Smad1 and 5 but Not Smad8 Establish Stem Cell Quiescence Which Is Critical to Transform the Premature Hair Follicle During Morphogenesis Toward the Postnatal State
Eli and Edythe Broad CIRM Center for Regenerative Medicine and Stem Cell Research and Department of Biochemistry and Molecular Biology, USC Norris Cancer Center, University of Southern California, Los Angeles, California, USA
Department of Pathology, Department of Biochemistry and Molecular Biology, USC Norris Cancer Center, University of Southern California, Los Angeles, California, USA
Eli and Edythe Broad CIRM Center for Regenerative Medicine and Stem Cell Research and Department of Biochemistry and Molecular Biology, USC Norris Cancer Center, University of Southern California, Los Angeles, California, USA
Department of Pathology, Department of Biochemistry and Molecular Biology, USC Norris Cancer Center, University of Southern California, Los Angeles, California, USA
Correspondence: Krzysztof Kobielak, M.D, Ph.D., The Broad CIRM Center for Regenerative Medicine and Stem Cell Research, Department of Pathology, University of Southern California, 1425 San Pablo Street, BCC-513, Los Angeles, California 90033, USA. Telephone: +1-323-442-3208; Fax: +1-323-442-7832; e-mail: firstname.lastname@example.org
Hair follicles (HFs) are regenerative miniorgans that offer a highly informative model system to study the regulatory mechanisms of hair follicle stem cells (hfSCs) homeostasis and differentiation. Bone morphogenetic protein (BMP) signaling is key in both of these processes, governing hfSCs quiescence in the bulge and differentiation of matrix progenitors. However, whether canonical or noncanonical pathways of BMP signaling are responsible for these processes remains unresolved. Here, we conditionally ablated two canonical effectors of BMP signaling, Smad1 and Smad5 during hair morphogenesis and postnatal cycling in mouse skin. Deletion of Smad1 and Smad5 (dKO) in the epidermis during morphogenesis resulted in neonatal lethality with lack of visible whiskers. Interestingly, distinct patterns of phospho-Smads (pSmads) activation were detected with pSmad8 restricted to epidermis and pSmad1 and pSmad5 exclusively activated in HFs. Engraftment of dKO skin revealed retarded hair morphogenesis and failure to differentiate into visible hair. The formation of the prebulge and bulge reservoir for quiescent hfSCs was precluded in dKO HFs which remained in prolonged anagen. Surprisingly, in postnatal telogen HFs, pSmad8 expression was no longer limited to epidermis and was also present in dKO bulge hfSCs and matrix progenitors. Although pSmad8 activity alone could not prevent dKO hfSCs precocious anagen activation, it sustained efficient postnatal differentiation and regeneration of visible hairs. Together, our data suggest a pivotal role for canonical BMP signaling demonstrating distinguished nonoverlapping function of pSmad8 with pSmad1 and pSmad5 in hfSCs regulation and hair morphogenesis but a redundant role in adult hair progenitors differentiation. Stem Cells2014;32:534–547
The hair follicle (HF) is a unique regenerative miniorgan that provides a highly instructive model system to study the underlying mechanisms that govern hair and skin appendage development as well as hair follicle stem cell (hfSCs) homeostasis and progeny differentiation. During early morphogenesis, HFs first develop into a placode and then as a small germ down growth following underlying mesodermal cues to the overlying ectoderm [1, 2]. The epidermal placode then reciprocates signals to the mesoderm instructing the condensation of cells into a dermal papilla (DP). At the hair peg stage, the DP transmits signals to the surrounding follicular epithelium which produces proliferating bulb progenitors, the precursors of matrix . The matrix cells adjacent to the DP then begin to differentiate into the hair shaft (HS) at the center and into surrounding inner root sheath (IRS) layers. Surrounding the IRS is the outer root sheath (ORS) which is a continuation of the basal layer of interfollicular epidermis (IFE), and both layers are attached to the basement membrane, the interface separating the epidermis and dermis. Basal keratinocytes of the IFE proliferate and their progenies detach from the basement membrane, progressively differentiating while moving upward to generate the overlying layers of stratified epidermis .
Mature HFs are fully established during the growth phase (anagen), the lower epithelial portion of HFs then undergo cycles of degeneration (catagen) and rest (telogen) with subsequent regrowth at the next anagen . The upper permanent portion of each HF forms at the base a bulge, a reservoir of hfSCs containing label-retaining, slow cycling cells (LRCs) marked by CD34 expression [6-10]. However, before the morphological bulge is fully established postnatally at the end of the first hair cycle (by postnatal day 18, P18), a prebulge region that contains a LRCs population expressing hfSCs markers: Lhx2, Sox9, Tcf3, and Nfatc1, but not yet CD34, is already specified during early HF morphogenesis [11-15]. So far, several important inductive signaling pathways between epithelial and dermal components during early HFs morphogenesis and, then later during HFs homeostasis and differentiation, have been characterized . Recent studies identified that bone morphogenetic protein (BMP) signaling is used reiteratively throughout the body and is involved in processes including general embryonic patterning and development of limbs, palate, tongue, and also skin appendages such as nails and sweat glands . Canonical BMP signaling is initiated following the interaction of BMP ligands to the transmembrane BMP receptor complex formed by type I and type II BMP receptors (BMPRs) . Upon ligand binding, BmprII-mediated phosphorylation of the BMPPRI instigates the recruitment and phosphorylation of intracellular receptor-activated Smad1, Smad5, or Smad8 proteins which then complex with Smad4. The Smad co-complexes then enter the nucleus and transcriptionally regulate canonical BMP-responsive target genes expression by binding to sequence-specific motifs within the promoters of target genes termed “smad binding elements” . The BMP ligands, their antagonists (such as noggin and follistatin), BMPRs, and canonical BMP-Smad proteins have complex spatial expression patterns within developing epidermis, HFs as well as in the underlying dermis . The pattern of BMP ligand expression in the developing epidermis is very complex as different BMP ligands are expressed in proliferating basal and differentiating suprabasal layers, suggesting a role for BMP signaling in both of these processes in vivo [18-20]. Moreover, during HFs morphogenesis, there is abundant expression of different BMP ligands in the epithelial progenitor cells, the differentiating HF layers, and also by the mesenchymal cells and DP [18, 20-24].
