Loss of a quiescent niche but not follicle stem cells in the absence of bone morphogenetic protein signaling

Abstract

During the hair cycle, follicle stem cells (SCs) residing in a specialized niche called the “bulge” undergo bouts of quiescence and activation to cyclically regenerate new hairs. Developmental studies have long implicated the canonical bone morphogenetic protein (BMP) pathway in hair follicle (HF) determination and differentiation, but how BMP signaling functions in the hair follicle SC niche remains unknown. Here, we use loss and gain of function studies to manipulate BMP signaling in the SC niche. We show that when the Bmpr1a gene is conditionally ablated, otherwise quiescent SCs are activated to proliferate, causing an expansion of the niche and loss of slow-cycling cells. Surprisingly, follicle SCs are not lost, however, but rather, they generate long-lived, tumor-like branches that express Sox4, Lhx2, and Sonic Hedgehog but fail to terminally differentiate to make hair. A key component of BMPR1A-deficient SCs is their elevated levels of both Lef1 and β-catenin, which form a bipartite transcription complex required for initiation of the hair cycle. Although β-catenin can be stabilized by Wnt signaling, we show that BMPR1A deficiency enhances β-catenin stabilization in the niche through a pathway involving PTEN inhibition and PI3K/AKT activation. Conversely, sustained BMP signaling in the SC niche blocks activation and promotes premature hair follicle differentiation. Together, these studies reveal the importance of balancing BMP signaling in the SC niche.

Adult stem cells (SCs) typically reside in specialized niches where they receive microenvironmental cues that govern normal homeostasis and wound repair (1). Pulse–chase experiments with labeled nucleotides (2) or fluorescent histone proteins (3) have designated the hair follicle (HF) bulge as the residence of skin's slow cycling, label-retaining epithelial cells (4). Clonal analyses reveal that bulge cells exhibit the properties of self-renewing SCs in vitro, and when transplanted onto the backs of hairless (Nude) mice, they generate epidermis, HFs, and sebaceous glands (5–7).

Initiated during embryonic development, HFs continue to generate hair until approximately postnatal day 14 (P14), when they enter a destructive phase, known as catagen (8, 9). During catagen, the lower two-thirds of the HF degenerates until the specialized dermal papilla (DP) cells at the base of the follicle come to rest below the bulge. In the resting (telogen) phase of the hair cycle, the bulge exists in a quiescent state in contact with the DP. The first telogen is brief, lasting only a few days before a new growth (anagen) phase begins, but the second resting phase lasts 3–4 weeks. With the end of telogen and the onset of the anagen, bulge SCs become activated to form the outer root sheath (ORS) and matrix cells that fuel HF regeneration [see supporting information (SI) Fig. 5]. Maintaining their association with the DP throughout anagen, matrix cells proliferate and differentiate upward to produce the hair shaft and its channel, the inner root sheath (IRS).

Although the mechanisms controlling the hair cycle are still unfolding, bone morphogenetic protein (BMP) signaling is likely to be involved. Interaction with exogenous BMPs stimulates the transmembrane receptor BMPR1A to phosphorylate Smad1, Smad5, and Smad8, which then heterodimerize with Smad4. These complexes translocate to the nucleus where they transactivate their target genes (10).

When keratin 14 (K14)-driven Cre recombinase is used to ablate Bmpr1a gene function in embryonic skin, the early stages of HF morphogenesis occur, but matrix cells fail to differentiate (11, 12). Although these later differentiation steps rely on active BMP signaling, earlier steps in the lineage appear to require the impairment of the pathway. Noggin, an extracellular BMP inhibitor, is expressed by mesenchyme, where it induces follicle morphogenesis in the embryo and promotes new HF growth (anagen) postnatally (13, 14). Interestingly, once embryonic HFs have initiated, they express BMP4, suggesting a negative feedback loop to prevent new HF initiation in the vicinity (9).

