Trps1 activates a network of secreted Wnt inhibitors and transcription factors crucial to vibrissa follicle morphogenesis

The family of trichorhinophalangeal syndromes (TRPS I, OMIM 190350; TRPS II, OMIM 150230; TRPS III, OMIM 190351) is characterized by sparse and slow-growing scalp hair, as well as craniofacial and skeletal abnormalities (Giedion et al., 1973). Heterozygous germline mutations in TRPS1 on chromosome 8q23 result in autosomal dominant inheritance of TRPS types I and III (Momeni et al., 2000; Ludecke et al., 2001). TRPS1 encodes a vertebrate protein with nine zinc-finger domains, including a GATA-type zinc finger and two C-terminal Ikaros-like zinc fingers (Momeni et al., 2000).

Monoallelic nonsense, missense or in-frame splice site mutations in TRPS1, as well as disruption or complete deletion of the gene, have been reported in cases of TRPS I (Momeni et al., 2000; Ludecke et al., 2001; Seitz et al., 2001; Hilton et al., 2002; Kaiser et al., 2004; Rossi et al., 2007; Piccione et al., 2009). Alternatively, missense mutations in exon 6 that alter the GATA-type zinc-finger domain were reported in patients classified as having the more severe TRPS III (Ludecke et al., 2001; Kobayashi et al., 2002; Hilton et al., 2002; Piccione et al., 2009). Ludecke et al. proposed that the mutant proteins resulting from these missense mutations have a decreased affinity to DNA and act in a dominant-negative manner in multiprotein complexes (Ludecke et al., 2001).

Electrophoretic mobility shift assays have demonstrated that both human and murine TRPS1 specifically bind the consensus GATA sequence (WGATAR) in DNA (Chang, G. T. et al., 2002; Malik et al., 2001). Furthermore, murine Trps1 has been shown to function as a sequence-specific transcriptional repressor in both reporter gene assays and Xenopus embryo explant experiments, with the ability to repress GATA-dependent activation in a dose-dependent manner (Malik et al., 2001). This repression was dependent on the integrity of the Trps1 GATA-type zinc-finger domain and also required the C-terminal 119 amino acids of the protein, which harbor the two Ikaros-like zinc-finger domains (Malik et al., 2001). Consistent with their crucial role in mediating the transcriptional activity of Trps1, the sequences of the GATA-type zinc-finger motif and the neighboring basic regions, as well as the Ikaros-type zinc fingers, are 100% conserved at the amino acid level between Xenopus, mouse and human TRPS1 proteins (Malik et al., 2001).

TRPS1 has been shown to transcriptionally repress five target genes through specific binding to consensus GATA sites in their promoters, four of which are expressed in bone. These transcriptional targets include prostate-specific antigen (PSA; also known as KLK3) (van den Bemd et al., 2003), runt-related transcription factor 2 (RUNX2) (Napierala et al., 2005), signal transducer and activator of transcription 3 (Stat3) (Suemoto et al., 2007), parathyroid hormone-related protein (Pthrp; also known as Pthlh) (Nishioka et al., 2008) and osteocalcin (BGLAP) (Piscopo et al., 2009). In each case, mutation of the consensus GATA binding sites abolished TRPS1-mediated repression of target gene transcription.

To investigate the role of Trps1 during mammalian development, Malik et al. generated mice with a targeted in-frame deletion of the GATA-type zinc-finger domain (Malik et al., 2002). The resulting Trps1^Δgt/Δgt mice displayed a number of hair follicle, craniofacial and skeletal defects that mirror the phenotypic characteristics of human TRPS patients. Homozygous Trps1^Δgt/Δgt mice died within 6 hours of birth due to respiratory failure stemming from thoracic skeletal defects. Homozygous mutant mice were further reported to completely lack vibrissae follicles during late gestation, with no histological evidence of earlier follicle formation. In addition, neonatal Trps1^Δgt/Δgt mice exhibited a ∼50% reduction in dorsal pelage follicle density compared with their wild-type littermates (Malik et al., 2002). More recently, Trps1^–/– mice were generated and reported to display ‘severe hair follicle abnormalities’ without further elaboration (Suemoto et al., 2007). While these studies revealed that Trps1 is necessary for proper hair follicle formation, they did not address the molecular mechanisms by which Trps1 regulates hair follicle development.

Here, we performed a detailed histological analysis of early vibrissa follicle development in Trps1^Δgt/Δgt mouse embryos, revealing mutant vibrissae hair germs that were reduced in number, irregularly spaced and markedly smaller than those of their wild-type counterparts. To gain insight into the role of Trps1 as a transcriptional regulator in the hair follicle, we performed microarray hybridization analysis, comparing expression patterns in whole whisker pads of wild-type versus Trps1^Δgt/Δgt embryos. Our findings uncovered a network of transcription factors, Wnt inhibitors and extracellular matrix proteins regulated by Trps1 during early vibrissa follicle morphogenesis and demonstrated, for the first time, that Trps1 is capable of acting as a transcriptional activator.