Studies using in vivo genetic modulation of BMP signaling have offered significant insight into the diverse functions of BMP signaling during skin and HF morphogenesis, postnatal homeostasis and differentiation. BMP overexpression studies (BMP4 and BMP6) or BMP antagonist (Noggin) ablation have highlighted the importance of BMP inhibition in normal HFs' initiation during early development, induction of the hair cycle , and basal layer hyperproliferation [26, 27]. The role of BMP signaling and its importance during skin and hair development as well as postnatally was further defined by ablation of BMP receptor 1A (BMPR1A) in vivo [28-31]. Although BMPR1A is coexpressed with another BMP receptor 1B (BMPR1B) in epidermis, it is exclusively expressed in developing and postnatal hair [26, 30-32]. Thus, these loss-of-function experiments emphasized the key role of BMPR1A in BMP signal transduction for proper hair matrix progenitor differentiation into the IRS and HS [30, 31, 33] and highlighted its essential role in maintaining quiescent hfSCs . These findings were in agreement with BMP signaling inhibition experiments using noggin activation in the skin, which reported an overall increase in HFs density and aberrant epidermal differentiation [32, 34, 35].
Although BMP signaling has been implicated as a crucial player in these biological processes, precisely whether BMP signaling is specifically transduced from the receptor in a canonical or noncanonical manner remains unresolved. Here, we address this question by using conditional gene targeting to specifically ablate the downstream effectors of canonical BMP signaling, Smad1 and Smad5 during skin and hair morphogenesis and postnatally in adult HF cycling. Surprisingly, during development, by using our dKO HFs model we revealed distinct activation patterns of pSmads, with pSmad1 and pSmad5 exclusively activated in HFs and pSmad8 restricted to IFE but absent in HFs, which resulted in precluded prebulge and bulge formation and hair differentiation during morphogenesis. In contrast, adult postnatal ablation of both Smad1 and Smad5 revealed that pSmad8 was no longer limited to epidermis but its activation was inclusive to dKO hfSCs and matrix progenitors, which in control (Con) HFs also physiologically expressed pSmad1 and pSmad5. However, this postnatal rearrangement of pSmad8 activation did not prevent dKO hfSCs from precocious anagen activation but instead restored matrix progenitor differentiation into visible hair. Thus, our results reveal very unexpected findings that, in contrast to morphogenesis, postnatally there is functional redundancy between pSmad8 and pSmad1/5 only in matrix progenitor differentiation which significantly was excluded during morphogenesis and hfSCs quiescence in the prebulge and bulge. Collectively, these findings provide new insights into the functions of different downstream components of the canonical BMP pathway in establishment and maintenance of hfSCs quiescence.
Materials and Methods
Generation, Genotyping, and Manipulation of Transgenic Mouse Lines
Keratin 14 (K14) promoter-driven Cre recombinase mice and K15-CrePR transgenic mice, inducible by topical administration of RU486 were obtained from Jackson Laboratories (Bar Harbor, ME, http://www.jax.org) [10, 36]. Conditional floxed alleles of Smad1 and Smad5 were generated by Huang et al. and Umans et al., respectively [37, 38]. Rosa26-STOP-eYFP (Yellow Fluorescent Protein) reporter mice were obtained from Jackson Laboratories . To generate mice lacking Smad1 and Smad5 in the skin epithelium during development (morphogenesis), Smad1fl/fl and Smad5fl/fl mice were crossed with K14Cre mice or, alternatively, with K15CrePR mice to generate offspring with inducible, targeted Smad1 and Smad5 ablation (dKORU) in the hfSCs population following RU486 treatment postnatally. All mice were genotyped from tail snips. YFP labeling and targeted deletion of Smad1 and/or Smad5 in hfSCs was achieved by topical application of RU486 (25 mg/mL, VWR), to shaved dorsal back skin during the second postnatal telogen from P43 to P59. Following recombination events, Smad-deficient cells (and all progeny) were permanently marked by YFP expression. All animal procedures were performed with the appropriate approval of the IACUC committee.
Histology, Immunostaining, and BrdU Labeling
Cryosections for immunostaining were briefly fixed in 4% paraformaldehyde (PFA), washed in phosphate buffered saline (PBS), and incubated in 0.1% Triton X-100 (v/v in PBS) for 10 minutes. Sections were then blocked in 10% appropriate serum (v/v in 0.1% Triton X-100/PBS) for 30 minutes at room temperature then incubated in primary antibodies (Abs) overnight at 4°C. Antibodies were diluted in blocking solution at the following dilutions: AE13 (1:100; Santa Cruz, Santa Cruz, CA, http://www.scbt.com), BrdU (1:100; Abcam, Cambridge, U.K., http://www.abcam.com), CD34 (1:100; ebioscience), Gata3 (1:75, Santa Cruz), K1 (1:300; Gifted by C. Jamora), K5 (1:300; gifted by C. Jamora), K6 (1:200; gifted by P. Coulombe), K15 (1:100; Thermo Scientific), Ki67 (1:300; Novocastra, Newcastle upon Tyne, U.K., http://www.novocastra.co.uk), Loricrin (1:300; gifted by C. Jamora), Nfatc1 (1:100; Santa Cruz), pSmad1/5 (1:50; Cell Signaling), pSmad1/5/8 (1:50; Cell Signaling, Beverly, MA, http://www.cellsignal.com), Sox9 (1:100; Santa Cruz), YFP (1:2,000; anti-GFP, Abcam). Sections were washed in PBS and incubated with secondary Abs, diluted at 1:300 in appropriate blocking solution for 45 minutes at room temperature in the dark. Sections were washed thoroughly with PBS, counterstained with 4′,6-diamidino-2-phenylindole (DAPI) to label nuclei, mounted, and stored at 4°C until ready for analysis. Additional tissue processing was performed to obtain defined YFP staining (for lineage tracing experiments) where skin samples were fixed in 4% PFA for 2 hours and then immersed in 30% sucrose (w/v in PBS) overnight at 4°C before embedding in OCT for sectioning and staining as described above. For bromodeoxyuridine (BrdU) pulse experiments, BrdU (50 µg/g; Calbiochem, San Diego, CA, http://www.emdbiosciences.com) was administered by i.p. injection 4 hours before analysis. BrdU incorporation was detected by BrdU immunostaining which was performed after fixed sections were first incubated in 1 M HCl for 1 hour at 37°C. For detection of sebaceous glands (SGs), oil red O staining was performed by fixing sections in 4% PFA for 10 minutes, washing with PBS then immersing in oil red O staining solution for 10 minutes at room temperature. Sections were washed in isopropanol and counterstained with hematoxylin. To localize the DP, alkaline phosphatase (AP) staining was performed using the TRACP and ALP Assay Kit (Takara, Otsu, Japan, http://www.takara.co.jp) kit according to the manufacturer's instructions. Stained sections were examined using a Zeiss upright fluorescent microscope with images processed with Adobe Photoshop CS3.
