BMP signaling in dermal papilla cells is required for their hair follicle-inductive properties

Unveiling the importance of BMP signaling in regulating the molecular identity of DP cells

To explore what constitutes the hair-inducing cell fate of DP cells, we evaluated our previously acquired hair bulb molecular signatures for surface receptors and growth factors that could play a role in conferring DP cell fate specification in hair induction and growth. Out of 184 annotated DP signature genes, eight were receptors and eight were growth factors. An additional nine receptors and six growth factors were expressed by DP and also by other hair bulb cell types, while five further growth factors, including Wnt3a and Shh, surfaced as potential DP signature receptor ligands produced by other hair bulb cell types but not DP (Table 1).

To directly test the role of these pathways, we performed an in vitro screen in which we used FACS to freshly isolate pure DP cells, which in culture lose their hair-inducing activity within the first passages (Kishimoto et al. 2000). We then asked whether ligand-mediated pathway activation could preserve DP’s molecular identity in vitro and prolong hair-inducing capacity in vivo (Fig. 1A). We tested alkaline phosphatase (AP) expression as our initial readout, since it is both a well-established DP marker (Fig. 1B; Handjiski et al. 1994) and also among the most highly expressed genes within the DP signature (Akp2) (Rendl et al. 2005). Nearly all freshly isolated DP cells expressed AP activity, and during the first few days in culture, the majority remained positive (Fig. 1C). However, after the first passage, DP cells began to lose AP expression (Fig. 1C, left graph), and by the second passage, <5% were AP-positive (data not shown). When primary DP cells were colabeled with bromodeoxyuridine (BrdU), 80%–90% of DP cells were proliferative in culture (Fig. 1C, right graph). This number remained high with passage, despite a steady decline of BrdU/AP double-positive cells (Fig. 1C, red bars).

We next tested signature growth factors of major signaling pathways to assess whether they affected proliferation and/or could prevent the loss of AP activity (Fig. 1D). None of the ligands caused gross alterations in either morphology (data not shown) or proliferation of the cultures (Fig. 1D, top graph). However, four different BMPs had a marked effect in maintaining AP activity in DP cells in vitro (Fig. 1D, bottom graph). Real-time PCR detection of AP gene expression (Akp2) showed that the BMP-mediated effects were at the mRNA level (Fig. 1E). Moreover, the effects on Akp2 expression were highly specific to members of the BMP family, since 23 other factors—including SHH, WNTs, FGFs, and others that are expressed within the DP microenvironment—scored negatively in this assay. The effects of BMP were not limited to Akp2 as a similar BMP-specific pattern of regulation was also observed when we analyzed the mRNA expression levels of another DP signature gene, Wif1 (Fig. 1F). Moreover, the BMP effects extended to a third DP signature gene, Alx4, and the effects observed were dose-dependent with a sensitivity extending down to 25–50 ng/mL (Supplemental Fig. S2A). Since the potential importance of BMP signaling on DP cells has not been explored previously, and since BMP6 had the most pronounced effects of the BMPs, we concentrated on BMP6 for most of our subsequent studies.

To determine the extent to which other DP signature features are lost upon culture and the degree to which BMPs can alleviate these effects, we treated DP cells with BMP6 and performed semiquantitative PCR on a broad range of DP signature genes expressed by freshly isolated DP cells prior to culture (Rendl et al. 2005). As shown in Figure 2A, many DP signature features were retained in vitro. Notable examples included Fgf7 and Fgf10 encoding key DP factors that are essential for Mx cell maintenance (Guo et al. 1996; Grose et al. 2007), and in many of these cases the genes were not influenced by exogenous BMP6. In contrast to Fgf7 and Fgf10, some signature genes were markedly down-regulated. This included not only Akp2 and Wif1, but also genes encoding Hhip, a Sonic hedgehog inhibitor; Hey1, a downstream target of Notch signaling; and the secreted BMP receptor inhibitors Nog (Noggin) and Sostdc1 (ectodin) (Fig. 2A). Some of these, like Hhip and Sostdc1, were not maintained by exogenous BMP6. Others, including Akp2, Wif1, Alx4, Hey1, and Nog, were up-regulated by BMP6. Interestingly, Bmp6 gene expression itself was down-regulated, suggesting that Bmp6 is regulated by a negative feedback loop.