Trps1^+/tm1Shiv mice (Malik et al., 2002), which are referred to in the text as Trps1^+/Δgt, were a gift of Dr R. Shivdasani (Dana-Farber Cancer Institute at Harvard Medical School). All mouse experiments were performed under approval of the Institutional Animal Care and Use Committee of Columbia University.

Whole muzzle skin dissected from embryonic day (E) 12.5-13.5 embryos (day of plug considered E0.5) was dissected in PBS, fixed in 10% formalin for up to 72 hours, washed through an ethanol series and embedded in paraffin. After deparaffinization and rehydration, sections (8 μm) were stained with Hematoxylin and Eosin and permanently mounted with Permount (Thermo Fisher Scientific, Waltham, MA, USA). Sections were photographed using an HRc AxioCam fitted onto an Axioplan2 fluorescence microscope (Carl Zeiss, Thornwood, NY, USA). Vibrissa follicle quantifications were performed using the measurement tool in AxioVision 4.5 software (Carl Zeiss).

Whole muzzle skin dissected at multiple timepoints from E12.5 to postnatal day (P) 1 was mounted in O.C.T. compound (Sakura Finetek, Torrance, CA, USA) and frozen in liquid nitrogen. Sections (8 μm) were fixed in 4% paraformaldehyde (PFA)/0.1% Triton X-100 for 10 minutes at room temperature or in methanol for 15 minutes at –20°C followed by acetone for 2 minutes at –20°C and washed in PBS. Sections were blocked for 1 hour in 10% heat-inactivated goat or donkey serum in PBS and incubated overnight at 4°C in primary antibody diluted in 1% serum in PBS. Primary antibodies and dilutions were as follows: anti-Trps1 [1:5000; gift of Dr R. Shivdasani (Malik et al., 2001)]; anti-P-cadherin (1:100; Invitrogen, Carlsbad, CA, USA); anti-syndecan 1 (1:200; BD Pharmingen, San Jose, CA, USA); anti-Lhx2 (1:10,000; gift of Dr T. Jessell, Columbia University, New York); anti-Tnc (1:200; Sigma-Aldrich, St Louis, MO, USA); and anti-Lef1 (1:25; Cell Signaling Technology, Danvers, MA, USA). After washing in PBS, sections were incubated in either an Alexa Fluor 594 donkey anti-rat IgG, Alexa Fluor 488 goat anti-rabbit IgG or Alexa Fluor 594 goat anti-rabbit IgG (Molecular Probes, Invitrogen) secondary antibody (1:500) diluted in 1% serum in PBS for 1 hour. Sections were mounted in VECTASHIELD mounting medium with DAPI (Vector Laboratories, Burlingame, CA, USA) and photographed using an HRc AxioCam fitted onto an Axioplan2 fluorescence microscope or an LSM 5 laser-scanning Axio Observer Z1 confocal microscope (Carl Zeiss).

Whole whisker pads were dissected from E12.5 embryos and total RNA was isolated using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. Triplicate RNA samples from three independent embryos of each genotype were amplified and labeled for hybridization to Affymetrix GeneChip MOE430A microarrays using Affymetrix reagents and protocols (Affymetrix, Santa Clara, CA, USA). The data output was analyzed using GeneSpring GX 10.0 software (Agilent Technologies, Palo Alto, CA, USA). P-values were calculated using an unpaired t-test. Expression values with P≤0.05 and a fold difference of at least 1.5 relative to wild-type baseline expression levels were considered significant. Microarray data have been deposited at the Gene Expression Omnibus with the Accession Number GSE33766.

First-strand cDNA was synthesized from total RNA isolated from whole muzzle skin dissected from E12.5 embryos using a 2:1 ratio of random primers to oligo(dT) primer and SuperScript III reverse transcriptase (Invitrogen) according to the manufacturer’s instructions. qRT-PCR was performed on an ABI 7300 machine and analyzed with ABI Relative Quantification Study software (Applied Biosystems, Foster City, CA, USA). Primers were designed according to ABI guidelines and all reactions were performed using Power SYBR Green PCR Master Mix (Applied Biosystems), 250 nM primers (Invitrogen) and 100 ng cDNA in a 20 μl reaction volume. The following PCR protocol was used: step 1, 50°C 2 minutes; step 2, 95°C 10 minutes; step 3, 95°C 15 seconds; step 4, 60°C 1 minute; repeat steps 3 and 4 for 40 cycles. All samples were run in quadruplicate for three independent runs and normalized against an endogenous internal control, B2m. PCR products were electrophoresed on a 1% agarose TBE gel containing ethidium bromide and photographed with an Electrophoresis Documentation and Analysis System 120 camera (Kodak, Rochester, NY, USA) to confirm amplicon size. qRT-PCR primers are listed in supplementary material Table S1.

