Defining BMP functions in the hair follicle by conditional ablation of BMP receptor IA

During embryonic development, a layer of multipotent stem cells gives rise to the epidermis and its appendages, including hair follicles. Hair follicle morphogenesis arises from a series of epithelial–mesenchymal cues, initiating an epithelial down-growth, which then proliferates and differentiates to form first the channel, or inner root sheath (IRS), and then the hair itself (Fig. 1). Postnatally, the mature hair shaft is composed of a core, or medulla, cloaked by a concentric ring of cortical cells, which in turn are surrounded by a layer of hair shaft cuticle (the hair surface). Beneath the skin surface, the hair is surrounded by an IRS composed of a cuticle, Huxley's layer and Henle's layer. IRS cuticle interlocks with hair cuticle to guide the hair, but near the skin surface, the IRS degenerates to release the shaft (Hardy, 1992). Above the base (bulb) of the follicle, the Henle's layer is encased by the companion cell layer and outer root sheath (ORS), a structure contiguous with and biochemically similar to the epidermal basal layer.

Both hair shaft and IRS form concomitantly through upward terminal differentiation of transiently amplifying progenitor (matrix) cells, which maintain contact with dermal papilla (DP), the follicular mesenchyme at the core of the hair bulb (Oliver and Jahoda, 1988). It is thought that a reservoir of stem cells, the bulge, resides at the base of the noncycling segment of the hair follicle and replenishes matrix cells as they differentiate (Cotsarelis et al., 1990). Contact with DP cells appears to be essential, although not necessarily sufficient, to convert stem cells to matrix cells (Taylor et al., 2000; Oshima et al., 2001).

The spatially defined differentiation programs of the hair follicle make it an excellent model to study the mechanisms that underlie stem cell lineage determination. A myriad of signal transduction pathways have been implicated in the process (for reviews see Fuchs et al., 2001; Panteleyev et al., 2001; Millar, 2002; Niemann and Watt, 2002; Alonso and Fuchs, 2003). Sonic hedgehog (Shh) is expressed in a subset of matrix cells and is thought to play a role in follicle proliferation (Oro and Higgins, 2003). Hair shaft differentiation is dependent on Wnt signaling, and alterations in this pathway generate a number of follicle abnormalities (Gat et al., 1998; Chan et al., 1999; Millar et al., 1999; Huelsken et al., 2001; Andl et al., 2002). In mice, the Wnt reporter gene TOPGAL is most active in the differentiating cells (cortex) of the developing hair shaft (DasGupta and Fuchs, 1999).

Canonical Wnt signaling results in the stabilization of β-catenin, which activates members of the Lef1/Tcf family of DNA-binding proteins (for review see Moon et al., 2002). For matrix progeny to respond to Wnt signaling (Reddy et al., 2001), they must express Lef1. Recently, it was shown that Lef1 expression in keratinocytes is dependent on Noggin, a DP-secreted inhibitor of BMP signaling (Botchkarev et al., 1999; Jamora et al., 2003). When Lef1-positive hair progenitor cells respond to Wnts, they activate hair-specific keratin genes, which possess Lef1 DNA-binding sites in their 5′ regulatory regions (Zhou et al., 1995; Merrill et al., 2001). Although Lef1–β-catenin complexes appear to be required for hair keratin gene expression, other transcription factors, including Foxn1 and Hoxc13, may also be involved (Prowse et al., 1999).

At first glance, the recently identified role for BMP inhibition in the hair follicle seems straightforward. However, Lef1 mRNA expression is highest in proliferating matrix cells, and there is an unexplained lag before Lef1 and β-catenin concentrate in the nucleus and cells progress toward hair differentiation (Zhou et al., 1995; DasGupta and Fuchs, 1999). Moreover, a positive role for BMPs has also been postulated in the follicle, where both proliferation and differentiation seem to be affected by BMP activation (Blessing et al., 1993; 1996). In addition, whereas Noggin has a positive effect on Lef1 expression, it has a negative effect in cortex, where ectopic expression of Noggin diminishes expression of a number of transcriptional regulators Msx1, Msx2, Foxn1, and Hoxc13 as well as blocking hair differentiation (Kulessa et al., 2000; Ma et al., 2003).

The difficulties in understanding BMP-mediated regulation in the follicle are compounded by the complex expression patterns involved. BMP2 and BMP4 are in epithelial hair progenitor cells; BMP4 and BMP7 are in the DP; BMP7 is in ORS; and BMP7, BMP8a, and BMP8b are in IRS (Zhao and Hogan, 1996; Takahashi and Ikeda, 1996; Kratochwil et al., 1996; Wilson et al., 1999). Although BMPs seem to be important for hair follicle development and differentiation, sifting through their potential functions has posed a substantial challenge.

