A Novel Mechanism for Activation of GLI1 by Nuclear SMO That Escapes Anti-SMO Inhibitors

Basal cell carcinoma (BCC) is the most common skin cancer, with several clinical subtypes. BCC commonly manifests as a nonaggressive and slow-growing tumor histologically categorized as nodular or superficial. Micronodular and infiltrative subtypes are more aggressive and may be locally destructive if left untreated (1). BCC incidence is associated with exposure to ultraviolet radiation, commonly developing in elderly people especially on the head and neck (1). Hedgehog (HH) signaling was first linked to BCC due to loss-of-function mutations in the Ptch1 gene in patients with nevoid BCC syndrome (NBCCS) or Gorlin syndrome, an autosomal dominant disorder predisposing sufferers to early onset BCC, among other developmental defects (2–4). Loss of PTCH1 function leads to constitutive activation of the GLI1 and GLI2 transcription factors via a pathway requiring the transmembrane protein Smoothened (SMO; ref. 1). Up to 70% to 80% of sporadic BCCs have loss-of-function mutations in PTCH1, 6% to 21% harbor activating mutations in SMO and GLI1, and GLI2 is rarely mutated (3, 5, 6).

Our knowledge of HH signaling in BCC stems mainly from mouse models, but this is constrained as Ptch1^−/− mice are embryonic lethal. However, Ptch1^+/− mice develop follicular hamartomas with BCCs arising after ultraviolet irradiation (7), whereas conditional Ptch1 knockout mice develop BCC-like lesions with tumors displaying nuclear GLI1 expression, indicative of HH pathway activation (8). In addition, overexpression of Gli1 leads to BCC-like epidermal tumor formation, and targeted expression of an active mutant of Gli2 to the basal layer of murine epidermis is sufficient to sustain BCC-like lesions (9, 10). Regarding SMO, conditional expression of an active SMO mutant (M2) induces BCC-like tumors when activated in the murine basal layer (11). Intriguingly, reports show that SMO-M2 induced follicular hamartomas and not BCC-like lesions, concluding that, by comparison with the K5-GLI2ΔN mouse model (10), GLI1 and GLI2 expression was insufficiently high to initiate the more malignant tumor form (12). To our knowledge, GLI1 and GLI2 expression levels have not been compared between SMO-M2 follicular hamartomas and irradiated Ptch1^+/− or conditional Ptch1^−/− BCC-like tumors, but although it is difficult to compare different mouse studies, if loss of PTCH1 alone is sufficient to induce BCC-like lesions (8) while ectopic SMO-M2 is not (albeit at low levels; ref. 12), this suggests that SMO-independent mechanisms may contribute to carcinogenesis in tumors associated with loss of PTCH1 function. Indeed, such a hypothesis was proposed to explain why, unlike irradiated Ptch1^+/− mice, SMO-M2 tumors do not arise from the hair follicle (13). However, the same study also concluded that tumors may not arise in ears or tails of irradiated Ptch1^+/− mice because of low SMO expression, and thus, the canonical signal cannot be transduced.

The importance of HH signaling in BCC development is supported by FDA approval of the anti-SMO compounds vismodegib and sonidegib to treat advanced/metastatic BCC. Complete responses have been documented, and the overall response rate for locally advanced and metastatic BCC is around 50% (14). However, this also suggests that in the absence of a drug bioavailability issue, many BCCs are driven by SMO-independent mechanisms or that some tumors are heterogeneous such that their progression is not solely driven by SMO-dependent signaling.

To address these issues and to investigate the HH signaling pathway further in BCC, we analyzed the effects of suppressing the PTCH1 gene in immortalized human keratinocytes. Our results show that the majority of transcriptome changes observed upon PTCH1 suppression are independent of SMO activity. In addition, we present evidence that the increase of GLI1 transcription in PTCH1 knockdown cells is associated with nuclear SMO function, a mechanism insensitive to pharmacologic inhibition. Nuclear SMO was also observed in human and, notably, mouse BCCs, although the pattern of expression was heterogeneous. In summary, this study identifies a novel mode of SMO-GLI1 signaling as well as revealing the presence of SMO-independent signaling mechanisms downstream of PTCH1 that may be relevant to BCC biology.

