|Home | About | Journals | Submit | Contact Us | Français|
Hair shafts are produced from stem cells located in the bulge. Our knowledge of the genetic pathways regulating cell fate acquisition in the immediate descendents of these stem cells, and fate maintenance in their committed progeny, is still incomplete. One pathway involved in fate maintenance within the hair matrix is the Notch pathway. Here we use compound genetic mutants to demonstrate that two transcription factors, Msx2 and Foxn1, are both required to maintain Notch1 expression in the hair follicle matrix. In their absence, Notch1 is markedly reduced in hair matrix; as a consequence, medulla and inner root sheath (IRS) differentiation is impaired. Our studies also suggest that Foxn1 is a direct activator of the Notch1 promoter activity through one or more putative Foxn1 consensus binding sites located within the 4.7 kb of mouse Notch1 promoter. Since recombinant human BMP4 can induce Foxn1 expression in Msx2-deficient hair follicles, and that their effect on cortical keratin expression appears synergistic, we suggest that these two genes function in parallel pathways downstream of BMP signaling and upstream of Notch1. Independent from their role in Notch activation, Msx2 and Foxn1 also contribute to the expression of several cortical and cuticle keratins. The impact of these additional defects is the complete loss of all visible external hairs, not seen in Notch1 mutants. Our results position Msx2 and Foxn1 upstream of Notch1 within the hair matrix and demonstrate that together these factors play a pivotal role in IRS, cortex and medulla differentiation.
The mammalian hair follicle is among many ectodermal organs (e.g., hairs, teeth, nails and exocrine glands) that originate from sequential and reciprocal interactions between two apposing tissue layers: the epidermis and dermis (Hardy, 1992; Millar, 2002; Pispa and Thesleff, 2003). This signal exchange between the two tissue layers leads to the downgrowth of the epidermal placode and the formation of the hair follicle, which contains a dermal papilla (DP) encapsulated by matrix cells and a bulge region containing slow cycling hair follicle stem cells (Cotsarelis et al., 1990; Tumbar et al., 2004). Once the basic finger-like structure of the hair follicle is established, rapidly proliferating progenitor cells respond to signals from the DP and generate transient amplifying daughters, which move upwards, exit the cell cycle and differentiate into concentric cylinders of post-mitotic keratinocytes: the medulla, cortex and cuticle of the hair shaft, the cuticle, Huxley's layer and Henle's layer of the inner root sheath (IRS), and the companion layer that separates the IRS and the outer root sheath (ORS) (Ito, 1986; Rothnagel and Roop, 1995; Stenn and Paus, 2001). The ORS lines the outermost layer of the hair follicle, which is continuous with the basal layer of epidermis.
A better understanding of how these cell lineages are generated from the multipotent progenitor matrix cell population in the hair follicle is beginning to emerge. Recently Legue and Nicolas provided evidence that lineage-restricted precursor cells of the IRS and of the hair shaft were organized into proximo-distal clonal columns adjacent to the DP (Legue and Nicolas, 2005). After an asymmetric cell division that produces a transient progenitor in the next radial layer, the progenitor cell will divide symmetrically to produce two post-mitotic cells which undergo terminal differentiation. According to this model, patterning of concentric layers of the hair follicle is predetermined by the arrangement of progenitor cells contacting the DP in the matrix. Thus, matrix cell fate decision and subsequent differentiation are tightly coupled to signals from the DP.
Several signaling molecules and/or their antagonists are reported to be expressed in the DP as well as in the hair matrix. Perturbations of these signaling pathways by either gain- or loss-of-function mutation lead to defective hair follicle differentiation. Fgf7 is expressed in the DP of anagen hair follicles (Rosenquist and Martin, 1996). Fgf7-deficient mice exhibit a greasy and matted hair phenotype similar to the rough mice (Guo et al., 1996). Loss-of-function mutations in Fgfr2-IIIb either by knockout or by over-expression of a dominant negative form lead to abnormal medulla differentiation (Petiot et al., 2003; Schlake, 2005). BMP signaling is absolutely required for hair follicle differentiation. Expression of both the ligands (BMP2, 4, and 8) and the BMP2/4 type I receptor is detected in hair follicles (Wilson et al., 1999; Zhao and Hogan, 1996). Among them, BMP4 is expressed in the DP and the hair matrix (Kulessa et al., 2000; Wilson et al., 1999). Attenuation of BMP signaling by either ectopic expression of Noggin using the Msx2 promoter or conditional ablation of BMPR1a impairs hair shaft and IRS differentiation and downregulates Msx1, Msx2, Foxn1 and Hoxc13 (Andl et al., 2004; Kobielak et al.,2003; Kulessa et al.,2000; Yuhki et al., 2004). Finally, Wnts and their receptors are expressed in postnatal hair follicles (Reddy et al., 2001; Reddy et al., 2004). Specifically, Wnt5a and Fz10 are detected in the DP. Lef-1, the transcription factor mediating Wnt signaling, is expressed in hair shaft precursor cells where Wnt activity is detected using the Wnt-responsive TOPGAL reporter (DasGupta and Fuchs, 1999). Alterations of Wnt signaling by ectopically expressing Wnt3 or DVL2 in the ORS lead to shortened and fragile hair shafts (Millar et al., 1999). BMP and Wnt signaling also coordinates to regulate hair regeneration (Huelsken et al., 2001; Ito et al., 2007; Kobielak et al., 2003; Ma et al., 2003).