An early consequence of Noggin ablation in vivo is a loss of expression of Lef1, a DNA binding protein essential for efficient HF morphogenesis (14, 15). Conversely, keratinocytes treated with Noggin or deficient in BMP receptor 1A display nuclear Lef1 in vitro (11, 13), and Lef1-positive matrix cells persist after Bmpr1a is ablated in vivo (11, 12). Lef1 acts as a bipartite transcription factor in conjunction with stabilized β-catenin, which is known to stimulate de novo HF morphogenesis and increase follicle density when overexpressed (16).

In adult follicle SCs, Smad1 is phosphorylated and BMP6 levels are elevated, suggesting that BMP signaling is active in the bulge (5, 12) (see SI Fig. 6A, P19). Conversely, in the early hair germ that emerges from the activated bulge, nuclear P-Smad1 is diminished (see SI Fig. 6A, P20) and remains low in the developing ORS and in the lower part of the matrix. The strongest BMP signaling is in the cells that differentiate to produce the IRS and hair shaft (11) (see SI Fig. 6A, P24). These studies suggest a possible explanation for how bulge SCs might exist in a quiescent state and how this characteristic might change during subsequent lineage determination and differentiation.

The possible role for BMPs in follicle SCs takes on greater interest given recent reports that conditional inactivation of Bmpr1a leads to an increase in the number of stem and/or progenitor cells in the hematopoietic lineage (17) and intestinal epithelium (18). That said, BMPs don't always function to inhibit SC self-renewal. BMPs can act to sustain self-renewal of both murine and human embryonic SCs (19, 20), and in the fly embryo, dpp (vertebrate BMP2/4 homolog) defines the niche that maintains anterior germline SCs (21).

Here, we use an inducible conditional targeting strategy to uncover roles for BMP signaling in balancing quiescence and activation of hair follicle SCs in adult mice. We find that inhibition of BMP signaling in follicle SCs is essential for promoting the transition between quiescent bulge SCs to proliferating progeny. We further show that in the absence of BMPR1A function, premature anagen occurs, and although niche characteristics are perturbed, the SCs are not lost. Rather, both Lef1 and β-catenin are up-regulated in the SC niche, associated with signs of inactivation of PTEN and activation of the PI3K/AKT pathway in the Bmpr1a tamoxifen inducible conditional knockout (cKO^TM) niche. These findings provide a mechanistic link for the convergence of BMP and Wnt pathways in follicle SC activation.

Results

BMP Signaling Is Necessary for Maintaining Quiescence of the SC Niche.

Conditional knockout mice targeted by K14-Cre for Bmpr1a gene ablation die shortly after birth (11, 12). To evaluate how ablation of Bmpr1a affects postnatal homeostasis of the HF SC niche, we mated Bmpr1a fl/fl floxed mice (22) and inducible K14-Cre^TM mice (23). Topical application of tamoxifen (TM) was then used to control the inactivation of BMP signaling.

Offspring from the matings of Bmpr1a(fl/+)/K14-Cre^TM and Bmpr1a(fl/+) mice yielded litters of the expected numbers, genotype, and Mendelian ratios. Before TM treatment, Bmpr1a(fl/fl)/K14-Cre^TM were indistinguishable from control (CON) animals, which included wild type (WT), Bmpr1a(fl/+); Bmpr1a(fl/fl); Bmpr1a(fl/+)/K14-Cre^TM, and Bmpr1a(+/+)/K14-Cre^TM. At P44, HFs were just entering the extended second telogen, and both Bmpr1a(fl/fl)/K14-Cre^TM and Bmpr1a(fl/fl) HFs exhibited a resting-phase morphology (Fig. 1A and SI Fig. 6B). We shaved and applied TM topically at this time, and within 16d the Bmpr1a gene was efficiently targeted as judged by genotyping (SI Fig. 6C), anti-P-Smad-1 immunofluorescence (SI Fig. 6D), and in situ hybridization with a cRNA probe to the deleted exon 2 (SI Fig. 6E).

Fig. 1.

