Control of hair follicle cell fate by underlying mesenchyme through a CSL–Wnt5a–FoxN1 regulatory axis

Compromised hair follicle differentiation in mice with mesenchymal deletion of the RBP-Jκ gene

Little is known about the function of Notch signaling in cells of the mesenchymal lineage. “Canonical” Notch signaling is critically dependent on the DNA-binding protein CSL/RBP-Jκ (Bray 2006). To delete this gene specifically in the mesenchyme, transgenic mice with Cre expression driven by the promoter of the collagen type I (α2 chain) gene (ColI-Cre) (Florin et al. 2004) were crossed with mice with the RBP-Jκ gene flanked by loxP sites (Han et al. 2002). Further crossing of mice into a ROSA26R background confirmed that, as reported previously (Florin et al. 2004), Cre activity in the skin was confined to dermal papilla cells and dermal fibroblasts (Supplemental Fig. 1). Specificity of deletion of the RBP-Jκ gene in the mesenchymal compartment of the skin (dermal papilla cells and dermal fibroblasts) was further verified by physical separation of various skin components, followed by real-time PCR analysis of genomic DNA (Supplemental Fig. 2). Analysis of embryonic skin showed specific deletion of the RBP-Jκ gene in the mesenchymal compartment as early as embryonic day 15.5 (E15.5) (data not shown).

Phenotypically, homozygous (ColI-Cre × RBP-Jκ^loxP/loxP) mice were recognizable after birth by their “kinky” tail and smaller size, with areas of decreased hair density (Fig. 1A). They also exhibited a skeletal phenotype that will be presented elsewhere. Histologically, hair follicles were structurally normal at birth (Supplemental Fig. 3), but, by 4–10 d, a significant percentage started to show structural alterations, with disrupted inner root sheath (IRS) and cuticle layer organization, followed by hair shaft/cortex alteration and signs of cystic degeneration (Fig. 1B,C; Supplemental Fig. 4). Immunofluorescence analysis showed substantially decreased expression of 44-kDa/46-kDa acidic hair keratins, a marker of cortex and cuticle layers of the hair shaft, and trichohyalin, a marker of medulla and IRS (Lynch et al. 1986), while, conversely, expression of markers normally confined to the interfollicular epidermis (cytokeratins 1 and 10) was detectable in hair follicles (Fig. 1D,E). This “switch” between hair follicle and interfollicular marker expression was further documented and quantified by real-time RT–PCR analysis of physically isolated hair follicles from the (ColI-Cre × RBP-Jκ^loxP/loxP) mice and age-matched controls (Fig. 1F). Degeneration of hair follicles into cysts became widespread by 2–3 wk, with the mature cysts exhibiting high levels of interfollicular marker expression (Supplemental Fig. 5A–C). Vibrissae hair follicles were also normal after birth, but were then lost in the mutant mice by 3–4 wk, undergoing similar cystic degeneration as pelage hair follicles (Supplemental Fig. 5D,E).

Versican and alkaline phosphatase staining showed that dermal papillae remained in contact with hair follicles as well as the cysts that formed over time (Fig. 1G,H). Expression of another dermal papilla cell marker, Sox2, was also similar in control and mutant mice (Fig. 1I). Sox2-positive dermal papilla cell populations were shown recently to be essential for production of auchene and awl hair fibers (Driskell et al. 2009). Consistent with our expression data, plucked hair analysis showed that mutant mice produced all four types of hair in an approximately normal ratio (Supplemental Fig. 6A). Reflecting the hair follicle structural defects, a large percentage of all hair types showed abnormal packing and irregular fiber profile, with aberrant restrictions and bending (Supplemental Fig. 6B).