The effects of BMP6 regulation manifested early during culture initiation and without changes in proliferation and cell morphology (Supplemental Fig. S2B). A similar range of DP signature gene regulation was found with BMP2, BMP4, and BMP7 (Supplemental Fig. S3A). Taken together, these data suggest that DP cells in culture lose certain features of their molecular identity that BMP signaling can restore. Interestingly, these DP signature genes regulated by BMP signaling were not similarly affected by SHH, WNTs, FGF2, FGF7, FGF10, or other growth factors (Fig. 1E,F; data not shown). This was surprising for factors such as FGF2, as DP cells express Fgfr1 and Fgfr2IIIc encoding its receptors.

To further probe the significance of BMPs on the DP signature, we retrovirally overexpressed Bmp6 and a constitutively active form of the BMP receptor 1a (Bmpr1a*) (Akiyama et al. 1997). While overexpression of Bmp6 in DP cells would mimic endogenous expression, overexpression of the active receptor would bypass the need for cytokine activation and suggest that signaling occurs through this receptor. We also tested Wnt3a, expressed by hair bulb epithelial cells and known to extend hair-inducing activity of cells enriched for DP (Kishimoto et al. 2000). All viruses infected with comparable efficiency based on GFP expression derived from the bicistronic expression vectors, and in each case, the expected coding sequences were detected (Supplemental Fig. S4A). Additionally, Wnt3a functionality was verified by nuclear translocation of β-catenin and by activation of the Wnt reporter TOPflash (Supplemental Fig. S4B; Korinek et al. 1998).

As shown in Figure 2B, overexpression of Bmp6 and Bmpr1a* both led to robust maintenance of AP activity that was even stronger when compared with BMP6 cytokine treatment. Similarly, Bmp6-mediated and Bmpr1a*-mediated up-regulation of gene expression was also observed for a number of DP signature genes (Fig. 2C). In contrast, Wnt3a expression did not influence AP activity or expression of most BMP-regulated genes, although it did activate expression of Twist, a known Wnt target gene (Howe et al. 2003) that is also part of the DP signature (Fig. 2C). These data extended our earlier findings (Fig. 1D,E) and indicate that BMP signaling maintains parts of the DP signature that are not regulated by Wnt signaling.

Evidence for BMP signaling in vivo and its regulation in vitro

Given our observations thus far, we turned to addressing whether BMP signaling occurs in the DP in vivo and whether the prerequisites for such signaling are met in vitro. Bmp2 and Bmp4 are both expressed in the hair matrix, while Bmp4 and Bmp6 are expressed in the DP (Fig. 3A; see also Kulessa et al. 2000; Rendl et al. 2005). Bmpr1a is expressed in all different hair bulb populations, although it was especially prominent in the DP (Fig. 3B; see also Botchkarev and Sharov 2004; Rendl et al. 2005). Thus, the potential for both autocrine and paracrine regulatory loops exists within the microenvironment of the bulb. When we evaluated the array data for all BMP ligands and inhibitors expressed in the hair bulb microenvironment, the complexity of possible negative and positive BMP signaling loops was even greater than those emerging from the signature BMP genes alone (Supplemental Fig. S5).

The outcome of BMP signaling is cytoplasmic phosphorylation of SMAD1/5/8, enabling their interaction with SMAD4 and translocation to the nucleus, where their target genes are then transcribed (Massague et al. 2005). As judged by specific immunofluorescence staining for phosphorylated (active) SMAD1, BMP signaling was active in the DP (Fig. 3C; see also O’Shaughnessy et al. 2004). Moreover, within the DP (Fig. 3C, arrows), a gradient of anti-pSMAD1 staining was observed, suggesting the strongest signaling in the cells closest to the differentiating cells of the hair and IRS (Fig. 3C, arrowheads), where BMP signaling is known to be required. Marked by c-Kit (KIT), Mc also exhibited intense pSMAD1 labeling in the upper DP niche. Conversely, the weakest anti-pSMAD1 labeling was at the base of the DP adjacent to the matrix cells, where, intriguingly, BMP inhibition is thought to be required for maintaining matrix character (Kulessa et al. 2000; Kobielak et al. 2003; Andl et al. 2004).