Two immortalized keratinocyte cell lines were used for this study: NEB1 and N/Tert cells (obtained in 2009). Cells were cultured in Alpha MEM and 10% (v/v) FBS (Lonza), 2 nmol/L l-glutamine, 2% (v/v) penicillin–streptomycin (PAA Laboratories), and keratinocyte growth medium supplement consisting of 10 ng/mL EGF, 0.5 μg/mL hydrocortisone, 5 μg/mL insulin, 1.8 × 10⁻⁴ mol/L adenine, and 1 × 10⁻¹⁰ mol/L cholera toxin (Sigma). SMO inhibitors KAAD-Cyc and SANT-1 (Merck) were used at 100-nmol/L and 1-μmol/L concentrations from 24 to 96 hours. Cells were negative for Mycoplasma, tested using MycoAlert (Lonza). Cells were retrovirally transduced with PTCH1 small hairpin RNAs (shRNA) targeting exon 3 (AAGGTGCTAATGTCCTGACCA) and exon 24 (AAGGAAGGATGTAAAGTGGT). A nontargeting control sequence was used (GCGCGATATATAGAATACG). The sequences were cloned into the pSUPERIOR.retro.puro vector (OligoEngine). Retrovirally transduced cells were then selected with 1-μg/mL puromycin (Sigma) for a week and cultured normally thereafter, from which clonal cell lines were derived.

To generate the EGFP-SMO/M2 constructs, SMO-M2 was excised from the PRK-SMO vector (6). SMO-M2 contains an activating point mutation of W535L (6). Purified DNA products were cloned into pEGFP-C2 (Clontech Laboratories). To generate the EGFP-NLS-SMO construct, the predicted N(o)LS sequence of SMO was cloned into pEGFP-C2. SMO-NLS-F-5′ GAGGCAGAGATCTCCCCAGA 3′ and SMO-NLS-R-5′ ATATGGATCCGTGTGGAGGAAGAAGGAG 3′ oligonucleotides were used. The purified SMO N(o)LS DNA was ligated into pEGFP-C2. Site-directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies) was utilized to modify the N(o)LS of SMO according to the manufacturer's instructions. Mutagenesis primers used were as follows: ESMOmNLS-F-5′ CTTCTTCCGGGCCAGGGCCGCCTGCAGTCGAGATCTGA 3′ and ESMOmNLS-R-5′ TCAGATCTCGACTGCAGGCGGCCCTGGCCCGGAAGAAGA 3′ that mutate K679A and K680A, respectively (Supplementary Fig. S1B). GLI1 firefly luciferase reporter pGL3-6GBS (15) combined with a pCMV-Renilla luciferase normalization vector was transfected into keratinocytes using FuGENE 6. Cells were harvested 24 hours after transfection and RNA extracted for qPCR analysis.

RNA was extracted from cells using the RNeasy Mini Kit (Qiagen). cDNA was then synthesized using the Superscript VILO cDNA Synthesis kit (Invitrogen). Note that 100 ng of cDNA per reaction was used for qPCR analysis using the Rotor-Gene SYBR Green PCR Kit (Qiagen). The Rotor-Gene 2000 machine (Qiagen) was used for the analyses with the GAPDH reference gene. Primer sequences for qPCR were: AGR2: F-5′ GGGATGGAGAAAATTCCAGTG 3′, R-5′ GGGTACAATTCAGTCTTCAG 3′; AMOT: F-5′ CATGGAGGGCAGGATTAAGA 3′, R-5′ TCGTCTCGCTTTTCTTCCAT 3′; CXCL11: F-5′ GTGCTACAGTTGTTCAAGGC 3′, R-5′ CTAGGTTTTTCAGATGCTCT 3′; FBN2: F-5′ AATGTGGGTCTCAACCTTCG 3′, R-5′ CTGTAGCCACCCAGGATGTT 3′; GAPDH F-5′ GTGAACCATGAGAAGTATGACA 3′, R-5′ CATGAGTCCTTCCACGATACC 3′; GFP: F-5′ TATATCATGGCCGACAAGCA 3′, R-5′ GAACTCCAGCAGGACCATGT 3′; GLI1: F-5′ GAAGACCTCTCCAGCTTGGA 3′, R-5′ GGCTGACAGTATAGGCAGAG 3′; GLI2: F-5′ TGGCCGCTTCAGATGACAGATGTTG 3′, R-5′ CGTTAGCCGAATGTCAGCCGTGAAG 3′; luciferase: F-5′ AGTGCTCATCATCGGGAATC 3′, R-5′ CATCCAACATTTTCGTGTCG 3′; MMP2: F-5′ AGGGCACATCCTATGACAGC 3′, R-5′ ATTTGTTGCCCAGGAAAGTG 3′; MMP9: F-5′ TTGACAGCGACAAGAAGTGG 3′, R-5′ TCACGTCGTCCTTATGCAAG 3′; PTCH1: F-5′ ACTCGCCAGAAGATTGGAGA 3′, R-5′ TCCAATTTCCACTGCCTGTT 3′; SMO: F-5′ GGCTGCTGAGTGAGAAG 3′, R-5′ CTGGTTGAAGAAGTCGTAGAAG 3′; SNAI1: F-5′ TTTACCTTCCAGCAGCCCTA 3′, R-5′ CCCACTGTCCTCATCTGACA 3′; THY1: F-5′ CCCAGTGAAGATGCAGGTTT 3′, R-5′ GACAGCCTGAGAGGGTCTTG 3′; VIM: F-5′ CCCTCACCTGTGAAGTGGAT 3′, R-5′ CAACCAGAGGGAGTGAATCC 3′.

