|Home | About | Journals | Submit | Contact Us | Français|
Hair follicle (HF) regeneration begins when communication between quiescent epithelial stem cells (SCs) and underlying mesenchymal dermal papillae (DP) generates sufficient activating cues to overcome repressive BMP signals from surrounding niche cells. Here, we uncover a hitherto unrecognized DP transmitter, TGF-β2, which activates Smad2/3 transiently in HFSCs concomitant with entry into tissue regeneration. This signaling is critical: HFSCs that cannot sense TGF-β exhibit significant delays in HF regeneration, whereas exogenous TGF-β2 stimulates HFSCs in vivo and in vitro. By engineering TGF-β- and BMP-reporter mice, we show that TGF-β2 signaling antagonizes BMP signaling in HFSCs but not through competition for limiting Smad4-coactivator. Rather, our microarray, molecular, and genetic studies unveil Tmeff1 as a direct TGF-β2/Smad2/3 target gene, expressed by activated HFSCs and physiologically relevant in restricting and lowering BMP thresholds in the niche. Connecting BMP activity to an SC’s response to TGF-βs may explain why these signaling factors wield such diverse cellular effects.
Tissue homeostasis and regeneration are regulated through balancing quiescence and activation of SCs. HFs offer a unique opportunity to explore this process. Throughout adult life, they undergo dynamic, synchronized cycles of degeneration (catagen), quiescence (telogen), and regeneration (anagen) (Millar, 2002; Schmidt-Ullrich and Paus, 2005; Blanpain and Fuchs, 2009). During telogen, which can last for months, HFSCs are quiescent and reside within a specialized microenvironment called the bulge (Cotsarelis et al., 1990). Within this niche, HFSCs surround the hair shaft produced in the previous cycle. Throughout telogen, the base of the bulge, called the secondary hair germ (HG), directly abuts the underlying DP, a key signaling center for HFSCs.
The telogen→anagen transition relies upon DP-HFSC cross-talk to generate the necessary threshold of activating factors. In addition to Wnt-activating cues (Greco et al., 2009; Enshell-Seijffers et al., 2010; Rabbani et al., 2011), bone morphogenetic protein (BMP) inhibitory factors play a central role in the anagen-promoting process (Kulessa et al., 2000; Botchkarev et al., 2001; Zhang et al., 2006). Upon activation, HFSCs in the HG are the first to proliferate and initiate HF regeneration, whereas HFSCs within the bulge become active several days later (Greco et al., 2009).
As the new HF emerges, the DP stimulus is pushed increasingly further from niche SCs, which return to quiescence (Hsu et al., 2011). By contrast, throughout anagen, relatively undifferentiated bulge cell progeny along the outer root sheath (ORS) accelerate proliferation as they approach the DP. This fuels a steady production of transiently amplifying (TA) matrix cells, which undergo a few divisions while in contact with DP and then terminally differentiate to form the hair and inner root sheath (IRS). At the anagen→catagen transition, matrix cells apoptose and the DP retracts upward along with the dying/differentiating epithelial strand. As the HF reenters telogen, BMPs from the inner layer of non-SC niche cells (Hsu et al., 2011) and from surrounding dermal tissue (Plikus et al., 2008) impose a threshold, which must be overcome to initiate the next cycle.
BMPs belong to a superfamily that includes transforming growth factor βs (TGF-βs). TGF-βs function in tissue morphogenesis, homeostasis, and cancer by regulating diverse biological processes including proliferation, apoptosis, differentiation, and extracellular matrix (ECM) production (Siegel and Massagué, 2003). Skin epithelial cells express distinct Ser/Thr kinase receptors for both BMP and TGF-β pathways. These differentially propagate their respective signals by phosphorylating Smad1/5/8 (BMP) or Smad2/3 (TGF-β), which form distinct bipartite transcription factors with Smad4 (ten Dijke and Arthur, 2007). Although BMP’s inhibitory effects are well documented, the effects of TGF-β signaling on HFSCs remains largely unexplored.