Fluorescence-Activated Cell Sorting Analysis of HF Bulge Markers
Analysis of bulge, hfSCs from adult mouse dorsal back skin were performed as described previously . For hfSCs CD34 marker expression analysis, telogen (P59) and anagen (P130) YFP+ ConRU and dKORU cell suspensions were stained with the following antibodies, anti-α6-integrin conjugated to Phycoerythrin (PE) (1:200; BD Pharmingen, San Diego, CA, http://www.bdbiosciences.com/index[lowen]us.shtml) and anti-CD34 coupled to Alexafluor-700 (1:50; BD eBioscience). Cells were gated first for live cells (absence of DAPI incorporation) and then through the YFP+ channel to specifically examine viable YFP+ hair cell populations. Basal hfSCs:YFP+/α6-integrinHigh/CD34High labeled cell fractions were then analyzed with a fluorescence-activated cell sorting (FACS) Aria cell sorter (BD Biosciences equipped with FACS DiVa software).
Data Quantification of Telogen and Anagen hfSCs Marker Immunofluorescent Staining
Immunofluorescent (IF) stainings for telogen and anagen YFP+ ConRU and dKORU HFs were performed using K15 and Sox9 Abs. The number of double positive YFP+/K15+ (telogen, n = 10; anagen n = 10 HFs counted) and YFP+/Sox9+ (telogen, n = 9; anagen n = 10 HFs counted) cells per field of view (FOV) were quantified, and representative images were presented in the Results section. For quantification of pSmads activity in anagen HFs, YFP+/pSmad1/5/8+ (n = 20) and YFP+/pSmad1/5+ (n = 20) immunostaining was performed on serial sections of YFP+ ConRU and dKORU HFs, and the number of double YFP+/pSmad+ cells were counted per FOV, and representative images were provided in the Results section. Analysis of the second postnatal telogen hair cycle was performed on ConRU, S1-KORU, S5-KORU, and dKORU dorsal skin samples which were sectioned and stained with H&E at time points P85, P90, and P100 (n = 15 HFs counted per HF cycle stage for each sample).
Full Thickness Skin Grafting Experiments
The full thickness dorsal skin from K14Cre/Smad1fl/fl/Smad5fl/fl (dKO) mice sacrificed at P0 along with skin from Con littermates were grafted onto either side of the back of recipient athymic (nude) mice as previously described . Graft regions from the dKO and Con grafts were frequently monitored for hair growth and harvested for analysis after 3 days, 4 weeks, and 6 weeks postengraftment to examine hair morphogenesis.
Scanning Electron Microscopy
Freshly isolated newborn whiskerpad samples were collected and fixed overnight (4% PFA, 2.5% glutaraldehyde; 0.1 M cacodylate buffer) at 4°C prior to processing for routine scanning electron microscopy as described previously .
Generation of Mice Lacking Smad1 and Smad5 (dKO) in the Skin Epidermis and Its Derivatives
Although Smad8-null mutant mice develop normally, are viable and fertile , ablation of either Smad1 or Smad5 during embryonic development results in lethality [42-44], thereby impeding the study of canonical BMP-mediated, Smad function in skin and epidermal appendages. To overcome this obstacle, we used a Cre-LoxP approach to inactivate Smad1 and/or Smad5 in the developing epithelium in a tissue-specific manner using conditional floxed alleles [37, 38]. We generated transgenic mice that were conditionally deficient for epithelial Smad1 and Smad5 (dKO) by using a well-characterized keratin-14 (K14) promoter-driven Cre recombinase . The K14 promoter activity begins in developing epithelia from approximately embryonic day (E) 9.5 and then continue to drive Cre expression in adult epidermis, including HFs, teeth, oral palate, and tongue [30, 31, 45]. To visualize and compare potential phenotypic changes arising from targeted Smad1 and Smad5 deletion, dKO and Con mice were also crossed on the ROS26-STOP-eYFP reporter background to permanently YFP label K14-derived tissue (Fig. 1A) . Primers specific for the floxed and wild-type alleles of Smad1 and 5 were used to genotype mice (Fig. 1B). At birth (P0), dKO and Con pups were of similar size (Fig. 1C) and displayed comparable levels of YFP expression in the epidermis (Fig. 1C′). However, when compared to Con littermates (Fig. 1C, 1D), dKO pups (Fig. 1C, 1C′, 1D′) presented a range of phenotypic abnormalities that included open eyes (Fig. 1C, 1C′), fewer visible whiskers (Fig. 1D′ vs. 1D), and severely deformed paws (Fig. 1C, 1C′).
Smad1 and Smad5 Deletion Results in Normal Vibrissae Patterning but Aberrant Differentiation
At birth, Con pups displayed numerous vibrissae which protruded from the surface whiskerpads (Fig. 1D), in contrast, dKO mice exhibited an almost complete absence of visible whisker formation (Fig. 1D′). However, dKO pups (Supporting Information Fig. S1A′, S1B′) displayed normal patterning and comparable numbers of whiskerpads to Con littermates (Supporting Information Fig. S1A, S1B). Scanning electron microscopy confirmed the number and positioning of mouse whiskers in both Con (Supporting Information Fig. S1C, with vibrissae shaft production, Supporting Information Fig. S1C′) and dKO (Supporting Information Fig. S1D, with lack of visible vibrissae formation, Supporting Information Fig. S1D′). Morphological analysis revealed normal initiation of vibrissae development with SGs and DP formation in dKO (Supporting Information Fig. S1E′, S1F′) and Con (Supporting Information Fig. S1E, S1F), however in dKO no defined shaft formation was observed. The vibrissae phenotype in dKO animals is similar to that observed in conditional Bmpr1a-null mice  which also exhibit compromised whisker formation. Thus, these data suggest that canonical BMP signaling through Smad1/5 signaling activity is important for normal vibrissae differentiation but independent from whisker patterning.