In vitro, BMPs also stimulated nuclear phospho-SMAD1 localization, substantiating the effects of exogenously added BMP6 on DP cells (Fig. 3D). Since the DP signature includes a plethora of negative BMP regulators (Rendl et al. 2005), and since some of these inhibitors are also expressed by other hair bulb cells (see Supplemental Fig. S5; Kulessa et al. 2000), we examined the effects of exogenously added Noggin (NOG). In all cases tested, DP signature genes that were strongly up-regulated by recombinant BMP6 were also potently inhibited by NOG (Fig. 3E). Interestingly, while BMP6 resulted in a dose-dependent down-regulation of Bmp6, it caused an up-regulation of Nog mRNAs (Fig. 3E; see also Fig. 2A). A similar pattern of regulation was also observed for BMP4 (Supplemental Fig. S4B). These findings suggest that in the hair bulb, BMP signaling in the DP is under tight regulation by multiple positive and negative regulators.

BMP regulation of DP signature genes is DP-specific and not recapitulated in dermal fibroblasts or during osteogenic differentiation

Since DP cells are specialized mesenchymal cells that are related to dermal fibroblasts and since BMP signaling plays an important role in generating AP-positive osteoblasts for bone formation and mineralization (Leboy 2006), we wondered how specific BMP-mediated regulation of DP signature genes is to DP cells. To address this question, we first compared the effects of BMP6 on dermal (3T3) fibroblasts with that of freshly isolated DP cells. Of the DP signature genes tested, none was expressed comparably in untreated fibroblasts, nor did they change appreciably in response to BMP6 (Fig. 4A). This was true whether we cultured the cells in fibroblast medium (F) or in DP medium (A, AmnioMAX).

We also contrasted our results with those obtained with C2C12 myoblast cells, known to differentiate toward the osteoblast lineage upon exposure to BMPs (Cheng et al. 2003). In C2C12 cells Nog, Hey1, and Akp2 were all up-regulated by BMP6 (Fig. 4B). In contrast, some other BMP6-sensitive DP signature genes were not induced in BMP6-treated C2C12 cultures (Fig. 4B, bottom row). Moreover, certain established osteoblast genes (Peng et al. 2003) such as Bglap (osteocalcin), Sox9, and Idb2 were induced by BMP6 only in C2C12 cells and not in DP cells (Fig. 4B, top row). These findings underscore the intrinsic features of DP cells and their distinctive response to BMPs.

BMP treatment of DP cells in vitro enhances their ability to induce hair formation in vivo

Since DP cells diminish their molecular signature in vitro, and since this effect can be ameliorated by BMP treatment, we wondered how this might have an impact on the loss of hair-inducing activity of DP cells when cultured in vitro (Kishimoto et al. 2000). To address this issue, we performed chamber-graft hair growth assays in which epidermal KC and fresh dermal cell preparations (also containing DP cells) are transferred into silicone chambers that had been placed over full-thickness wounds on the backs of Nude mice (Lichti et al. 1993; Weinberg et al. 1993). Here we modified this assay to combine GFP-expressing KC and FACSorted pure cultured DP cells that allowed monitoring the graft performance and ensured that the hair-inducing effects are derived from DP cells, respectively (Fig. 5A).

As expected, when KC were grafted alone, they provided a covering to the wound site but no hairs developed (Fig. 5A, left panel). In contrast, KC engraftments supplemented with cultured DP cells exhibited both re-epithelialization and sparse hair formation (Fig. 5A, middle panel). Notably, when DP cells were cultured in medium containing BMP6, their capacity to induce hair was improved compared with the control DP cells (Fig. 5A, right panel). We qualitatively (Fig. 5B) and quantitatively (Fig. 5C) assessed this effect over a range of 10 different experiments. Even when effects of BMPs were moderate, HF production was always enhanced when DP cells were treated with BMPs (Fig. 5B). Interestingly, although DP cells usually became senescent and stopped growing between passages four and six, in one experiment, BMP-treated DP cells could be expanded until passage eight, where they still induced HF formation (data not shown). That said, while BMPs delayed, they did not completely avert the loss of hair-inducing ability, which otherwise typically occurred within the first three passages of control DP cultures.