Immunohistochemistry was performed on sections (8 μm) of paraffin-embedded whole muzzle skin dissected from E16.5 embryos using an anti-Prdm1 mouse monoclonal antibody (gift of Dr K. Calame, Columbia University, New York) based on a previously published protocol (Chang, D. H. et al., 2002). Sections were photographed using an HRc AxioCam fitted onto an Axioplan2 fluorescence microscope (Carl Zeiss).

An 852 bp probe template spanning from exon 2 to the 3′ UTR of Sox18 was amplified by PCR from an E15.0 dermal cDNA stock using the following primers: mSox18-F, 5′-CGCGACCATCCCAACTACAAGTAC-3′; and mSox18-R, 5′-AAAGATGCCATTTCTGTCGCCTCC-3′. The PCR product was cloned into the pCRII dual promoter (T7 and SP6) vector (Invitrogen) and standard procedures were followed for the preparation of digoxigenin-labeled cRNA (Roche Applied Science, Indianapolis, IN, USA) antisense and control sense probes. We previously reported the Apcdd1 (Shimomura et al., 2010) and Dkk4 (Bazzi et al., 2007) probes. In situ hybridization was performed on sections (16 μm) of sucrose-infiltrated frozen whole muzzle skin dissected from E12.5 embryos based on our previously published protocol (Shimomura et al., 2010). Sections were photographed using an HRc AxioCam fitted onto an Axioplan2 fluorescence microscope (Carl Zeiss).

The list of transcripts generated by microarray hybridization analysis was analyzed using the Babelomics 4.2.0 suite (http://babelomics.bioinfo.cipf.es). Single enrichment analysis was performed with the FatiGO tool, using a two-tailed Fisher’s exact test to identify over-represented functional annotations in the transcript list compared with the entire genome. Results with P<0.05 were considered significant.

HEK 293T cells were seeded onto 10 cm dishes and cultured to 80-90% confluency in Dulbecco’s Modified Eagle’s Medium (Gibco, Invitrogen) supplemented with 10% fetal bovine serum (FBS) (Gibco), 100 IU/ml penicillin and 100 μg/ml streptomycin (P/S). The cells were treated with 1% formaldehyde for 10 minutes at 37°C, washed twice with cold PBS containing protease inhibitors and harvested. ChIP was carried out using the Chromatin Immunoprecipitation Assay Kit (Millipore, Billerica, MA, USA) according to the manufacturer’s instructions. Cell lysates were precipitated with 3 μg of anti-Trps1 rabbit polyclonal antibody (gift of Dr R. Shivdasani) or normal rabbit IgG (Santa Cruz Biotechnology, Santa Cruz, CA, USA) as a negative control. After elution, DNA was recovered using the Rapid PCR Purification System (Marligen Biosciences, Ijamsville, MD, USA). PCR reactions were performed using input, IgG-precipitated and Trps1-precipitated DNA samples, Platinum PCR SuperMix (Invitrogen) and 0.67 μM primers (Invitrogen) in a 30 μl reaction volume. The following PCR protocol was used: step 1, 94°C 5 minutes; step 2, 94°C 45 seconds; step 3, 55°C 30 seconds; step 4, 72°C 1 minute; repeat steps 2-4 for 36-40 cycles; step 5, 72°C 10 minutes. PCR products were electrophoresed on a 1% agarose TBE gel containing ethidium bromide and photographed with an Electrophoresis Documentation and Analysis System 120 camera (Kodak). Positive immunoprecipitation results were confirmed in at least two independent trials. PCR primers are listed in supplementary material Table S1.

The Sox18 promoter (2510 bp upstream of the translation initiation site) was amplified by PCR from a C57BL/6 DNA stock using the following primers: mSox18p-F-XhoI, 5′-CAACTCGAGCTCACTTTGGCCAAAGCTAG-3′; and mSox18p-R-HindIII, 5′-GACAAGCTTGATCTCTGCATTCCAGCTC-3′. The amplified product was subcloned into the XhoI and HindIII sites of the luciferase reporter vector pGL3-Basic (Promega, Madison, WI, USA). We previously reported the Dkk4 promoter construct (Bazzi et al., 2007).

Saos-2 cells were seeded onto 6-well dishes 24 hours before transfection. At 80-90% confluency, a mouse Trps1 expression plasmid or pCXN2.1 backbone vector (1 μg) were transfected into each well in combination with the mouse Sox18 promoter reporter plasmid, mouse Dkk4 promoter reporter plasmid or pGL3 backbone vector (1 μg) using FuGENE HD (Roche Applied Science). A plasmid encoding a β-galactosidase reporter (0.5 μg) was also transfected for normalization of transfection efficiency. The cells were cultured for 24 hours after transfection in McCoy’s 5A Medium Modified (Invitrogen) with 15% FBS and P/S, harvested and lysed. Luciferase and β-galactosidase signals were measured using the Luciferase Assay System (Promega) and β-Galactosidase Enzyme Assay System with Reporter Lysis Buffer (Promega), respectively, according to the manufacturer’s instructions. All assays were performed in triplicate for three independent trials.