Different pSmads Are Activated in Distinct Skin Compartments During Skin Morphogenesis
To determine the expression pattern of active canonical BMP-Smad signaling components in newborn Con and dKO mouse skin, we performed pSmad1/5/8 IF staining. First, antibodies that recognize pSmad1/5/8 labeled the nuclei in Con IFE and HFs throughout the ORS, matrix, and in all differentiated layers of the IRS (Fig. 1I, 1I′, 1E, 1E′, inset). In contrast, in dKO HFs nuclear pSmad1/5/8 staining was only visible in the IFE and in the infundibulum/upper part of ORS adjacent to IFE (Fig. 1J, 1J′, 1F, 1F′ arrow). Next, we examined the pattern of activated pSmad1/5. Surprisingly, in Con skin we observed distinct, nuclear pSmad1/5 expression restricted to the upper matrix region, differentiated layers (IRS and HS) and weakly in the infundibulum (Fig. 1G, 1G′, arrow, inset), whereas IFE was negative (Fig. 1K, 1K′). The efficient ablation of both Smads in dKO skin was confirmed by the complete absence of nuclear pSmad1/5 staining in developing HFs and IFE (Fig. 1H, 1H′, 1L, 1L′). Thus, using a combination of specific pSmad Abs we discriminated distinct patterns of pSmads activity in Con and dKO skin with pSmad8 restricted to the IFE and pSmad1 and/or pSmad5 exclusively activated in HFs during morphogenesis.
Next, to determine whether canonical BMP signaling transduced by Smad1 and Smad5 is required for normal HF development, we analyzed newborn Con and dKO skin. Morphologically, Con skin displayed developing HFs in early anagen with evidence of early differentiation and HS production (Supporting Information Fig. S2A). In contrast, malformed bulbous hairpeg-like structures lacking HS formation were observed in dKO skin (Supporting Information Fig. S2A′). HFs density measurements (Supporting Information Fig. S2B) revealed that dorsal dKO skin displayed a marked reduction in HFs number compared with Con littermate skin (4 ± 1 dKO vs. 7 ± in Con HFs/FOV). The DP was localized to the base of Con (Supporting Information Fig. S2C) as well as dKO (Supporting Information Fig. S2C′) HFs at P0. We observed similar BrdU incorporation levels in Con (Supporting Information Fig. S2D, inset) and dKO (Supporting Information Fig. S2D′, inset) hair matrix progenitors. Furthermore, although dKO epidermis was marginally thinner than Con epidermis, we demonstrated that deletion of Smad1 and Smad5 did not perturb epidermal differentiation as the expression patterns of K1 and loricrin were comparable in Con (Supporting Information Fig. S2E, S2F) and dKO (Supporting Information Fig. S2E′, S2F′) back skin epidermis.
Lack of Prebulge Formation and HF Differentiation in Smad1- and Smad5-Deficient Mice During Morphogenesis
To further characterize the consequences of Smad1 and Smad5 ablation during HF morphogenesis, we examined the expression of several HF differentiation markers. AE13, a marker of the HS specific keratins  was strongly expressed in the cortical layer of Con (Fig. 2A) but not in dKO HFs (Fig. 2A′). We observed strong nuclear expression of GATA-3, which is a transcription factor essential for the differentiation of IRS progenitor cells , in Con HFs (Fig. 2B), however, dKO follicles were devoid of GATA3 marker expression (Fig. 2B′). K6 is a marker of the companion layer separating the ORS and IRS , Con HFs displayed strong expression (Fig. 2C) but no K6 expression was visible in dKO HFs, only ectopic K6 expression in dKO epidermis (Fig. 2C′).
The prebulge is formed during HFs morphogenesis and is thought to contribute to the formation of the mature postnatal HF bulge where hfSCs reside . To examine whether dKO HFs displayed normal expression of prebulge markers, we labeled Con and dKO skin with K15, Sox9, and Nfatc1 Abs [10, 11, 13, 14]. Con skin displayed distinct K15 expression in the basal layer of the epidermis adjacent and extending to the upper ORS layer of HFs (Fig. 2D). In comparison, dKO skin exhibited reduced K15 staining in the corresponding upper infundibulum-hinge region (Fig. 2D′). Strong nuclear Sox9 expression was observed in the Con prebulge region (Fig. 2E), however, expression was considerably reduced in dKO HFs (Fig. 2E′). NFATc1, a marker specific to the quiescent hfSCs  marked the prebulge of Con HFs (Fig. 2F) but was not detected in dKO HFs (Fig. 2F′). Together, these data suggest that Smad1 and Smad5 are both essential for establishment of prebulge region and HFs differentiation.
Grafted dKO Skin Displays Absence of Hair Differentiation and Failure to Develop the Postnatal Hair Bulge
As dKO mice die within the first 24 hours after birth, we performed full thickness dorsal skin grafts from newborn Con and dKO mice onto recipient athymic (nu/nu) mice to determine the long-term consequences of Smad1 and Smad5 ablation during morphogenesis. At 4 weeks (wks) postengraftment, visible pelage hair was observed in Con (Fig. 3A, left) but not on dKO grafted skin (Fig. 3A, right). At 3 days postgrafting, Con HFs were in anagen with differentiating follicles that extended into the dermis (Fig. 3B). In contrast, dKO HFs were much smaller, displaying delayed morphogenesis such that they remained superficial to the epidermis (Fig. 3B′). Measurements of HF density revealed that dKO graft skin exhibited considerably fewer HFs compared with the Con graft region (Fig. 3C). By 4 weeks postgrafting, Con follicles had completed morphogenesis, established the postnatal bulge, and had transitioned into early anagen (Fig. 3D, 3E). At the same time point, dKO HFs grafts were aberrant, presenting immature follicles that lacked evidence of HS differentiation (Fig. 3D′, 3E′). Although dKO HFs were delayed, they showed continued down-growth and formation of SGs (Fig. 3D′, 3E′). Strikingly, in dKO HFs there was no visible bulge region suggesting that they did not progress through hair cycle involution (Fig. 3D′, 3E′) and instead persisted in a prolonged anagen-like state. Con grafts displayed normal DP and dermal sheath (DS) localization identified with AP staining at the base of anagen follicles adjacent to hair matrix (Fig. 3F). In the dKO grafts, although the atypical bulb matrix was visible, it was still encased by the DP and DS-like mesenchymal cells (Fig. 3F′). SGs structures, identified with oil red O staining, were associated with both Con and dKO HFs (Fig. 3G, 3G′).