Notch signaling is known to regulate cell fate specification and pattern formation (Lai, 2004). It is also required to maintain lineage-specific differentiation of the hair follicle. Notch1, 2, 3 and the ligands, Jagged1, 2 are expressed in mature hair follicles (Kopan and Weintraub, 1993; Pan et al., 2004; Powell et al., 1998). Upon activation by its ligand, Notch1 undergoes two proteolytic events leading to the release of Notch1 Intra-Cellular Domain (NICD), which translocates to the nucleus and binds to RBP-Jκ, converting it from a transcriptional repressor into an activator (Mumm and Kopan, 2000). Both Notch1 mRNA and activated Notch1 protein (NICD) are present in the precursor cells but not in oligo-lineage progenitor cells that retain direct contacts with the DP. Both Notch1 loss- and gain-of-function mutations in the hair follicle result in differentiation defects in neighboring cells, reflecting a role for Notch1 in maintaining hair follicle differentiation in a cell-nonautonomous fashion (Lin et al., 2000; Pan et al., 2004). So far, several upstream regulators of Notch1 have also been characterized in mice and human. Sox2 directly regulates Notch1 expression in retinal progenitor cells (Taranova et al., 2006). P53 binds Notch1 promoter in human epithelial cells (Lefort et al., 2007; Yugawa et al., 2007). However, direct regulators of Notch1 in hair follicles have not been identified.
In addition to signaling molecules, a multitude of transcription regulators also control hair follicle differentiation. Mutation in a winged-helix transcription factor, Foxn1, causes the nude phenotype and congenital athymia in both mice and humans (Frank et al., 1999; Nehls et al., 1994). In nude mice, impaired keratinization and structural defects were found in the hair shaft and IRS, resulting in the formation of shorter, broken hair shafts that seldom penetrate the skin surface (Kopf-Maier et al., 1990). Hoxc13 gain- and loss-of-function mutations in mice also result in severe hair differentiation defects (Godwin and Capecchi, 1998; Tkatchenko et al., 2001). Gata3, a Notch target and key regulator of T-cell lineage determination (Amsen et al., 2007a,b), is expressed in the IRS and inactivation of Gata3 leads to a failure in the differentiation of IRS precursors (Kaufman et al., 2003). Mutation in Msx2, a mammalian homolog of the Drosophila msh (muscle segment homeobox), results in cyclic alopecia, defects in hair shaft differentiation and reduced expression of Foxn1 and its target gene, acidic hair keratin mHa3 (Ma et al., 2003).
In this study, we have further investigated the genetic circuitry linking these various factors using genetically modified mice. Our data show that Msx2 and Foxn1 function downstream of BMP in a partially redundant manner. Foxn1 is probably a direct activator of Notch1 expression in the hair matrix. Foxn1 and Msx2 cooperatively maintain keratin expression. These results place Foxn1 and Msx2 as important regulators of hair shaft differentiation acting upstream to the Notch signaling pathway.
Msx2 mutant mice were generated previously and maintained on CD-1 background (Satokata et al., 2000). Foxn1 mutant mice were purchased from Charles River Laboratories (Wilmington, MA) and maintained on BALB/c background. Msx2tm1Rilm/+;Foxn1nu/+ double heterozygous mice were generated by crossing the two mutant stocks. Since the Foxn1 mutant allele cannot be genotyped easily, we bred the double heterozygotes with Foxn1 mutant mice to generate Msx2tm1Rilm/+;Foxn1nu/Foxn1nu which can be identified by the nude phenotype and possession of an Msx2 mutant allele. These mice are phenotypically indistinguishable from the nude mice. Genotyping for the Msx2 mutant allele was done as previously described (Satokata et al., 2000). Subsequently, Msx2tm1Rilm/+;Foxn1nu/Foxn1nu males were used to breed with Msx2 mutant females and approximately half of the pups should be Msx2tm1Rilm/Msx2tm1Rilm;Foxn1nu/+ genotype which is confirmed by genotyping at the Msx2 locus. Msx2/Foxn1 double mutant mice were generated by intercrossing double heterozygotes. Completely hairless mice were recognized as double mutants and confirmed by genotyping the Msx2 locus and by amplification and subsequent sequencing of the mutated Foxn1 gene. N1CKO mice were generated as described previously (Pan et al., 2004).