In this paper, we use conditional knockout technology to explore the roles of BMP signaling in hair follicle differentiation. Given the potential for redundancy in BMP expression, we focused on the BMPRIA receptor, known to be expressed in hair follicle placodes (Botchkarev et al., 1999) and required for specification of limb ectoderm (Ahn et al., 2001; Soshnikova et al., 2003). By mating keratin 14 promoter-driven Cre mice (Vasioukhin et al., 1999) with Bmpr1a flox/flox mice (Mishina et al., 2002), we were able to drive Cre recombinase expression in the stem cell compartment of embryonic skin epithelium. Here, we identify and characterize the defects in hair morphogenesis that arise from a failure to undergo BMP signaling. We also evaluate how the loss of BMPRIA impacts on translation of Lef1/β-catenin–mediated gene expression in the developing follicles and in postnatal skin grafts. Our results provide strong support for the previously proposed role of inhibition of BMP signaling in producing and maintaining Lef1-positive matrix cells, and they also reveal a novel role for BMP receptor activation in priming the hair shaft progenitor cells to respond to Wnt signaling. The sequential inhibition and activation of BMPRIA are both essential for development of the hair shaft, and ironically, both steps precede Wnt signaling in the differentiation program.

To introduce conditional mutations into the Bmpr1a locus, we mated Bmpr1a flox/flox (fl/fl) mice (Mishina et al., 2002) with our transgenic mice that express Cre under the control of the K14 promoter (Vasioukhin et al., 1999). Offspring from matings of Bmpr1a(fl/+), K14Cre mice yielded litters of the expected numbers, genotype and Mendelian ratios. However, mice harboring two floxed alleles of Bmpr1a and K14-Cre were smaller in size (Fig. 2 A) and displayed aberrant whiskers (Fig. 2 B). Although the number of whiskers in knockout (KO) mice was nearly always less than their wild-type (WT) counterparts Fig. 2 B′), this varied, and some KO animals displayed no whiskers at all. As a measure of epidermal function, newborn KO animals and their WT counterparts were able to exclude blue dye, indicating that the epidermal barrier was intact (unpublished data). However, most KO animals died several days after birth, most likely due to defects in oral epithelium. Overall, neonatal mice conditionally targeted for disruption of Bmpr1a gene function could readily be sorted on the basis of phenotype (Fig. 2, A and B), and this was confirmed by genotype (Fig. 2 C).

As judged by immunofluorescence, the targeted ablation was nearly if not fully complete by E15.5. In contrast to WT skin, KO skin revealed few if any traces of residual BMPRIA protein in the developing epidermal basal layer (Fig. 2 D). By E17.5, BMPRIA-positive hair follicles were visible in WT backskin (Fig. 2 E), and by P1, WT follicles revealed BMPRIA-positive cells in most if not all differentiation stages (Fig. 2 F). Expression was highest in ORS, matrix, and precortex, in good agreement with Botchkarev et al. (1999). The absence of BMPRIA in embryonic skin verified the success of targeting, and importantly, revealed no unusual stability in Bmpr1a mRNA or protein that might interfere with our analyses.

The KO animals used for immunofluorescence exhibited few whiskers and no visible hair. In examining BMPRIA expression, we noticed a correlation between the extent of BMPRIA mosaicism and the number of whiskers and hair present. Furthermore, analyses revealed that the aberrant whiskers (Fig. 2 B) were genetically mosaic, whereas whiskers and hairs failed to form in complete absence of BMPRIA. For the present paper, we focused on conditional knockouts where the majority of the follicles displayed no traces of anti-BMPRIA staining.

To assess whether follicular defects extended to backskin, and to probe more deeply into the nature of these defects, we next examined the histology of WT and neonatal Bmpr1a-null follicles at postnatal day two (P2). Toluidine blue–stained semithin sections revealed gross abnormalities in these follicles (Fig. 3). By this age, some WT backskin hairs began to display a well-differentiated structure, replete with ORS, IRS, and hair shaft (Fig. 3 A, HS). Most P2 KO follicles exhibited an ORS and bulbar like structure, but lacked a typical IRS and hair shaft (Fig. 3 B). Overall, follicles appeared immature, resembling those of late-stage embryonic skin, rather than postnatal skin. However, as KO follicles extended downward, they wove in and out of the plane of sectioning. In contrast, KO epidermis retained a differentiation program that was morphologically similar to WT.

By P4, differences in follicles were even more dramatic (Fig. 3, C and D). P4 KO follicles still showed no signs of a dark blue–stained cortex, a cross-striated medulla or an IRS, all well-developed in WT P4 counterparts (Fig. 1). In addition, the distinction between DP and matrix was less obvious than in WT follicles. Overall, aberrations in KO follicles did not seem to be due to a developmental lag, but rather a molecular defect in differentiation.