Postnatal ablation of Bmpr1a affects homeostasis of the SC niche in adult hair follicles. (A) Chart illustrating the temporal progression through the first and second postnatal hair cycles and the regiment for administering TM and BrdU. (B) Phenotypes of CON^TM and Bmpr1a cKO^TM mice, after shaving back skins during telogen, targeting Bmpr1a and monitoring after initiation of the next hair cycle. cKO^TM mice did not regrow hair. (C and D) Skin sections stained with H&E or with antibodies that are color-coded according to the fluorescent tags of the secondary antibodies. Brackets denote bulge (Bu), and arrowheads denote follicle cells outside the niche. Note that the CD34 monoclonal antibody gives high dermal background in older animals; the hair shafts typically display autofluorescence. FACS analyses are of SC populations sorted by α6-integrin and high CD34. R3, basal SCs; R2, i.e., the large box including R3, total bulge SCs. Note that both CD34-positive bulge pools are diminished 2 weeks after Bmpr1a targeting in cKO^TM skin. (E and F) Skin sections as in C and D, but in this case, depicting analysis of bulge SCs after pulse–chase and pulse experiments with BrdU (see A for time schedule). FACS analyses are based on α6-integrin and incorporation of BrdU during a 4-h pulse. Cells that incorporated BrdU above limits of detection by immunofluorescence microscopy are boxed as R5. Note that the pool of BrdU-incorporating cells is substantially increased in cKO^TM skin.

Bmpr1a(fl/fl) mice carrying the TM-activated K14-Cre^TM transgene (cKO^TM) failed to regrow hair. Despite the lack of visible hair, their HFs were in anagen by P59 (Fig. 1C, arrowheads). By contrast, CON follicles after tamoxifen treatment (CON^TM) remained in telogen for >2 additional weeks (P75), after which they entered anagen and regrew their hair (Fig. 1B).

Before TM, both cKO and WT bulge cells consisted of an α6 integrin-rich basal layer and an α6-low suprabasal layer, both of which were CD34-positive (5) (SI Fig. 6F). Following TM, however, cKO^TM follicles displayed ≈9–10× fewer CD34(+) bulge cells by fluorescence-activated cell sorting (FACS) than CON^TM follicles (Fig. 1D). In addition, when mice were pulsed with BrdU 2×/d for 3d in the first postnatal anagen (P23–25), treated with TM from d44 to d55 and examined at 5 weeks after administering BrdU (see Fig. 1A), label-retaining cells were prevalent in CON^TM but not in cKO^TM bulges (2, 4) (Fig. 1E).

The absence of CD34-positive, bulge label-retaining cells upon the inactivation of BMP signaling could be reflective of an activated follicle SC niche. To test this possibility, we administered 4-h pulses of BrdU and processed the skins immediately thereafter (see Fig. 1A). At P57 and P59, CON^TM follicles were in telogen and only an occasional bulge cell incorporated BrdU during each 4-h pulse. By contrast, cKO^TM follicles displayed numerous BrdU-labeled bulge cells (Fig. 1F, P57, and SI Fig. 6G, P59, arrowheads). Quantification by FACS revealed >4× more BrdU-labeled cells within the cKO^TM versus CON^TM skin at this stage (Fig. 1F). At P77, CON^TM HFs had just passed the telogen-to-anagen transition, and labeled cells were now detected in the developing hair germs (SI Fig. 6H). By contrast, P77 cKO^TM HFs had progressed well beyond anagen initiation. Developing HFs appeared morphologically enlarged, misshapen, and hyperproliferative (SI Fig. 6H, arrowheads).

Expansion of Markers of Bulge and Proliferating Progeny Following Ablation of Bmpr1a.

To determine the extent to which SC properties are lost upon Bmpr1a targeting, we next examined the expression of DNA binding proteins Sox9 and Lhx2, which are essential for bulge SC maintenance (24, 25). Antibodies against each marker labeled the cKO^TM bulge but at reduced intensity over CON^TM (SI Fig. 7 A and B). For both cKO^TM at P59 and CON^TM at P77 follicles, Sox9 and Lhx2 levels were high in the P-cadherin(+) hair germs emerging after anagen induction (Fig. 2 A and B and SI Fig. 7 A and B). By P77, labeling Sox9 and Lhx2 in cKO^TM follicles was concentrated in the midzone below the cKO^TM bulge, while Sox9 was absent in the P-cadherin(+) cells encased by the DP (Fig 2 A and B). In their patterns of BrdU-incorporation and biochemical markers (24–26), these two cell populations bore a resemblance to the proliferative cells of the lower ORS and matrix, respectively.