A total of 15,000 cells per well were seeded in 12-well plates on 18-mm diameter glass coverslips (VWR International) for 3 days. Cells were fixed with 4% (w/v) paraformaldehyde and permeabilized with 0.1% (w/v) Triton X-100 (Sigma). Note that 3% (w/v) BSA (Fisher Scientific) was used for blocking and primary antibodies diluted to 1:1,000 in 3% (w/v) BSA. Primary antibodies used for this study were fibrillarin-C13C3 (Cell Signaling Technology), GLI1-C18, GLI1-H300, PTCH1-C20, SMO-C17, SMO-N19 (Santa Cruz Biotechnology), and SMO-ab72130 (Abcam). Fluorescence dye-labeled Alexa 488/568 (Invitrogen) secondary antibodies were diluted to 1:800 in PBS. Cell nuclei were stained with 0.1 μg/mL of DAPI (Sigma). Coverslips were mounted onto microscope slides using VECTASHIELD mounting medium (Vector Laboratories). Images were taken on the Zeiss LSM 510 META confocal microscope and quantified using ImageJ software.

SMO protein expression was examined in 29 biopsies of human BCC, 14 invasive and 15 noninvasive samples. Ptch1^+/− γ-irradiated mice tumors were kindly provided by Anna Saran (ENEA). Tissue sections (4 μm) were cut from paraffin blocks and transferred to microscope slides. Slides were run on the Ventana Discovery platform (Ventana Molecular Discovery Systems, Roche Diagnostics) with an automated protocol, including deparaffinization, antibody incubations, DAB application, hematoxylin and bluing reagent counterstaining, and finally slide cleaning. Following dehydration in a series of solvents, the slides were mounted with coverslips.

Gene expression microarray profiling was performed on NEB1-shCON and NEB1-shPTCH1 cells lines using Human Gene 1.1 ST Array Strip (Affymetrix). Each sample was repeated in triplicate, and the data normalized using the MetaCore pathway analysis software (GeneGo). Three Affymetrix chips were used for each sample, and the raw data were processed using the multiarray average (RMA) method. Probabilities of gene expression between the experimental and control group were generated using the Wilcoxon signed-rank test. The Tukey biweight method was used to obtain log ratios that were then antilogged to generate fold-change values comparing genes between samples. Genes with a P value at or below 5% and a fold change greater than ±2 fold are considered as differentially expressed genes (DEG). DEGs were analyzed in MetaCore using its GeneGo database of cellular pathways and process networks.