As with many other mutations, mice with keratinocyte-specific deletion of the RBP-Jκ gene exhibit also hair follicle degeneration and cyst formation (Blanpain et al. 2006). We therefore compared in detail alterations resulting from deletion of the RBP-Jκ gene by the Cre transgene driven by a ColI versus keratin 5 (K5) promoter, using mice of a young age and at an early stage of hair follicle degeneration (Supplemental Fig. 7). In both cases, deletion of the RBP-Jκ gene resulted in markedly decreased expression of acidic hair keratins and trichohyalin, and K6 up-regulation. However, while in the (ColI-Cre × RBP-Jκ^loxP/loxP) mice there was pronounced hair follicle expression of suprabasal interfollicular cytokeratins 1 and 10 (K1 and K10), no such expression was found in the (K5-Cre × RBP-Jκ^loxP/loxP) mice (Supplemental Fig. 7B). As reported previously (Blanpain et al. 2006), K1/K10 expression was decreased even in the interfollicular epidermis of mice with keratinocyte-specific RBP-Jκ deletion, while in mice with the mesenchymal deletion, expression of these genes was increased (Supplemental Fig. 7C). This was paralleled by the expected decrease of direct Notch target genes Hes1 and Hey1 in mice with the first genotype, but not in mice with the second genotype (Supplemental Fig. 7C). Besides differentiation marker expression, dermal papilla cells, as identified by versican and Sox2 staining, were normally present in the skin of mice with the ColI-Cre-driven deletion but not K5-Cre-driven deletion (Supplemental Fig. 7D), consistent with distinct phenotypes resulting from RBP-Jκ deletion in the two skin compartments.

Intrinsically defective hair-inducing capabilities of dermal papilla cells with RBP-Jκ deletion

To assess whether the phenotype of the (ColI-Cre × RBP-Jκ^loxP/loxP) mice is skin intrinsic or secondary to systemic alterations, we resorted to full-thickness skin grafting onto nude mice. Grafts of RBP-Jκ mutant skin showed aberrant hair follicle differentiation and cyst formation, as in the original mice with this mutation (Fig. 2A).

The interesting possibility exists that RBP-Jκ deletion in extrafollicle dermal cells contributes to the observed phenotype. To specifically establish whether the hair follicle alterations can be attributed to a defect in dermal papilla cells, we performed hair follicle reconstitution assays (Zheng et al. 2005) using freshly isolated dermal papilla cells with or without RBP-Jκ deletion in combination with wild-type keratinocytes. The latter were obtained from mice with a GFP transgene driven by an actin promoter (Okabe et al. 1997), which allowed unequivocal identification of hair follicles reconstituted by donor cells relative to endogenous hair follicles of the recipient mice. Efficient hair follicle formation was observed with GFP-positive keratinocytes admixed with control dermal papilla cells, while only few hair follicles (∼10-fold difference) were formed by the same keratinocytes admixed with RBP-Jκ-deficient papilla cells (Fig. 2B). Instead, the latter combination of cells resulted in widespread formation of cysts, which were rarely observed in the graft controls (Fig. 2C). Immunofluorescence analysis showed that the aberrantly reconstituted hair follicles and cysts formed by GFP-positive keratinocytes in combination with RBP-Jκ^−/− dermal papilla cells had—as in the (ColI-Cre × RBP-Jκ^loxP/loxP) mice—reduced acidic hair keratins and trichohyalin expression, and were positive for cytokeratins 1 and 10 (Fig. 2D).

The Wnt5a gene is a direct Notch/CSL target in dermal papilla cells

Gene expression studies were undertaken to probe into the molecular defects resulting from the RBP-Jκ deletion in dermal papilla cells. As expected, real-time RT–PCR analysis showed that the “canonical” Notch–RBP-Jκ target genes Hes1, Hes5, and Hey1 were down-modulated in dermal papilla cells isolated from mice with the deleted RBP-Jκ gene (Fig. 3A). Expression of signaling molecules that make part of the “dermal papilla-specific signature”—Wnt5a, Fgf7, Fgf10, and Noggin—was strongly decreased in the mutant mice, while expression of other dermal papilla “signature” genes, including Sox2 (Rendl et al. 2005; Driskell et al. 2009), was maintained or even increased (Fig. 3A). Differential expression of Wnt5a, Noggin, and Fgf7/10 was confirmed at the protein level by immunoblot analysis of dermal papilla cell extracts (Fig. 3B) and immunofluorescence of corresponding skin sections (Fig. 3C). Modulation of these genes, similar to that observed in vivo, was found with cultured dermal papilla cells from (ColI-Cre × RBP-Jκ^loxP/loxP) mice (Fig. 3D), as well as with dermal papilla cells with the floxed RBP-Jκ gene after “acute” deletion by adenovirally mediated Cre expression (Fig. 3E).