We next performed pSmad1/5/8 and pSmad1/5 Abs staining to identify the pattern of Smad1/5/8 activity in 6-week-old skin grafts. In Con grafts, anagen HFs displayed abundant nuclear pSmad1/5/8 in the hair matrix, HS and IRS, however, no nuclear staining was visible in dKO HFs (Fig. 3H vs. 3H′). Furthermore, we observed abundant nuclear pSmad1/5 staining in Con HFs, primarily localized to the matrix, HS and IRS, however, pSmad1/5 staining was absent in dKO HFs (Fig. 3I vs. 3I′). Normal HF development requires matrix cells to proliferate, therefore, we examined Ki67 expression in Con and dKO skin grafts and detected similar levels of nuclear Ki67 expression in the matrix and ORS (Fig. 3J vs. 3J′). Next, using IF directed against HF differentiation markers, we found that AE13 expression was present in Con but absent in dKO HFs (Fig. 3K vs. 3K′). Furthermore, the presence of GATA3 precursors of the IRS was observed in Con skin grafts HFs whereas GATA3 expression was markedly reduced in dKO hairs (Fig. 3L vs. 3L′). CD34 expression was observed in the HF bulge in Con graft follicles (Fig. 3M, inset arrow), however, CD34 expression was undetectable in dKO graft follicles suggesting perturbed slow-cycling characteristic (Fig. 3M′). Notably, single Smad mutants (either Smad1-KO or Smad5-KO) did not display perturbed hair differentiation (data not shown). Collectively, these results indicate an essential role for both Smad1 and Smad5 in the transduction of canonical BMP signaling in matrix progenitor differentiation and bulge formation during hair morphogenesis.
Postnatal Deletion of Smad1 and/or Smad5 Results in Precocious Anagen Onset of hfSCs
Since we observed that prebulge and bulge formation was disrupted in dKO HFs during morphogenesis and then later in skin grafting, we next investigated the role of pSmads in hfSC homeostasis after the bulge region was established postnatally. To this end, we used a conditional, inducible, K15-driven CrePR recombinase approach  to ablate floxed Smad1 and/or Smad5 genes following topical application of RU486 (RU) during the second postnatal telogen (P43-P59). By crossing onto a Rosa26-YFP reporter line, we were also able to simultaneously and permanently induce YFP expression in hfSCs and their progenies. First, we investigated the postnatal pSmads expression pattern in Con (ConRU), Smad1-null (S1-KORU), Smad5-null (S5-KORU), and double Smad1 and Smad5-null (dKORU) in the second postnatal extended telogen (Fig. 4A–4E). After 16 days of RU treatment (P43-P59), activated YFP expression was specifically observed in S1-KORU, S5-KORU, and dKORU HF bulges indicating high efficiency of recombination in vivo. Using anti-pSmad1/5/8 IF staining, we observed that ConRU HFs at P65 showed nuclear pSmad staining throughout the bulge (Fig. 4A). In single S1-KORU (Fig. 4B) and single S5-KORU (Fig. 4C) YFP+ hfSCs exhibited nuclear pSmad staining indicating that Smad1-null YFP+ hfSCs could express either nuclear pSmad5 and/or Smad8, similarly, S5-KORU YFP+ hfSCs could express pSmad1 and/or Smad8. Surprisingly, we observed nuclear pSmad staining in dKORU YFP+ hfSCs, which suggested that activated pSmad8 was expressed postnatally within the bulge hfSCs population (Fig. 4D). To verify this, we checked the efficiency of ablation of both Smad1 and Smad5 in the bulge in dKO YFP+ hfSCs, using the specific pSmad1/5 Abs staining and confirmed a lack of pSmad staining in postnatal dKO YFP+ labeled hfSCs (Fig. 4E). Next, we determined the postnatal consequences of loss of Smad1 and/or Smad5 on hfSC homeostasis. After the RU treatment, all HFs were in prolonged telogen from P65 to P85 (Supporting Information Fig. S3; Fig. 4V-P85). By P90, ConRU, S1-KORU, S5-KORU (Fig. 4F, 4J, 4N, respectively) remained in telogen, however, all dKORU HFs (Fig. 4R) displayed precocious anagen (Fig. 4V-P90). At this time point, Ki67 IF confirmed the quiescent nature of ConRU, YFP+ S1-KORU, and YFP+ S5-KORU HFs (Fig. 4G, 4K, 4O and insets, respectively), in contrast, positive nuclear Ki67 staining was observed in the anagen dKORU hair matrix (Fig. 4S, and insets). By P100, ConRU HFs remained morphologically in telogen (Fig. 4H, 4I, 4V-P100), however, YFP+ S1-KORU and YFP+ S5-KORU displayed enlarged hair germs (HG) indicative of telogen-anagen transition (Fig. 4L, 4M, 4P, 4Q, 4V-P100) whereas YFP+ dKORU HFs progressed in anagen (Fig. 4T, 4U, 4V-P100). Quantification of the late telogen postnatal HF cycle in ConRU and Smad-deficient HFs confirmed that between P65-P85, 100% HFs remained in telogen (Fig. 4V-P85; Supporting Information Fig. S3). At P90, 100% of dKORU HFs had started anagen but ConRU, YFP+ S1-KORU, and YFP+ S5-KORU HFs had not (Fig. 4V-P90). At P100, where all dKORU HFs progressed in anagen, approximately 70% of S1-KORU and S5-KORU HFs were observed in early anagen while all ConRU HFs remained in telogen (Fig. 4V-P100).