Fig. 2.

Expansion of the proliferating populations and their ability to repair epidermal wounds when BMP signaling is ablated in hair follicles postnatally. cKO^TM and CON^TM mice were targeted for Bmpr1a ablation at P44 and analyzed at P77 (A–D) or P90-P230 (E and F). Frozen skin sections (10 μm) were subjected to indirect immunofluorescence for Sox9, P-cadherin, and Lhx2 (A, B, and E), or to in situ hybridizations (ISH) with digoxygenin cRNA probes specific for Sox4 and Shh (C, D, and F). Note that Sox9^hiLhx2^hi- and P-cadherin^hiLhx2⁽⁺⁾-positive cells mark two distinct populations emerging from the bulge lacking BMPR1A. These populations developed into tumor masses over time. (G) A 4-h pulse at P120 shows that the tumor borders are highly proliferative. (H) Ablation was performed at P44, and a 3-day pulse was administered at P67. Skin was analyzed after a 3-week chase. Note the absence of label-retaining cells in tumor masses. (I) Wound healing. cKO^TM mice were mated to Rosa26-floxed-lacZ mice; after Bmpr1a ablation (LacZ+), skins were scraped to remove the epidermis, and wound repair was monitored for the periods indicated. Note the reepithelialization (arrows) by BMPR1A-deficient follicle cells. Arrowhead denotes border between scab/blood clot and the epidermis, dotted lines denote dermo–epithelial borders, and brackets denote bulge (Bu). hg, hair germ; SG, sebaceous gland.

In normal HFs, the Sox4 and Sonic hedgehog (Shh) genes are not expressed in the bulge but are coexpressed in the developing hair germ (26, 27). While Sox4 and Shh in situ hybridizations of P59 and P77 CON^TM follicles gave the expected patterns, hybridizations were markedly expanded in the cKO^TM counterparts. Sox4 was atypically detected in both cKO^TM bulge and progeny (Fig. 2C and SI Fig. 7C). Shh positive cells were prevalent throughout the lower cKO^TM follicle (Fig. 2D and SI Fig. 7D).

Several months after TM treatment, each cKO^TM HF appeared as a branch of amorphous tumor-like masses of cells consisting largely of the same two populations of cells seen at P77 (Fig. 2 E and F). Multiple branches of these masses formed over time. The P-cadherin and Shh-positive edges of the tumor masses remained proliferative many months after Bmpr1a targeting, and pulse–chase experiments showed no traces of BrdU label retention (Fig. 2 G and H). Additionally, cells from these structures were able to replenish the epidermis when scraped or enzymatically removed (Fig. 2I and SI Fig. 8 B and A, respectively).

Lef-1/β-Catenin Stabilization and PI3 Kinase/AKT Pathway Activation Upon Bmpr1a Targeting.

β-Catenin stabilization promotes the transition from quiescent bulge cells to proliferating Sox4-positive progeny (26). As expected, nuclear Lef1 and β-catenin were detected in the hair germs of P77 CON^TM HFs but not in the bulge (SI Fig. 9 A and B). By contrast, nuclear Lef1 and β-catenin were expressed not only in precocious hair germs of P59 cKO^TM HFs but also in the bulge (Fig. 3 A and B, respectively). By P77, most cells within down-growing masses were positive for these markers (SI Fig. 9 A and B).

Fig. 3.