HEK 293T cells were seeded onto 12-well dishes 24 hours before transfection. At 80% confluency, a TOP-flash Wnt reporter plasmid (50 ng) was transfected into each well in combination with expression plasmids for human LEF1 (100 ng), human β-catenin (500 ng) (both gifts of Dr J. Kitajewski, Columbia University, New York) and/or mouse Trps1 (500 ng) using Lipofectamine 2000 (Invitrogen). A plasmid encoding a β-galactosidase reporter (100 ng) was also transfected for normalization of transfection efficiency. The cells were cultured for 36 hours after transfection in Opti-MEM (Gibco), harvested and lysed. Luciferase and β-galactosidase signals were measured as described above. All assays were performed in triplicate for three independent trials.

We began by performing histological analysis of vibrissa follicle development during the earliest stages of morphogenesis in Trps1^Δgt/Δgt mouse embryos. Mutant vibrissae placodes (Fig. 1B) were reduced in number and irregularly spaced at E12.5 compared with those found in wild-type littermates (Fig. 1A). These abnormalities were even more pronounced at E13.5 (Fig. 1C,D; supplementary material Fig. S1A-C), when Trps1^Δgt/Δgt vibrissae were noticeably shorter than their wild-type counterparts (Fig. 1E). Furthermore, whereas 68.00±4.90% of Trps1^+/+ vibrissae had reached the peg stage of morphogenesis by E13.5, the mutant vibrissae were developmentally delayed, with only 12.38±6.04% (P<0.001) having reached the peg stage by this timepoint (Fig. 1F).

We next compared the levels of P-cadherin and syndecan 1 between wild-type and Trps1^Δgt/Δgt vibrissae to examine the relative size and location of the epithelial and mesenchymal compartments, respectively. P-cadherin was expressed throughout the follicular epithelium of both wild-type and mutant vibrissae, with highest expression levels at the leading edge of the invaginating follicles (Fig. 1G,H). Similarly, syndecan 1 staining was relatively consistent in the mesenchyme surrounding wild-type and Trps1^Δgt/Δgt developing vibrissae (Fig. 1I,J).

As the development of Trps1^Δgt/Δgt vibrissae is dysregulated from the earliest stages of morphogenesis, we next performed immunofluorescence analysis to examine the expression of Trps1 during vibrissa follicle induction, specifically at E12.5 when the vibrissae placodes are first discernible, in an attempt to identify the cell population(s) responsible for the mutant phenotype. Trps1 expression was intense and nuclear in the mesenchymal cells underlying the anlagen of the developing vibrissae follicles, while more diffuse in the overlying epithelial cells of the placode (Fig. 1K,L), predicting a role for this transcription factor in both compartments of the developing vibrissa.

In order to identify genes that are transcriptionally regulated by Trps1 in both the epithelium and mesenchyme of the follicle during early morphogenesis, we performed microarray hybridization analysis, comparing expression patterns in whole whisker pads of wild-type versus Trps1^Δgt/Δgt embryos at E12.5. Expression values with P≤0.05 and a fold difference of at least 1.5 relative to wild-type baseline expression levels were considered significant. A box-and-whisker plot of the microarray data revealed a similar distribution of normalized intensity values across all of the samples used (supplementary material Fig. S2A), and non-biased hierarchical analysis demonstrated clustering of the biological replicates (supplementary material Fig. S2B).

A relatively short list of differentially expressed transcripts was generated (Table 1), many of which have known expression patterns in the skin or hair follicle. Eighteen transcripts were found to be downregulated in the mutant samples (supplementary material Fig. S2C). These included a number of transcription factors, such as Prdm1, Sox18, Lhx2 and Sox21, among others, and secreted or membrane-bound Wnt inhibitors, including Wif1, Apcdd1 and Dkk4. Conversely, 32 transcripts were found to be upregulated in the mutant samples (supplementary material Fig. S2C), many of which encode extracellular matrix proteins, including Epyc, Dcn, Lum and Tnc.

We validated the expression changes of selected transcripts by quantitative real-time PCR (qRT-PCR) analysis and found that the transcriptional trends were upheld, with a significant change (P<0.05) in target gene expression levels in the Trps1^Δgt/Δgt samples (Fig. 2).

Beginning with transcripts in the upper portion of the list which we hypothesized are activated (directly or indirectly) by Trps1, we next compared the expression of a subset of these potential Trps1 target genes between wild-type and Trps1^Δgt/Δgt embryos in vivo. Immunohistochemistry performed on E16.5 whisker pads revealed decreased Prdm1 expression in the dermal papillae and inner root sheath precursors of Trps1^Δgt/Δgt vibrissae follicles (Fig. 3A,B). In situ hybridization analysis on E12.5 whisker pads demonstrated decreased Sox18 expression in the dermal condensates underlying the developing vibrissae placodes in Trps1^Δgt/Δgt embryos (Fig. 3C,D). Furthermore, immunofluorescence analysis performed on E12.5 whisker pads revealed decreased Lhx2 expression throughout the epithelial placodes of Trps1^Δgt/Δgt vibrissae, particularly in the basal layer (Fig. 3E,F). These in vivo expression differences were all consistent with the trends predicted by the microarray hybridization analysis.