To generate N1(4.7)-EYFP mice, the 4.7 kb Notch1 promoter was released from clone pBSK-N1(RI) by EcoRI and NaeI digestion. After Klenow fill-in, this promoter sequence was inserted into the SmaI site of pEYFP-N1 (Clontech, Palo Alto, CA) to generate pEYFP-N1/N1P. SV40 intron was amplified from pcDNAI/Neo (Invitrogen, Carlsbad, CA), cloned into the pGEM T-Easy (Promega, Madison, WI) vector and sequenced. The primers for amplification were as followed: forward primer, 5′-(EagI)TGCTAGAGGATCTTTGTGAAGG-3′; reverse primer, 5′-(NotI)AGACATGATAAGATACATTG-3′. Restriction enzyme sites were designed to facilitate subsequent cloning. SV40 intron was then released from pGEM T-Easy vector by EagI digestion and inserted into the NotI site of pEYFP-N1/N1P. The final construct containing Notch1 promoter, EYFP and SV40 intron was released by XhoI and NotI digestion. Transgenic founders were generated by pronuclear injection of this construct into C57Bl/6×CBA embryos. N1(4.7)-EYFP founders were identified by PCR genotyping and yellow fluorescence in tail biopsies. For PCR, the forward primer is in Notch1 promoter: 5′-TGCCTTGTAGGGCCCAGCGC-3′; the reverse primer is in EYFP sequence: 5′-AGCTCAGGTAGTGGTTGTCG-3′. To generate Foxn1nu/Foxn1nu;N1(4.7)-EYFP, we first bred N1(4.7)-EYFP with Foxn1 mutant mice to produce Foxn1nu/+;N1(4.7)-EYFP mice. Foxn1nu/Foxn1nu;N1 (4.7)-EYFP mice were generated by further crossing Foxn1nu/+;N1(4.7)-EYFP with Foxn1 mutant mice. Mutagenesis of the Foxn1 consensus binding sites in the 4.7 kb Notch1 promoter was generated as specified (Supplemental Table S2) and N1(4.7)13mu-EYFP transgenic mice containing the mutated Notch1 promoter, EYFP and SV40 intron were generated as described above.
Dorsal skin samples were collected from postnatal day 7 (P7) mice, fixed overnight in 4% paraformaldehyde in PBS, embedded in paraffin and sectioned at 10 μm. Digoxigenin (DIG)-UTP-labeled cRNA probes were generated from the following templates: pCMV-Msx2 containing the Msx2 cDNA sequence (Ma et al., 2003), pBSK plasmid containing the Foxn1 cDNA sequence (Nehls et al., 1994), pNM1.2 plasmid containing the mHa3 cDNA sequence (Meier et al., 1999), clone CJ4a containing the 1.4 kb Jagged2 cDNA sequence (Jiang et al., 1998) and pBKS plasmid containing the Notch1 cDNA sequence (Kopan and Weintraub, 1993). To prepare cRNA probes for mHb6, mHa2, K2-19, K6irs, K17 and Jagged1, total RNA was isolated from P7 wild type mouse back skin and first strand cDNA was synthesized. PCR was then performed with gene-specific primers listed in Supplemental Table S1. PCR products were cloned into PCR4-TOPO (Invitrogen Life technologies, Carlsbad, CA), and antisense probes synthesized as specified (Supplemental Table S1). In-situ hybridization was carried out as previously described (Ma et al., 1998). Signals were visualized using anti-DIG antibody coupled to alkaline phosphatase (AP) conjugate (Roche, Indianapolis, IN) and AP substrates NBT and BCIP (Sigma, St. Louis, MO). To quantify K2-19 expression level in wild type and various mutant hair follicles, SYBR green-based real-time PCR was performed as described (Yin et al., 2006). The relative amount of PCR products was obtained after normalization with Gapdh. Primers used for real-time PCR were listed in Supplemental Table S1. t-Test with two-sample assuming equal variances were performed.
P7 skin sections were prepared at 5 μm as described above. Primary antibodies used were mouse anti-trichohyalin (AE15) (O'Guin et al., 1992) (1:500); rabbit anti-NICD (V-1744, Cell Signaling, Beverly, MA) (1:200), anti-Notch1 (Abcam, Cambridge, MA) (1:1000), anti-Cadherin6 (Cho et al., 1998) (1;100), anti-GFP (FL, Santa Cruz Biotechnology, Santa Cruz, CA), anti-Dsc2 (M-55, Santa Cruz Biotechnology) (1:50); goat anti-Foxn1 (G-20, Santa Cruz Biotechnology) (1:100) and anti-Jagged1 (C-20, Santa Cruz Biotechnology) (1:2000). Secondary antibodies used were fluorescein-coupled goat anti-mouse IgG (Jackson Laboratory, West Grove, PA), Alexa594-coupled anti-rabbit IgG (Molecular Probes, Eugene, OR), Alexa647-coupled anti-goat IgG (Molecular Probes), horse radish peroxidase (HRP)-coupled anti-rabbit IgG (GE Healthcare Bio-Sciences Corp., Piscataway, NJ) followed by tyramide-Cy3 (Cadherin6) (PerkinElmer Life And Analytical Sciences, Wellesley, MA) and biotin-coupled anti-sheep IgG followed by ABC reagent (Vector Laboratories, Burlingame, CA) and tyramide-fluorescein (Jagged1) (PerkinElmer). Sections were counter-stained with bis-benzimide (Sigma, St. Louis, MO). For Western blot, primary antibodies used were rabbit anti-β-tubulin (H-235, Santa Cruz Biotechnology) (1:400) and goat anti-Foxn1 (G-20, Santa Cruz Biotechnology) (1:500). Secondary antibodies used were HRP-coupled anti-rabbit IgG (GE Healthcare) and HRP-coupled anti-goat IgG (Sigma).