TGF-βs potently inhibit proliferation of interfollicular epidermis (IFE), and postnatal loss of TGF-β signaling renders skin prone to tumorigenesis (Bierie and Moses, 2006; Guasch et al., 2007; Massagué, 2008). Conditional loss of TGF-β receptor II (TβRII-cKO) results in loss of canonical TGF-β signaling in the skin epithelium, but overall tissue morphology remains intact (Guasch et al., 2007). Intriguingly, full gene knockout studies of the receptor ligands show three different effects on embryonic HF development: delay (TGFβ2), enhancement (TGFβ1), or none (TGFβ3) (Foitzik et al., 1999). Given these complexities, we turned to the following key questions. Do TGF-βs play a role in adult HF regeneration and if so, how? Which TGF-β(s) are involved, which cells transmit the signals, and what cells are affected? How might TGF-β signaling within the niche influence the balance between HFSC quiescence and activation, and by what mechanism? Here we address these points and in so doing, shed important new light on the signaling crosstalk within the SC niche.
To determine whether and how TGF-βs might affect HF regeneration, we first verified the specificity of our anti-phospho-Smad2 antibodies. Upon TGF-β exposure in vitro, primary wild-type (WT) mouse epidermal keratinocytes (1°MKs) exhibit phosphorylation and nuclear translocation of Smad2, which can be detected by immunoblotting of nuclear extracts and immunofluorescence. This band and nuclear staining were absent in the equivalent 1°MKs from Tgfbr2fl/fl;Tg(KRT14-Cre)1Efu (TβRII-cKO) mice whose skin epithelium lacks TβRII, an essential component of the TGF-β receptor (Figures S1A and S1B available online; Guasch et al., 2007). Using WT and TβRII-cKO mice, we then turned to in vivo analyses.
During normal homeostasis, epidermis displayed little or no signs of active TGF-β signaling (not shown). Postnatally, the first signs of TGF-β signaling appeared in telogen, when pSmad2+ nuclei were detected in HG cells adjacent to the DP (Figure 1A). As follicles began cycling, nuclear pSmad2+ cells also appeared in CD34+ HFSCs at the bulge base. By the time full anagen was reached, pSmad2 immunofluorescence was no longer detected in HFSCs (Figure S1C).
The second telogen (marked by two rather than one club hairs) lasted about 4 weeks in control mice. During early and mid-telogen phases, TGF-β activity was weak. As telogen ended, pSmad2 reappeared in the P-cadherin+ HG and lower bulge (Figure 1B). Notably, pSmad2 was detected ~5 days before this proliferative activity was seen within the HG (Figure 1C). Similar to the first hair cycle, most HG cells in early anagen exhibited TGF-β signaling, but as the expanding HG engulfed the DP and formed the matrix, pSmad2 waned.
To further examine TGF-β signaling, we engineered a lentivirus harboring a canonical TGF-β-reporter driven by Smad2/3-Smad4 binding sites (Figure 1D). In vitro, reporter expression (NLS-mRFP+) was detected only in TGF-β-stimulated, transduced (H2B-CFP+) 1°MKs from WT and not in TβRII-cKO mice (Figure 1E; shown are data for TGF-β2). When transduced by in utero infection into embryonic skin epithelium (Beronja et al., 2010) and then monitored in adult mice, the TGF-β-reporter was active selectively in the WT HG, beginning at late telogen and continuing into early anagen (Figure 1F). Underscoring reporter specificity, similarly transduced TβRII-cKO animals showed no signs of reporter activity. These data revealed that nuclear pSmad2 at the telogen→anagen transition is accompanied by pSmad2-mediated active transcription in the HG.
The TGF-β/pSmad2 signaling pattern upon anagen onset paralleled the two-step activation of HFSCs described previously (Greco et al., 2009). To determine whether TGF-β signaling functions in HFSC activation, we analyzed this process in TβRII-cKO mice. By 22 days, control HFs had already entered the first anagen and showed proliferating (Ki67+) HG cells, while TβRII-cKO HFs were still in telogen (Figure 2A). Because the first WT telogen lasts only 1–2 days, the phenotypic consequences of losing TGF-β signaling to SC activation and tissue regeneration were best visualized by waiting until HFs had entered the extended second telogen. At this time, we clipped hair coats of sex-matched littermates and then followed the emergence of new hairs as the next cycle began (Figure 2B).
In the second telogen, hair growth was so delayed in TβRII-cKO mice that 75% coat recovery occurred ~1 month later than normal (Figure 2C). Despite this delay, the final lengths of the four hair types were normal (Figure 2D), as were HF morphologies and established markers (Figure 2E). Thus, once TβRII-deficient SCs were eventually activated, they executed what appeared to be an otherwise normal hair cycle.
