|Home | About | Journals | Submit | Contact Us | Français|
Constitutive Hedgehog (Hh) signaling underlies several human tumors1, including basal cell carcinoma (BCC) and basaloid follicular hamartoma in skin2,3. Intriguingly, superficial BCCs arise as de novo epithelial buds resembling embryonic hair germs4–6, collections of epidermal cells whose development is regulated by canonical Wnt/β-catenin signaling7,8. Similar to embryonic hair germs, human BCC buds showed increased levels of cytoplasmic and nuclear β-catenin, and expressed early hair follicle lineage markers. We also detected canonical Wnt/β-catenin signaling in epithelial buds and hamartomas from mice expressing an oncogene, M2SMO9, leading to constitutive Hh signaling in skin. Conditional overexpression of the Wnt pathway antagonist Dkk1 in M2SMO-expressing mice potently inhibited epithelial bud and hamartoma development without affecting Hh signaling. Our findings uncover a hitherto unknown requirement for ligand-driven, canonical Wnt/β-catenin signaling for Hh pathway-driven tumorigenesis, identify a new pharmacological target for these neoplasms, and establish the molecular basis for the well-known similarity between early superficial BCCs and embryonic hair germs.
Hair follicle development is orchestrated by secreted signaling molecules that act on intracellular effector pathways in epithelial and mesenchymal progenitors10. The canonical Wnt signaling pathway initiates hair bud formation, whereas Hh signaling subsequently promotes the proliferative expansion of follicle epithelium required to assemble a mature follicle7,8,11,12. BCC is the most common type of cancer in light-skinned individuals, and several mutations leading to constitutive activation of the Hh pathway have been identified in these tumors, including loss-of-function mutations in PTCH1 and gain-of-function mutations in SMO2. Nearly all human BCCs show elevated Hh pathway activity, and several animal models support the notion that uncontrolled Hh signaling is sufficient to drive BCC- or BCC-like tumorigenesis in mice2. The morphological similarity between early superficial BCCs and hair germs was first noted over 70 years ago4,5, suggesting the possibility that canonical Wnt signaling is involved in pathological responses to deregulated Hh signaling in skin. Although some prior studies have identified coordinate changes in the Hh and canonical Wnt pathways in BCCs and other neoplasms13–19, direct evidence establishing the functional significance of Wnt signaling in Hh pathway-driven pathology in vivo is lacking. In this report we combined analysis of human tissues with in-depth studies of genetic mouse models to test the role of Hh-Wnt crosstalk in the setting of constitutive Hh signaling in skin.
Both human superficial BCCs and embryonic hair buds (Fig. 1) comprise a focal grouping of epidermal cells protruding into the underlying dermis (Fig. 1a, b) and express early-stage follicle lineage markers, including the outer root sheath markers K17 and Sox9, and the hair matrix/inner root sheath marker CDP (Fig. 1c, d, g–j). Epithelial cells in both the superficial BCCs and embryonic hair buds were more proliferative, based on Ki67 immunostaining, than adjacent epidermis (Fig. 1k,l) and did not express the epidermal differentiation marker K1 (Fig. 1e, f). In contrast to embryonic hair buds (Fig. 1b, arrowhead), superficial BCCs did not show a morphologically recognizable mesenchymal papilla (Fig. 1a), which is required for later stages of hair follicle morphogenesis.
We then examined epithelial bud development in mice with focally activated Hh signaling in skin, achieved using the M2SMO oncogene9 (Fig. 2). We restricted much of our analysis to a triangular region of volar skin completely devoid of follicles or other skin appendages, allowing us to study the effects of deregulated Hh signaling in a morphogenetically naive epidermis (Fig. 2a). Constitutive activation of Hh signaling using M2SMO resulted in de novo epithelial bud initiation in this normally hairless region (Fig. 2b–e). Similar to human superficial BCC and embryonic hair germs (Fig. 1), immunophenotyping revealed multiple similarities between M2SMO-induced epithelial buds and embryonic mouse hair buds (Fig. 2h–u). However, downregulation of E-cadherin, a characteristic alteration in embryonic hair buds that may contribute to a shift from membrane-bound to cytoplasmic and nuclear β-catenin, was not apparent in M2SMO-induced buds (Fig. 2v, w), and M2SMO-induced ectopic buds (like human superficial BCCs) were not associated with a morphologically or biochemically detectable mesenchymal condensate (Fig. 2g,h)).