As characterization of Lhx2 protein expression in the hair follicle had thus far been described only for pelage hairs (Rhee et al., 2006; Törnqvist et al., 2010), we performed immunofluorescence analysis of the expression of Lhx2 during vibrissa follicle morphogenesis. Compared with its expression in the developing pelage follicle, we found expansion of the Lhx2 expression domain into additional layers of the follicular epithelium, in particular the inner root sheath layers (supplementary material Fig. S3). Furthermore, we detected faint Lhx2 expression in the dermal papilla as early as E14.5, as well as in the dermal cells of the vibrissae-specific collagen capsule surrounding the developing follicles (supplementary material Fig. S3). Importantly, this pattern of Lhx2 expression overlapped with that reported for Trps1 in developing vibrissae (Fantauzzo et al., 2008b).

Next, we focused on the lower portion of the microarray list and performed single enrichment analysis using the Babelomics 4 suite to identify over-represented functional annotations compared with the genome as a whole. Proteinaceous extracellular matrix (GO:0005578) was the most significant GO term detected in the cellular component analysis, with an adjusted P-value of 9.96×10^–7. Transcripts within our list associated with this term included Epyc, Egr1, Dcn, Lum, Tnc, Col8a1, Fus and Tinag (supplementary material Fig. S4).

Immunofluorescence analysis of one of these extracellular matrix proteins, Tnc, revealed striking increases in expression in the mesenchyme underlying developing Trps1^Δgt/Δgt vibrissae follicles at E12.5 and E13.5 (Fig. 3G-J). Our findings thus demonstrate that Trps1 regulates the expression of extracellular matrix proteins throughout the whisker pad microenvironment, in addition to its role as a transcriptional regulator within the developing vibrissa follicle.

To determine whether Trps1 directly binds the promoters of the potential target genes identified by microarray hybridization analysis in this study, we located consensus GATA-binding sites within 3 kb of the transcriptional start site of the human WIF1, PRDM1, SOX18, LHX2, APCDD1, DKK4 and SOX21 promoters and performed endogenous chromatin immunoprecipitation (ChIP) experiments in HEK 293T cells. We found that Trps1 bound up to ten sites in the WIF1 promoter out of 15 potential binding sites in the DNA. Similarly, Trps1 bound one site each in the SOX18 and LHX2 promoters, up to four sites each in the APCDD1 and DKK4 promoters, and up to three sites in the SOX21 promoter (Fig. 4A). Despite the presence of two consensus GATA-binding sites within 3 kb of the transcriptional start site of PRDM1, Trps1 binding was never observed in this promoter (data not shown), indicating that Trps1 regulates PRDM1 expression in an indirect manner.

To date, the prevailing view asserts that Trps1 acts exclusively as a transcriptional repressor in various contexts (van den Bemd et al., 2003; Napierala et al., 2005; Suemoto et al., 2007; Nishioka et al., 2008; Piscopo et al., 2009). In light of our in vivo expression and ChIP results, we revisited this notion and postulated that Trps1 can also function as a transcriptional activator. We performed luciferase reporter promoter assays in the human epithelial osteosarcoma cell line Saos-2 and demonstrated that Trps1 activated both Sox18 (30.13±15.88%; P=0.0759) and Dkk4 (28.09±11.32%; P=0.0254) transcription (Fig. 4B). These results reveal a novel role for Trps1 as a direct transcriptional activator of downstream target genes.

Notably, three Wnt inhibitors, Wif1, Apcdd1 and Dkk4, were detected among the transcripts downregulated in the Trps1^Δgt/Δgt whisker pad samples by microarray hybridization analysis, leading us to postulate that Trps1 may repress the Wnt signaling pathway. In situ hybridization analysis at E12.5 revealed that Apcdd1 expression was decreased in both the epithelial placode and underlying mesenchyme of Trps1^Δgt/Δgt vibrissae (Fig. 5A,B), and that Dkk4 expression was decreased throughout the epithelial placodes of the developing mutant vibrissae and the surrounding interfollicular whisker pad epidermis (Fig. 5C,D).

To examine whether the decreased expression of these Wnt inhibitors in Trps1^Δgt/Δgt vibrissae led to a concomitant increase in canonical Wnt signaling in these follicles, we examined the expression of Lef1 by immunofluorescence analysis. Whereas Lef1 was strongly expressed in the nuclei of mesenchymal cells in both the wild-type and Trps1^Δgt/Δgt whisker pad dermis, its expression was markedly increased in the epithelial cells of the placode in the mutant follicles (Fig. 5E,F).