Because pSmad2+ cells appeared at the interface between HG and DP, and after the extended crosstalk period that is required for HFSC activation, we sought to identify the TGF-β ligands involved and the cell population expressing them. A survey of existing microarray data indicated that in adult skin, TGFβ2 mRNAs might be enriched in DP relative to HG or HFSCs (Greco et al., 2009). Support for this came from real-time quantitative polymerase chain reaction (RT-qPCR) of mRNAs isolated from purified cell populations of late-telogen phase, Lef1-RFP transgenic skins. TGFβ2 mRNAs selectively were enriched in DP relative to other IFE and HF populations, whereas TGFβ1 and TGFβ3 mRNA expression was low and showed no specific enrichment (Figures 3A and S2A).
Immunofluorescence analyses were consistent with these findings. Of the three TGF-βs, only TGF-β2 protein appeared in HFs toward the end of telogen (Figures 3B and S2B). Intercellular TGF-β2 first accumulated at the ECM interface between DP and HG, but by anagen onset, it intensified in the DP, with weaker staining in HGs. Although latent TGF-β complexes can reside within ECM (ten Dijke and Arthur, 2007), TGF-β2 signal intensity in the DP coincided with nuclear pSmad2 in the HG.
To test whether TGF-β2 signaling can precociously activate resting HFSCs, we coinjected recombinant, active TGF-β2 with fluorescent beads into mid-telogen skin dermis, i.e., well before the normal appearance of endogenous TGF-β2 in DP (see experimental design in Figure 3B). In TβRII-heterozygous (WT) mice, TGF-β2 injections resulted in Smad2 phosphorylation not only in the HG, but also in IFE and some dermal cells (Figure S2C). By contrast, TβRII-cKO skin epithelium was refractory to TGF-β2- mediated Smad2 activation, while dermal cells responded. This result indicated that the effects of DP-derived TGF-β2 on precocious pSmad2 induction rely upon the HG’s ability to respond through the canonical receptors. With our TGF-β-reporter mice, we further demonstrated that the signaling involves canonical pSmad2-mediated transcriptional activation (Figure S2D).
Bromodeoxyuridine (BrdU) administration and S-phase cell quantification revealed that HGs of WT HFs reacted to TGF-β2 by displaying both nuclear pSmad2 and proliferation (Figures 3C and S2C). Although this further substantiated the sequential order of events we had observed in the normal hair cycle (see Figure 1C), it contrasted with TGF-β’s established negative growth effects on epidermal cells, which was also reflected in decreased BrdU incorporation within IFE (Figure 3C). These results indicate that IFE and HF progenitors interpret TGF-β2 differently and further underscore the specificity of the proliferative response of HFSCs to TGF-β2.
TGF-β2’s stimulatory effects depended upon close proximity of HFs to the bead injection site and were followed by a precocious hair cycle (Figures 3D, 3E, S2E, and S2F). This differed from similar injections with FGF7, another DP-derived factor that stimulated HG proliferation, but did not kindle a new hair cycle (Greco et al., 2009). Taken together, our findings indicate that in normal homeostasis, DP utilizes TGF-β2 in short-range signaling to stimulate HFSCs at the bulge base. The outcome is canonical TGF-β receptor/pSmad2 transcription that in some way enables them to achieve the activating threshold necessary to flip the telogen→anagen transition switch.
Throughout the resting phase, high BMP signaling maintains HFSCs in a quiescent state and must be overcome to promote new tissue growth (Andl et al., 2004; Blanpain et al., 2004; Kobielak et al., 2007; Plikus et al., 2008; Rendl et al., 2008; Greco et al., 2009; Hsu et al., 2011). The appearance of TGF-β2 protein in the DP toward the end of telogen coincided with the timing of enhanced mRNAs encoding DP-derived BMP inhibitory factors, including Sostdc1 and Bambi (Greco et al., 2009). Given the seemingly opposing effects of TGF-β2 and BMPs on HFSC quiescence, we wondered whether TGF-β2 might be impacting the BMP pathway, and if so how.
To address this possibility, we first evaluated the consequences of loss of TGF-β signaling on BMP receptor signaling in HFSCs (Figure 4A). In WT HFSCs, nuclear anti-phospho-Smad1/5/8 immunostaining, a sign of active BMP signaling, was observed throughout early and mid-telogen, but waned toward the end of telogen, coincident with the appearance of pSmad2 (Figure 4A; see also Figure 1B). However, in TβRII-cKO HFSCs, pSmad1/5/8 remained high throughout this time (Figure 4A), consistent with the notion that TGF-β signaling negatively affects BMP signaling.