The Wnt5a, Noggin, and Fgf7/10 genes code for diffusible factors that may be under parallel RBP-Jκ control, or regulation of one of them might play a primary role on the others. To address this question, dermal papilla cells with the RBP-Jκ deletion were treated with each of these factors individually, in a purified form, followed by analysis of expression of the others. As shown in Figure 4A, treatment of the RBP-Jκ^−/− dermal papilla cells with Wnt5a rescued, to a partial extent, expression of Noggin, Fgf7, and Fgf10 expression. The converse treatment with Noggin, Fgf7, or Fgf10 had no effect on Wnt5a levels (Fig. 4B), suggesting that Wnt5a is a primary target of RBP-Jκ function.

Consistent with this conclusion, activated Notch1 expression by adenoviral vector infection, or tamoxifen-dependent activation of a retrovirally transduced Notch1–ER fusion protein (Schroeder and Just 2000), induced Wnt5a in wild-type dermal papilla cells, but not in RBP-Jκ-deficient cells (Fig. 4C). Induction of Wnt5a expression was similarly observed upon endogenous Notch activation by treatment of dermal papilla cells with purified Jagged 1 ligand (Fig. 4D). Nucleotide sequence analysis of the Wnt5a gene promoter and first intronic regions showed the presence of “canonical” RBP-Jκ-binding sites, to which the endogenous Notch1 protein was found to bind by chromatin immunoprecipitation (ChIP) assays with control dermal papilla cells but not cells with the RBP-Jκ deletion (Fig. 4E). In parallel, luciferase reporter assays showed that activated Notch1 expression induces activity of a Wnt5a promoter construct containing RBP-Jκ-binding sites, but not one with the sites deleted (Fig. 4F).

Compromised hair follicle differentiation in mice with deletion of the Wnt5a gene

Mice with homozygous deletion of the Wnt5a gene die at birth due to a number of developmental defects (Yamaguchi et al. 1999; Yang et al. 2003). So far, the consequences of Wnt5a deletion in the skin have not been analyzed. Histological analysis of E18.5 embryos showed that loss of Wnt5a, as the mesenchymal RBP-Jκ deletion, causes no detectable defects in hair follicle morphogenesis and differentiation at this time (data not shown).

To assess the role of Wnt5a after birth, E17.5 embryo skins were grafted onto nude mice. Hair follicle abnormalities started to be seen by 1 wk after grafting of the Wnt5a mutant skin, becoming more substantial by 2 and 4 wk (Fig. 5A,B; Supplemental Fig. 8). In parallel with these alterations, immunofluorescence analysis showed consistent down-modulation of acidic hair keratins and inappropriate cytokeratin 10 expression, with dermal papilla cells (as detected by versican staining) still present (Fig. 5C). In the latter, however, there was strong reduction of Noggin and Fgf10 expression, consistent with our other results pointing to Wnt5a as an upstream regulator of these molecules (Fig. 5D).

While Wnt5a expression is limited to dermal papilla cells during development, after birth it also extends to the keratinocyte compartment (Reddy et al. 2001; Rendl et al. 2005). To determine to what extent the hair follicle alterations of the Wnt5a^−/− grafted skin can be attributed to defective dermal papilla cell function, hair follicle reconstitution assays were undertaken with dermal papilla cells isolated from Wnt5a^−/− versus control littermates in combination with keratinocytes harboring an intact Wnt5a gene and expressing GFP. Wnt5a-deficient dermal papilla cells, like those with the RBP-Jκ deletion, were similarly impaired in their hair follicle-inductive properties, inducing instead extensive cyst formation (Fig. 5E,F). Even in this case, immunofluorescence analysis showed down-modulation of hair follicle differentiation markers and inappropriate cytokeratin 1 expression (Fig. 5G).

Functional rescue of defective RBP-Jκ^−/− hair follicles by Wnt5a

To assess whether the defects caused by deletion of RBP-Jκ can be attributed to decreased Wnt5a expression, intact skin of postnatal day 4 (P4) mice was placed in culture and incubated for 4 d in either the presence or absence of exogenous Wnt5a. The fraction of hair follicles undergoing cystic degeneration over a 4-d incubation period was significantly decreased as a consequence of Wnt5a treatment (Fig. 6A,B). In parallel, hair follicle differentiation-related genes were up-regulated and interfollicular markers were down-modulated in hair follicles isolated from the Wnt5a-treated skins (Fig. 6C). Similar rescue of gene expression defects was observed when purified follicles from the mutant mice, instead of skin explants, were placed in culture and directly treated with Wnt5a (Fig. 6D).