Skins were obtained from P3 wild type and Msx2 mutant mice. The dermis was separated from the epidermis and dissociated as previously described (Rogers et al., 1987). The dermal suspension containing hair follicles was collected by centrifugation at 1000 g for 3 min, washed three times in PBS and resuspended in serum-free DMEM. Recombinant human BMP4 (R & D systems, Minneapolis, MN) was added to the medium to a final concentration of 100 ng/ml before culturing in an incubator supplied with 5% CO2 at 37 °C for 24 h. The dermal suspension was then centrifuged and subjected to total RNA extraction and reverse transcription PCR (RT-PCR). For β-actin, Msx2 and Gata3, PCR products are shown after 30 cycles. Foxn1 PCR products are shown after 36 cycles. Foxn1 expression before and after hrBMP4 induction in both wild type and Msx2 mutants was quantified by real-time RT-PCR. Primers used for RT-PCR and real-time PCR were listed in Supplemental Table S1. t-Test with two-sample assuming equal variances were performed.
Previous studies showed that Msx2 and Foxn1 mRNA were detected both in the matrix and in the hair shaft of anagen hair follicles (Ma et al., 2003; Meier et al., 1999; Schlake and Boehm, 2001). Foxn1 proteins were detected in the distal part of the matrix above the line of Auber and the cortex and cuticle of the hair shaft (Johns et al., 2005). Foxn1 was downregulated in Msx2 mutant hair follicles at both mRNA and protein levels (Ma et al., 2003; Meier et al., 1999; Schlake and Boehm, 2001). We compared expression patterns of these two genes in P7 anagen hair follicles by in-situ hybridization using DIG-labeled cRNA probes. We detected expression of both Msx2 and Foxn1 in the hair matrix and differentiating hair cortex (Supplemental Fig. S1A). Since expression of both Msx2 and Foxn1 is dramatically reduced when BMP signaling is perturbed (Andl et al., 2004; Kobielak et al., 2003; Kulessa et al., 2000; Yuhki et al., 2004) and that Msx2 is upstream of Foxn1 (Ma et al., 2003; Meier et al., 1999; Schlake and Boehm, 2001), it's possible that BMP4 might regulate Foxn1 expression through Msx2. We directly tested this hypothesis by treating P3 wild type and Msx2-deficient hair follicles in suspension with recombinant human BMP4 proteins (rhBMP4) and examined the regulation of Foxn1 expression. In both wild type and Msx2-deficient follicles, rhBMP4 upregulated Msx2 and Foxn1 expression significantly, and Gata3 to a lesser extent, as assayed by RT-PCR (Supplemental Fig. S1B). Because quantitative real-time RT-PCR analysis showed that rhBMP4 induced a two-fold upregulation of Foxn1 in both wild type and Msx2 mutants (Supplemental Fig. S1C), we conclude that BMP4 regulates Foxn1 independently of Msx2. However, given that the overall Foxn1 level were reduced to less than half that of wild type in Msx2 mutant, Msx2 is required to maintain basal Foxn1 expression level in the hair follicles.
The overlapping expression of Msx2 and Foxn1 in the matrix and the cortex raises the possibility of functional interactions. If the two factors contribute to the same process, genetic interactions are expected. However no genetic interactions exist as Msx2tm1Rilm/+; Foxn1nu/+ mice displayed no detectable hair phenotype. To test whether Foxn1 is a genetic modifier of Msx2, we reduced Foxn1 expression in Msx2 mutants with the expectation that removing one functional Foxn1 allele should exacerbate the Msx2 mutant hair phenotype. Indeed, hair loss in Msx2tm1Rilm/Msx2tm1Rilm;Foxn1nu/+ compound mice was accelerated compared to that in Msx2 mutants: while Msx2 mutant mice began to lose hair at P14 (Satokata et al., 2000), hair loss in Msx2tm1Rilm/Msx2tm1Rilm;Foxn1nu/+ mice was detectable as early as P12 (data not shown). By P15, when Msx2 mutants started to lose hair in the head and neck region (Fig. 1A.a, c), Msx2tm1Rilm/Msx2tm1Rilm;Foxn1nu/+ mice were already bald and the few remaining hairs on their neck and back were sparse and curly (Fig.1A. b, d, f). The accelerated hair loss phenotype in compound mutants correlated with a further reduction of mHa3 and mHb6 transcripts in the hair cortex (Fig. 1B.b, c, f, g), but not as severe as in Foxn1 mutant hair follicles where their expression was completely abolished (Fig.1B. d, h). These results indicate that Msx2 and Foxn1 cooperatively control cortical keratin expression during hair differentiation.