To further explore TGF-β2’s effects, we cultured epidermal 1°MKs, where the antiproliferative effects of TGF-βs are well established (Siegel and Massagué, 2003). Indeed, all three TGF-βs promoted comparable Smad2 phosphorylation and induced growth arrest in a monophasic, dose-dependent fashion (Figures S3A–S3E). Consistent with its lower binding affinity to TβRII (Cheifetz et al., 1987; Massagué et al., 1992), ~10× more TGF-β2 (10 versus 1 pM) was necessary to achieve growth inhibition (Figure S3E). Recombinant TGF-βs had no effects on TβRII-cKO 1°MKs, thereby confirming the specificity of the response.
Next, we tested cultured HFSCs, purified by fluorescence-activated cell sorting (FACS) of mid-telogen HFs (Figure 4B). At higher concentrations (50–100 pM), TGF-βs reduced both number and size of HFSC colonies. However, at lower concentrations (~10 pM) where epidermal 1°MKs were still growth restricted, HFSC colonies grew larger. This effect was highly reproducible and not seen in TβRII-null HFSCs (Figure 4C). In good agreement with our in vivo observations, these results further unveiled the dose dependency of TGF-β’s stimulatory effects on HFSCs.
Consistent with HFSC quiescence in and out of the niche in vivo, nuclear pSmad1/5/8 staining was also greater in nascent colonies of HFSCs than in 1°MKs (Figure S3F). This was true for both WT and TβRII-null cells, but not Bmpr1a-cKO cells, which showed no signs of BMP signaling. Moreover, pSmad1/5/8 staining in WT HFSCs was reduced by exposure to TGF-β2 (Figure S3G).
Further substantiating the antagonistic effects of TGF-β2 on the BMP pathway, we found that BMP4’s effects were monophasic and could be partially relieved by TGF-β2 in a TβRII-dependent manner (Figures 4B and 4C). Importantly, neither BMP4 nor TGF-β2 affected the colony size in Bmpr1a-cKO HFSCs (Figure 4D). Moreover, even though TGF-β2 on its own no longer exerted effects on colony size as HFSCs were passaged, TGF-β2 still counteracted the negative effects of BMPs that were added exogenously to HFSC cultures (Figure S3H). Thus, the stimulatory growth effects of TGF-βs appeared to be restricted to cells exhibiting active BMP signaling.
To further pursue the notion that TGF-β2 acts positively to overcome BMP inhibitory thresholds and activate HFSCs, we compared TGF-β2’s effects to Noggin, an established and potent BMP antagonist (Botchkarev et al., 2001; Plikus et al., 2008). In early telogen, neither Noggin nor TGF-β2 were sufficient to overpower the exceedingly high BMPs within the HFSC niche and dermis. However, in mid-telogen, both Noggin and TGF-β2 precociously diminished nuclear pSmad1/5/8 and induced HG proliferation (Figures 4E and 4F). As endogenous TGF-β2 signaling and BMP signaling blockers rose and dermal BMP waves waned at telogen’s end, these effects were less pronounced. Notably, although probably not physiologically relevant, TGF-β1 and TGF-β3 behaved similarly, indicating that it is TGF-β availability and not TGF-β2 per se that is important in the process (Figure 4G).
In contrast to Noggin, TGF-β2’s effects were seen only in WT and not TβRII-cKO mice. This result was important as it demonstrated that the opposing effects of TGF-β2 on BMP signaling are mediated directly through the TGF-β receptor pathway rather than cross-counter effects on their respective receptors. Moreover, these data provided compelling complementary evidence that the atypically long telogen of TβRII-cKO mice arises from protracted BMP signaling in HFSCs.
To further probe the antagonistic interaction between TGF-β2 and BMP signaling, we engineered and tested a BMP-reporter lentivirus (Figure 5A). In vitro, transduced WT and TβRII-null, but not Bmpr1a-null, 1°MKs displayed BMP-reporter activity in response to BMP4. In vivo, WT HFSCs showed concomitant downregulation of both BMP-reporter activity and pSmad1/5/8 at telogen’s end. By contrast, TβRII-cKO HFSCs sustained BMP-reporter activity and pSmad1/5/8 during this time.