Precocious and sustained activation of Lef1/β-catenin and P-PTEN/PI3K/AKT signaling in hair follicles lacking BMPR1A. (A–E) cKO^TM and CON^TM mice were targeted for Bmpr1a ablation at P44 and analyzed at P59. Frozen skin sections (10 μm) were subjected to indirect immunofluorescence or immunohistochemistry by using the antibodies indicated in the lower right of each frame. Color coding is according to the fluorescent tags of the secondary antibodies. Arrowheads denote aberrant expression. (F) Immunoblot analyses of protein extracts isolated from P59 skin preparations as described (26). Antibodies used are indicated at the right. (G and H) TOPFlash assays. MKs were transfected with vector encoding the reporter gene TOPFlash or the control FOPFlash containing mutations in the multimerized Lef1 DNA binding sites. MKs were then treated with the recombinant factors or with the PI3K inhibitor Ly as indicated, and luciferase assays were conducted as described (13). Experiments were conducted in triplicate. Lam5, laminin5; β-cat, β-catenin; β4, β4-integrin; hg, hair germs; DP, dermal papilla; Nogg, Noggin; Wnt3, Wnt3a.

Most pathways that stabilize β-catenin inhibit GSK3β kinase, which phosphorylates and targets β-catenin for turnover (28). In CON^TM HFs, GSK3β was not phosphorylated (active) at Ser-9 during telogen (Fig. 3C) but becomes phosphorylated as anagen ensues (SI Fig. 9C). In contrast, P59 and P77 cKO^TM follicles showed expanded anti-P-GSK3β labeling, corresponding to the anti-β-catenin pattern (Fig. 3C Right and SI Fig. 9C Right). These findings were consistent with a role for GSK3β inactivation in generating the β-catenin-rich tumor masses when Bmpr1a was targeted.

GSK3β-Ser-9 is an established target for AKT phosphorylation, and given the link between BMP signaling and PTEN activation, we tested whether PTEN might be altered in BMPR1A-deficient follicle SCs. From telogen to anagen in CON^TM (P77, 2nd cycle) or WT (P21 and P23, 1st cycle) HFs, antibodies against phosphorylated (inactive) PTEN labeled weakly in developing hair germs (SI Fig. 9D). No signal was detected in telogen CON^TM (P59) or WT (P19) HFs (Fig. 3D Left and SI Fig. 9D Left). By contrast, anti-P-PTEN staining was prevalent throughout the bulge and emerging progeny of cKO^TM HFs (Fig. 3D Right and SI Fig. 9D Right).

AKT's targets include Ser-552 of β-catenin, a mark which facilitates 14–3-3ζ association and nuclear translocation (29). In cKO^TM follicles, anti-14–3-3ζ staining was intense in developing tumor-like masses. Consistent with a role for AKT/14–3-3ζ in SC activation, staining was strong in the hair germs of cKO and CON HFs (Fig. 3E and SI Fig. 9E). Immunoblot analyses further suggested that without BMPR1A, PTEN remained inactive and PI3K/AKT/GSK3β pathway was active (Fig. 3F).

Finally, we subjected cultured WT and KO keratinocytes (MK) to assays with the TOPFlash reporter gene, active when β-catenin and Lef1 are stabilized (30). WT MK required Wnt3a and Noggin for optimal activity, and although Wnt3a was permissive for TOPFlash, expression was inhibited by BMP6 (Fig. 3G). In all cases, expression depended on the Lef1 enhancer element, as judged by assays with the mutated FOPFlash.

In low serum-containing medium, TOPFlash was higher in KO than WT MK. Noggin elevated WT but not KO levels, suggesting that low levels of endogenous BMPs accounted, in part, for these differences (13) (Fig. 3H). Differences in PI3K signaling were also involved, given that Ly294002, a chemical inhibitor of PI3K, markedly reduced TOPFlash activity and minimized the differences between WT and KO MK. Noggin did not override the inhibitory effects of PI3K, consistent with the notion that inhibition of BMP signaling operates through PI3K activation. However, Wnt3A enhanced TOPFlash activity in both WT and KO MK, and this observation was not markedly repressed by PI3K inhibition (Fig. 3H). In summary, TOPFlash activity was optimal when WNT signaling was activated and BMPR1A activity was impaired. WNT signaling and PI3K signaling were both able to stabilize β-catenin, but they appeared to do so by parallel, and at least partially independent, pathways.

Inducible Activation of BMPR1A Results in Premature Differentiation of Follicle SC Progeny.