To monitor the boundaries and origins of epidermal and DP cells, we grafted male cells onto female Nude mice, and conducted fluorescence in situ hybridization (FISH) with Y-chromosome (Ychr) probes (Fig. 5D). Underscoring probe specificity, no Ychr-positive dermal cells were detected in either host skin (shown) or KC control skin (data not shown). In contrast, grafts containing BMP6-treated DP displayed Ychr-positive cell clusters at centers of newly formed HFs (Fig. 5D). These clusters labeled with antibodies against a number of DP markers and were also positive for AP activity (Fig. 5D; additional data not shown). Thus, properties of the DP signature that were otherwise lost without BMP treatment in culture were restored when cells were exposed to their natural environment.

Interestingly, a number of AP-negative cells within dermis also scored positive for Ychr (Fig. 5D, arrows). These cells appeared to be fibroblasts, suggesting that DP cells can contribute to other dermal cells in a wound setting. Recently, similar behavior was shown for dermal sheath cells that normally line the ORS but that contribute to dermal regeneration upon skin injury (Gharzi et al. 2003). Together, these findings suggest that under the conditions used here, BMP6-treated DP cultures not only improved HF-inducing capabilities, but also adopted at least one alternative lineage.

BMP signaling is required for HF induction by DP cells

The data thus far indicated that BMP signaling favorably regulated DP signature genes in vitro and improved HF induction in vivo. While BMP signaling did not appear to be sufficient on its own to fully maintain DP’s ability to induce HF formation, we wondered whether it nevertheless might be an indispensable component of the process. To address this question, BMP signaling had to be specifically abolished in the DP. This is presently not possible by conventional mouse genetics, since no known promoter uniquely targets expression to DP. To circumvent this caveat, we devised the in vitro/in vivo hybrid knockout strategy outlined in Figure 6A. We first crossed the Bmpr1a floxed mouse strain (Bmpr1a^fl/fl), targeted for conditional loss of the BMP receptor (Mishina et al. 2002), with Rosa-Cre^ERtam mice, encoding a tamoxifen-inducible Cre recombinase under the control of the ubiquitously active Rosa26 promoter (Badea et al. 2003). To isolate pure populations of DP cells, we then bred these mice with our K14-H2B-GFP/Lef-RFP reporter mice and employed FACSorting (Rendl et al. 2005). We then activated Cre-mediated Bmpr1a gene ablation by hydroxytamoxifen (OHT) treatment of pure DP cells in vitro and tested their inducing activity in the chamber-graft hair growth assay described above (Fig. 6A). In this case, to specifically track the fate of DP cells, we isolated male DP cells, combined them with female GFP-positive KC, and engrafted them onto female Nude mice.

Real-time PCR with primers specific for the ablated exon 2 confirmed that after 3 d of OHT treatment in vitro, wild-type Bmpr1a expression was abolished (Fig. 6B). Prior to grafting, DP cells were expanded for 7–12 d (one passage), and during this time, no obvious morphological and growth differences were observed (Fig. 6C). Additionally, as judged by annexin V staining and by propidium iodide uptake, the spontaneous apoptosis rates of Bmpr1a-null DP and control cells were comparable (Fig. 6D).

While Bmpr1a ablation did not alter DP growth or morphology, it had a marked effect on expression of DP signature genes sensitive to BMPs. Under early culture conditions, when expression of BMP-sensitive signature genes could still be detected, Akp2, Wif1, and other signature mRNA levels were diminished compared with control cultures (Fig. 6E). Interestingly, only Bmp6 was up-regulated, again highlighting the circuitry of the integrated negative feedback loop that we uncovered earlier. Overall, the effects appeared to be specific for the BMP signaling pathway, as genes that had scored previously as not responsive to BMP treatment similarly showed no effect on Bmpr1a ablation.

Loss of BMP receptor activity also compromised the ability of DP cells to induce HFs in full-thickness chamber-graft assays in vivo, where GFP-positive KC were combined with cultured DP cells as the only source of dermal cells. Thus, while mock-treated Bmpr1a^flfl/Rosa-Cre^ERtam DP control cells generated robust hair growth, OHT-treated Bmpr1a^flfl/Rosa-Cre^ERtam DP cells showed little or no ability to induce hair formation (Fig. 7A). OHT on its own had no effect on hair-forming activity, indicating that the effects were specific for loss of BMP signaling in the DP (data not shown).