To understand the role of PTCH1 in keratinocyte biology and BCC formation, we used shRNA to suppress PTCH1. Human skin keratinocyte cell lines (NEB1 and N/Tert) were retrovirally transduced with PTCH1 shRNA targeting exons 3 or 24 and normalized against a nontargeting control vector (shCON). Clonal cell lines were generated from heterogeneous populations of the PTCH1 knockdown cells and validated by qPCR; exon 24 clone 1 (C1) of NEB1 cells displayed the strongest level of PTCH1 suppression and the highest level of GLI1 mRNA expression (Supplementary Fig. S1A), which was confirmed at the protein level as well as by increased GLI reporter activity (Fig. 1A and C). PTCH1 protein expression was predominantly nuclear in shCON cells, which may represent the C-terminal fraction previously described (16). In addition to the upregulation of GLI1, suppression of PTCH1 reduced SHH and IHH mRNA levels, whereas SMO was increased, but there was no significant effect upon GLI2 expression (Fig. 1B). To confirm that increased GLI1 resulted directly from reduced PTCH1 activity, we made use of the fact that the shPTCH1 construct targets downstream of the STOP codon at the end of exon 23. Ectopic expression of the coding sequence of only the PTCH1-1B (PTCH1_L) isoform suppressed increased GLI1 mRNA expression by approximately 50% in NEB1-shPTCH1 cells (Fig. 1D); higher suppression was probably hampered by PTCH1-1B appearing to induce an apoptotic response in some cells, as described in 293T cells (17). Morphologically, shPTCH1 cells formed holoclone-like colonies, characteristic of nodular/micronodular BCC tumor islands with smaller nuclei, compared with shCON cells also containing more nucleoli (Fig. 1E and F).

In addition to the increase in SMO mRNA (Fig. 1B), an increase of SMO protein expression was observed by immunocytochemistry in shPTCH1 cells compared with shCON using a commercial N-terminal antibody (N-19; Santa Cruz Biotechnology). More specifically, SMO displayed cytoplasmic and nuclear localization in shCON cells, with a comparable but more intense fluorescent pattern in shPTCH1 cells (Fig. 2A). A similar pattern of fluorescence was observed using another antibody targeting the C-terminus (C-17; Santa Cruz Biotechnology), which provides further evidence for the presence of nuclear SMO in human keratinocytes (Fig. 2A). SMO N-19 antibody specificity was confirmed by a reduction of the fluorescent signal in cells transfected with SMO siRNA (Fig. 4D).

To our knowledge, nuclear SMO has not been described before; however, in contrast to the Drosophila homolog, human and mouse SMO contain a C-terminal region predicted to harbor a monopartite nuclear/nucleolar localization signal, or N(o)LS (Supplementary Fig. S1B). Tagging EGFP with the putative N(o)LS sequence targeted the protein to the nucleolus based on fluorescent colocalization with fibrillarin. A full-length EGFP–SMO fusion protein also localized to the nucleus and occasionally to the nucleolus. Mutational inactivation of residues within the N(o)LS region mNLS2 (Supplementary Fig. S1B) localized EGFP-SMO exclusively to the cytoplasm (Fig. 2B) and impaired its ability to increase GLI1 mRNA levels (Fig. 2C).

In addition to the full-length SMO above, we also expressed the constitutively active SMO-M2 mutant which showed much higher levels of GLI1 expression (Supplementary Fig. S1C). Mutation of the NLS2 in SMO-M2 also significantly downregulated GLI1. Interestingly, mutation at another site, mNLS1, within the N(o)LS (Supplementary Fig. S1B) did not influence the ability of EGFP-SMO or EGFP-SMO-M2 to activate GLI1 expression (thus providing an internal negative control). The fact that GLI1 levels were reduced with mNLS2 in both shCON and shPTCH1 cells and for both EGFP-SMO and EGFP-SMO/M2 provides very strong evidence for a specific role of nuclear SMO in GLI regulation (Supplementary Fig. S1C).

SMO protein expression was examined in a panel of BCCs to determine its subcellular localization: The SMO-N19 antibody was employed, having confirmed its specificity for SMO by immunocytochemistry (Figs. 2A and 4D). In human BCCs, the staining pattern was predominantly cytoplasmic and membranous with some evidence of nuclear expression (Fig. 3A). In contrast, mouse BCC tumors derived from Ptch1^+/− γ-irradiated mice showed clear evidence of nuclear SMO, although heterogeneous staining was also observed (Fig. 3B).