If these two genes function in parallel pathways during hair differentiation, Msx2/Foxn1 double mutant hair phenotype should be more severe than that of either single mutant. On the other hand, Msx2 and Foxn1 could also function in a linear genetic pathway with Foxn1 executing functions downstream of Msx2. In that case, the double mutant phenotype should resemble that of Foxn1. To differentiate between these two possibilities, we generated Msx2/Foxn1 double mutant mice. As shown in Fig. 3, Msx2-deficient, Foxn1-deficient and compound (Msx2tm1Rilm/Msx2tm1Rilm;Foxn1nu/+) mice all have vibrissa and pelage hairs at P11, with Foxn1 mutants displaying short escaper pelage hairs (Figs. 2A–C). In contrast, Msx2/Foxn1 double mutant mice have no external hairs of any kind at P11 including pelage, vibrissa and eyelash hairs, and none appear throughout their adult lives (Fig.2D and data not shown). We characterized more than 20 double mutants and all displayed this phenotype whereas their littermates carrying other genotypes still possessed pelage hairs, indicating that the double mutant phenotype was not the result of mixed genetic background. Because this phenotype is more severe than either of the single mutant, it reflects additive effect from two parallel pathways. Notably, Msx2 and Foxn1 act downstream of patterning processes: double mutant animals induced a similar number of hair follicles compared to that in wild type mice (Fig. 3A.a, d).
To characterize the hair differentiation defects in Msx2/Foxn1 double mutants, we examined the expression of multiple terminal differentiation markers, such as keratins that mark one or more layers of hair shaft and IRS in P7 anagen hair follicles. A previously uncharacterized keratin, K2-19, was strongly expressed in the differentiating cortex of wild type anagen hair follicles (Fig. 3A.a). Its expression was visibly decreased in Msx2 mutants, barely detectable in Foxn1 mutants and absent in double mutant hair follicles (Fig. 4A.b–d, Supplemental Fig. S4). Quantitative real-time PCR revealed that compared to wild type, K2-19 expression level was reduced to 27.27%±4.54% (p=5.0×10−5, n=3) in Msx2 mutants, 14.91%±1.93% (p=8.86×10−8, n=3) in Msx2tm1Rilm/Msx2tm1Rilm; Foxn1nu/+ mice, 10.27%±1.14% (p=8.54×10−9, n=3) in Foxn1 mutants and 0.03%±0.02% (p=3.0×10−16, n=3) in double mutant hair follicles (Fig. 3B), indicating a synergistic effect of Msx2 and Foxn1 on K2-19 expression. Quantification of another cortical keratin mHa3 expression showed similar findings (data not shown). mHa2 is an acidic hair keratin that is specifically expressed in the hair shaft cuticle (Winter et al., 1994) (Fig. 3A.e). Its expression was not altered in Msx2 mutant hair follicles but was absent in both Foxn1 and the double mutant hair follicles (Fig. 3A.f–h). Expression of an IRS marker, K6irs (Aoki et al., 2001), was not altered in Msx2 mutant but is reduced in Foxn1 and the double mutant hair follicles (Fig. 3A.i–l). Staining with a medulla and IRS marker, monoclonal antibody AE15 (stains trichohyalin granules), showed that cells comprising both layers are present in all three mutants examined except that the double mutant medulla was disorganized (Fig. 3A.m–p, insets). We also detected a reduction in K17 expression in the double mutant medulla, but not in the ORS (Supplemental Fig. S5). To exclude the possibility that the observed changes in keratin gene expression resulted from the shortened anagen in Msx2 mutants (Ma et al., 2003), we repeated these experiments in both P3 and P5 double mutant hair follicles and similar findings were obtained at both time points (Supplemental Fig. S6 and data not shown). Altogether, the double mutant mice showed differentiation defects in both the hair shaft and the IRS.
Another pathway acting downstream of patterning to regulate maintenance of follicular fates is the Notch signaling pathway. Medulla and IRS differentiation is impaired in Notch1-deficient hair follicles (Pan et al., 2004). We therefore asked whether the defects in the Msx2 and/or Foxn1 mutants reflected perturbation in Notch1 activity. In-situ hybridization revealed normal Notch1 expression and activation levels in Msx2 mutants but significantly reduced expression in Foxn1 mutants (Figs. 4A–C). Notch1 mRNA was further reduced in the double mutants (Fig. 4D). Staining with the antibody recognizing only the activated form of Notch1 (NICD, Lin and Kopan, 2003) provided confirmation that Notch1 activation is greatly affected by loss of Msx2 and Foxn1. In both wild type and Msx2 mutant hair follicles, NICD is excluded from lineage-restricted precursor cells but present in their differentiating descendants (Lin and Kopan, 2003; Pan et al., 2004) (Figs. 4E, F). In marked contrast, NICD is significantly reduced in Foxn1 mutants and barely detected in Msx2/Foxn1 double mutant hair follicles at P7 as well as at P5 (Figs. 4G, H and Supplemental Fig. S6). In contrast, the expression of two Notch ligands, Jagged1 and Jagged2, was largely unaffected in any of the mutants except for a moderate reduction in Jagged1 expression in the double mutant matrix (Figs. 4I–L, M–P).