Ultrastructure of WT and Bmpr1a-null skin helped to define the extent of abnormalities (Fig. 4). WT P2 follicles sometimes displayed an immature core, as the medulla was still forming in younger follicles of this age. However, most P2 follicles did exhibit a keratinized cortex, packed with dense keratin filaments, and a well-developed IRS, whose appearance precedes the hair shaft (Fig. 4 A; Robins and Breathnach, 1970). At early stages of normal terminal differentiation, all IRS layers displayed trichohyalin granules (Th), which were larger and more abundant than in medulla. Progressively upward, the Henle's layer of IRS became keratinized, revealing it as an electron dense row of flat, vertical cells (Fig. 4 A). The companion cell layer abutted the Henle's layer.

In P2 Bmpr1a-null follicles, ORS appeared largely normal (Fig. 4 B). Internally to ORS was a group of cells, which possessed morphology and orientation similar to the companion cell layer. The follicle was encased by a dermal sheath (DS; Fig. 4, B and F, arrowheads), but the morphological distinction between surrounding mesenchymal cells and KO follicle cells was otherwise difficult to make. With the exception of occasional Henle cells, only a few P2 follicles (∼1:100) displayed morphological signs of IRS cells (not depicted; and Fig. 3), and the number of such follicles correlated with the rare BMPRIA mosaic follicles.

At the base (bulb) of both WT and KO follicles, the relatively undifferentiated cells of the matrix were present (Fig. 4, C and D). The matrix displayed numerous mitoses, which are consistent with its proliferative status. In addition, a basal lamina of ECM separates the inner strand of DP, necessary for maintaining matrix as undifferentiated hair progenitors (Fuchs, et al., 2001; Millar, 2002). Otherwise, the difficulties faced earlier in distinguishing mesenchymal and epithelial cells persisted at the ultrastructural level. Because the keratin 14 promoter is only active in skin epithelial cells and not in DP (Vasioukhin et al., 1999), DP defects seemed to be indirectly due to BMPRIA loss.

Finally, the differences noted in P2 follicles were exaggerated by P4 (Fig. 4, E and F). At this stage, WT medulla was well developed, and its organized row of cells at the follicle center were packed with trichohyalin and melanin granules (Fig. 4 E, Mel). In contrast, the P4 KO follicle stem consisted predominantly of ORS and companion cell layer cells (Fig. 4 F). Although melanin granules were present in cells of KO follicles, the lack of medulla cells may account for the perturbations seen in their distribution.

To further characterize the abnormalities in Bmpr1a-null follicles, we turned toward immunofluorescence microscopy. In WT follicles, IRS and medulla label with an AE15 mAb, specific for trichohyalin (O'Guin et al., 1992). In contrast, AE13 marks the hair-specific keratins, the major structural proteins of hair (Lynch et al., 1986). In WT neonatal P1 skin, follicles are often positive for both AE15 and AE13 (Fig. 5, A and B). However, in P1 Bmpr1a-null skin, no AE15 or AE13 staining was observed (Fig. 5, C and D).

Similar differences were seen in P8 follicles, a stage where WT follicles were nearing their height of growth and differentiation. By this age, all WT backskin follicles displayed AE15 and AE13 staining whenever sectioning was in the plane of these structures (Fig. 5, E and F). In contrast, >90% of P8 KO follicles lacked AE15 and AE13 reactivity (Fig. 5, G and H). The occasional KO backskin follicle that exhibited histological and biochemical signs of IRS and hair shaft production correlated with the low levels of BMPRIA mosaicism observed.

Antibodies against keratin 5 (K5) are specific for the epidermal basal layer and ORS of the normal hair follicle (Fig. 5, A and E; Byrne et al., 1994). In addition, certain keratin 6 (K6) proteins are specific for the companion cell layer, separating ORS and IRS (Fig. 5, B and F; Winter et al., 1998). Both of these markers were maintained in KO follicles (Fig. 5, C, D, G, and H). As judged by immunofluorescence, K6 expression appeared to be increased and expanded in KO versus WT follicles.

To probe further into the aberrations in IRS and hair differentiation, we examined the expression of the transcription factors known to regulate these programs. The transcription factor GATA-3 plays an essential role in the differentiation of IRS progenitor cells (Kaufman et al., 2003), whereas hair-specific keratin genes possess regulatory sequences for the Lef1 transcription factor (Zhou et al., 1995). GATA-3 was markedly diminished in KO follicles relative to their WT counterparts (Fig. 6, A and B). In contrast, although KO follicles exhibited a loss of AE13 immunoreactivity, they retained Lef1 (Fig. 6, C and D). The result supported the hypothesis that Lef1 expression in skin keratinocytes may depend on inhibition of BMP signaling (Botchkarev et al., 1999; Jamora et al., 2003).