We assessed the canonical PTCH–SMO–GLI1 signaling axis in shPTCH1 cells by employing SMO pharmacologic inhibitors. Twenty-four hours after exposure, GLI1 protein and mRNA expression were strongly suppressed in shCON cells, whereas no changes were observed in shPTCH1 cells treated with 100 nmol/L KAAD-Cyc (Fig. 4A and B). The same lack of response was observed in shPTCH1 cells treated with 100 nmol/L SANT-1 (Fig. 4C); in addition, no change in GLI1 expression was observed upon prolonged exposure to either inhibitor (up to 96 hours). To validate these results, other NEB1 and N/Tert-shPTCH1 clones were examined, targeting exons 3 and 24. Again, elevated GLI1 expression was suppressed by neither KAAD-Cyc nor SANT-1, indicating that the increase of GLI1 observed upon PTCH1 suppression in NEB1 and N/Tert human keratinocytes is insensitive to SMO pharmacologic inhibition (Supplementary Fig. S1D and S1E). To investigate the SMO–GLI1 signaling axis in more detail, shPTCH1 cells were treated with siRNA-targeting SMO (siSMO). Intriguingly, GLI1 protein expression was reduced in both shCON and shPTCH1 cells, suggesting that SMO is required for GLI1 signaling (Fig. 4D). When comparing the effects of KAAD-Cyc and siSMO in shPTCH1 cells, GLI1 mRNA expression was reduced by siSMO, but not KAAD-Cyc, whereas GLI2 expression was reduced by both methods (Fig. 4E). Although GLI2 expression was not elevated in shPTCH1 cells, GLI2 levels were suppressed by KAAD-Cyc in both shCON and shPTCH1 cell lines, thus supporting the concept of a canonical SMO–GLI2 signaling axis independent of PTCH1 control.

To characterize the global effects of PTCH1 suppression, shCON and shPTCH1 cells were subject to cDNA microarray profiling. Furthermore, to determine if the effects of PTCH1 suppression are sensitive to pharmacologic SMO inhibition, shPTCH1 cells were additionally exposed to 100 nmol/L KAAD-Cyc for 24 hours prior to RNA harvesting. In comparison with shCON cells, expression levels of 213 transcripts were altered ≥2-fold (137 increased and 76 decreased, P < 0.01) in shPTCH1 cells (Fig. 5A). However, of the 76 downregulated genes, the expression of 41 genes was further reduced by KAAD-Cyc treatment, whereas the expression of 47 of the 137 upregulated genes was further increased (Fig. 5B). Exposure to KAAD-Cyc failed to normalize the expression of the majority of transcripts to basal levels, with the inhibitor often exacerbating the effects of PTCH1 suppression. Only two upregulated DEGs, CXCL10 and SNORA38B, were reduced by 50% upon KAAD-Cyc treatment.

Figure 5A contains transcripts associated with tumor biology, including the cancer stem cell marker THY1 (↑6.3-fold), epithelial–mesenchymal transition (EMT) markers VIM (↑3.6-fold) and SNAI1 (↑2.4-fold), and matrix metalloproteases MMP7 (↑9.5-fold), MMP2 (↑4.1-fold), and MMP9 (↑2.2-fold). In addition, the chemokines CXCL10 (↑8.2-fold) and CXCL11 (↑4.9-fold) proved to be of interest, as both are known to be elevated in BCC (18). Of the suppressed transcripts, FBN2 (↓10.2-fold), AMOT (↓3.7-fold), and AGR2 (↓3.3-fold) have been associated with tumor biology. Excluding MMP7 and CXL10 (not detectable with two primer sets each), the differential expression of the transcripts listed above was confirmed by qPCR in shPTCH1 cells (Fig. 5C). However, elevated SNAI1 expression was partially suppressed by KAAD-Cyc, whereas there was a potent (albeit potentially undesirable) increase of MMP9. Finally, a similar but reduced pattern of differential expression was observed for these transcripts in the shPTCH1-ex3 clone as well as in one N/Tert-shPTCH1 clonal cell line (Supplementary Fig. S1F).

The purpose of this study was to identify novel targets downstream of PTCH1 relevant to BCC biology and regulated by SMO-dependent and SMO-independent mechanisms. PTCH1 was stably suppressed in immortalized human keratinocytes, and clonal cell lines were generated. The fact that clonal cell lines with strong PTCH1 suppression adopt a holoclone-like morphology characteristic of BCC nodular islands supports the validity of the model. We have previously shown that retroviral GLI1 expression promotes tight N/Tert colony formation, and endogenous GLI1 may help promote this phenotype in the shPTCH1 model (19). However, the level of ectopic GLI1 is considerably higher than that induced upon PTCH1 suppression, and it is likely that additional factors regulate colony formation in shPTCH1 cells. Interestingly, while validating the shPTCH1 model with regard to canonical PTCH/SMO/GLI signaling, some observations do not correlate with current dogma, i.e., PTCH1 suppression led to an increase of GLI1, but not GLI2, and an increase of GLI1 correlated with increased nuclear SMO, which was unresponsive to anti-SMO pharmacologic agents.

