping. For both methods of Cre induction, histological analysis of induced mutant skin revealed the initiation of hair follicle development from existing hair follicle outer root sheaths and epidermis, resulting in the development of multiple, tightly packed hair follicles (see Figs S5, S6 in the supplementary material). The mutant follicles were associated with sebaceous glands,expressed hair follicle matrix and outer root sheath markers (see Fig. S6 in the supplementary material), and produced functional hair shafts (Fig. S5 in the supplementary material). Induced KRT5-rtTA tetO-Cre Ctnnb1^(Ex3)fl/+ footpads did not display an obvious increase in the number of sweat glands, but contained ectopic, abnormal hair follicles developing from the epidermis (see Fig. S5H-K), suggesting that stabilizedβ-catenin may preferentially induce hair follicle development in this context.

These data indicate that the stabilized β-catenin mutation produces distinct phenotypes in embryonic and adult skin(Fig. 7). Nuclear localizedβ-catenin was detected prominently in adult induced hair follicles (see Fig. S6D′-F′ in the supplementary material), but was not as uniform as in induced embryonic ectoderm(Fig. 1L), possibly accounting in part for the different phenotypic outcomes. Semi-quantitative PCR analysis of floxed and recombined floxed Ctnnb1 alleles revealed that the efficiency of recombination in induced adult epidermis is similar to, or slightly greater than, that in embryonic surface ectoderm (see Fig. S5L in the supplementary material), suggesting that differential β-catenin accumulation is not primarily due to differences in recombination efficiency. Instead, stage and context dependent activation of the β-catenin promoter, positive feedback signaling in embryonic and hair follicle epithelium, and/or inhibitory effects in established, adult epidermis may affect β-catenin expression and function in different subsets of cells. Lack of an embryonic skin phenotype in previously described KRT14-ΔNβcat transgenic mice(Gat et al., 1998) could be due to β-catenin-mediated downregulation of the KRT14 promoter,in contrast to positive feedback regulation of epithelial β-catenin mRNA levels by stabilized β-catenin (Liu et al., 2007), or to subtle activity differences in theΔNβcat and exon 3-deleted β-catenin mutant proteins.

We have demonstrated here that an activating mutation in epithelialβ-catenin had dramatic effects on embryonic surface ectoderm. Hair follicle placodes were expanded, but failed to develop normally, instead precociously expressing hair shaft keratins(Fig. 7). Mutant embryonic skin was prematurely pigmented and innervated, and epidermal stratification was disrupted, resulting in a total failure of epidermal barrier function. Our data support a model in which Wnt/β-catenin signaling plays multiple roles in the embryonic surface ectoderm, promoting placode and hair shaft fate, activating signals that attract nerve fibers and enhance pigmentation,and blocking squamous differentiation.

Although these experiments rely on a gain-of-function approach, many of the genes displaying differential expression in our microarray profiling experiments are known from previous studies to be expressed in developing hair follicles, and/or are suppressed in embryonic skin expressing the secreted Wnt inhibitor DKK1 or lacking β-catenin(Andl et al., 2002; Huelsken et al., 2001). Furthermore, we were able to verify placodal expression in wild-type embryos and decreased expression in Wnt-inhibited skin for Sp5, a gene not previously associated with embryonic hair follicle development.

Placode induction occurs at least two days earlier in gain-of-functionβ-catenin mutants than in wild-type ectoderm, indicating that stabilizedβ-catenin overrides the requirement for an initiating dermal message. Uncontrolled ectodermal β-catenin signaling induces the formation of a broadly condensed dermis associated with widespread acquisition of dermal condensate markers. Thus, dermal condensates can be either initiated, or stabilized and expanded, by signaling from the epithelium. These data suggest a model in which the final positioning of hair follicles is determined not by a fixed pattern of dermal condensates, but rather by the interplay of epithelial and dermal signals. Consistent with this model, stable dermal condensates are not observed in mice lacking epithelial Eda signaling(Schmidt-Ullrich et al.,2006).