Finally, we tested for possible secondary abnormalities in DP expression. Despite a relatively uncondensed DP, both Lef1 and the mesenchymal transcriptional regulator Runx3 were expressed in both WT and KO DP (Fig. 6, C, D, G, and H). WT and KO follicle bulbs also exhibited a line of labeling with antibodies against β4 integrin, a component of hemidesmosomes which help to adhere matrix cells to the basement membrane at the DP boundary (Fig. 4, G and H). With this said, other DP aberrations are noted below.

To examine proliferation status of KO follicle cells, we used an antibody against Ki67, a proliferating nuclear antigen that is active in cycling cells. This antibody strongly stains the proliferating matrix cells of WT P8 follicles (Fig. 6 A). Importantly, the antibody also labeled matrix in KO follicles, although the pattern of labeling was different from that in WT (Fig. 6 compare B with A). In addition, as judged by in situ hybridization, expression of Sonic hedgehog (Shh) was still seen in KO follicle bulbs. However, rather than the highly restricted, asymmetric pattern typical of Shh in WT bulbs (Fig. 6 E), Shh expression was expanded and expressed symmetrically in matrix cells near the DP boundary (Fig. 6 F). When together, with our ultrastructural studies, these data suggested that matrix cells were not growth arrested, but rather disorganized and blocked in their differentiation. For the IRS, the block appeared to be at the level of elevated GATA-3 expression, although it is not clear whether the decrease was a cause or consequence of BMP signaling; for the hair shaft, the block appeared to be past the level of Lef1 expression but before the Wnt-mediated activation of hair-specific keratin genes. Further experiments addressing this issue are provided below.

Bmpr1a-null mice were typically frail, and only mosaic animals survived beyond several days after birth. Consequently, to assess the long-term consequences of ablating BMPRIA expression in skin epithelium, we grafted full thickness neonatal skin from Bmpr1a-null and WT littermate mice onto the hairless backs of immunocompromised, recipient nu/nu mice. To confirm the uniformity of targeting within the skin selected for grafting, we used KO skin that exhibited spots of organized melanin granules, but no signs of body hairs (Fig. 7, A and A′). At 24 d after engraftment, hairs were evident only on WT grafts (Fig. 7 B). Histologically, these follicles appeared normal (unpublished data). In contrast, follicles formed in KO grafts were aberrant (Fig. 7, C and C′). In some cases, odd follicle-like structures had grown downward but displayed only wisps of IRS cells (Fig. 7 C, left). These follicles lacked a hair shaft and IRS, but possessed a well-developed sebaceous gland. In most cases, cystlike structures developed (Fig. 7, C and C′). At the light microscopy level, the cells within the cysts appeared morphologically similar to those of embryonic follicles, but ultrastructural analyses revealed their peculiar organization (Fig. 7 D). The cysts typically contained invaginations of mesenchymal cells, which were separated from the epithelial cells by a basal lamina (Fig. 7 D, inset). Each cyst was also surrounded by a basal lamina, on the other side of which was a dermal sheath of mesenchymal cells (Fig. 7 D). At the central core, some cyst cells possessed melanin granules (unpublished data). Their presence was interesting, as these granules are taken up by early stage, differentiating hair cells. With this possible exception, cyst cells appeared to be a mixture of matrix, ORS, and companion cell layer cells. They displayed numerous mitoses, reflective of a proliferative state, but they also exhibited some apoptotic figures, in contrast to their WT counterparts.

At the immunohistochemistry level, some rare follicle-like structures were seen in KO skin (Fig. 8 A, center structure). These possessed sebaceous glands and ORS, which were positive for K5; they also displayed a cystlike structure at the base (Fig. 8 A, example shown). However, the stem of these rare structures were hollow, with no signs of AE15 or AE13 staining (Fig. 8, A and B). Most cysts expressed K5 toward their periphery and K6 at their centers. Many of the more peripheral cells were also positive for Ki67, confirming the proliferative state of the cysts (Fig. 8 C). Antibodies against GATA-3 weakly labeled a small subset of cells within most cysts (Fig. 8 C, inset). In contrast, many cyst cells labeled with antibodies against Lef1 (Fig. 8 D).