To ask if FOXN1 directly regulated Notch1 expression in hair follicles, we first cloned a ~4.7 kb Notch1 proximal promoter fragment upstream of EYFP. Three N1(4.7)-EYFP transgenic lines show identical EYFP expression pattern in various tissues but differ in intensity; we used a line with the strongest EYFP expression for subsequent analysis. This 4.7 kb Notch1 sequence is sufficient to drive tissue-specific reporter gene expression in the skin and kidney but not in the vascular component (data not shown). In the skin, EYFP was detected in the interfollicular epidermis (including ectopic expression in the basal layer), the ORS and the matrix of the hair follicle (Fig. 5 and Supplemental Fig. S8). If the proximal Notch1 promoter is directly or indirectly regulated by Foxn1, expression of transgene should be reduced in Foxn1 mutants. We therefore generated Foxn1nu/Foxn1nu;N1(4.7)-EYFP mice and compared level of EYFP expression to that in the N1(4.7)-EYFP hair follicles by both immunohistochemistry and in-situ hybridization. EYFP antibody detected expression in the ORS and in matrix cells (Fig. 5A.c, d). As observed with the endogenous locus, EYFP expression in matrix cells was dramatically downregulated in Foxn1nu/Foxn1nu;N1(4.7)-EYFP hair follicles as assayed by both immunohistochemistry and in-situ hybridization; expression in the ORS and in the collecting ducts of the kidney remained unaffected (Fig. 5A.a–f). Endogenous membrane-associated Notch1 protein was also significantly reduced in Foxn1nu/Foxn1nu;N1(4.7)-EYFP hair follicles (Fig. 5A.g, h).
Previous studies showed that FOXN1 protein binds to an 11-bp consensus sequence with a core sequence of 5′-ACGC (Schlake et al., 1997). We surveyed the 4.7 kb promoter sequence and found a total of 13 ACGCs. To eliminate any possibility of FOXN1 binding, we mutated all 13 ACGCs and generated N1(4.7)13mu-EYFP transgenic mice (Fig. 5B). Again, three transgenic lines were characterized and all showed identical EYFP expression patterns with variable expression levels. We selected a line that exhibited similar EYFP level to the N1(4.7)-EYFP mice in the kidney (Fig. 5B.a–c) and compared their EYFP expression in hair follicles. Similar to what was observed in Foxn1nu/Foxn1nu;N1 (4.7)-EYFP mice, both EYFP mRNA and protein levels were dramatically downregulated in N1(4.7)13mu-EYFP hair matrix (Fig. 5B.d, e, g, h). In contrast, EYFP expression in the ORS of N1(4.7)13mu-EYFP hair follicles was unaffected (Fig. 5B.d, e). Live fluorescent images confirmed the above observations and showed that EYFP fluorescence was greatly reduced in N1(4.7)13mu-EYFP hair matrix but not ORS compared to those in N1(4.7)-EYFP (Supplemental Fig. S8). If Foxn1 regulates Notch1 solely through one or more of the 13 putative Foxn1 binding sites, then N1(4.7)13mu-EYFP transgene activity in the matrix should not be further reduced by Foxn1 mutation. We therefore tested this by crossing N1(4.7)13mu-EYFP into Foxn1 mutant mice. As predicted, we did not observe a further reduction in EYFP expression in Foxn1nu/Foxn1nu;N1(4.7)13mu-EYFP hair follicle matrix compared to that in N1(4.7)13mu-EYFP hair follicles (Fig. 5B.f, i). EYFP expression in the kidney and ORS was affected neither by mutations in the binding sites nor by Foxn1 mutation (Fig. 5B.a–f). One the other hand, endogenous Notch1 expression was greatly reduced in Foxn1nu/Foxn1nu;N1(4.7)13mu-EYFP hair follicles (Fig. 5B.j–l). Together, these results strongly suggest with Foxn1 directly regulating Notch1 expression in the hair matrix by binding to one or more of the 13 putative binding sites.
Previous studies showed that Notch1-deficient hair follicles exhibit shortened anagen that can be observed as early as P11 (Vauclair et al., 2005). However, catagen onset was not accelerated when Notch1-deficient follicles reside close to wild type follicles (Lee et al., 2007). To examine whether Notch1 downregulation in our double mutant mice can lead to shortened anagen, we investigated double mutant hair follicle histology during late anagen. We found that anagen hair follicles were still present on the back skin of double mutant mice as late as P15, with a histology similar to that in wild type mouse (Figs. 6A–D), indicating that these hair follicles have not prematurely entered catagen. Likewise, we did not observe a premature catagen entry phenotype in Msx2 mutant as previously described (Ma et al., 2003) which could be due to differences in mouse strain used and/or the region of skin sampled (dorsal back vs. cervical region). On the other hand, the double mutant hair follicles developed cysts underneath the skin epidermis, similar to Foxn1 mutant hair follicles (data not shown). To determine whether the double mutant hair follicles can complete a hair cycle, we examined second anagen induction in wild type and various mutants. Both wild type and Msx2, Foxn1 single mutant mice regenerated hair follicles during the second anagen and exhibited mature anagen hair structure at P37 (Fig. 6E and data not shown). Unexpectedly, the double mutant hair follicles were not regenerated. Only abnormal follicle remnants were seen under the cysts, indicating that Msx2 and Foxn1 may also have important functions during hair follicle regeneration.