With regard to GLI2, it has been shown to activate GLI1, but using an artificial active mutant and not the wild-type protein (20). GLI2 mouse models of BCC also employed the active mutant (21), and although elevated GLI2 levels have been observed in human tumors, the role of the wild-type protein remains unclear. GLI2 was not increased in shPTCH1 cells (Fig. 1B), which indicates that GLI2 expression is PTCH independent. Despite this, GLI2 was suppressed upon KAAD-Cyc treatment, revealing that it is controlled by SMO (Fig. 4E). It is also possible that PTCH1 suppression permits GLI2 nuclear localization and subsequent activation of GLI1, but this does not correlate with the fact that GLI2 was suppressed by anti-SMO inhibitors. Another consideration is the nuclear localization of PTCH1 in NEB1-shCON cells (Fig. 1A), which was detected using a C-terminal antibody. PTCH1 can be cleaved with the C-terminal fragment (PTCH1-C) residing in the nucleus to negatively regulate GLI1 (16). Based upon our immunohistochemistry data, it is likely that suppression of PTCH1-C permits the increase of endogenous GLI1 (Fig. 1A–D). Irrespective of the role of GLI2 or PTCH1, SMO control of GLI1 was confirmed by a reduction of GLI1 in the presence of SMO siRNA in both shCON and shPTCH1 cells (Fig. 4D and E). By contrast, anti-SMO inhibitors only suppressed GLI1 levels in shCON cells. It is unlikely that failure to suppress GLI1 was due to the increase of SMO in shPTCH1 cells, as GLI1 levels remained elevated in the presence of high drug concentrations, and therefore, we hypothesize that nuclear SMO regulates GLI1 through a mechanism that is unresponsive to anti-SMO inhibition.

Indeed, the most intriguing aspect of the shPTCH1 model concerns the localization of SMO. SMO expression for both RNA and protein was increased. The reliability of anti-SMO antibodies has been much debated in the HH field; however, a similar expression pattern was observed with three antibodies (SMO-N19, C-17, and ab72130; Supplementary Fig. S1G), which was suppressed upon SMO siRNA treatment (Fig. 4D). We are unaware of any studies that have tested anti-SMO antibody specificity using siRNA, but nuclear SMO expression was recently described in cancers of unknown primary origin, and this significantly correlated with nuclear GLI1 and nuclear β-catenin (22). Strong SMO expression was observed in nodular BCC, which was predominantly cytoplasmic but with clear areas of nuclear staining (Fig. 3A and B). SMO expression was reduced in morphoeic tumors, and although this may appear paradoxical, it correlates with data showing that GLI1 expression is reduced in morphoeic BCC (23). Nuclear SMO was most prevalent in mouse BCC-like nodular tumors, which may reflect the more homogenous nature of the system, but, as for human tumors, expression was also reduced in morphoeic-like BCC (Fig. 3B).

BCC tumors that become resistant to vismodegib most commonly acquire point mutations affecting their ability to bind SMO (24). How and if nuclear SMO expression influences drug effectiveness remains to be determined, but the overall response to anti-SMO compounds in clinical trials was approximately 50% (14). As such, many BCCs are classified as SMO independent because they are dependent upon other mechanisms that correlate with the lack of SMO expression observed in some tumors (Fig. 3B). Alternatively, some tumors may still be SMO dependent, but an anti-SMO compound lacks efficacy because it is influenced by SMO subcellular localization. Indeed, it has been postulated that more selective targeting of subcellular compartments may improve G-protein–coupled receptor (GPCR) drug selectivity (25).