The formation of de novo epithelial buds in response to ectopic Hh signaling in epidermis was intriguing, given the compelling evidence pointing to the canonical Wnt pathway as the initiator of skin appendage development7,8,20–22. We therefore determined whether pathological Hh signaling in this setting was associated with activation of the canonical Wnt/β-catenin pathway (Fig. 3). Increased levels of β-catenin in the cytoplasm and nucleus were seen in human embryonic hair buds (Fig. 3b), consistent with the notion that Wnt/β-catenin signaling is activated during early stages of human follicle development, as already shown in mouse hair follicles (Fig. 3d). We also detected increased levels of β-catenin in the cytoplasm and nucleus of neoplastic cells in human superficial BCCs and in M2SMO-induced mouse epithelial buds (Fig. 3a, c), suggesting activation of the canonical Wnt pathway in response to ectopic Hh signaling. Furthermore, nonphosphorylated (active) β-catenin was detected in lysates from M2SMO-transgenic mouse skin by immunoblotting, with little or no expression in control, nontransgenic volar skin (Fig. 3e). Epithelial buds arising in M2SMO-expressing mice progress to form follicular hamartomas (Supplementary Fig. 1), a type of benign tumor associated with deregulated Hh signaling in mice and humans3,23,24, but not BCCs. Follicular hamartomas expressed the same set of lineage markers as buds, but with Ki67, β-catenin and cyclin D1 largely restricted to the outermost cell layers (Supplementary Fig. 1 and Supplementary Note).
To determine if aberrant Hh signaling may be influencing the canonical Wnt/β-catenin pathway through alterations in Wnt ligand expression, we performed semiquantitative reverse transcriptase–polymerase chain reaction (RT-PCR) (Fig. 3f). As expected, mRNA encoding the Hh target genes Gli1 and Ptch1, indicating Hh pathway activation, was detected in samples from M2SMO volar skin, with negligible levels in controls. In addition, transcripts encoding multiple Wnt ligands (Wnt3, 4, 5a, 7b, 10a and 10b) and the transcriptional coactivator Tcf4, were coordinately induced in M2SMO volar skin. Expression of endogenous Wnt target genes25 Axin2 and Sp5, and the indirect target Ccnd1 was also upregulated in M2SMO-expressing skin, indicating activation of a transcriptional program associated with canonical Wnt signaling.
Using a transgenic mouse model wherein doxycycline-regulated activation of the secreted Wnt inhibitor Dkk1 could be achieved in skin26, we next tested whether canonical Wnt signaling was required for development of epithelial buds and their expansion to form hamartomas (Fig. 4). We first induced Dkk1 in M2SMO-expressing mice (M2SMO + Dkk1) during embryogenesis, by administering doxycycline to pregnant dams at embryonic day 16.5 (Fig. 4a). Blockade of canonical Wnt signaling at this stage led to nearly complete inhibition of M2SMO-driven ectopic bud development in volar skin, as assessed by whole-mount analysis and histology (Fig. 4b) of skin on postnatal day 7 (P7). Dkk1 induction at P1 (Fig. 4c) also led to a marked inhibition of M2SMO-mediated hair bud development and a profound suppression of follicular hamartoma formation, both in volar (Fig. 4d) and in tail (Fig. 4e) skin examined at P35. Although expression of Wnt genes was similarly elevated in M2SMO and M2SMO + Dkk1 skin (Fig. 4g), translocation of β-catenin to the cytoplasm and nucleus was suppressed in M2SMO + Dkk1 mice (Fig. 4f), and expression of the Wnt target genes Axin2 and Sp5 was negligible, indicating effective blockade of canonical Wnt signaling by Dkk1 (Fig. 4g). In contrast, expression of the Hh target genes Gli1 and Ptch1 was similar in M2SMO and M2SMO + Dkk1 skin, indicating that M2SMO-driven constitutive Hh signaling was unaffected by Wnt pathway inhibition (Fig. 4g). Similarly, ectopic expression of the Hh-responsive keratin, K17 (ref. 27) was seen in both M2SMO and M2SMO + Dkk1 volar skin (Fig. 4h), whereas expression of hair bud and follicular hamartoma markers Sox9 and CDP, and increased cell proliferation, were no longer detected in M2SMO + Dkk1 mice (Fig. 4h). The profound blockade of M2SMO-induced hair bud and follicular hamartoma development by Dkk1 establishes that biological responses to ectopic Hh signaling in skin are largely mediated indirectly, via the canonical Wnt/β-catenin signaling pathway.
Notably, buds and hamartomas also developed in dorsal skin of M2SMO mice, but in this location they were not suppressed in M2SMO + Dkk1 mice, and β-catenin remained localized to the nucleus, suggesting that Dkk1 was not effectively inhibiting canonical Wnt signaling at this site. This may be due to insufficiently high expression of Dkk1 in dorsal skin or a relative deficiency in the level of Kremens 1 and 2 (ref. 28), which facilitate Dkk1's ability to block Wnt signaling. Bud and hamartoma development was also not inhibited in small regions of volar skin near footpads (arrowhead in Fig. 4d), and here again the presence of nuclear β-catenin indicated inefficient blockade of canonical Wnt signaling. These results further underscore the tight correlation between inhibition of canonical Wnt/β-catenin signaling and suppression of Hh pathway-induced bud and hamartoma development.