The hair follicle is a powerful system to study epithelial– mesenchymal interactions during organogenesis. Msx genes play important roles during organogenesis, and are specifically involved in signal transmission between opposing tissue layers (Maas et al., 1996). Msx2 function is required in the epithelial differentiation of various ectodermal organs, such as the tooth (Bei et al., 2004), the hair (Ma et al., 2003) and the mammary gland (Satokata et al., 2000). During hair differentiation, a BMP/Msx2/Foxn1/acidic hair keratin genetic pathway was proposed that regulates cortex differentiation. In this paper, we further refined the genetic relationship between Msx2 and Foxn1 during hair differentiation and identified Notch1 as a possible direct transcriptional target of Foxn1 in the matrix.
Both Msx2 and Foxn1 are required during hair differentiation and Foxn1 appears to be genetically downstream of Msx2 as its expression is reduced in Msx2 mutant hair follicles (Ma et al., 2003). Msx2tm1Rilm/Msx2tm1Rilm;Foxn1nu/+ mutant hair phenotype suggests that hair differentiation is sensitive to Foxn1 and Msx2 gene dosage. Foxn1nu/+ mice do not have any hair differentiation defect and mHa3 expression is not changed in these mice (Inoue et al., 2004). Western blot showed that Foxn1 protein level in Foxn1nu/+ was indeed reduced to approximately 50% that of wild type (Supplemental Fig. S2). Therefore, just reducing Foxn1 level by 50% is not sufficient to cause a reduction in cortex keratin expression and subsequent abnormal differentiation of the hair, perhaps due to functional redundancy with other factors. The reduced mHa3 expression observed in Msx2 mutant mice most likely results from reduced activity of these redundant proteins. Since a 50% reduction of Foxn1 in Msx2tm1Rilm/Msx2tm1Rilm;Foxn1nu/+ mice results in a more severe hair differentiation defect than Msx2 mutants alone, we interpret this phenotype to mean that the Msx2 mutation offers a sensitized genetic background which brings out Foxn1nu/+ phenotypes and as a result further enhances the Msx2 mutant phenotype. Foxn1 expression is only reduced by 50% in the absence of Msx2 protein, suggesting that other factors also regulate Foxn1 expression. Using in vitro follicle culture we demonstrated that BMP4, a known regulator of Msx2 and Foxn1 expression, could induce Foxn1 expression in the absence of Msx2 (Supplemental Fig. S1). This indicates that Msx2 contributes to the basal Foxn1 expression level and that BMP signaling activates Foxn1 independent of Msx2. Moreover, the Msx2/Foxn1 double mutant hair phenotype was more severe than that of either single mutant, never developing any external vibrissae, eyebrows and pelage hairs throughout their lives. Thus, these two genes do not appear to function in a simple linear genetic pathway but rather in parallel pathways that interact during hair differentiation.
In the hair matrix, Msx2 appears to regulate Fgfr2 expression independent of Foxn1 (He, Cai and Ma, unpublished observation); Foxn1 regulates K6irs expression in the IRS independent of Msx2. Although we have not identified unique Msx2 targets in the hair cortex, the effect of these two proteins on common targets such as hair keratin expression appears synergistic which also support a parallel rather than a linear pathway. One possible explanation for our observed genetic relationship between Msx2 and Foxn1 is that the two proteins physically interact and form a transcriptional complex. However, no interactions were observed in either GST pull down or co-immunoprecipitation experiments (data not shown). Therefore, the cooperativity we report might be explained by co-regulation on common targets, such as cortical keratins and Notch1. Altogether, we deduced the existence of a genetic pathway in which BMPs regulate Msx2 and Foxn1 in a parallel fashion, Msx2 regulates basal Foxn1 expression, and together these factors ensure cortical keratin and Notch1 expression in the hair follicle (Fig. 7).
Investigation of terminal differentiation marker expression in Msx2 and Foxn1 single mutants and compound mutants revealed that Msx2 and Foxn1 control lineage specific differentiation of postnatal hair follicles. Msx2 specifically controls the expression of cortical keratins and keratin-associated proteins. In addition to the markers presented, we found that expression of all known cortical keratins and over a dozen keratin-associated proteins were also downregulated in Msx2 mutant mice by RT-PCR (Supplemental Fig. S3). Foxn1 mutant mouse has a much more severe hair phenotype than Msx2 mutant; consistently, expression of many cortical keratins is nearly lost in Foxn1 mutant follicles. In addition, Foxn1 also contributes to cuticle and IRS differentiation, as mHa2 and K6irs expression is affected in Foxn1 mutant hair follicles. More severe differentiation defects in all three layers of the hair shaft and IRS were seen in Msx2/Foxn1 double mutant hair follicles with cortical differentiation being affected the most, reflecting partial redundancy between these proteins. It is interesting that the double mutant hair phenotype is very similar to that in BMPR1a-deficient mouse, suggesting that Msx2 and Foxn1 are the major effectors downstream of BMP signaling in the hair follicle. Since Msx2 and Foxn1 are not expressed in the medulla, they must regulate medulla differentiation in a cell-nonautonomous fashion. Our data suggest that this effect reflects attenuation of Notch1 signaling in Msx2/Foxn1 double mutants. Since Msx2 and Foxn1 expression is preserved in Notch1-deficient hair follicles (Supplemental Fig. S7), and Foxn1 is likely a direct regulator of Notch1 expression in the matrix (Fig. 5), we place Msx2 and Foxn1 upstream of Notch1 in the signal hierarchy leading to hair differentiation (Fig. 7). In this scenario, BMP signals can indirectly activate Notch1 expression through Msx2 and Foxn1. However, since Notch1 appears to function only in IRS and medulla differentiation, overexpressing NICD under the control of a proper promoter in the hair matrix may only rescue the medulla phenotype of the BMPR1a-deficient or Msx2/Foxn1 double mutant hair follicles, but not the cortex phenotype.