In situ hybridization revealed Shh expression, but in an unusual pattern, paralleling the number of invaginations of mesenchymal cells seen within each cyst (Fig. 8 E). Closer inspection revealed that Shh was restricted to a subset of cells maintaining close contact with invaginating DP-like, mesenchymal cells. Even at this stage, the follicle-like structures from grafts continued to express Wnts, as shown here by the marked expression of mRNAs for Wnt10b, a major Wnt of the hair follicle (Fig. 8 F; Reddy et al., 2001). Wnt3, another Wnt prominent in the hair follicle, has recently been shown to be expressed in the developing limb epithelium of Bmpr1a-null animals (Soshnikova et al., 2003). Thus, expression of both Shh and Wnt ligands seemed to persist in the aberrant KO follicles.

The corresponding epithelial invaginations surrounding the mesenchymal inlets of cysts were distinguished by anti-β4 staining (Fig. 8 G). In addition, a ring of anti-β4 staining marked the epithelial boundaries of cysts, which were encased by a basal lamina. On the opposite side were several layers of mesenchymal cells. They differed from a normal DS in that they were positive for Runx3 (Fig. 8 G) and alkaline phosphatase activity (Fig. 8 H). Curiously, both Runx3 and alkaline phosphatase are found in dermal condensates including DP (Handjiski et al., 1994; Yamashiro et al., 2002).

Thus, in the absence of BMPRIA, follicles appeared to develop, but the ORS could not sustain its normal architecture in the absence of an IRS and hair shaft. The bulbar structures continued to grow, but reorganized into cystlike structures. Interestingly, changes in the biochemistry of the KO cyst cells appeared to impact indirectly on the organization/specification of both epithelium and mesenchyme.

Finally, GATA-3, which is expressed in WT epidermis (Kaufman et al., 2003), was down-regulated in KO epidermis as it was in the KO IRS progenitors (Fig. 8 C). In addition, normally expressed in embryonic epidermis, Lef1 was induced in postnatal KO epidermis, albeit at reduced levels relative to KO cysts (Fig. 8 D). Some epidermal regions in grafted skin were also thicker than WT, and exhibited K6 expression and increased Ki67 labeling, reflective of hyperproliferative epidermis (Fig. 8, B and C). For the purposes of the present paper, we focused on understanding the role of BMPRIA on hair follicles, although clear epidermal abnormalities were noted.

To further explore the molecular defects underlying BMPRIA-related defects in hair shaft differentiation, we tested the ability of Bmpr1a-null matrix cells to activate Lef1/β-catenin–regulated genes. For these studies, we generated mice conditionally null for Bmpr1a and transgenic for TOPGAL, driving β-galactosidase under the control of the Wnt-responsive TOP promoter. In vivo, TOPGAL is faithfully expressed in skin, wherever Lef1 and nuclear β-catenin are present, e.g., as in precortical and cortical cells (DasGupta and Fuchs, 1999). Genotyping was conducted to confirm that the desired genetics were successful (unpublished data).

For analyses, we took frozen sections of neonatal skin (test) and cartilage (control) and subjected them to X-gal assays. As shown in Fig. 9 A, WT follicles scored positive for TOPGAL activity in the cells known to respond to Wnt signaling. In contrast, follicles from Bmpr1a-null, TOPGAL-positive mice did not (Fig. 9 A′). That TOPGAL was present in KO animals was confirmed by its presence in tissues where the K14-Cre recombinase was not, e.g., in cartilage (Fig. 9 A′, inset).

In WT skin, TOPGAL expression correlated well with cells that exhibit nuclear Lef1 as well as nuclear β-catenin (Fig. 9 B). In KO skin, even though Lef1 was expressed, nuclear β-catenin was not detected in the corresponding region of developing follicles (Fig. 9 B′). These findings suggested that the defect resided in a failure of BMPRIA-deficient follicles to progress to the stage where they respond to Wnts, stabilize β-catenin and activate Lef1, and their downstream targets, including the hair keratin genes.

That BMPRIA signaling normally occurs at this time was verified by immunostaining with antibodies against phospho-Smad1, the transcriptional regulator, which is phosphorylated and activated upon ligand engagement of BMPRIA (for review see Massague, 1998). In WT follicles, these antibodies stained nuclei in the developing IRS, as well as in some layers of the developing hair (Fig. 9 C). In contrast, no nuclear staining was detected in either WT ORS, or in the hair bulb of KO follicles, although it was still detected in the DP, where BMP receptors were not targeted for ablation (Fig. 9 C′). These findings suggest that BMP signaling normally occurs in both the developing IRS and hair, and both are compromised in KO follicles.

To explore these differences in greater detail, we cultured primary keratinocytes isolated from skins of WT and KO animals. PCR was used to confirm the genotype of the cell populations. Two independently derived cultures were prepared for each genotype, and analogous behavior was observed within a particular genotype.

Because keratinocytes produce and secrete high levels of BMP2 and BMP4 (Jamora et al., 2003), we first verified that loss of BMPRIA was sufficient to prevent BMP signaling in these cultures. As shown in Fig. 9 D, WT keratinocytes displayed prominent nuclear anti–P-Smad1 staining. In contrast, KO keratinocytes exhibited little if any staining (Fig. 9 D′).