Expansion of K15 Marked Bulge hfSCs Follows Postnatal Ablation of Both Smad1 and Smad5
To determine postnatally how precocious anagen entry resulting from loss of Smad1 and Smad5 affected hfSCs marker expression over time, we examined the expression of known hfSC markers, CD34, K15, and Sox9 during the telogen and anagen phases of the postnatal HF cycle. In early telogen, similar CD34 expression levels were observed using IF and FACS with only a slight decrease observed by FACS analysis in YFP+ dKORU HFs (Fig. 5A′, 5B′, 12.2%) compared with YFP+ ConRU (Fig. 5A, 5B, 18.2%). At P130 in anagen HFs, CD34 expression in the YFP+/α6integrinHighCD34High population, labeling basal hfSCs, was significantly decreased in dKORU cells (Fig. 5F′, 2.8%) in contrast to ConRU (Fig. 5F, 14.3%). Interestingly, the reduction of the YFP+/α6integrinHighCD34High population was comparable to that previously observed in YFP+ Bmpr1a cKORU HFs at P59 (Supporting Information Fig. S4C, S4C′, 1.8%) . These findings were also confirmed by decreased IF staining for CD34 in the YFP+ dKORU HFs which already displayed perturbed bulge architecture (Fig. 5E′ vs. ConRU5E). In these FACS analyses, all CD34 and α6 populations were gated through YFP positive cells and we observed a similar percentage of total YFP+ cell populations for ConRU and dKORU HF-derived cells (Supporting Information Fig. S4A, S4A′, telogen and Supporting Information Fig. S4B, S4B′, anagen). Moreover, K15 (Fig. 5C–5C′′) and Sox9 (Fig. 5D–5D′′) expression was present and comparable in telogen YFP+ ConRU (Fig. 5C, 5D) and dKORU HFs (Fig. 5C′, 5D′). At P130, when both ConRU and dKORU HFs were in anagen, although we observed comparable nuclear Sox9 expression in the bulge and ORS of YFP+ dKORU HFs (Fig. 5H′) and ConRU HFs (Fig. 5H, 5H′′), interestingly, K15 expression was expanded in the YFP+ dKORU hfSCs (Fig. 5G′ arrows, 5G′′) when compared to ConRU HFs (Fig. 5G).
pSmad8-Mediated BMP Signaling Can Sustain Postnatal Hair Differentiation
We observed distinct activation patterns of pSmads during hair morphogenesis with epidermal-restricted pSmad8 expression (Fig. 1F, 1F′) and pSmad1 and pSmad5 exclusively activated in HFs (Fig. 1G, 1G′). This spatial distinction of pSmad1/5 activation resulted in the blocking of HF terminal differentiation and visible hair formation in dKORU (Figs. 2A′–2C′, 3A). As we observed pSmad8 expression in the bulge of dKORU hfSCs (Fig. 4D), we next determined whether postnatal pSmad8 expression was expanded to all bulge-derived descendants and, if so, how it affected the overall HF architecture. To determine whether Smad8 alone was activated in postnatal anagen dKORU HFs, we labeled dKORU serial sections of HFs with pSmad1/5/8 (Fig. 6A) and pSmad1/5 Abs (Fig. 6B). We observed no pSmad1/5 staining in YFP+ dKORU cells (Fig. 6B), however, there was strong nuclear pSmad1/5/8 expression in the HFs localized to YFP+ matrix progenitors, ORS, HS, and IRS (Fig. 6A). Thus, after efficient ablation of Smad1 and Smad5 in dKORU HFs there was still some pSmad activity present in postnatal anagen YFP+ dKORU HFs (Fig. 6A, 6C) suggesting pSmad8 activity in YFP+ dKORU cells.
Next, we addressed whether pSmad8 activity in YFP+ dKORU bulge descendent progenies affected the overall HFs architecture. To this end, after topical RU treatment from P43 to P59, we waxed a region of dorsal skin from ConRU (Fig. 6D) and dKORU (Fig. 6D′) mice to remove most of the HS. The removal of hair also synchronized new anagen activation in both YFP+ ConRU and dKORU HFs. After 16 days postwaxing (pw), we observed visible hair production in ConRU (Fig. 6E) and, surprisingly, in dKORU (Fig. 6E′) waxed regions. In the Con, the hair generated appeared normal (Fig. 6F), however, the hair regenerated in dKORU appeared somewhat disheveled (Fig. 6F′). We next examined YFP expression within the regenerating HFs by whole mount visualization of skin biopsies from waxed regions and observed that all regenerating HFs, both in ConRU and dKORU, were in anagen and displayed YFP expression throughout all HF layers (Fig. 6G–6H′). In contrast to the ConRU, which generated HFs with a straight orientation (Fig. 6H), dKORU HFs displayed a slightly wavy structure (Fig. 6H′). YFP examination of sections from 16d pw skin biopsies of both ConRU and dKORU confirmed that HFs were in anagen and were efficiently marked with YFP (Fig. 6I, 6I′). Together, these data suggest that the YFP+ dKORU hair phenotype was attributable to remaining pSmad8–mediated BMP signaling in postnatal HFs which lacked Smad1 and Smad5 expression. To test this hypothesis, we labeled serial sections of ConRU and dKORU YFP+ skin biopsies at 16 days pw with pSmad1/5/8 (Fig. 6J, 6J′) and pSmad1/5 Abs (Fig. 6K, 6K′). We quantified the number of double positive YFP+/P-Smad1/5/8+ cells in anagen HFs and found abundant nuclear pSmad1/5/8 staining throughout the hair matrix, HS, and IRS in both ConRU and dKORU with comparable activity levels (with an average of approximately 70 double positive YFP+ and pSmads+ cells per FOV, Fig. 6J′′). Although, the quantified amount of double positive cells in serial sections of ConRU YFP+/P-Smad1/5 was lower (with an average of approximately 14 double positive YFP+/pSmads+ cells per FOV, Fig. 6K′′ and Fig. 6K), double positive YFP+/pSmad1/5+ cells were undetectable in YFP+ dKORU HFs (Fig. 6K′′ and Fig. 6K′) and confirmed efficient ablation of Smad1/5 in YFP+ dKORU anagen follicles. Sporadically we observed rare YFP-/Smad1/5+ cells (2–3 cells per FOV, Fig. 6K′′). Together, the comparable level of nuclear pSmad1/5/8 activity in YFP+ ConRU and YFP+ dKORU HFs and lack of nuclear pSmad1/5 activity in YFP+ dKORU HFs suggested that pSmad8-mediated BMP signaling activity alone in dKORU HFs could restore hair shaft production during postnatal hair regeneration.