SMO is a seven-pass transmembrane protein closely related to the Frizzled family of GPCR proteins. Drosophila and vertebrate SMO share considerable homology, although the former has a longer C-terminal tail. The C-terminal has been studied comprehensively, particularly the region that is phosphorylated, and determines the extent of pathway activation (26). However, unlike human SMO, the Drosophila homolog does not contain a putative NLS (homology with dSMO extends to residue 632 for human SMO), and although such a region has not been described for SMO, NLS domains have been recognized in other GPCRs, including the Bradykinin and Apelin receptors (25). Nuclear translocation of GPCRs is mediated by importins and can be constitutive or agonist induced; for example, Frizzled2 is cleaved upon Wingless stimulation, with the C-terminus entering the nucleus to regulate postsynaptic neuron development in Drosophila (27). Future work will determine if SMO is similarly cleaved, but the fact that nuclear SMO is detected with an N-terminal antibody suggests that this is not the case and not all GPCRs are cleaved before entering the nucleus. Other proteins that regulate GPCR signaling in the nucleus and harbor NLS sequences include β-arrestins and the GPCR kinases (28, 29). Therefore, the identification of an NLS in SMO as a GPCR family member is not without precedent and represents a mode of signaling already established for these proteins.

With regard to the NLS sequences, the SMO region predicted to be an N(o)LS targets EGFP to the nucleolus (Fig. 2B). Indeed, although NoLS (nucleolar) are normally distinct to NLS (nuclear), they may occur in the same region, which can make it difficult to delineate the role of specific subdomains. Studies suggest that shared motifs create genetic stability and that a combined N(o)LS region may facilitate active nuclear import (30). We have not functionally delineated the SMO N(o)LS region but determined that it is functional and that nuclear import is important for GLI1 regulation (Fig. 3C). Whether SMO nucleolar localization regulates GLI1 requires further investigation, but to our knowledge, nucleolar function has not been described for other GPCRs. The nucleolus is emerging as a target for cancer therapy and is regulated by oncogenic signaling pathways, including RAS/ERK and PI3K/AKT/mTOR, which influence GLI activity (31). Nucleoli are often more abundant in tumors compared with normal cells, and this has been observed in BCC (32). Interestingly, PTCH1 suppression results in an increase in the number of nucleoli (Fig. 1F).

Although nucleoli numbers decrease during keratinocyte terminal differentiation, an increase in number is seen in fibroblasts of patients with X-ray–treated Gorlin syndrome (33, 34).

Regarding a nuclear role for GPCRs, control of gene regulation has been reported for various proteins that may represent signaling from the nuclear membrane or more directly; for example, it was shown that F2rl1 forms part of a transcriptional complex with Sp1 that modulates Vegfa expression (35). Delineation of the N(o)LS region may help determine if SMO localizes to the nuclear membrane, and regarding DNA binding, it is intriguing that SMO has a putative leucine zipper region from residues 405 to 426 that was not previously predicted. Microarray data highlight the downregulation of zinc-finger proteins commonly localized to the nucleus/nucleolus, with more genes downregulated upon KAAD-Cyc treatment in shPTCH1 cells (Supplementary Table S1). In particular, ZNF750 is known to be involved in psoriasis, squamous cell carcinoma, and esophageal, lung, and cervical cancers (36). The role that SMO plays in promoting nucleolar changes requires further investigation as well as the consequences of reduced zinc-finger protein expression in shPTCH1 cells.

We also aimed to identify novel targets downstream of PTCH1 that may be important for BCC biology, particularly those that are SMO independent. Microarray analysis identified 213 transcripts that were altered ≥2-fold in shPTCH1 cells versus the control (137 increased and 76 decreased). Of these, less than 20% returned to basal levels in the presence of KAAD-Cyc, revealing that the majority of DEGs are SMO independent in shPTCH1 cells, or that if they are SMO dependent, their mechanism of control is unresponsive to conventional anti-SMO inhibition.

CXCL11 is reported to increase proliferation of and enrich for BCC-derived epithelial cells as well as providing immunoprotection by increasing indoleamine 2,3-dioxygenase (IDO) expression through its receptor, CXCR3, which is itself implicated in tumor biology (18). MMPs are associated with tumor invasion and metastasis by degradation of the extracellular matrix. MMP2 and MMP9 degrade collagen IV, and the levels of MMP2 and MMP9 mRNA are significantly higher in nodular and infiltrative BCCs compared with normal adjacent tissue (37). In addition, MMP9 expression was significantly higher in infiltrative compared with nodular tumors. MMP7 expression has also been described in BCC for aggressive and recurrent forms, although we could not confirm the microarray result by qPCR (38).