FoxN1 as a target and mediator of RBP-Jκ/Wnt5a signaling in hair follicles

Wnt5a has been implicated as an activator of noncanonical Wnt signaling, although, in some contexts, it can also induce the canonical pathway (Mikels and Nusse 2006). In the functional rescue experiments above, real-time RT–PCR analysis of isolated hair follicles from cultured skin showed that Wnt5a treatment caused no modulation of canonical Wnt target genes like Lef1, Axin2, or DKK1. In contrast, expression of FoxN1, a keratinocyte-specific transcription factor reported previously to be under Wnt control (Balciunaite et al. 2002), was substantially induced (Fig. 7A). Immunofluorescence analysis confirmed that Wnt5a treatment caused a significant increase of FoxN1 expression in hair follicles of cultured (ColI-Cre × RBP-Jκ ^loxP/loxP) skin (Fig. 7B). Levels of FoxN1 expression were similarly analyzed in hair follicle reconstitution assays with dermal papilla cells with or without deletion of the RBP-Jκ and Wnt5a genes. The results indicated that sustained FoxN1 expression is under dermal papilla cell control through a RBP-Jκ- and Wnt5a-dependent mechanism (Fig. 7C,D).

A number of transcriptional targets of FoxN1 have been reported previously, including Phospholipase C-δ1 (Plcd1) (Janes et al. 2004; Nakamura et al. 2008) and desmocollin 2 (DSC2) (Johns et al. 2005). Similar down-modulation of these genes, as well as trichohyalin, was found in cultured hair follicles with siRNA-mediated knockdown of FoxN1 expression as in hair follicles derived from mice with the RBP-Jκ deletion (Fig. 7E,F). Conversely, Wnt5a treatment led to induction of these genes in wild-type hair follicles, with no such induction occurring in similarly treated hair follicles derived from FoxN1 mutant mice (Fig. 7G).

Involvement of the Wnt5a receptors ROR2 and Fzd6 in control of FoxN1 expression

Wnt5a can function through multiple receptors, like the Frizzled family member Fzd6 and ROR2, a transmembrane tyrosine kinase with an extracellular Fzd-like cystein-rich domain (van Amerongen et al. 2008). Real-time RT–PCR analysis showed that expression of both receptors is reduced in hair follicles as a consequence of RBP-Jκ or Wnt5a deletion (Supplemental Fig. 9A), and is induced by Wnt5a treatment (Supplemental Fig. 9B), consistent with a positive feedback regulatory mechanism reported in other systems (O'Connell et al. 2010). Functionally, knockdown of ROR2 in hair follicles led to down-modulation of FoxN1 (Supplemental Fig. 9C) and suppressed induction of this gene, as well as its downstream targets, by Wnt5a treatment (Supplemental Fig. 9D). Similar results were observed after Fzd6 knockdown (Supplemental Fig. 9E,F).

We complemented the above findings by assessing the consequences of loss of ROR2 function in vivo. Mice with homozygous disruption of the ROR2 gene die at birth due to breathing defects (Takeuchi et al. 2000). Immunofluoresence analysis of the skin of these mice after full-thickness grafting onto nude mice revealed that expression of FoxN1 as well as trichohyalin was reduced consistently in the hair follicles of these mice (Supplemental Fig. 10A,B). However, this was associated with only slight alterations in hair follicle structures (in IRS and/or hair shaft organization) that did not proceed to overt cyst formation, consistent with ROR2 functioning in concert with Fzd6 and possibly other Wnt5a receptors.

Mice

RBP-Jκ^loxp/loxp mice (Han et al. 2002) were crossed with transgenic mice expressing a Cre recombinase under control of the promoter/enhancer unit of the α2 chain of the collagen type I gene (Florin et al. 2004) or the keratin 5 promoter (Berton et al. 2000) to generate (ColI-Cre × RBP-Jκ^loxP/loxP) and (K5-Cre × RBP-Jκ^loxP/loxP) mice, respectively. (ColI-Cre × RBP-Jκ^loxP/wt) mice were also crossed into the ROSA 26R reporter strain, an in vivo reporter of Cre activity (Soriano 1999). Transgenic mice with EGFP expression driven by a ubiquitous β-actin promoter (C57BL/6-Tg [Act-EGFP]) were described (Okabe et al. 1997). B6;129S7-Wnt5a^tm1Amc/J mice (Yamaguchi et al. 1999) were purchased from The Jackson Laboratory, and NMRI wild-type and Nude mutant (heterozygotes and homozygotes) were from Center d'Elevage R. Janvier. ROR2 knockout mice were described previously (Oishi et al. 1999). All mouse work except for ROR2 mice was performed according to the Swiss guidelines and regulations for the care and use of laboratory animals. For ROR2 mice, work was performed according to International Animal Care and Use Committee guidelines and regulations of Tulane University.