To test this hypothesis, we performed triple fluorescent staining to mark YFP+, pSmad1/5/8 along with the HF differentiation marker AE13 for pre- and cortical layers. We observed that all AE13 positive cortical cells were positive for pSmad1/5/8 staining in dKORU saggital (Fig. 6L–6L′′) and transverse sections (Supporting Information Fig. S5A, S5A′), however, YFP+/AE13+ dKORU cells were negative for pSmad1/5 staining in serial sections of the same follicles (Fig. 6M–6M′′ and Supporting Information Fig. S5B, S5B′) demonstrating the efficiency of ablation in dKORU HFs. Moreover, we also observed colocalization of YFP+ dKORU cells with nuclear GATA3 expression, an important HF differentiation mediator, in the differentiating layers of YFP+ dKORU anagen HFs (Supporting Information Fig. S5C–S5C′′). Together, these data support the hypothesis that postnatal dKORU cells are capable of proper HF differentiation events and mediate HS production through canonical BMP-pSmad8 activity. Interestingly, although dKORU skin was able to regenerate hair, we also observed that the mutant HFs did not progress through the normal hair cycle. Whole mount analysis revealed that 4 weeks after waxing, YFP+ ConRU HFs had progressed to telogen (Supporting Information Fig. S6A, S6A′) and H&E staining of ConRU sections confirmed that the YFP+ HFs were in a resting state (Supporting Information Fig. S6B, S6B′). In contrast, 4 weeks pw YFP+ dKORU HFs persisted in an anagen-like state (Supporting Information Fig. S6C, S6C′). H&E staining confirmed dKORU HFs in full anagen (Supporting Information Fig. S6D) and all HFs remained efficiently YFP+ labeled (Supporting Information Fig. S6D′). These findings could suggest that inactivation of Smad1 and Smad5 in the postnatal HF bulge promotes the loss of hfSCs quiescence and prolonged HF activation.
Here, genetic approaches allowed us to specifically ablate the downstream components of canonical BMP signaling, Smad1 and Smad5, to characterize the functions of specific pSmad complexes in HF morphogenesis and postnatal hfSCs homeostasis.
pSmad-Complex Compartmentalization During Skin Morphogenesis
Surprisingly, during morphogenesis our dKO HFs model revealed distinct activation patterns of pSmads, with remaining pSmad8 restricted to the IFE but absent in HFs which physiologically expressed only Smad1 and/or Smad5 (Fig. 7B). This finding allowed us to propose a model whereby abundant, pSmad8 activation is predominant throughout the normal IFE, and pSmad1/5 was exclusively associated with the HF lineage (Fig. 7A). During morphogenesis, we confirmed efficient recombination events in dKO skin where Smad1 and Smad5 were effectively deleted (Fig. 1H, 1H′, 1L, 1L′), and our results indicated that both Smad1 and Smad5 are not required for proper epidermal differentiation (Supporting Information Fig. S2E–S2F′). Furthermore, our results suggest that epidermal pSmad8 activity is sufficient to maintain a significant level of canonical BMP signaling required for normal epidermal differentiation (Fig. 7B).
During skin appendage morphogenesis, we observed that pSmad1/5 activation is restricted to the matrix and differentiating HF layers. Furthermore, loss of pSmad1/5 during HF morphogenesis and skin engraftment had severe consequences for hair differentiation and the production of visible HS (Figs. 2A–2C′, 3A). Thus, these data indicate that canonical BMP signaling, acting via Smad1/5 components, is necessary to transmit signals critical for HFs differentiation (Fig. 7D vs. 7C). Moreover, Smad1 and Smad5 appear to be functionally redundant, since single Smad1or Smad5 mutants did not manifest any hair differentiation phenotype (data not shown). However, as Smad8 was not present or re-expressed upon Smad1/5 ablation in dKO HFs during morphogenesis, we could not rule out a possible redundancy in function between pSmad1/5 and pSmad8.
Requirement of Canonical BMP-pSmad-Mediated Signaling for HF Cycle Regulation
We demonstrate that, in addition to defective HF differentiation, dKO HFs fail to transition through the normal hair cycle and instead persist in a prolonged anagen-like state (Supporting Information Fig. S6D, S6D′). These data are consistent with previous reports of prolonged anagen-like follicles described in BMPRIA-null HFs [29-31]. Ablation of BMPRIA results in the development of follicular cysts and matricomas, however, dKO HFs displayed typical follicle structure albeit with an immature appearance. One possible explanation could be that the time points examined (the longest at 6 weeks postengraftment) were too early to observe HF cyst or tumor development. Alternatively, the BMPR1A receptor may direct aspects of its function by activating noncanonical pathways which remain intact in dKO Smad1/5 mice, these possibilities highlight interesting avenues to investigate in the future.
Canonical BMP-pSmad Signaling As Prerequisite for HF Bulge Establishment
The culmination of HF morphogenesis is the establishment of the mature, hair bulge containing quiescent hfSCs. Even before the mature bulge appears, the expression of prebulge markers pre-figure the formation of this structure . Our data revealed that Smad1/5 were required for the formation of prebulge and bulge structures. In dKO engrafted skin, HFs lacked a defined, morphological bulge region and CD34, a marker of postnatal hfSCs, was not present in dKO follicles (Fig. 7D vs. 7C). Thus, to further investigate the existence of the hfSCs population in dKO HFs (despite the absence of prebulge and bulge structures), we examined the expression of hfSCs markers like Lhx2, Sox9, Tcf3, and Nfatc1 which are present during early HFs morphogenesis [11-15]. Indeed, although the hfSC quiescence marker CD34 was reduced in adult postnatal dKO HFs, we still observed Sox9 expression in conjunction with expanded K15 expression (Fig. 7F vs. 7E). Furthermore, we also observed the hfSC quiescence marker, Nfatc1 , was also missing in dKO follicles (Fig. 7B vs. 7A).
Thus, our findings suggest that bulge morphology is attributed with a quiescent state of hfSCs marked by CD34 and Nfatc1 expression, but hfSCs are not lost in dKO HFs and are still able to generate long-lived HFs expressing K15 and Sox9 markers. These data indicate that canonical BMP signaling is required to establish the prebulge and then the bulge region by pSmad1 and pSmad5 activity (Fig. 7B, 7D vs. 7A, 7C). Together, our results emphasize the critical role of pSmad1 and pSmad5 both during HF morphogenesis and postnatally in the formation and maintenance of the HF bulge with quiescent hfSCs.