Our results further indicate that controlled downregulation ofβ-catenin signaling is required to generate a regular array of placode and interplacode fate, consistent with a reaction-diffusion model for establishment of the placodal array (Jiang et al., 2004), involving Wnt/β-catenin as the primary placode promoting signal, and predicting an important function in placode spacing for secreted Wnt-inhibitory factors whose expression is promoted within placodes by β-catenin. Microarray analysis revealed increased expression in gain-of-function mutant β-catenin skin of several secreted Wnt inhibitors, including Dkk4, Dkk1 and Wif1, consistent with prior reports and mathematical modeling analyses(Bazzi et al., 2007; Maini et al., 2006; Sick et al., 2006; Stark et al., 2007). Secreted Wnt inhibitors act at the level of Wnt receptors and/or co-receptors, and so would be ineffective at modulating the activity of epithelial gain-of-function mutant β-catenin. Thus, the balance between activating and inhibitory signals is shifted in favor of placode formation in stabilized β-catenin mutant ectoderm, resulting in an expanded adoption of placode fate(Fig. 7).

The precocious development of placodes, the increase in their density, and their formation in normally hairless areas, suggests that the broad distribution of placodes in mutant embryos is due, at least in part, to a switch in fate of the embryonic surface ectoderm, rather than being wholly accounted for by an expansion of cells normally destined to adopt placode fate. Deletion of the intracellular Wnt pathway inhibitor Apcproduces a milder phenotype (Kuraguchi et al., 2006), possibly due, in part, to the central, integrating role of β-catenin in the Wnt/β-catenin signaling pathway and to compensation of Apc functions by other pathway inhibitors.

Our phenotypical analyses at successive developmental stages indicate thatβ-catenin first promotes ectodermal placode fate, and indirectly enhances the attraction of nerve fibers to the skin. Rather than proliferating,penetrating the dermis, and developing into a mature multilayered hair follicle, mutant placodes precociously express hair shaft markers and exhibit pigmentation. The predominance of hair shaft differentiation is consistent with the hypothesis that, at certain developmental stages, β-catenin acts as a specific activator of hair shaft fate(Fig. 7). Our data further suggest that β-catenin signaling either directly or indirectly suppresses the expression of Gata3, a transcription factor required for normal development of the inner root sheath.

Our results indicate that Sp5 is a direct transcriptional target of β-catenin in epithelial cells, and may suppress the expression of a subset of epidermal differentiation genes, including Krt10 and involucrin, contributing to the adoption of placode rather than stratified epidermal fate. Sp1-binding sequences in the involucrin promoter are required for normal expression levels in vivo (Crish et al., 2006), and Sp5 competitively suppresses Sp1 target genes in other developmental systems (Fujimura et al., 2007), providing a possible mechanism for its effects on involucrin expression. Sp5 is expressed in wild-type placodes,suggesting a role in normal placode development. Sp5-null mice show no overt phenotype (Harrison et al.,2000); however, we speculate that Sp5 may act redundantly with Sp8 or other related genes in the skin. Activated β-catenin appears to directly or indirectly inhibit Sp1 transcript accumulation(Fig. 6A), which could provide an additional mechanism by which epidermal differentiation gene expression is suppressed in placodes.

Although the activating β-catenin mutation promotes hair follicle fate in the skin, previous analyses of KRT14-Cre Ctnnb1^(Ex3)fl/+ mutant embryos revealed enhanced and accelerated development of fungiform taste papillae and taste buds in the tongue, and ectopic tooth development in the maxilla and mandible(Jarvinen et al., 2006; Liu et al., 2008; Liu et al., 2007). Thus, the effects of stabilized ectodermal β-catenin exhibit a remarkable regional specificity. These observations suggest the existence of pre-existing regional signals that differ between the skin, and the oral epithelia of the tongue and jaws, and that determine the precise nature of the cellular responses downstream of β-catenin.

While this manuscript was under revision, Nähri et al. published data similar to those described here (Nähri et al., 2008).

We thank Elaine Fuchs for Topgal reporter mice, Adam Glick for K5-rtTA mice, Don Baldwin for Affymetrix array analysis, Qian-Chun Yu for TEM, Jean Richa for transgenic mouse production, Henry Sun, Pierre Coulombe and Lutz Langbein for keratin antibodies, Leroy Ash and Tzvete Dentchev for histology, and Uri Gat and George Cotsarelis for helpful discussions. This work was supported by NIH grants RO1-AR47709, R01-DE015342,RO1-HD053829 (S.E.M.) and RO1-AR45973 (A.A.D.).