To further explore the role of BMP signaling in follicle SC behavior, we engineered mice harboring TRE-BmpR1A-CA-HA (BmpR1A-CA, constitutively active BMP receptor 1A), a tetracycline (Dox) regulatable transgene encoding a constitutively active and epitope-tagged form of BMPR1A (31). Mice were mated to be doubly transgenic (dTg) for TRE-BmpR1A-CA-HA and K14-rtTA, encoding a Dox-inducible transactivator (32) (Fig. 4A and SI Fig. 10A). Before Dox administration, single transgenic CON and dTg mice were indistinguishable.

Fig. 4.

Mice expressing a dominant active BMPR1A transgene display precocious hair follicle differentiation. Bitransgenic mice harboring a Dox-regulatable, superactive BMPR1A (dTg), and noninducible CON mice were treated with Dox at P19, after shaving the lower back skin areas (brackets). Analyses of mice were performed at P35, following a 4-h pulse with BrdU. (A) Transgene schematics. (B) Phenotypic abnormalities in the regrowth of the hair coat of dTg^Dox-treated mice. (C–I) P35 skin sections subjected to either H&E staining or immunofluorescence with the antibodies indicated. Color-coding is according to the fluorescent tag of the secondary antibodies. AE15 recognizes trichohyalin, an IRS marker. Note that following BMPR1A superactivation, dTg^Dox hair follicles develop cysts of terminally differentiated cells located just below the bulge (Bu, bracket). Dotted lines demarcate dermo–epidermal boundaries. Mx, matrix cells. Boxed areas are magnified in Insets.

At P19, during the short telogen of the first hair cycle, posterior back skins were shaved and Dox was administered. CON mice, after doxycycline treatment (CON^Dox), developed a new hair coat, reflecting the normal transition into the second hair cycle. By contrast, doxycycline inducible double transgenic for K14-rtTA and TRE-BmpR1A-CA-HA (dTg^Dox) mice displayed defects in hair coat regeneration (Fig. 4B). Immunofluorescence confirmed efficient transgene induction (Fig. 4C), and reflective of BMPR1A-CA activity, anti-phospho-Smad1 staining was prevalent, with the exception of what appeared to be keratinized follicular masses (SI Fig. 10B).

BmpR1A-CA expression severely perturbed the hair cycle. By P35, overall bulge morphology still appeared to be intact in dTg follicles, but a large keratinized cyst had developed beneath the bulge, where a mature HF was seen in WT counterparts (Fig. 4 D and D′). BrdU labeling confirmed the paucity of proliferating keratinocytes within the P35 dTg cyst-like structures, but P35 CON^Dox HFs displayed BrdU-labeled cells throughout the ORS and matrix (Fig. 4 E and E′).

As expected from prior studies, GATA-3 (33) and Lef1 (34) differentially marked the IRS and hair shaft precursor cells of CON^Dox HFs (Fig. 4F). In dTg^Dox skin, however, GATA-3 and Lef1 were precociously expressed in the follicle cysts that emanated from the bulge (Fig. 4F). In some cases, the same cells showed coexpression (Fig. 4F Right Inset). Cyst cells were also immunolabeled with monoclonal antibodies specific for the IRS protein trichohyalin (AE15) and hair keratins (AE13) (Fig. 4G and SI 10D). dTg^Dox cysts did not label with antibodies against epidermal differentiation proteins, including keratin 10 (K10) and filaggrin (Fig. 4H and SI Fig. 10E). Thus, although dTg^Dox cysts did not exhibit the organized architecture typical of WT HFs, they displayed features of follicle, rather than epidermal, differentiation.

While CD34 marked the bulge SC niche in P30 dTg^Dox HFs, its expression was reduced compared with P30 CON^Dox HFs, and by P35, it was barely detectable (Fig. 4 I and I′; see also SI Fig. 10C). Similar results were obtained with Sox9 antibodies (SI Fig.10 F and F′). No traces of proliferative bulge progenitors, marked by Sox 4 and Shh, were seen (SI Fig. 10 G, G′, H, and H′, respectively). These collective results were consistent with the view that follicle SCs may still be triggered to execute an HF fate, but when they do, they terminally differentiate in the presence of constitutive BMPR1A activity.