Skin cell preparation and purification

The whole procedure is illustrated in Supplemental Figure 2A, along with expression analysis of various markers to assess the effectiveness of the separation procedure. For separation of interfollicular epidermis from the dermis, back skin from P1 mice was treated with 1% Dispase II (Roche) overnight at 4°C. For hair follicle separation, the dermis was further digested by 0.3% type I collagenase (Sigma) in HBSS (Gibco) for 1 h at 37°C. After centrifugation in a tabletop centrifuge (1000 rpm for 5 min, followed by 400 rpm for 5 min twice), the sedimented material was subjected to two rounds of density gradient centrifugation—the first in 10% Percoll (d = 1.01), and the second in 33% Percoll (d = 1.05)—collecting the bottom fraction each time. Microscopic examination showed >90% hair follicle purification by this approach. The recovered hair follicles were further dissociated into cell suspension by mechanical disruption (20 strokes through a 23-gauge and then a 26-gauge needle). Cells were then centrifuged and resuspended into 500 μL of culture medium and incubated with rat anti-mouse c-Kit (eBioscience) and rat anti-mouse CD133 (eBioscience) antibodies (at a 1:100 dilution) for 1 h at room temperature. Cells were then washed twice in culture medium, followed by incubation with goat anti-rat microbeads (Miltenyi Biotech) for 30 min at 4°C. Magnetic cell sorting, using a MACS midi separator and LS columns (Miltenyi Biotech), was then used to deplete cells positive for c-Kit (melanocytes) and CD133 (dermal papilla cells), and the remaining cells (hair follicle keratinocytes) were used for further analysis.

For separation of dermal papilla cells and dermal fibroblasts, cells from 1% collagenase-treated dermis were collected by centrifugation, followed by resuspension in 500 μL of culture medium and incubation with rat anti-mouse c-Kit and rat anti-mouse CD34 (Abcam) antibodies (1:100). Magnetic cell sorting was used to remove the c-Kit-positive (melanocytes) and CD34-positive (endothelial) cells. Cells recovered after this depletion procedure were further incubated with rat anti-mouse CD133 antibodies, and magnetic cell sorting was used to separate them in into a CD133-positive fraction (dermal papilla cells) and CD133-negative fraction (dermal fibroblasts). For hair follicle reconstitution and dermal papilla cell culture assays, dermal papilla cells were directly isolated from dermal cell preparations by magnetic cell sorting with anti-CD133 antibodies.

For the isolation of mature hair follicles (from P10 mice) and cysts (from 3-wk-old mice), the total skin was at first digested in 1% type I collagenase for 2 h at 37°C. Hair follicles and cysts were then purified by the two-step density gradient centrifugation described above. Hair follicles with relatively normal versus abnormal structures from the (ColI-Cre × RBP-Jκ^loxP/loxP) mice, as well as the cysts, were further manually isolated under a dissecting stereomicroscope (Leica MZ 16F).

Cell culture, viral infection, and plasmids

For culturing, the sorted dermal papilla cells from newborn mice were plated at an initial density of 2 × 10⁵ per 10-cm dish and cultured in DMEM (Gibco) and 10% fetal bovine serum (FBS) (Sigma) for 3 d (to 60%–70% confluence) prior to further analysis. For in vitro deletion of the RBP-Jκ gene, cultured dermal papilla cells from (RBP-Jκ^loxP/loxP) mice were infected with an adenovirus expressing activated Notch1 or the Cre recombinase together with GFP, in parallel with a control virus expressing GFP alone, at a multiplicity of infection of 100, which resulted in >99% of GFP-positive infected cells. Adenoviruses and our conditions for adenovirus production and infection were as reported previously (Rangarajan et al. 2001). As an alternative approach, dermal papilla cells were infected with a retrovirus expressing constitutively active Notch1 fused to an estrogen receptor domain (rNERT-neo) (Schroeder and Just 2000) and empty vector control (rneo), followed 4 d later by treatment with 1 μM 4-hydroxytamoxifen (OHT) for 48 h to induce nuclear translocation and activation of Notch-dependent transcription.