There are multiple reports describing interactions between the Wnt and Hh pathways at various levels, and two of these are particularly noteworthy in light of our findings. Expression of Wnt5A, 7B, 7C, 8, 8B, and 11 has been reported in Xenopus animal cap explants injected with Gli2 and Gli3 mRNA, and blockade of Wnt signaling inhibits the morphogenetic response to Gli2 in this system15. Also, in E1A-immortalized RK3E rat kidney cells, expression of Wnt2b, Wnt4 and Wnt7 was induced by GLI1, and dominant-negative TCF4 inhibited GLI1-mediated focus formation in cell culture19. Although these data demonstrate that Hh-Wnt crosstalk is necessary for an embryonic process and in vitro transformation, respectively, our findings are the first to establish a stringent requirement for canonical Wnt/β-catenin signaling in Hh pathway-driven neoplasia.
Previous studies in normal skin have shown that in developing hair buds, canonical Wnt signaling precedes, and is required for, subsequent activation of Hh signaling7,8,20,21. Our data indicate that this temporal relationship is reversed in the setting of Hh-driven pathology in epidermis, where ectopic activation of Hh signaling leads to canonical Wnt signaling with resultant formation of de novo epithelial buds and follicular hamartomas. Several earlier reports have described links between the Hh and Wnt pathways in BCC, including upregulation of one or more Wnt genes15 and localization of β-catenin to the cytoplasm and/or nucleus16–18. Our results are in keeping with these observations and provide the first direct evidence that canonical Wnt signaling is essential for a tumorigenic response to deregulated Hh signaling in skin, but additional studies are required to test the significance of Hh-Wnt crosstalk in full-blown BCC. Interestingly, a recent report described a role for β-catenin in cutaneous squamous cell carcinoma29, which is biologically and pathogenetically distinct from BCC and follicular hamartomas, and not linked to aberrations in the Hh pathway, but it is not known whether signaling in squamous tumors is driven by Wnt ligands.
Taken together, our findings suggest that blockade of canonical Wnt/β-catenin signaling may be a useful strategy for treatment of neoplasms currently considered to be caused by uncontrolled Hh signaling. Because deregulated Hh signaling influences β-catenin signaling primarily at the level of Wnt ligands, the range of potential therapeutic strategies is considerably greater than it is for colorectal and other cancers with mutational defects in APC or β-catenin, and would likely include antibodies or other recombinant proteins that antagonize the interaction of Wnt ligands with Frizzled and LRP receptors. Future work will better clarify the utility of targeting proximal Wnt pathway components for the prevention or treatment of Hh-dependent neoplasms and other disorders.
Generation of transgenic mice expressing M2SMO in skin, and triple-transgenic mice expressing M2SMO and doxycyclineinducible Dkk1 expression in skin (ΔK5-M2SMO; K5-rtTA; TRE-Dkk1, designated M2SMO + Dkk1), is described in Supplementary Methods. To induce Dkk1 expression in M2SMO + Dkk1 mice, doxycycline (20 mg ml−1) was administered in drinking water with 5% sucrose, and in doxycycline-containing chow (Bio-serve, 200 mg/kg). After 3 d mice were maintained on doxycycline-containing chow but received normal drinking water. All mice were housed and maintained according to University of Michigan institutional guidelines, as stipulated by the University Committee on the Use and Care of Animals.
Human superficial BCC cells were obtained from the Cutaneous Surgery and Oncology Unit, Department of Dermatology, University of Michigan Medical School, according to an institutional review board–approved protocol (IRBMED 2000−0015). Embryonic human skin was obtained from Advanced Bioscience Resources. For hematoxylin and eosin (H&E) staining and immunohistochemistry, human and mouse skin samples were fixed in neutral-buffered formalin overnight, transferred to 70% ethanol, processed and embedded in paraffin. Mouse skin was also embedded in OCT Compound (Tissue-Tek) for frozen sections. To prepare whole mounts of volar skin, mice were euthanized and ventral hind limb skin was removed. Volar skin was microdissected and whole-mount preparation was performed essentially as described for tail skin30. Transilluminated whole-mount images were captured with a digital camera (Spot RT3; Diagnostic Instruments) mounted on a dissecting stereomicroscope (Leica MZFL3), using Spot Software Version 4.6 (Diagnostic Instruments).