Importantly, our BMP-reporter was specific, and in the absence of BMPs, TGF-β2 had little or no effect on its expression (Figure 5B). In contrast, BMP4 strongly stimulated BMP-reporter (ZsGreen) expression. TGF-β2 dampened this activity specifically in WT and not in TβRII-cKO 1°MKs. Moreover, similar results were obtained with cultured, FACS-purified HFSCs. However, HFSCs displayed active BMP signaling in vitro, and hence TGF-β2 alone was sufficient to reduce BMP-reporter activity in HFSCs (Figure 5C). Based upon these results, TGF-β2 appeared to counteract not only extrinsic but also intrinsic BMP signaling.
The dampening effects on BMP signaling were paralleled at the mRNA level and appeared relatively late (12–24 hr) after TGF-β2 treatment (Figure 5D). With 1°MK, this effect was even more obvious when they were pretreated with TGF-β2 prior to adding BMP4 (Figure 5E). Because the reporter differs from control vector only by multimerized Smad1/5/8 binding sites, these results further bolstered the notion that TGF-β2 was specifically affecting canonical BMP signaling at the transcriptional level.
One way of achieving these effects might be through competition between pSmad2/3 and pSmad1/5/8 for a limiting amount of the Smad4 cofactor. Such mechanisms should be refractory to drugs such as cycloheximide, which effectively blocks protein synthesis within minutes in epidermal keratinocytes (Rice and Green, 1979). However, 10–100 μg/ml cycloheximide nearly abrogated the inhibitory effects of TGF-β2 on BMP-reporter expression (Figure 5F; shown are data for 10 μg/ml). Thus, TGF-β2’s antagonistic effects appeared to require new protein synthesis.
Further evidence against a Smad4 competition mechanism came from testing whether simultaneous activation of Smad2/3 interfered with the availability of transcriptional cofactor Smad4 for Smad1/5/8. Immunoblot analysis of nuclear and cytoplasmic lysates from TGF-β2, BMP, and TGF-β2/BMP-costimulated (short-term) cells showed that BMP4 stimulated phosphorylation and nuclear accumulation of Smad1/5/8 and that this did not change upon TGF-β2 costimulation (Figure 5G). Conversely, TGF-β2-enhanced phosphorylation and nuclear accumulation of Smad2/3 did not change with BMP4 costimulation. Notably, although nuclear:cytoplasmic ratios of Smad4 increased upon BMP4/TGF-β2 costimulation, there appeared to be ample Smad4 to accommodate both pathways. This was further demonstrated by coimmunoprecipitating Smad4 with Smad1 antibodies. As shown in Figure 5H, complex levels were not diminished by BMP4/TGF-β2 costimulation.
In searching for an alternative transcriptional mechanism that could explain TGF-β2’s antagonistic effects on BMP signaling, we wondered whether one of pSmad2/3-Smad4’s target genes might encode a negative regulator of BMP signaling. To test this hypothesis, we purified and transcriptionally profiled HFSC mRNAs at times prior to and during peak TGF-β2 expression in the DP.
Microarray profiling and comparative analyses were performed on duplicate sets of HG (YFP+Pcadhi), bulge HFSCs (YFP+ CD34+), and total (YFP+) cells FACS-purified from two pairs of late-telogen/early-anagen female TβRII-Het (WT) and cKO littermates (Figures S4A and S4B). As expected, the HG displayed more changes over these times than the other cell populations. Concomitant with TGF-β2 expression, 341 probes (292 genes) were specifically elevated by ≥2× in the WT HG relative to the bulge or total YFP+ cells. Shown in Table S1, this “activated HG signature” differed from prior HG array data (Greco et al., 2009) in that it was rich in, e.g., cell cycle genes associated with stem cell activation.
In comparing “preactivated” TβRII-null and WT HGs, only 40 of these genes were downregulated in TβRII-null HGs. By contrast, in comparing their “activated” states, 252 genes were downregulated by ≥2× in the TβRII-null HG (Table S2), and 239 of these were among the “activated HG signature.” Although some changes could be more closely linked to HG activation and not specifically to the absence of TGF-β2 signaling, 63 of these genes have evolutionally conserved Smad2/3-Smad4 sequence motifs within the body of the gene or 5,000 bp 5′ upstream (ECR browser [Ovcharenko et al., 2004]) (Table S3).
Several interesting findings emerged from these comparisons (Figures 6A, S4C, and S5D). TβRII ablation did not affect the master transcription factors required for HFSC/HG maintenance. This was true even for Nfatc1, which is known to play a role in HFSC quiescence downstream of BMP signaling (Horsley et al., 2008). Rather, genes relating to (1) BMP inhibition, (2) cell cycle stimulation, and (3) HF lineage-associated transcription were featured among the genes downregulated in late-telogen/early-anagen TβRII-cKO versus WT HG. Notably and in contrast to cell cycle genes, inhibitors of BMP signaling remained on the shortlist of putative TGF-β target genes: Bambi (4.5×), Bmper (3.7×), and Tmeff1 (6.7×).