Previously, we showed that cultured neonatal keratinocytes require conditioned medium containing the BMP inhibitor Noggin in order to induce the expression of Lef1 mRNA and protein (Jamora et al., 2003). In the absence of added Noggin, WT keratinocytes displayed little if any nuclear staining of anti-Lef1 (Fig. 9 E). In contrast, KO keratinocytes displayed strong nuclear anti-Lef1 staining (Fig. 9 E′). Western blot analyses confirmed a marked increase in Lef1 expression in the absence of BMPRIA (Fig. 9 F). Overall, these data were consistent with a block in BMP signaling in response to the loss of BMPRIA. As importantly, the data demonstrated convincingly that the action of Noggin-conditioned medium on Lef1 expression observed previously in keratinocytes (Jamora et al., 2003), is mediated through effects on BMP receptor activation.

To test whether the Lef1 produced by Bmpr1a-null keratinocytes was able to activate Wnt-regulated genes, we conducted in vitro studies using two well-established Wnt-responsive promoters: TOP, containing multimerized Lef1 DNA-binding sites (Korinek et al., 1997); and HK1, one of the hair-specific keratin promoters containing functional Lef1 DNA-binding sites (Zhou et al., 1995; Merrill et al., 2001). As a control, we used the FOP promoter, harboring mutations in the Lef1 binding sites of TOP. All promoters were used to drive expression of luciferase (Flash).

Under normal media conditions, both cultures exhibited low activation of TOPFlash and HK1Flash, relative to FOPFlash (Fig. 9, G and H). In the presence of an NH₂-terminally truncated, stable β-catenin, WT keratinocytes still showed minimal activation of either TOPFlash or HK1Flash. In contrast, KO cells displayed a greater than fivefold up-regulation of TOPFlash and HK1Flash in response to transfected ΔNβ-catenin (Fig. 9, G and H). The results obtained with Bmpr1a-null keratinocytes were similar to those which we had observed with WT keratinocytes treated with Noggin (Jamora et al., 2003). Together, these data demonstrated that (a) Bmpr1a-null keratinocytes are deficient in Smad activation; (b) these KO keratinocytes produce Lef1, whereas WT keratinocytes require Noggin treatment for activation; and (c) Wnt-responsive promoter activity in Bmpr1a-null keratinocytes can be achieved by transfection with a constitutively stabilized β-catenin.

The first obvious morphological defect in K14-Cre, Bmpr1a fl/fl skin was the failure of the IRS to develop. Our studies also show clearly that BMPRIA signaling occurs in the IRS, and in the absence of the receptor, developmental defects in the IRS arise under conditions where the ORS appears relatively unperturbed and the hair shaft has not yet developed. Thus, it is curious that ectopic expression of an Msx2-Noggin transgene in the epithelial cells of the hair bulb of >P2-aged mice resulted in primarily hair shaft defects without IRS abnormalities (Kulessa et al., 2000). One explanation is that differential expression of BMP4 in the IRS but not the hair shaft spares the IRS from Msx2-noggin action. In this regard, BMP signaling molecules have been reported to have relatively short-range effects (Nikaido et al., 1999).

Precisely how IRS is affected is not yet clear. Our results suggest an inability of Bmpr1a-null skin to sustain elevated levels of GATA-3, the transcription factor that we recently implicated in IRS differentiation (Kaufman et al., 2003). Intriguingly, links between GATA factors and BMP signaling have been reported in other cell types (Patient and McGhee, 2002). Thus, in cardiomyocytes, Noggin suppresses expression of GATA-4, required for terminal differentiation (Monzen et al., 1999), whereas conversely in embryonic blood, BMP4 induces GATA-2 (Xu et al., 1999). In addition, GATA factors have recently been shown to induce and/or sustain the expression of BMPs (Nemer and Nemer, 2003; Peterkin et al., 2003), thereby providing a potential feedback loop to promote terminal differentiation. When considered with our present findings on BMPRIA, the IRS lineage becomes a new candidate for a cell type whose differentiation may be regulated by BMP-GATA circuitry.

For more than a decade, BMPs and BMP antagonists have been implicated in hair follicle morphogenesis (for reviews see Botchkarev, 2003; Alonso and Fuchs, 2003). During embryonic development, increased BMP signaling inhibits the initiation of follicle morphogenesis, whereas noggin activity stimulates the process (Noramly and Morgan, 1998; Botchkarev et al., 1999; Botchkarev, 2003). Postnatally, noggin still seems to block hair shaft differentiation and hair keratin expression (Botchkarev et al., 1999; Kulessa et al., 2000; Jamora et al., 2003). The fact that hair progenitor cells form in the absence of BMPRIA provides compelling evidence that BMP signaling is not required for hair progenitor formation. However, the defect in hair formation in these follicles suggests that BMP signaling is necessary for the differentiation of these progenitors.