Different Threshold of Canonical BMP Signaling via pSmad1 and 5 but not pSmad8 Alone Is Required for Maintenance of Quiescence hfSCs and Hair Proper Cycling
Precocious anagen entry following postnatal ablation of Smad1/5 in hfSCs was in agreement with our previous studies, which demonstrated premature anagen and loss of quiescent hfSCs in conditional, BMPRIA-null postnatal HFs [29, 48]. However, here we observed that even the deletion of individual Smads: Smad1 KORU or Smad5 KORU, was sufficient to prime hfSCs for earlier activation (P100) which was significantly slower than observed in double Smad1/5 dKORU (P90) when compared to ConRU (P120). Therefore, it suggests that quiescent hfSCs are able to sense a different threshold of canonical BMP signaling which could modulate their different responsiveness during the hair cycle. The presence of this canonical BMP signaling “fine-tuning” within the hfSCs suggests that Smad1 and Smad5 might functionally overlap and regulate the same BMP-responsive elements in targeted genes. This is in agreement to other studies, which have demonstrated that increasing expression of DP-derived BMP inhibitors such as noggin, Sostdc1, or Bambi modulate the BMP signaling threshold and promote the transition from the “refractory telogen” to the “competent telogen” where hfSCs become primed for activation to initiate the new hair cycle [26, 32, 49, 50]. Interestingly, single Smad1-KORU and Smad5-KORU HFs were capable of transitioning through the hair cycle and could progress through catagen, successfully degenerating to form a postnatal telogen follicle (data not shown). In contrast, dKORU follicles failed to undergo programmed hair involution and remained in a prolonged anagen-like state (Supporting Information Fig. S6D′ vs. S6B′). These results indicate that loss of at least two Smads is sufficient to disturb hfSCs quiescence and the structural integrity of the bulge leading to loss of proper hair cycling. Moreover, it also indicates that the solo presence of activated pSmad8 in dKORU HFs cannot maintain quiescence of adult postnatal hfSCs and proper HFs cycling. Furthermore, although quiescence of hfSCs (as indicated by CD34 expression) is lost in the absence of at least two Smads (Fig. 5E′, 5F′vs. 5E, 5F), stem cells persist and become expanded in vivo reflected by K15 expansion and Sox9 staining (Fig. 5G′, 5H′ vs. 5G, 5H). Thus, our results emphasize the critical role of Smad1 and Smad5 during postnatal maintenance of HF bulge morphology, hfSCs quiescence and hair cycling (Fig. 7F vs. 7E). However, it still remains to be elucidated if the different threshold of canonical BMP signaling is limited to these two specific Smads: Smad1 and Smad5, or, whether the effect could be similar by the deletion of any two Smad combinations within the postnatal hfSCs population.
Postnatal Hair Differentiation Can Be Maintained by Smad8 Alone After Smad1/5 Ablation
Surprisingly, 2 weeks after depilation, both YFP+ ConRU and dKORU HFs were in anagen and, in both cases, visible hair shafts were present (Fig. 6D–6F′). These results suggested that pSmad8-mediated signaling alone in postnatal dKO hfSCs is able to compensate for the lost of Smad1/5 in the differentiation of matrix progenitors but it is not sufficient to maintain the quiescence of undifferentiated hfSCs (Fig. 7E vs. 7F). In single Smad1 KO or single Smad5 KO, only partial rescue of hfSCs quiescence is observed (precocious anagen in single KOs HFs still occurred when compared to Con HFs but significantly delayed compared to dKO HFs, Fig. 4F–4U), however, full rescue of matrix progenitors differentiation (data not shown). One possible explanation could be the existence of a different threshold for activation of Smad target genes, higher for quiescent hfSCs and lower for differentiated matrix progenitors. Moreover, this lower threshold for activation of Smads target genes was also observed in developing epidermis. Alternatively, another scenario could be that pSmad8 has different transcriptional specificity, perhaps binding to target gene promoters differing from Smad1 and/or Smad5 in quiescent hfSCs but overlapping with Smad1/5 target genes in matrix progenitors during differentiation. The latter would further suggest that stem cells, in comparison to progenitor cells, exist in different epigenetic stages  that could modulate different accessibility to the pool of Smads target genes during quiescence and differentiation. Surprisingly, the complete lack of hair progenitor differentiation in morphogenesis could not be restored, even following long-term dKO skin engraftment (Fig. 7B, 7D vs. 7A, 7C). Interestingly, these results were contrary to those observed in adult mice when Smad1/5 were ablated postnatally after the quiescent bulge hfSCs pool was established and full hair differentiation was rescued (Fig. 7F vs. 7E). Thus, it further suggests that canonical BMP signaling (via pSmad1 and pSmad5), is crucial during hair morphogenesis for HF to reach the adult mature state by promoting the changes in Smad8 expression pattern to occur in postnatal HFs (Fig. 7G). Therefore, without pSmad1 and pSmad5 hfSCs still persist but cannot reach quiescent homeostasis to form the prebulge and subsequent bulge regions during development and this event is critical to enable Smad8 expression in adult HFs including stem and progenitors cells (Fig. 7G).
Together, our data demonstrate a pivotal role of canonical BMP signaling, with Smad8 redundancy only with Smad1 and 5 in postnatal HF differentiation, but not during development and hfSCs regulation. Furthermore, our findings emphasize a role of pSmad1 and pSmad5 activity in switching the premature HF characteristic from the developmental toward the mature state, demonstrating that this is an important event required to reach postnatal homeostasis and would be of great interest to investigate in the future.
We thank Dr. Anita B. Roberts (National Cancer Institute, NCI) for floxed-Smad1 and Dr. An Zwijsen (University of Leuven) for Smad5 mice; USC Animal Facility for mouse husbandry, USC FACS Core Facility for cell sorting, and the USC Imaging facility for their electron microscopy assistance. This study was supported initially by the Donald E. and Delia B. Baxter Foundation Award for K.K. and National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health Grants R01-AR061552 (to K.K.), and R03-AR061028 (to K.K.), E.K is a fellow of the California Institute for Regenerative Medicine (CIRM)—Research Training Program II in Stem Cell Biology. S.J.B is supported by a grant from the NINDS (NS063999) and A.K. by a grant from NIDCR (R21-5351568360).
E.K.: conception and design, collection and/or assembly of data, data analysis and interpretation, and manuscript writing; V.M.H. and S.J.B.: provision of study material or patients; A.K.: data analysis and interpretation; K.K.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.