After validating the microarray data by RT-qPCR, we examined temporal expression of BMP inhibitor genes after TGF-β2 stimulation in vitro (Figures 6B and 6C). Bambi has been reported as a direct target of TGF-β-Smad signaling (Sekiya et al., 2004) and has previously been identified as a possible participant in DP-HG crosstalk (Greco et al., 2009). However, of the three, Tmeff1 mRNA was not only the sixth most changed putative TGF-β target gene in the TβRII-cKO HG signature, but it also displayed more pronounced and sustained induction upon TGF-β2 treatment (Figure 6C). Therefore, we focused on this hitherto unexplored gene in skin for our remaining studies.
To test whether Tmeff1’s putative Smad2/3 binding site is indeed utilized, we conducted chromatin immunoprecipitation (ChIP) with Smad2/3 antibodies. Notably, Smad2/3 binding was enriched on the 5′ UTR of the Tmeff1 gene of WT but not TβRII-null 1°MK (Figure 6D). Moreover, this enrichment was observed only when WT cells were stimulated with TGF-β2.
Further evidence linking Tmeff1 to TGF-β2 signaling came from protein analyses, where Tmeff1 protein displayed kinetics that paralleled Tmeff1 gene induction in vitro (Figure 6E). In vivo, Tmeff1 protein appeared concomitantly with pSmad2+ cells in (1) HGs activated during normal homeostasis and (2) HGs precociously activated upon intradermal TGF-β injections (Figures 6F, 6G, and S4E). Finally, when TβRII-cKO HFSCs eventually initiated their next hair cycle, the activated HGs showed no signs of either Tmeff1 or pSmad2 (Figures 6H, S4F, and S4G). These findings link Tmeff1 directly to TGF-β2 signaling and not merely SC activation.
To investigate whether Tmeff1 is functionally relevant to the TGF-β2-mediated downregulation of BMP signaling, we performed lentiviral Tmeff1-shRNA-mediated gene knockdown experiments in BMP-reporter-transduced 1°MK. 1°MK were exposed to BMP4/TGF-β2 as before, so that control cells would maintain TGF-β2 signaling-dependent suppression of BMP-reporter activity. In striking contrast to scrambled-shRNA where BMP-reporter (ZsGreen) remained silent, cells transduced with Tmeff1-shRNA were green (Figure 7A). By contrast, albeit somewhat less effective at knocking down their targets, Bambi and Bmper-shRNA were considerably less potent at rescuing BMP-reporter activity (Figures S5A and S5B).
In BMP4/TGF-β2-treated cells transduced with Tmeff1-shRNA, endogenous Tmeff1 expression was kept low, and BMP-reporter activity remained elevated. In contrast, Tmeff1 levels did not affect TGF-β-reporter activity or Smad2 phosphorylation (Figures S5C and S5E). Moreover, Tmeff1 appeared to be a key mediator of the specific antagonistic effects of TGF-β2 on BMP-reporter activity, because introducing mCherry- or FLAG-tagged Tmeff1 into BMP4-stimulated cells resulted in strong suppression of both Smad1/5/8 phosphorylation and BMP-reporter activity (Figures 7B, S5D, and S5F).
To assess the physiological relevance of Tmeff1, we employed our powerful in utero lentiviral delivery system, which enables stable, efficient, and selective shRNA-knockdown in skin epithelium (Beronja et al., 2010). We transduced mice with two different Tmeff1-shRNAs and a scrambled control shRNA. Both Tmeff1-shRNAs behaved similarly, ruling out off-target effects. For the purposes here, we show representative skin sections from mice transduced with the most effective Tmeff1-shRNA.
At P21, when HFs were in their short first-telogen, Tmeff1 protein was detected only in HGs of scrambled-shRNA- and not in HGs of Tmeff1-shRNA-transduced mice (Figure 7C). When adult mice were shaved at the start of their first postnatal anagen (P23–P24), scrambled-shRNA-transduced hair coats regrew with kinetics similar to uninfected littermates, whereas Tmeff1-shRNA-transduced animals exhibited a significant delay in hair regrowth (Figures 7D and 7E). By P43, when scrambled-shRNA-transduced HFs had already entered into their second-telogen (two hairs/HF), Tmeff1-shRNA-transduced HFs were still in the midst of their first anagen (Figure 7F). Although the extent of these differences varied somewhat with individual infections, the trend was consistent, underscoring Tmeff1 as a downstream target of TGF-β signaling and important mediator of SC activation at the telogen→anagen transition.