BMPRIA is the only BMP receptor known to be expressed in the follicle (Botchkarev et al., 1999), and the pattern of phospho-Smad1 antibody staining was consistent with the activation of this receptor not only in IRS but also in some cells of the developing hair. With this said, because the IRS develops before the hair shaft and is required for its proper differentiation (Kaufman et al., 2003), we cannot be unequivocally certain that the defects in hair differentiation are direct. However, the impairment in hair differentiation and hair keratin gene expression seemed to be stronger than in the GATA-3 knockout, leading us to surmise that at least some of the defects in matrix to cortical conversion are attributable directly to the absence of BMPRIA activation in this lineage.

Although phospho-Smads have not been implicated in hair keratin gene expression, several other DNA-binding proteins are critical, including Lef1, which functions as a positive regulator upon Wnt signaling of the precortical cells (Zhou et al., 1995; DasGupta and Fuchs, 1999; Merrill et al., 2001), Hoxc13, a homeobox gene expressed in matrix, precortex, cortex, and cuticle (Jave-Suarez et al., 2002); and Foxn1, a forkhead/winged-helix transcription factor (Schlake et al., 2000). When defective, all of these genes yield marked hair follicle phenotypes (Nehls et al., 1994; van Genderen et al., 1994; Segre et al., 1995; Godwin and Capecchi, 1998).

Foxn1 and Hoxc13 are down-regulated in Msx2-noggin transgenic follicles (Kulessa et al., 2000). A priori, because Lef1 is up-regulated, both in the presence of noggin (Kulessa et al., 2000; Jamora et al., 2003) and in the absence of BMPR1 (this paper), Foxn1 and Hoxc13 might seem more likely candidates for BMP control. In fact, we have found that Foxn1 is down-regulated in our Bmpr1a-null skin (unpublished data). With this said, Hoxc13- and Foxn1- null mutations both permit development of cortex and medulla, whereas Lef1-null mutations act considerably earlier in hair follicle morphogenesis (van Genderen et al., 1994).

Because the TOPGAL transgene is only responsive to Lef1/β-catenin and not to Foxn1 or Hoxc13, the absence of its activity in the postnatal follicles of Bmpr1a-null mice was informative. Lef1 expressed by KO keratinocytes was clearly functional, as evidenced by the ability of a constitutively stabilized ΔNβ-catenin to rescue TOP as well as hair keratin promoter activity in Bmpr1a-null keratinocytes. In addition, at least some Wnts were still expressed by Bmpr1a-null skin. Thus, although other explanations are possible, a defect in Wnt responsiveness could explain the failure of hair keratin genes to be activated in Bmpr1a-null skin. Interestingly, in thymus, Foxn1 expression was recently shown to be dependent on Wnt signaling (Balciunaite et al., 2002), providing a possible explanation as to why Foxn1 expression was also down-regulated in the Bmpr1a-null skin.

Our results are particularly intriguing in light of recent works by Soshnikova et al. (2003), who targeted conditional ablation of the BMPRIA receptor in the apical ectodermal ridge. Not only was limb development severely impaired in these animals, but in addition, Wnt responsiveness was markedly impaired. Experiments using gain or loss of function mutations in β-catenin revealed that β-catenin stabilization acts downstream of BMPRIA receptor activation in the AER (Soshnikova et al., 2003). Although BMP signaling need not always occur upstream from Wnt signaling (Barrow et al., 2003), our findings suggest that it appears to act earlier than Wnts in hair shaft differentiation, in a fashion more similar to the AER.

In the context of our prior studies on hair shaft lineage determination, our findings suggest a new twist on the interplay between Wnts and BMP signaling. Ironically, in the postnatal hair follicle, inhibition and activation of BMP signaling both seem to be required to prime cells in order to execute Wnt signaling and activate its differentiation program. Fig. 10 provides a model, which may explain this paradox. In the hair shaft lineage, DP cells secrete noggin, which may inhibit BMPRIA activation in matrix. This inhibition may then allow matrix cells to produce elevated levels of Lef1 mRNA (Zhou et al., 1995; Botchkarev et al., 1999; Jamora et al., 2003) along with the other transcription factors such as Msx1/2 (Kulessa et al., 2000). However, without BMP signaling, the matrix cells may be prevented from activating Lef1/β-catenin and differentiating.