The Msx2/Foxn1 double mutant hair follicle revealed a previously unappreciated role for both proteins in regulating Notch1 expression. Notch1 expression overlaps with both Msx2 and Foxn1 in the matrix. Notch1 expression is significantly reduced in Foxn1 mutant hair matrix and barely detected in the double mutants. Our transgenic studies indicate that the 4.7 kb Notch1 promoter recapitulates endogenous Notch1 expression in some organs, including the hair follicle. Foxn1, Notch1 and EYFP transcripts are co-localized in the matrix cells that are not in direct contact with the dermal papilla (Supplemental Fig. S11). Furthermore, the transgene can also recapitulate endogenous Notch1 regulation by Foxn1 in the hair matrix. Site-directed mutagenesis of all 13 putative Foxn1 binding sites led to a significantly downregulated transgene expression specifically in the hair matrix but not in other sites of expression such as the kidney, the interfollicular epidermis and the ORS. Finally, Foxn1 mutation did not further reduce transgene reporter expression level carrying the 13 mutations indicating that these 13 putative Foxn1 binding sites are sufficient to confer Foxn1 regulation on Notch1. Together, these data strongly suggest a direct regulation of Notch1 by Foxn1. However, Foxn1 proteins were not detected in the proximal hair bulb (Johns et al., 2005) (Supplemental Fig. S7), although its transcripts were present (Ma et al., 2003; Meier et al.,1999; Schlake and Boehm, 2001) (Supplemental Figs. S1 and S11). One explanation is that the antibody is not sensitive enough to detect low levels of Foxn1 protein. Alternatively it is also possible but less likely that factors other than Foxn1 can regulate Notch1 expression through one or more of these 13 sites in the hair matrix. Nonetheless, it is clear that Notch1 functions downstream of Foxn1 in the hair matrix, and thus it is likely that some of the effects of Foxn1 on medulla differentiation may be mediated through Notch1. Consistently, Desmocolin 2 (Dsc2), a Foxn1 target in the medulla may also be Notch1 target. Medulla Dsc2 is expressed in the interfaces between two vertically adjacent medulla cells and between cortex and medulla cells, both of which are regulated by Foxn1 (Johns et al., 2005) (Supplemental Figs. S9A, B). Dsc2 expression is significantly reduced in the first interface but not the latter in Notch1-deficient follicles possibly reflecting a defective medulla but a normal cortical structure (Supplemental Figs. S9C, D).
At this point, we are unclear how Msx2 contributes to Notch1 expression in the matrix. Msx2 clearly contributes to the maintenance of Notch1 expression as Notch1 expression is further reduced in the double mutant matrix compared to Foxn1 mutant alone. The hairless phenotype of the double mutants is seen at birth and must reflect the combined effect of the two genes on cortical keratin expression (e.g. K2-19) as well as of their respective targets (e.g. Fgfr2 and K6irs), as residual Notch1 expression is still present at that time. Consistently, Cdh6, a Notch1 target, is still expressed in the double mutants but not in Notch1-deficient hair follicles (Supplemental Fig. S10). Likewise, no premature catagen entry was observed in the double mutant hair follicles in contrast to Notch1-deficient hairs (Vauclair et al., 2005). These results demonstrate that residual Notch1 expression in Msx2/Foxn1 double mutant hair follicles is still sufficient to maintain Cdh6 expression and prevent early entry of the hair follicles into catagen. We should expect Msx2/Foxn1/Notch1 triple mutant mice to exhibit a more severe hair phenotype than that of the Msx2/Foxn1 double mutants alone. Although the first anagen appears normal in the double mutant hair follicles, the second anagen is not induced correctly (Fig. 6). At this point, we are not sure whether this reflects a delay in second anagen induction or a complete failure of the second hair germ formation in the double mutants. Further investigation into the roles of these two transcription factors during hair follicle regeneration is warranted.
In summary, our studies describe the genetic circuitry operating downstream of BMPs. Msx2 and Foxn1 are independently induced to function synergistically to control hair shaft keratin expression. Foxn1 contributes to the maintenance of Notch1 expression which contributes indirectly to medulla and IRS differentiation. Together all three genes control hair shaft differentiation.
This work was supported by National Institutes of Health grants HD41492 to L.M. and GM55479 to R.K. We thank the Digestive Diseases Research Core Center Murine Models Core (supported by NIH P30DK052574) for production of transgenic mice. The authors would like to thank Dr. Tung-Tien Sun for AE15 antibody and Dr. Gregory Dressler for Cadherin6 antibody.
Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ydbio.2008.11.021.