THY1 (CD90) is a GPI-anchored cell-surface protein expressed in various tissues and cancers. Although it is a putative cancer stem cell marker associated with multiple oncogenic processes, including proliferation and metastasis, there is evidence supporting its role as a tumor suppressor (39); therefore, due to the cell-specific role of CD90, it is difficult to speculate about its role in BCC.

Increased expression of SNAI1 and VIM was of interest, as both are associated with the cancer stem cell phenotype including EMT (40). SNAI1 mRNA has been described in BCC and shown to be a target of GLI1 in the murine epidermis (41). Although we found no evidence that SNAI1 or VIM mRNAs are direct GLI1 targets in human keratinocytes, both proteins were more highly expressed in GLI keratinocytes that survived genotoxic insult (19, 42). Indeed, VIM expression is associated with a more aggressive myofibroblast phenotype in BCC (43).

We have not examined the migratory/invasive potential of shPTCH1 cells in vitro, but although BCCs may be locally invasive, they are rarely metastatic, and this correlates with the holoclone-like colonies observed in culture (Fig. 1E). How PTCH1 regulates cell–cell adhesion warrants investigation, but AMOT (angiomotin) positively regulates cell migration, and increased expression has been described in metastatic breast cancer (44). In addition, AGR2 is associated with tumor progression and metastasis (particularly hormone-dependent cancers), and its suppression impairs the migration of non–small cell lung cancer cells (45). As such, AMOT and AGR2 suppression may represent a negative feedback loop to limit tumorigenesis and the spread of transformed cells. It is likely that exposure to ultraviolet irradiation or other form of genetic insult will be required to transform shPTCH1 cells, and the complexity by which BCCs arise and progress is highlighted by the extent of genomic aberrations identified in these common tumors (46).

Finally, FBN2 was the most potently suppressed gene in shPTCH1 cells. FBN2 is a glycoprotein that forms part of connective tissue microfibrils in the extracellular matrix, often with low expression in tumors (47). How reduced FBN2 expression influences tumor biology is unclear, but it was hypothesized that, as for FBN1, this may relate to increased TGFβ signaling (48). Indeed, it has been previously reported that integrin-mediated TGFβ activation modulates the stromal microenvironment in morphoeic BCC (23, 49).

Since the discovery of PTCH1 in BCC (2–5), several models have been created attempting to identify how HH signaling contributes to BCC tumorigenesis. To our knowledge, this is the first in vitro model to sustain PTCH1 suppression in immortalized human keratinocytes. The model enables us to show that PTCH1 regulates SMO expression and that SMO, as for other GPCRs, localizes to the nucleus, where it regulates GLI1 expression through a mechanism that is the subject of ongoing research. Whether or not nuclear SMO influences primary or acquired resistance to anti-SMO compounds warrants further investigation, but these data suggest that drugs should be tested for efficacy against nuclear SMO. In addition, this model complements other studies implying that PTCH1 regulates pathways independently of SMO (50), which may involve transmembrane as well as nuclear control. Future studies aim to further characterize these SMO-independent signaling pathways to help realize new targets for treating BCC.

G.W. Neill is Medical Science Liaison Team Lead UK and Ireland at Sanofi Genzyme. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.M. Rahman, M.P. Philpott, G.W. Neill

Development of methodology: M.M. Rahman, A. Hazan, C.A. Harwood, D.P. Kelsell, K.J. Linton, G.W. Neill

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.M. Rahman, J.L. Selway, D.S. Herath, C.A. Harwood, M.S. Pirzado, R. Atkar, G.W. Neill

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.M. Rahman, A. Hazan, J.L. Selway, M.S. Pirzado, K.J. Linton, M.P. Philpott, G.W. Neill

Writing, review, and/or revision of the manuscript: M.M. Rahman, D.P. Kelsell, M.P. Philpott, G.W. Neill

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.M. Rahman, D.S. Herath, M.P. Philpott

Study supervision: M.P. Philpott, G.W. Neill

The authors gratefully acknowledge the funding from the British Skin Foundation. The authors also thank Anna Saran (ENEA, Rome, Italy) for providing mouse BCC tissue for immunohistochemistry and Dr. Monika Cichon for the shRNA control sequences. The authors also thank the Dr. Hadwen Trust (DHT) for Humane Research for additional funding to develop human models for human disease and confirm that no funding from the DHT was used for any animal research in this article.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.