For treatment of dermal papilla cells with Jagged 1 ligand, 2 mL of goat anti-rat IgG (R&D Systems) (diluted in PBS) were added to nontreated 10-cm plastic dishes at a concentration of 10 μg/mL and incubated for 30 min at 37°C. Dishes were washed five times with PBS, followed by blocking with 2 mL of 2% BSA in PBS for 16 h at 4°C. The blocking medium was then replaced with 2 mL of ligand solution. For this, recombinant rat Jagged 1 protein (R&D Systems) was diluted in 2% BSA in PBS to a final concentration of 10 or 20 μg/mL. Dishes were incubated with the Jagged 1 ligand solution for 2 h at 37°C, followed by five washes with PBS. Freshly isolated dermal papilla cells from P1 wild-type mice were then seeded at a concentration of 200,000 cells per dish and cultured in DMEM (Gibco) and 10% FBS (Sigma) for 24 h.

For Wnt5a promoter activity assays, the mouse Wnt5a promoter and first intronic region (3.943 kb, spanning from −1.66 to +2.29 kb relative to the transcription start site) was amplified by PCR of genomic DNA using the following primers: 5′-CCTTCTACCCTTGAACGTTG-3′ (Forward) and 5′-TTGAAGGACTGGGACTCAGA-3′ (Reverse). The product of the PCR was cloned into the pGL4 vector (Promega) at the BglII/HindIII sites to generate the Wnt5a-luc reporter plasmid. For removal of the two RBP-Jκ-binding sites, located at positions −1.64 and +1.81 kb relative to the transcription start site, the Wnt5a-luc plasmid DNA was digested by BglII/SmaI to remove 368 base pairs (bp) of 5′ upstream sequence, enclosing the first site, and by BsaI/MluI, for an internal deletion of 233 bp (+1.78 to +2.01 kb), enclosing the second site. The Wnt5a-luc plasmid and its Wnt5(ΔRBP-Jκ)-luc derivative were cotransfected into Saos-2 cells (in 12-well plates; 1 μg of DNA per well) together with an expression vector for activated Notch1 (2 μg of DNA per well) (Rangarajan et al. 2001) or empty vector control and the renilla phRL-TK reporter (Promega) for internal normalization. Cells were harvested 48 h after transfection, and were assayed with Dual Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions. All experiments were performed in triplicate.

Skin and hair follicle organ cultures

Full-thickness back skin from mice of the indicated age and background was placed in semisolid agarose medium (0.25% agarose [Sigma] in DMEM + 10% FBS), with the upper surface of the skin exposed to the air interface. For Wnt5a treatment experiments, two halves of the same skin samples were placed in culture in separate wells, followed, after 24 h of adaptation to culture conditions, by transfer to fresh semisolid medium containing either Wnt5a (200 ng/mL) or BSA alone. Samples were further transferred to freshly supplemented semisolid medium 48 h later. Recombinant Wnt5a protein was purchased from a commercial source (R&D Systems) and dissolved in 1% BSA-PBS to achieve 100 μg/mL stock solution.

For hair follicle cultures, hair follicles physically separated from the skin of P0 mice of the indicated background were placed onto 12 wells with organotypic inserts (Millipore) (∼1,000 follicles per well), with 2 mL of DMEM + 10% FBS per well and 0.5 mL of the same added to the insert. Samples were adapted to culture conditions for 24 h prior to treatment with Wnt5a (200 ng/mL final concentration) or BSA alone.

For siRNA-mediated knockdown experiments, P0 hair follicles, isolated as described above, were resuspended in HBSS (without Ca²⁺ and Mg²⁺) (Gibco) and transferred to six-well plates (Ultra-Low attachment microplates; Corning). About 1000 hair follicles were added per well, in 1.8 mL of culture medium, and kept on ice. siRNA transfection was performed with INTERFERin reagent (Polyplus) according to the manufacturer's recommendations. For each sample, 20 μL of siRNAs (from 10 μM stock) was added to 80 μL of 5% glucose and mixed with 12 μL of INTERFERin diluted into 88 μL of 5% glucose by gentle vortex for 10 sec. The mix was then incubated for 10 min at room temperature before adding to the hair follicle HBSS suspension, with a final siRNA concentration of 100 nM. The transfected hair follicle suspension was kept for 24 h at 4°C, followed by replacing HBSS with 2 mL of low-calcium mouse keratinocyte culture medium (Dotto et al. 1988) and culturing for 4 d at 34°C in 7.5% CO₂ condition. Medium was changed after 48 h (by hair follicle centrifugation at 450 rpm for 5 min at 4°C). For Wnt5a treatment, culture medium was replaced for the last 24 h of the experiment with 2 mL of fresh culture medium containing recombinant Wnt5a (200 ng/mL final concentration) or BSA alone (0.05%).