Interestingly, exposing cells to noggin results in BMP4 up-regulation (Kulessa et al., 2000). Thus, in time, matrix cells may balance BMP inhibitors with BMP activators, thereby activating BMPRIA. This process could also be accelerated as maturing matrix cells sever their connection with DP and move upward. As long as the hair progenitor cells have stockpiled sufficient Lef1, they should subsequently be able to respond to Wnt signaling, and activate the hair keratin genes (Merrill et al., 2001), Foxn1 (Balciunaite et al., 2002), the Wnt reporter gene TOPGAL (DasGupta and Fuchs, 1999), and presumably other genes, which then drive the cells into terminal differentiation. Although additional studies will be necessary to test this model, our current results take us one important step closer to understanding the intricacies that underlie the complex spatiotemporal patterns of hair follicle growth and differentiation.

The generation and characterization of the Bmpr1a fl/fl mice has been described previously (Mishina et al., 2002). The mice were mated with previously generated K14-Cre mice (Vasioukhin et al., 1999) to generate mice homozygous for the loss of BMPRIA function in skin epithelium. K14-Cre is active by E9 of skin development, and is effective at quantitative ablation by E15. These mice were also mated on the background of TOPGAL transgenic mice, driving expression of β-galactosidase only under conditions where cells are responsive to Wnt signaling and already express a member of the Lef1/Tcf family of DNA-binding proteins (DasGupta and Fuchs, 1999).

Tissues for immunofluorescence and hematoxylin and eosin staining were embedded and frozen immediately on dry ice. For transmission EM, tissues were fixed in 2% glutaraldehyde, 4% formaldehyde, and 2 mM CaCl₂ in 0.05 M sodium cacodylate buffer for >1 h at room temperature and processed for Epon embedding as described previously (Kaufman et al., 2003). Samples were visualized using a transmission electron microscope (Tecnai G2).

Digoxygenin probe synthesis was performed according to the manufacturer's instructions (Roche). In situ hybridizations on 10-μm frozen sections were performed as described previously (Schaeren-Wiemers and Gerfin-Moser, 1993).

OCT sections of indicated thickness were briefly fixed (30 s) in 0.1% glutaraldehyde, washed three times in PBS, and incubated in X-gal staining solution for 1–2 h at 37°C. Standard procedures were used for immunofluorescence staining. In short, OCT sections of indicated thickness were fixed for 10 min in 4% PFA in PBS and washed four times for 5 min in PBS. When staining with mouse mAbs, we used the reagents and protocol from the MOM Basic kit (Vector Laboratories). In other cases, the following block/diluent was used 2.5% NDS, 2.5% NGS, 1% BSA, 2% gelatin, and 1% Triton X-100 in PBS. The following primary antibodies were used at the indicated concentrations: AE13 (mouse, 1:50–1:100) (Lynch et al., 1986); AE15 (mouse, 1:10) (O'Guin et al., 1992); K5 (gp, 1:250) (Fuchs laboratory); GATA-3 (mouse, 1:100) and BMPRIA (1:100) (Santa Cruz Biotechnology, Inc.); β-galactosidase (rabbit, 1:400; mouse, 1:100) (Harlan; Sigma-Aldrich); Lef-1 (rabbit, 1:250) (Fuchs laboratory); K6 (rabbit, 1:500) (Fuchs laboratory); and Ki67 (rabbit, 1:1000) (NovoCastra Laboratories Ltd.). Relevant FITC or TxR-conjugated donkey or goat antibodies (1:100) (Jackson Laboratories) were used for detection of primary antibodies. For phospho-Smad1, we used the TSA™ Plus Fluorescing System from Perkin Elmer Life Sciences.

Neonatal Bmpr1a conditionally null animals were identified by their lack of whisker follicles, and were killed along with their WT littermates. Full-thickness skins were removed from the torsos of wild-type and null embryos, spread on a sterile plastic dish, and stored briefly at 4°C. During this time, each skin graft recipient site was prepared by removal of a patch of full-thickness skin on an anesthesized female nu/nu CD-1 mouse. Donor skin was placed on the graft bed and secured by sterile gauze and cloth bandages. Each showed a consistent phenotype dependent on the presence or absence of BMPRIA in the donor skin. Primary keratinocytes were cultured as described previously (Jamora et al., 2003).

We thank Yuji Mishina (National Institutes of Health, Bethesda, MD) for the Bmpr1a floxed mice and for careful reading of our manuscript before submission. We thank T.T. Sun (New York University School of Medicine, New York, NY) for the gift of AE13 and AE15 antibodies; Y. Groner (Weizmann Institute, Rehovot, Israel) for the Runx3 antibodies; and C. Tabin (Harvard Medical School, Boston, MA) for the Shh probe.

E. Fuchs is an HHMI Investigator. This research was funded by a grant from National Institutes of Health.