To further define the effect of Trps1 expression on canonical Wnt signaling, we performed a series of Wnt reporter assays in HEK 293T cells using the TCF/LEF luciferase reporter plasmid TOP-flash. As expected, TOP-flash activity was modestly increased over baseline levels with the addition of either LEF1 or β-catenin, and dramatically increased in the presence of both proteins (Fig. 5G). The addition of Trps1 significantly repressed (53.81±5.30%; P<0.001) the LEF1- and β-catenin-mediated activation of TOP-flash activity (Fig. 5G). Taken together, these findings identify Trps1 as a novel regulator of the canonical Wnt signaling pathway through its direct activation of several Wnt inhibitors.

In this study, we identified a number of downstream transcriptional targets of Trps1, including several Wnt inhibitors and extracellular matrix proteins. Moreover, we demonstrated that Trps1 upregulates the expression of multiple transcription factors in the vibrissa follicle, including Prdm1, Sox18, Lhx2 and Sox21, and, furthermore, that Trps1 directly binds the promoters of all but the first of these target genes to activate their transcription. These data establish a crucial role for Trps1 as an upstream regulator of transcriptional hierarchies in both compartments of the hair follicle. Importantly, we demonstrate, for the first time, the ability of Trps1 to function as a transcriptional activator, expanding upon its reported role as a transcriptional repressor.

Each of the transcription factors identified as targets of Trps1 in this microarray study have expression domains that colocalize in vivo with Trps1 in at least one compartment of the skin or hair follicle (Fantauzzo et al., 2008a) (Table 2). Notably, Trps1^Δgt/Δgt mice share phenotypic similarities with mutant mouse models of many of these transcription factors (Table 2). For instance, late gestation embryos with epiblast-specific deletion of the gene encoding the Krüppel-type zinc-finger transcription factor Prdm1 (Sox2-Cre;Prdm1^BEH/fl) lack vibrissae follicles completely (Robertson et al., 2007), which is comparable to the absence of vibrissae follicles observed at birth in Trps1^Δgt/Δgt mice. Moreover, embryos null for the LIM homeobox transcription factor Lhx2 (Lhx2^–/–) display a ∼40% reduction in pelage follicle density (Rhee et al., 2006), analogous to the reduced pelage follicle density in Trps1^Δgt/Δgt embryos.

Two members of the SRY box-containing family of transcription factors, Sox18 and Sox21, were also identified as targets of Trps1 in this study and both have corresponding mutant mouse models that display vibrissae follicle defects. Point mutations in Sox18 have been shown to underlie the ragged (Ra) mouse phenotype, which is characterized by varying degrees of coat sparseness (Pennisi et al., 2000a). Sox18^Ra/+ heterozygous neonates have short vibrissae and a thin, ragged pelage coat that lacks secondary auchene and zigzag hairs, whereas homozygous mutants almost completely lack vibrissae and coat hairs (Carter and Phillips, 1954). This vibrissa follicle phenotype is consistent with that observed in Trps1^Δgt/Δgt mice, while the pelage coat defect is more severe. In humans, mutations in SOX18 result in hypotrichosis-lymphedema-telangiectasia syndrome (OMIM 607823), which is typified by an absence of eyebrows and eyelashes, and by sparse scalp hair (Irrthum et al., 2003) reminiscent of the scalp hypotrichosis found in patients with TRPS types I and III (Giedion et al., 1973).

Sox21^–/– mice exhibit cyclic alopecia beginning at P11, resulting in a transient absence of both pelage and vibrissae follicles by P20-25 (Kiso et al., 2009). While the hair follicle phenotypes observed in Sox21^–/– mice have a later age of onset than those observed in Trps1^Δgt/Δgt embryos during morphogenesis, these results demonstrate a requirement for both Sox21 and Trps1 in proper vibrissa and pelage follicle maintenance.

In addition to the target genes discussed above, further examination of the lower portion of the microarray list consisting of transcripts upregulated in Trps1^Δgt/Δgt whisker pads revealed a striking enrichment for extracellular matrix proteins and provided additional insight into the mutant phenotype. Of the transcripts associated with the proteinaceous extracellular matrix identified as potential Trps1 targets, Egr1, Dcn, Lum, Tnc, Col8a1 and Fus are reportedly expressed in the skin (Larsen et al., 1994; Bianco et al., 1990; Chakravarti et al., 1995; Chiquet-Ehrismann et al., 1986; Muragaki et al., 1992; Champliaud et al., 1998). Transcripts encoding three members of the small leucine-rich proteoglycan family, Epyc, Dcn and Lum, were upregulated in Trps1^Δgt/Δgt mutant whisker pads at E12.5, indicating that Trps1 may repress their expression. Lum has been shown to localize to the mesenchyme surrounding the developing vibrissae follicles during morphogenesis (Chakravarti, 2002). Knockout models of both Dcn and Lum are viable, with a dermal collagen defect resulting in thin, fragile skin that exhibits reduced tensile strength (Danielson et al., 1997; Chakravarti et al., 1998). The increased expression of Epyc, Egr1, Dcn, Lum, Tnc and Col8a1 in Trps1^Δgt/Δgt mutant whisker pad samples signifies defects in connective tissue assembly. Defective formation of the collagen capsules in the whisker pads of these mutant embryos may disrupt the structural stability of the developing vibrissae, perhaps contributing to the degeneration of these follicles after peg downgrowth has been initiated. Furthermore, because pelage follicles are not associated with collagen capsules, aberrant expression of these extracellular matrix proteins may in part explain the more severe phenotype observed in vibrissae follicles in Trps1^Δgt/Δgt embryos.