The degree of GFP epifluorescence on the graft surface varied somewhat with different experiments, but it was generally comparable between control and knockout grafts (Fig. 7A, right panels; Supplemental Fig. S6A). In addition, OHT-treated, Bmpr1a-targeted DP cells showed some variability in the degree to which they impaired hair growth, ranging from complete absence of externally visible hair (Fig. 7A) to a few hairlike protrusions (Supplemental Fig. S6A). Overall, effects were substantial.

Histological analyses revealed abundant well-formed HFs in skins with grafted control DP cells (Fig. 7B, “Mock”), but large areas of either completely absent or aberrant downgrowths with DP knockout cells (Fig. 7B, “OHT,” arrowheads). For fair comparison, the amount of grafted area in these sections was judged by the integration of nuclear GFP-positive KCs, which was similar in control and knockout samples (Fig. 7C). Importantly, the paucity of HFs in these engraftments did not appear to be attributable to a failure of Bmpr1a-null DP cells to grow in vivo, since underlying dermis generated in the graft displayed Ychr-positive cells (Fig. 7D [“OHT,” arrowheads], E [panels ii,iii]; Supplemental Fig. S6B). However, control DP cells additionally gave rise to HFs, each encasing a pocket of Ychr-positive DP (Fig. 7D [arrows, “Mock”], E [panel i]) and surrounded by a Ychr-positive dermal sheath. In contrast, the few HF-like downgrowths generated from Bmpr1a-null DP cells lacked signs of an organized DP (Fig. 7D [“OHT”], E, [panel ii]). A summary of the outcome of DP-specific ablation of Bmpr1a in eight experiments is presented in Figure 7F. When we quantified the hair growth area within the GFP-positive graft, a robust reduction of 87% was observed compared with control (Fig. 7G).

Finally, we addressed whether the DP signature changes in vivo when BMP signaling is compromised. Unexpectedly, the Ychr-positive dermal cells in the Bmpr1a-targeted DP grafts displayed residual AP activity (Fig. 8A, “OHT,” arrowheads). The strongest AP activity was found at the base of aberrant HF-like downgrowths (Fig. 8C, arrow in panel ii) at levels similar to control DPs (Fig. 8A [“Mock,” arrows], C [panel i]). However, knockout cells that integrated into the dermis also exhibited aberrant, low-level AP activity (Fig. 8C, arrowheads in panels ii,iii). pSMAD1 staining confirmed the absence of BMP signaling within dermal cells. In addition, BMP signaling was compromised within the few HF-like epithelial downgrowths (Fig. 8D). Interestingly, in contrast to AP, WIF1 and NCAM were not detected in the dermal fibroblasts but were found to varying extents in dermal cells surrounding aberrant HF downgrowths (Fig. 8B,C,G; Supplemental Fig. S6C,F). Since DP signature proteins such as NCAM and WIF1 are not expressed by normal dermal sheath, we surmise that these positive cells reflected a failure of the Bmpr1a-null cells to organize into a proper DP structure and stimulate HF formation. Finally, HOXA9 and HHIP, which are exclusively expressed in normal DP (Fig. 8E,F, “Mock”), were not detected in dermal cells within knockout grafts (Fig. 8E,F, “OHT”; Supplemental Fig. S6D,G). Together, these data suggest that the ability of DP cells to respond to BMPs is required for them to orchestrate the appropriate microenvironment to establish proper DP signature and function.

DP cell fate specification: the impact of the microenvironment on DP character

The goal of our study was to exploit our recent genomic characterization of the molecular identity of the DP and its surrounding niche neighbors and systematically dissect the special features of hair bulb microenvironment that enable DP cells to exert their inductive powers. By analyzing how the molecular signature changes when DP cells are removed from their niche and placed in culture, we were able to identify not only a subset of genes whose expression is intrinsic to DP, but also another gene set whose expression is lost upon culture. The loss occurred concomitantly with a reduced ability of DP cells to promote HF morphogenesis in skin engraftments, suggesting that signature maintenance is functionally important.