Hair follicle reconstitution assays

For each injection, 1 × 10⁶ freshly isolated CD133⁺ dermal papilla cells from P1 mice (for [ColI-Cre × RBP-Jκ^loxP/loxP] and [RBP-Jκ^loxP/loxP]) and E18.5 embryos (for Wnt5a control and knockout mice) were admixed with 0.5 × 10⁶ freshly isolated P1 keratinocytes from actin-EGFP transgenic mice. The mixed cells were resuspended in 50 μL of DMEM + 10% FBS, and were injected at the dermal–hypodermal junction into NMRI nude mice as described previously (Zheng et al. 2005). Each mouse was injected in parallel, on the left and right flank, with keratinocytes admixed with either (ColI-Cre × RBP-Jκ^loxP/loxP) or (RBP-Jκ^loxP/loxP) dermal papilla cells, or with Wnt5a^−/− versus Wnt5a^+/+ dermal papilla cells. Eight to 12 mice were injected, being sacrificed at various times after injection as indicated.

Full-thickness skin grafting

Back skin of Wnt5a^tm1Amc/^tm1Amc and littermate wild-type control mice at E17.5, and (ColI-Cre × RBP-Jκ^loxP/loxP) and (RBP-Jκ^loxP/loxP) at P0, were cut into 5 × 5-mm pieces, washed thoroughly in PBS, and placed onto the recipient graft bed of nude mice (prepared by removing epidermis and superficial dermis with curved surgical scissors). To minimize individual animal variability, control and mutant skin were grafted in parallel on the left and right back flank, respectively, of each recipient mouse. Grafts were covered with a 16 × 57-mm wound bandage and further secured with adhesive tape. Bandage and adhesive tape were changed every day. Three days later, grafts were exposed to air until harvest.

Immunofluorescence staining and confocal analysis

The following primary antibodies were used: AE13 and AE15 (1:25; kindly provided by Dr. T.T. Sun), CD133 (1:200; eBiosciences), Cre (1:200; Novagen), cytokeratin 1 (1:500; Covance), cytokeratin 10 (1:500; Covance), cytokeratin 14 (1:1000; Covance), cytokeratin 6 (1:1000; Covance), FoxN1 (1:50; Santa Cruz Biotechnology), GFP (1:500; Abcam), Sox2 (1:50; R&D Systems), Versican (1:200; Chemicon), and Wnt5a (1:100; R&D Systems). The secondary antibodies were conjugated with Alexa Fluor 488 or 568 (Invitrogen). Counterstaining of nuclei was performed with DAPI.

Embryos or newborn pups were cryosectioned in sagittal direction at 20 μm at the midline of the body. Sections were fixed in precooled acetone for 20 min at −20°C, permeabilized for 20 min in PBST (PBS with 0.1% Triton X-100), and incubated in blocking solution (PBS with 2.5% donkey serum, 2.5% goat serum, 0.5% cold water fish gelatin, 0.5% BSA) for 60 min. Incubation with primary antibodies was for either 2 h at room temperature or 12 h at 4°C. For immunostaining of Cre and CD133, enzymatically dissociated total dermal cells were fixed in 4% PFA for 20 min on ice and then stained with the above protocol.

Images were acquired with a Zeiss LSM510 meta laser-scanning microscope (Carl Zeiss MicroImaging). For histological sections and whole-mount samples, Z stacks (1 μm) of six to eight and 20 planes, respectively, were acquired.

Quantitative real-time RT–PCR, immunodetection, and ChIP techniques

Conditions for real-time RT–PCR and immunoblotting were as described previously (Lefort et al. 2007). Immunoblotting was performed with the following antibodies: Wnt5a (R&D Systems), Fgf10 (Santa Cruz Biotechnology), Noggin (Santa Cruz Biotechnology), and β-actin (Santa Cruz Biotechnology).

ChIP assays were performed as described previously (Lefort et al. 2007). The list of gene-specific primers is provided in Supplemental Table S1.