Discussion

Fine-Tuning BMP Signaling to Balance Quiescence, Proliferation, and Differentiation of Follicle SCs.

Follicle SCs display signs of activated TGFβ and BMP signaling in vivo, and in vitro, ligands specific for each pathway cause proliferating keratinocytes to transiently withdraw from the cell cycle (3, 5, 6, 10). Thus, it was surprising that ablation of BMP signaling alone was sufficient to disrupt the quiescent state of the hair follicle SC niche. In the absence of BMPR1A, quiescent SCs were precociously activated to enter the proliferative phases associated with the new hair cycle, underscoring a role for BMP inhibition in governing the telogen to anagen transition.

Our loss-of-function studies also highlighted an essential role for BMP signaling in maintaining the biochemical and behavioral features of the SC niche. Our gain-of-function studies demonstrate that overactivation of BMPR1A signaling causes precocious terminal differentiation, revealing the importance of fine-tuning the levels of BMP signaling in the hair follicle lineage. In this regard, it is interesting that the intensity of nuclear anti-P-SMAD-1 within WT SCs is lower than that detected in the precursor cells of the hair shaft and IRS. Thus, the level of canonical BMP signaling in the SC niche may need to be sufficiently high to maintain quiescence, and yet not so high that terminal differentiation is triggered. We posit that the hair cycle may be activated when signaling to the SC niche dips below this threshold level, and when signaling rises above the threshold, proliferating progeny execute their terminal differentiation programs.

BMP6 and BMP4 seem to play autocrine and paracrine roles, respectively, in niche quiescence (5, 35). Conversely, BMP antagonists, including Noggin, Gremlin, and ectodysplasin, are made by the DP and, hence, are likely to play paracrine roles in SC activation (14, 36). LacZ knockin mice suggest that for at least one BMP inhibitor, namely Noggin, expression varies with the hair cycle, such that Noggin activity in the SC niche rises at the onset of anagen and wanes during telogen (35). Future studies will address how these competing patterns of BMPs and BMP inhibitors are temporally regulated to govern the hair cycle.

BMP Inhibition and SC Activation.

Several main features of follicle SCs, namely CD34 and label retention, were lost upon Bmpr1a targeting. Nevertheless, Sox9, Lhx2, Sox4, and Shh, which are characteristic of quiescent bulge cells and/or their early proliferating progeny, were expanded after inhibiting BMP signaling.

Given that this expansion of early progenitors persists for more than a year after Bmpr1a ablation, it suggests that the activated SCs of BMPR1A-deficient follicles are not lost. The ability of follicles to repair epidermal wounds further suggests that multipotent SCs survive Bmpr1a ablation. Our findings argue that BMPR1A inhibition is essential for the maintenance of proliferating bulge progeny but that CD34 and label retention are not essential features of follicle SCs.

Skin tumors arise when Bmpr1a is targeted by either K14-Cre^TM (12), Engrailed1-Cre (37), or Mx1-Cre (35). We observed that the numbers of branches of these proliferative tumor-like masses increased with age, and yet they did not progress to malignancy. This is consistent with what might be expected if the destructive phase (catagen) of the hair cycle does not occur properly in the absence of BMPR1A (12), but the follicles continue to undergo cycling.

We also noted that during the window of time when it was still recognizable, the Bmpr1a-null bulge expanded in size and was no longer label-retaining. We posit that in the absence of BMPR1A, slow cycling SCs convert to activated SCs, which retain their ability to self-renew and to commit to the early proliferative phases of the HF lineage. As shown for embryonic skin, HF differentiation beyond the Lef1-expressing matrix cell stage was blocked in the absence of BMPR1A (11, 12).

Mechanisms Involved in BMPR1A Signaling in the SC Niche.

The importance of β-catenin stabilization and activation of Lef/Tcf target genes in follicle formation is well established (15, 16, 38). This process involves signaling by ectodermal Wnts, which when blocked by overexpression of Dkk1, suppress follicle formation (39). The process also involves inhibition of BMP signaling, given that Lef1 and follicle formation are repressed in the absence of Noggin. Additionally, Lef1/β-catenin-mediated transcription is lost in Noggin-null mutants, and is elevated in the presence of excessive Noggin or in the absence of BMPR1A (11, 13, 14).