In systematically screening niche factors for their ability to maintain the DP signature in vitro, we discovered that BMPs had strong positive effects that could be counteracted by BMP inhibitors. When put to a functional test, the improved molecular signature gained by BMP treatment coincided with improved hair growth in Nude mice in vivo. While this improvement did not prevent cultured DP cells from eventually losing the hair-inducing activity, the partial benefit of BMP signaling correlated with a partial restoration of the DP signature.

Our findings lend strong support to the view that a combinatorial effect will likely be necessary to fully unleash the inductive powers of cultured DPs, as initially proposed by Shimizu and Morgan (2004) in their studies on Wnt signaling. With our recent acquisition of a molecular signature for the DP and its niche, we unveiled an additional requirement for a concerted activation of multiple pathways in maintaining DP’s molecular features. Many of the prominent DP signature genes that were sensitive to BMP signaling were distinct from those sensitive to Wnt pathway activation by either WNT3A treatment or Wnt3a expression. In the future it will be important to systematically dissect which DP signature genes are regulated through DP receptor-mediated interactions with Wnts, Shh, Fgfs, and other extracellular factors expressed by cells within the DP niche.

Another critical issue is how BMP signaling is regulated in vivo. In this regard, the effects of BMPs and BMP inhibitors on the DP signature in vitro were intriguing, and given that a plethora of BMPs and inhibitors are simultaneously expressed within the microenvironment of the hair bulb (Supplemental Fig. S5), it is tempting to speculate that it must be the overall balance of positive and negative BMP signaling molecules that determines the outcome and orchestrates the DP identity within this niche. In the bulb, this balance appeared to be tipped toward active BMP signaling in vivo, since we detected pSMAD1-positive DP nuclei.

The presence of multiple positive and negative regulators also suggests built-in redundancy, which is exemplified not only by the similar effects of various hair bulb BMPs on the DP signature but also by mice targeted for BMP genes encoding BMP6 (Solloway et al. 1998), BMP7 (Dudley et al. 1995), and BMP8a (Zhao et al. 1998), all of which are expressed in the hair bulb but none of which exhibit an apparent hair phenotype when ablated. In light of our studies, we examined skin sections of mice deficient for BMP6 since it had the most potent effects. We confirmed that Bmp6-null mice displayed normal HFs and also showed that BMP-regulated DP signature genes are still expressed (Supplemental Fig. S7A,B).

When placed within the framework of previously reported roles for BMP signaling in restricting proliferation of epidermal cells (Blessing et al. 1996) and HF cells, including bulge stem cells and matrix cells (Kulessa et al. 2000; Kobielak et al. 2003, 2007; Andl et al. 2004; Blanpain et al. 2004; Ming Kwan et al. 2004; Yuhki et al. 2004; for review, see Botchkarev 2003), our findings provide an important new role for BMP signaling in maintaining features of DP cells that are instrumental for HF formation. The regulatory feedback circuits governing Bmp6 and Noggin gene expression in DP cells and the directionality of BMP signaling in the niche provide further insights into the complexities of this mesenchymal–epithelial cross-talk. While further studies will be needed to fully understand this process, the apparent gradient of pSMAD1 in the nuclei of DP and its neighbors may be a reflection of this circuitry in action.

In vivo relevance: a novel DP-specific gene targeting approach

With the ability to purify DP niche cells (Rendl et al. 2005), the BMP-mediated effects on DP signature genes in culture and enhanced hair formation upon grafting could be directly attributed to DP cells. Moreover, our Rosa26-Cre^ERtam/Bmpr1a^fl/fl-inducible gene knockout experiments illustrate how this purification scheme can be applied to mice of different genetic backgrounds to circumvent the current lack of DP-specific gene targeting tools and still conduct DP-specific loss-of-function studies. Although a genetic loss-of-function study such as the one we described here is superior, this strategy could be applied to pure populations of wild-type DP cells infected with lentiviral vectors expressing either shRNAs for knockdowns or transgenes for gain-of-function analyses. Such approaches could expedite the speed at which we and others will be able to test the physiological relevance of DP signature genes. Ultimately, it will be important to manipulate DP signature genes using DP-specific Cre lines in a normal developmental context.

Another advantage of our in vitro/in vivo hybrid knockout method is the ability to use female GFP-tagged KC and purified male DP cells to conduct grafting experiments on female Nude mice. This enabled us to track the fates of our engrafted Bmpr1a-null DP cells and establish an essential role for BMP signaling in DP-induced hair formation. Moreover, it suggested that in the absence of Bmpr1a, DP cells can still survive and proliferate in skin, but they cannot maintain their proper identity or their proper fate.