Two additional transcripts of interest identified by our microarray analysis encode the transcription factor Egr1 and the extracellular matrix glycoprotein Tnc, which are expressed not only in the skin but also in the brain. Egr1 transcript expression is significantly increased in the vibrissae-related barrel field of the somatosensory cortex upon whisker stimulation (Patra et al., 2004). Tnc is expressed in the compact mesenchyme surrounding developing epithelia throughout the embryo, including the vibrissae follicles (Chiquet-Ehrismann et al., 1986), and, similar to Egr1, is also prominently expressed in various areas of the developing and adult brain, with specific expression in the vibrissae-related barrel field of the somatosensory cortex (Steindler et al., 1989; Crossin et al., 1989). Classic experiments have shown that injury of individual vibrissae results in the subsequent loss of the corresponding cortical barrel (Van der Loos and Woolsey, 1973). Notably, neuron projection (GO:0043005) was a significant GO term detected in our cellular component analysis of the transcripts upregulated in the Trps1^Δgt/Δgt whisker pad samples (supplementary material Fig. S4). While outside the scope of the present study, the misexpression of Egr1, Tnc and several neuron projection transcripts, as well as the sparse vibrissae follicles observed in Trps1^Δgt/Δgt mutant embryos, predict corresponding effects on the sensory function of the mystacial vibrissae and a decrease in cortical barrel density in the brain.

We have previously demonstrated that Trps1 directly represses the hair follicle stem cell specification gene Sox9 in the follicle (our unpublished results). Interestingly, Tnc expression is significantly upregulated in both mouse (Morris et al., 2004) and human (Kloepper et al., 2008) bulge stem cell compartments, where it has been proposed to play a role in maintaining niche quiescence (Morris et al., 2004). Similarly, Lhx2 has been shown to maintain hair follicle stem cells in an undifferentiated state (Rhee et al., 2006). In combination with our previous work with Sox9, the finding here that Trps1 regulates the expression of both Tnc and Lhx2 points to a novel role for this protein in the regulation of early hair follicle progenitors, and, potentially, adult follicle stem cells.

Of the molecules preferentially expressed in the bulge stem cell compartment of the adult follicle, Sox9 and Lhx2 are expressed earliest during follicle morphogenesis, beginning at the placode stage (Vidal et al., 2005; Rhee et al., 2006), when Sox9 is expressed in the suprabasal cells of the invaginating follicle and Lhx2 is detected in the basal layer (Nowak et al., 2008). Furthermore, they are the only known bulge markers that are additionally expressed in the bulge-derived secondary hair germ during anagen initiation (Nowak et al., 2008; Rhee et al., 2006), a site where Trps1 is also prominently expressed (Fantauzzo et al., 2008a). Using genetic marking techniques, Nowak et al. demonstrated that the early basal placode cells expressing Lhx2 are transient, whereas the suprabasal placode cells expressing Sox9 are longer lived, giving rise to the bulge stem cells and, eventually, to all the epithelial cells of the hair follicle (Nowak et al., 2008).

The differential regulation of these two transcripts by Trps1 indicates that Trps1 may be separately regulating the distinct cell populations that express these markers, both during the early stages of hair follicle morphogenesis and anagen initiation. We postulate that Trps1 activates Lhx2 to maintain the balance of quiescence in the epithelial cells directly adjacent to the underlying dermal papilla. Consistent with this hypothesis, similar to the hair phenotype observed in Lhx2^–/– skin grafts (Rhee et al., 2006), Trps1^Δgt/Δgt embryos exhibit increased proliferation throughout vibrissae follicles during the hair peg stage prior to their degeneration (our unpublished results), indicating an inability to maintain cells in a state of relative quiescence.

Interestingly, our results indicate that Trps1 upregulates the expression of three Wnt inhibitors, Wif1, Apcdd1 and Dkk4, in the vibrissa follicle and, furthermore, that this transcriptional regulation occurs through direct binding of Trps1 to the promoters of these target genes. Correspondingly, we have demonstrated that the reduced expression of these Wnt inhibitors in Trps1^Δgt/Δgt whisker pads results in an increase in Wnt signaling in the epithelial placodes of mutant vibrissae follicles. As Wnt signaling acts very early during hair follicle formation to promote the hair placode fate (Andl et al., 2002), the observation that Wnt inhibitors are dysregulated in Trps1^Δgt/Δgt mutant whisker pads suggests that the vibrissae phenotype in these embryos might be due, in part, to disruptions in Wnt signaling during early vibrissa follicle development.