TGF-β2 was previously shown to be positively involved in HF development during embryogenesis (Foitzik et al., 1999; Jamora et al., 2005). Our studies not only extend this to the postnatal hair cycle but further bolster the view set forth by Foitzik et al. (1999) that the responsiveness to TGF-β2 might be different between HFSCs and IFE keratinocytes. Even under the same culture conditions where TGF-β2 showed a positive effect on HFSCs, IFE 1°MKs were still growth restricted, suggesting that intrinsic features underlie these differences. Our results now show that extrinsic and intrinsic BMP signaling is higher in HFSCs than IFEs, constituting a physiological root of their strikingly different responses to TGF-βs.
At the start of telogen, intradermal BMP levels are high but then gradually decline (Plikus et al., 2008). Conversely, BMP inhibitors produced by the DP rise over telogen, thereby reducing the threshold that maintains SC quiescence. Our findings now show that TGF-β2 features prominently in dampening BMP signaling and facilitating HFSC activation. Because TβRII-cKO mice eventually enter into the next hair cycle after an unusually prolonged telogen phase, we are able to draw two additional important conclusions: first, that TGF-β2’s function is specific to the SC activation step; and second, that SC activation is not governed by a single factor or cell type but rather is exquisitely sensitive and responsive to the comprehensive environment of the HFSC niche.
The HFSC niche is also the residence for melanocyte stem cells (MSCs), which depend upon TGF-β signaling for their maintenance (Nishimura et al., 2010; Tanimura et al., 2011). TGF-β1/2 are present in the IRS cells that appear in early anagen (Tanimura et al., 2011), and we confirmed that pSmad2 is expressed in mature IRS (Figure S1C). However, our findings now add two intriguing new facets to this equation: first, TGF-β2 signaling acts in a short-range and concentration-dependent fashion, as judged by our bead injection and in vitro experiments; and second, DP is the major source of TGF-β2, expressed specifically and transiently at the quiescence→activation transition. Given that DP abuts both HG and MSCs at a stage when they both become activated, it will be interesting in the future to see how this dynamic interplay unfolds. Additionally, while our studies have concentrated on our discovery that TGF-β2 signaling negatively influences BMP signaling in the HFSCs, other signaling pathways, e.g., Wnts and Shh, could be impacted either directly or indirectly, and this could contribute to the TGF-β-related cell-type differences that we and others have observed in the skin.
BMP signaling negatively regulates SC proliferation in many niches including the bulge (Zhang et al., 2006; Kobielak et al., 2007; Plikus et al., 2008) and intestinal crypt (Haramis et al., 2004). In both of these cases, BMP signaling appears to suppress Wnt signaling. Although there are multiple means of antagonizing Wnt signaling through BMP signaling (He et al., 2004; Jian et al., 2006; Liu et al., 2006; Zhang et al., 2006; Kobielak et al., 2007), restricting BMP signaling appears to be a key step in promoting SC activation for tissue regeneration. Our discovery of a novel paracrine role for TGF-β2 in dampening BMP signaling unearths a newfound and early importance for TGF-βs at this critical crossroad in SC biology.
We were prompted to delve deeper into the underlying mechanisms when we realized that TGF-β2’s communication circuit is launched just prior to anagen onset when HFSCs undergo activation and proliferaton. We initially speculated that competition between pSmad1/5/8 and pSmad2/3 for limiting Smad4 might underlie the antagonistic effects we observed. As negative support for this hypothesis mounted, we turned to an alternative hypothesis, namely that a direct Smad2/3 target gene(s) might encode an inhibitor of BMP signaling.
Consistent with the tight link between TGF-β2 signaling and SC activation, our microarray and comparative analyses of late-telogen/early-anagen TβRII-cKO versus Het HGs strongly overlapped with our activated HG signature. Even though many of the shared differentially expressed genes possessed Smad2/3-Smad4 binding motifs, the list was not particularly enriched for such genes, suggesting that only a few may be relevant to TGF-β2’s antagonistic effects on BMP signaling. In this regard, Tmeff1 (also known as tomoregulin-1) surfaced and stayed at the forefront.