In the few cases in our knockout graft experiments where HF-like downgrowths did occur, they did not harbor a proper DP or a regularly layered architecture of epithelial cells, although at times they were accompanied by hair shaftlike protrusions. Since most graft regions were devoid of these rudimentary, aberrant formations, we surmise that they arose from the residual instructive ability of a few DP cells that may have been delayed in targeting both Bmpr1a alleles. Alternatively, these structures may reflect that even without BMP signaling, DP cells retain certain features of their signature, and this could be sufficient to induce aberrant mesenchymal–epithelial cross-talk.

Summary

In closing, we systematically screened growth factors from the DP niche microenvironment for the ability to regulate the molecular identity and hair-inducing qualities of DP cells. We identified BMP signaling as a determinant of stem cell-activating, hair-inducing cell fate. Finally, we showed that BMP signaling is not sufficient on its own to maintain the DP signature or DP’s HF-inductive properties. The novel assay we devised to conduct gain- and loss-of-function studies specifically in DP cells opens the door to future studies in dissecting the complex repertoire of epithelial–mesenchymal growth factor and receptor signaling that takes place in the DP niche.

Mice

DP cells were isolated from K14-H2B-GFP/Lef-RFP reporter mice by FACSorting as described previously (Rendl et al. 2005). For Bmpr1a ablation in DP cells, the reporter mice were crossed with Bmpr1a^flfl (Mishina et al. 2002) and Rosa-Cre^ERtam mice (Badea et al. 2003) (available at the Jackson Laboratory, stock no. 004847). For chamber grafts, immunocompromised Nude mice were obtained from Charles River Laboratories (Nu/nu) or Taconic (Swiss nude). Bmp6 knockout pups were kindly provided by the Robertson laboratory (Solloway et al. 1998).

Cell culture

FACS-isolated DP cells were cultured in AmnioMAX C-100 medium (Bratka-Robia et al. 2002). Growth factor, growth curve, and apoptosis experiments were performed at passage 1. For growth factor experiments, cells were cultured for 3–6 d and harvested for RT–PCR, AP activity, and immunofluorescence. A list of the growth factors and concentrations is provided in the Supplemental Material. For proliferation analysis 50 μM BrdU (Sigma) was administered. For induction of apoptosis, 100 nM TNFα (R&D Systems)/100 ng/mL cycloheximide (Sigma) was applied. Cre-mediated gene ablation was induced with 5 nM OHT (Sigma). Early passage 3T3-J2 dermal fibroblasts and C2C12 cells (a kind gift from the Greengard laboratory) were cultured in F medium (DMEM/10% FBS). For growth curve analyses, BMP6 and OHT-treated DP cells were plated at late passage 1 at low density (5000 cells per square centimeter), trypsinized, and counted in triplicate with a Coulter Counter (Beckman). For retroviral overexpression of a constitutively active form of Bmpr1a (ALK3-QD, kind gift from K. Miyazono) (Akiyama et al. 1997) or Bmp6 and Wnt3a (PCR-cloned from cDNA), sequences were subcloned into the MSCV-IRES-GFP retroviral vector (kindly provided by W. Pear, University of Pennsylvania, Philadelphia). Retroviral supernatants were produced in phoenix cells, and DP cells were infected following standard procedures.

Chamber-graft assay

Chamber grafts were performed as described (Lichti et al. 1993; Kishimoto et al. 2000; Blanpain et al. 2004). Experiments included a positive control of cell suspensions from freshly isolated wild-type dermis plus K14-GFPactin (Vaezi et al. 2002) or K14-H2B-GFP KC and a negative control of KC alone. Short-term cultured (4–7 d in KSFM; Invitrogen) newborn KC (2 × 10⁶) and BMP6 (400 ng/mL; R&D System) or control-treated DP cells (4 × 10⁶) in the first through fouth passages were used for maintenance grafts. For knockout experiments, DP cells were grafted after expansion in the first passage (∼9–12 d after sorting and plating ∼2 × 10⁶ cells). Hair typically appeared after 17–24 d.

See the Supplemental Material for additional experimental details.