The following primary antibodies were used for immunostaining: K1 (rabbit polyclonal, 1:1,000; Covance); K5 (rabbit polyclonal, 1:2,000; Covance); K17 (rabbit polyclonal, 1:4,000; gift); β-catenin (mouse monoclonal, 1:1,000; Sigma); Sox9 (rabbit polyclonal, 1:1,000; Chemicon); CDP (rabbit polyclonal, 1:100; Santa Cruz Biotechnology); Ki67 (rabbit polyclonal, 1:500; Vector Laboratories); AE13 (1:10; gift); AE15 (1:50; gift); P-cadherin (rat monoclonal, 1:500; Zymed); E-cadherin (rat monoclonal, 1:1,000; Zymed); cyclin D1 (rabbit polyclonal, 1:1,000; Neomarkers). For immunohistochemistry, 8-μm sections were cut in a parasagittal plane. For all antibodies except AE13, AE15, P- and E-cadherin, immunoreactivity of antigens was restored by immersing slides in boiling 0.01 M citrate buffer, pH ~6, for 10 min. Blocking was performed using 1.5% normal goat serum in phosphate-buffered saline, and tissue sections were incubated with primary antibodies diluted in phosphate-buffered saline containing 1% bovine serum albumin, typically for 1−3 h at room temperature (21−23 °C). Subsequent immunostaining procedures were performed using peroxidase Vectastain ABC kit (Vector Laboratories) and 3,3′-diaminobenzidine as a substrate, according to the manufacturer's protocol. A M.O.M. kit (Vector Laboratories) was used for immunostaining with mouse primary monoclonal antibodies according to the manufacturer's protocol. Sections were counterstained with hematoxylin and mounted using Permount (Fisher Scientific). For immunofluorescence staining, sections were cut from OCT-embedded blocks and fixed in cold acetone for 10 min. Fluorescein isothiocyanate–conjugated secondary antibodies (Jackson ImmunoResearch) were used at 1:75 dilution, and 100 ng ml−1 4-diamidino-2-phenylindole (Merck) was used for nuclear counterstaining. Endogenous alkaline phosphatase activity was visualized using Alkaline Phosphatase Substrate Kit I (Vector Laboratories) and incubating tissue sections with substrate solution for 2−4 h at room temperature.
Tissue was homogenized in RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.5 mM EDTA, 1 mM sodium orthovanadate, 200 mM sodium flouride) using a tissue microgrinder (Kontes 749520), vortexed vigorously, cleared by centrifugation for 15 min at 12,000g and denatured by heating at 98 °C for 5 min. A 10-μl aliquot of protein extract was resolved on a 10% (w/v) denaturing sodium dodecyl sulfate– polyacrylamide gel (BioRad) and transferred to nitrocellulose membrane. “Activated” β-catenin was detected by probing with α-activated β-catenin (αABC) antibody (mouse monoclonal, 1:1,000; Upstate). Pan-actin antibody (mouse monoclonal, 1:1,000; Labvision) was used to detect actin for loading control. Peroxidase-conjugated secondary antibodies were used and visualized by enhanced chemiluminescence using an ECL Plus kit (Amersham Biosciences).
See Supplementary Methods and Supplementary Table 1 online.
We thank Pierre Coulombe for providing K17 rabbit polyclonal antibody and Henry Sun for providing AE13 and AE15 antibodies; Eric Fearon and Deb Gumucio for constructive comments on the manuscript; and members of the Dlugosz lab for valuable input on this project. This work was supported by NIH grants R01-AR45973 and R01-CA87837 (A.A.D.), T32-HD007505 and T32-GM07863 (S.H.Y.), R01-AR47709 and R01-DE015342 (S.E.M.). For production of transgenic mice we acknowledge members of the Transgenic Animal Model Core of the University of Michigan's Biomedical Research Core Facilities, funded in part by P30-CA46592 (University of Michigan Cancer Center Core support).
Note: Supplementary information is available on the Nature Genetics website.
Experiments were designed by S.H.Y. and A.A.D. Tissue harvests, whole-mount analysis, semiquantitative RT-PCR and immunoblotting were performed by S.H.Y. T.S.W. performed collection of human BCC samples.
Immunohistochemistry and immunofluorescence staining were carried out by S.H.Y. and A.W. Genotyping was performed by S.H.Y., A.W., J.L., L.S. and J.F. M2SMO-expressing mice were initially generated by V.G. Animal maintenance and breeding were performed by S.H.Y., A.W., J.L. and J.F. TRE-Dkk1 mice were provided by T.A. and S.E.M., and K5-rtTA mice were provided by A.B.G. T.A. and S.E.M. participated in study design and discussion of the results. The manuscript was written by S.H.Y., with draft revisions by A.A.D. and input from S.E.M. and T.A.