Tmeff1 is a transmembrane protein containing two follistatin domains and an epidermal growth factor-like domain in its extracellular region. Although comparatively little is known about this protein, several tantalizing reports relate to our findings: first, Tmeff1 blocks BMP2-mediated mesoderm induction in the Xenopus embryo, and second, it inhibits signaling of another TGF-β superfamily member, Nodal, through direct binding to the Nodal coreceptor Cripto (Chang et al., 2003; Harms and Chang, 2003). Finally, although the epithelial-mesenchymal transition gene Snail, important in embryonic HF morphogenesis (Jamora et al., 2005), did not surface as a relevant TGF-β2 target in the adult HFSC niche, it is intriguing that like Snail family members, Tmeff1 is commonly upregulated during epithelial-mesenchymal transitions (Moreno-Bueno et al., 2006).
Our studies now demonstrate that Tmeff1 is a direct target of TGF-β signaling. They also uncover a hitherto unrecognized link between TGF-β signaling and the opposing effects of Tmeff1 on BMP signaling at least at Smad1/5/8 phosphorylation events. Given the widespread importance of BMP signaling in controlling adult SC quiescence and the need to restrict it upon SC activation, it will be interesting in the future to explore the extent to which the mechanisms we have uncovered here might represent a common theme in SC biology.
In closing, increasing numbers of reports have surfaced over recent years describing both synergistic and antagonistic effects between TGF-β and BMP signaling pathways (Wang and Hirschberg, 2003; Yew et al., 2005; Izumi et al., 2006; Giacomini et al., 2009; Wrighton et al., 2009; Keller et al., 2011). Indeed, these two cognate pathways seem to be intertwined frequently, although the outputs from their crosstalk are unpredictable. Our findings now provide a functionally important example of this interplay in SC biology and epithelial-mesenchymal crosstalk. In sifting through the possible mechanisms and dissecting physiologically relevant players, the pathway guided us to an interesting example of a target gene induced selectively by one of these signaling pathways that can potently impact the other signaling pathway.
TβRII floxed mice (Levéen et al., 2002) were crossed to K14-Cre (Vasioukhin et al., 1999) and/or Rosa26YFPlox/stop/lox (Srinivas et al., 2001) mice. Other mice were as published except for the Bmpr1afl/fl;K15-CrePGR;Rosa26YFP+/fl cross. CD1 mice were from Charles River laboratories. Recombinant human TGF-β1, -β2, and -β3 (50–100 ng, R&D systems), mouse Noggin (200 ng, R&D systems), or BSA control was intradermally injected with FluoSpheres (Invitrogen). BrdU (50 μg/g) was injected intraperitoneally 12 and 24 hr before lethal administration of CO2. Ultrasound-guided lentiviral injection and related procedures have been described (Beronja et al., 2010). Two controls were used as comparisons to knockdown mice: age- and sex-matched uninjected littermates and embryos infected with a nontargeting scrambled-shRNA, which activates the endogenous microRNA processing pathway but is not known to target any gene. All animals were maintained in an AAALAC-approved animal facility and procedures were performed with IACUC-approved protocols.
Data were analyzed and statistics performed (unpaired two-tailed Student’s t test) in Prism5 (GraphPad). Significant differences between two groups were noted by asterisks or actual p values.
We thank S. Karlsson for floxed-TβRII mice; D. Padua and S. Tavazoie for information about pSmad2 antibodies; N. Stokes and D. Oristian for in utero lentiviral injections and assistance in the mouse facility; S. Beronja and Y.C. Hsu for critical reading of the manuscript; and S. Williams, M. Schober, T. Chen, M. Kadaja, and other E.F. laboratory members for discussions. We appreciate the assistance of S. Mazel in the RU Flow Cytometry Resource Center, the Comparative Bioscience Center (an American Association of Accreditation of Laboratory Animal Care facility) for expert handling and care of the mice, and Memorial Sloan Kettering Genomics Core Facility for RNA and microarray processing. N.O. was supported by International Human Frontier Science Program Organization and is currently supported by the Japan Society for the Promotion of Science. E.F. is an investigator of the Howard Hughes Medical Institute. This work was supported by grants to E.F. from National Institutes of Health (R01-AR31737-27, -28, -29 and R01-AR050452-06, -07, -08) and Emerald Foundation.
Microarray data have been submitted to NCBI-GEO under accession number GSE33471.
Supplemental Information includes Supplemental Experimental Procedures, five figures, and three tables and can be found with this article online at doi:10.1016/j.stem.2011.11.005.