The importance of Wnt signaling in the vibrissa follicle is supported by the abnormal phenotypes observed in mice that are deficient for or overexpress members of this signaling pathway. For example, targeted disruption of Lef1 results in the absence of vibrissae follicles (van Genderen et al., 1994), whereas transgenic overexpression of Lef1 in the basal epidermis of the skin and hair follicle (K14-Lef1) leads to the development of curved, irregularly oriented vibrissae follicles in adult animals (Zhou et al., 1995). Furthermore, inhibition of Wnt signaling in the epidermis through overexpression of a Lef1 transgene lacking the β-catenin binding site (K14-ΔnLef1) results in sparse vibrissae (Niemann et al., 2002). Similarly, ablation of the effectors Tcf3/Tcf4 in the epidermis of the skin and hair follicle (K14-Cre;Tcf3^fl/fl;Tcf4^–/–) leads to a lack of vibrissae follicles (Nguyen et al., 2009). Epidermal-specific disruption of Ctnnb1 (K14-Cre;Ctnnb1^fl/fl), which encodes β-catenin, or overexpression of the secreted Wnt inhibitor Dkk1 in the epidermis (K14-Dkk1), inhibits vibrissa placode formation (Huelsken et al., 2001; Andl et al., 2002).

Both the Wnt effector Lef1 and the Wnt-responsive reporter gene TOPGAL are expressed in the epithelial placodes and dermal condensates of developing vibrissae follicles (van Genderen et al., 1994; DasGupta and Fuchs, 1999). Tissue recombination experiments between wild-type and Lef1^–/– embryos have demonstrated that dermal Lef1 expression is necessary for the initiation of murine vibrissa, but not pelage, follicle formation, while epithelial Lef1 expression is required for the completion of vibrissa follicle morphogenesis (Kratochwil et al., 1996), indicating a unique requirement for active Wnt signaling in both compartments of the developing vibrissa.

Similar to Lef1, Trps1 is also expressed in both compartments of the developing vibrissa follicle during embryogenesis. Subcellularly, nuclear localization of Trps1 is observed in the mesenchymal cells of the whisker pad before any histological evidence of vibrissa follicle development, signifying that Trps1 acts to alter target gene expression levels very early in the dermis. Expression of Trps1 in the nucleus of epithelial cells is observed only transiently at the onset of hair follicle morphogenesis (Fantauzzo et al., 2008a), indicating an additional role for Trps1 in the transcriptional regulation of placode formation.

These results point to a crucial role for Trps1 in both the epithelial and mesenchymal compartments of the skin at the onset of vibrissa follicle development. We propose that activation of secreted and membrane-bound Wnt inhibitors by Trps1 at the initiation of vibrissa follicle formation acts to attenuate Wnt signaling both in the epidermal placode and dermal condensate of the developing follicle, thereby allowing for proper vibrissa initiation. In Trps1^Δgt/Δgt whisker pads, a subset of Wnt inhibitors fail to be activated by Trps1 in the epithelial placode and dermal condensate at the initiation of vibrissa follicle morphogenesis, which is likely to contribute to defective follicle spacing and development. In support of this hypothesis, one of the Wnt inhibitors activated by Trps1, Dkk4, is expressed exclusively in the epithelial placodes of vibrissae and has been shown to function in negative-feedback inhibition in the hair follicle to attenuate Wnt signaling (Bazzi et al., 2007). The additional Wnt inhibitors upregulated by Trps1, Wif1 and Apcdd1, are expressed in the dermal papilla (Kansara et al., 2009; Jukkola et al., 2004) and may correspondingly mediate the effects of Trps1 in the mesenchymal compartment.

In summary, we have identified a number of transcriptional targets of Trps1 through which it regulates early vibrissa follicle organogenesis in the murine whisker pad, where, as we show for the first time, it functions not only as a transcriptional repressor but also as an activator. Notably, we have demonstrated that Trps1 orchestrates a complex morphogenetic process through subtle changes in the expression of a wide array of genes, each of which is likely to contribute in a combinatorial fashion to vibrissa development.

We thank Dr R. Shivdasani for the generous gifts of the Trps1^+/Δgt mice and Trps1 antibody; Dr C. Jahoda (University of Durham) for help with whisker pad dissections and continued discussions throughout this project; Dr V. Miljkovic at the Columbia University HICCC Microarray Service for assistance with microarray hybridization analysis; Drs K. Calame and E. Magnúsdóttir for sharing the Prdm1 antibody and immunohistochemistry protocol; Dr T. Jessell for providing the Lhx2 antibody; Dr J. Kitajewski for the gifts of the human LEF1 and β-catenin expression plasmids; members of the A.M.C. laboratory for helpful discussions, particularly Dr Y. Shimomura for advice on the ChIP protocol and Dr C. Higgins for counsel on the GeneSpring GX 10.0 software and Babelomics 4.2.0 analysis; Mr M. Zhang for technical assistance; and Ms E. Chang for mouse husbandry work.