Although Wnt signaling and BMP inhibition synergize during HF morphogenesis in the embryo, whether and how this synergy operates in controlling adult follicle SC behavior is less clear. Bulge cells display frizzled receptors, and their precocious activation in response to excess stabilized β-catenin underscores its importance in the telogen to anagen transition (26) and references therein. Thus, several lines of evidence suggest that Wnt signaling is not the sole regulator of β-catenin stabilization in adult HFs.

Wnt signaling is thought to disrupt the Axin-GSK3-β-catenin complex, thereby inhibiting the phosphorylation and destruction of β-catenin by GSK3β kinase (40). However, conditional gene targeting also suggests a role for the small GTPase Cdc42 and the PKCζ/Par3/Par6 polarity complex in regulating GSK3β activity in HFs (41). Our findings now link the inhibition of BMP signaling to β-catenin stabilization through the inactivation of PTEN, leading to the activation of the PI3 kinase and Akt. Our findings are consistent with a role for Akt both in inactivating GSK3β through phosphorylation at Ser-9 and in directly phosphorylating β-catenin at Ser 552, leading to its enhanced association with 14-3-3ζ and preferential nuclear location.

The first link between PTEN and BMPR1A emerged from genetic studies revealing mutations in these genes as causes of patients with Cowden syndrome, a hyperproliferative disorder of the colon. Subsequent studies revealed that BMP exposure resulted in increased levels of the PTEN tumor suppressor (42). PTEN, which acts by inhibiting PI3K-Akt, thereby antagonizes Wnt's effects on β-catenin stabilization (18). This PI3K pathway appears to be activated in several tissues where Bmpr1a has been inactivated, including the intestine (18) and Mx1-positive skin tumors (35). Our findings now extend this pathway to the follicle SC niche and provide a molecular explanation as to why PTEN inhibition and constitutive activation of AKT in skin lead to the induction of new hair in resting-phase follicles (43). Precisely how BMP signaling, Wnt signaling, and polarization/cytoskeletal cues are regulated to tip the intricate balance of β-catenin/Lef/Tcf transcriptional activity in SC activation is an interesting question which merits further studies.

Materials and Methods

Mice and BrdU labeling.

Bmpr1a-floxed mice (22) were mated to K14-CreER mice (23), and targeting was achieved by 2.5 mg/mouse TM (Sigma–Aldrich, St. Louis, MO) applied daily for 11 d to back skins of shaved mice at P44, corresponding to the start of the second postnatal telogen. For pulse–chase experiments, BrdU (Sigma–Aldrich) was administered by either peritoneal injection of P23 mice (50 μg/g/d) or by adding 0.8 mg/ml BrdU to the drinking water from P23 to P25 (5). Mice were then chased from day 23 to day 59. For pulse experiments, a 4-h BrdU pulse (50 μg/g) was administered peritoneally before analyses.

For gain-of-function studies, Fvb/n transgenic mice were made to express BMPR1A-CA under the control of a Dox regulatory element (TRE) and were mated to mice harboring the Dox-sensitive transactivator, rTA2SM2-VP16 (rtTA), under the control of the K14 promoter (32). Mice doubly transgenic for TRE-BmpR1A-CA-HA and K14-rTA2SM2-VP16 (rtTA) were shaved and induced by oral administration of Dox at P19 (telogen) of the first postnatal hair cycle.

Abbreviations

HF

hair follicle

SC

stem cell

DP

dermal papilla

ORS

outer root sheath

IRS

inner root sheath

K14

keratin 14

Pn

postnatal day n

TM

tamoxifen

CON

control

MKs

cultured primary keratinocytes

BMP

bone morphogenetic protein

cKO^TM

tamoxifen inducible conditional knockout

CON^TM

control after tamoxifen treatment

Bmpr1A-CA

constitutively active BMP receptor 1A

Dox

doxycycline

dTg^Dox

doxycycline inducible double transgenic for K14-rtTA and TRE-BmpR1A-CA-HA

CON^Dox

control after doxycycline treatment.