The Hedgehog (Hh) pathway has been implicated in a wide variety of human tumors, and early clinical trials with pathway antagonists have validated Hh signaling as a bona fide anticancer target. Despite these encouraging results, several issues surrounding the basic biology of the Hh pathway in human cancers remain unclear. These include the influence of specific oncogenic events on Hh signal transduction, the precise mode of Hh signaling (i.e., autocrine or paracrine) that occurs within human tumors, and the best means to inhibit aberrant pathway activity in the clinical setting. The cancer stem cell (CSC) hypothesis may explain a number of clinical phenomena, such as unchecked self-renewal and the development of metastatic disease, and to some extent, the Hh signaling pathway has been implicated in all of these processes. Therefore, Hh pathway inhibitors may also represent some of the first agents to formally examine the CSC hypothesis in the clinical setting. The diverse nature of Hh signaling in human cancers suggests that disease-specific factors must be carefully considered to identify the optimal use of novel pathway inhibitors. Clin Cancer Res; 16(12); 3130–40. ©2010 AACR.
Development and Hedgehog Pathway Signal Transduction
The Hedgehog (Hh) signaling pathway was initially identified in Drosophila as a critical mediator of segmental patterning during embryonic development, and it regulates the proliferation, migration, and differentiation of target cells in a spatial, temporal, and concentration dependent manner (1–3). Hh signaling is conserved in vertebrates and highly active during mammalian development, especially within the neural tube and skeleton, but subsequently silenced in most adult tissues. However, some postnatal organs, such as the central nervous system and the lung, rely on continued Hh signaling for tissue homeostasis and repair following injury (4–6).
Pathway activation is initiated by binding of one of the three secreted and lipid-modified ligands found in mammals, Sonic (SHh), Desert (DHh), and Indian Hedgehog (IHh), to Patched (Ptch1), a 12-pass transmembrane-spanning receptor (Fig. 1A, B). In the absence of ligand, Ptch constitutively represses the activity of Smoothened (Smo), a 7-pass transmembrane-spanning protein with homology to G-protein-coupled receptors. Following Hh ligand binding to Ptch, the repression of Smo is released, and the expression and/or post-translational processing of the three Gli zinc-finger transcription factors is modulated. Gli1 acts as a transcriptional activator and Gli3 as a repressor, whereas Gli2 can either activate or repress gene expression depending on post-transcriptional and post-translational modifications (7). The balance between the activating and repressive forms of the Glis results in the expression of target genes, including Ptch1 and Gli1 (8, 9).
Within this simplified schema, several other cellular components are required for Hh pathway activity. These include proteins involved in Hh ligand modification and cell surface binding [Hedgehog interacting protein (Hip), Hedgehog acyltransferase (Hhat), growth arrest-specific 1 (Gas1), and CDO], and Gli processing and localization [Suppressor of fused (SuFu), Protein kinase A (PKA), glycogen synthase kinase 3β (GSK3β), casein kinase 1 (CK1), G-protein-coupled receptor kinase 2 (GRK2), and βTRCP; refs. 10–14]. Moreover, increasing evidence has suggested that the subcellular localization of Hh pathway components is a major regulator of its activity. The examination of developmental defects arising in mice showed that mutations within the intraflagellar transport proteins Kif3a and IFT88 produce patterning defects that mimic Hh loss-of-function mutations (15). These proteins are required for the assembly and maintenance of primary cilia that are present on most cells of the body during interphase and involved in a wide variety of cellular processes including mechanosensation and the transduction of several signaling pathways. A number of studies have subsequently shown that pathway components translocate during activation, and in the absence of ligand, Ptch, but not Smo, is located within the primary cilia (16–20). Upon ligand binding, Ptch moves out and Smo moves into primary cilia to interact with Glis and their associated proteins that subsequently enter the nucleus to regulate gene expression (Fig. 1B).
Studies from a variety of experimental systems have identified the major components involved in Hh signal transduction, but extension of these results to human cancers should be approached with caution for several reasons. Many genetic studies have determined the role of specific pathway components by examining the effects of mutations on normal developmental programs, but the precise similarities between development and carcinogenesis are not fully understood. Meanwhile, detailed biochemical and cell biology studies have typically studied immortalized and nonmalignant embryonic fibroblasts (e.g., NIH 3T3 or CH310T1/2 cells) that exhibit robust pathway activity compared with most cancers. Additionally, these cells are mesenchymal in origin and can differentiate into mature cartilage, bone, and adipose cells. Thus, their relevance to Hh signaling within cancers derived from epithelial or hematopoietic tissues is unclear. Both genetic and epigenetic events underlie the formation and progression of human cancers, but the influence of these factors on Hh pathway signaling is largely unknown either in general or within specific tumors. Activating mutations in KRAS are found in a wide variety of epithelial carcinomas and may cause Gli1 activation and Hh target gene expression in a Smo-independent manner (21, 22), therefore, oncogenic events may directly induce pathway activity. The Hh signaling pathway has also been found to interact with other signaling pathways commonly activated in human cancers, such as PI3K/AKT (23, 24), and it is possible those also alter classical Hh signal transduction. It is also unclear whether primarily cilia are absolutely required for Hh signaling in human cancers because they are absent in Drosophila cells that are fully capable of pathway activation and may or may not be found at a high frequency within primary human basal cell carcinomas (BCC; ref. 25). Moreover, recent studies examining the role of primary cilia and aberrant Hh signaling in the formation of BCC and medulloblastoma have shown that mutations within Smo are not tumorigenic in the absence of cilia (25, 26). However, a constitutively activated form of Gli2 paradoxically induced tumor formation when cilia formation was blocked. Therefore, Hh signaling seems to be complex in cancers, and although several aspects of Hh signal transduction remain unclear, a better understanding of the mechanisms mediating pathogenic pathway activity within human cancer cells may identify additional therapeutic strategies.
Hedgehog Signaling in Cancer
The identification of somatic PTCH1 mutations in patients with Gorlin syndrome and basal cell nevus syndrome first implicated a role for the Hh pathway in cancer as these individuals are highly predisposed to developing advanced BCC, medulloblastoma, and rhabdomyosarcoma (27, 28). Additional evidence that aberrant Hh pathway activity may play a causal role in human cancers was provided by the identification of PTCH1 and SMO mutations in sporadic BCCs and medulloblastomas (29–31). Definitive proof that aberrant Hh signaling can induce cancers has come from transgenic mouse studies in which conditional loss of function of Ptch or gain of function of Smo or the Glis can recapitulate medulloblastoma and BCC (reviewed in ref. 32). Other Hh pathway components may also be genetically altered in human cancers including SUFU mutations in medulloblastoma, GLI1 and GLI3 mutations in pancreatic adenocarcinoma, as well as GLI1 gene amplification in glioblastoma (33–35).
In contrast to tumors harboring activating mutations, several human cancers display aberrant pathway activity in response to increased levels of HH ligand (Table 1). Within these ligand-dependent tumors, several modes of pathway activation have been described (Fig. 2). The simplest is autocrine and/or juxacrine signaling in which tumor cells both produce and respond to HH ligand (Fig. 2B). This mode of pathogenic HH activity has been described in a wide variety of human tumors including small cell lung, pancreatic, colon, and metastatic prostate cancers as well as glioblastomas and melanomas (5, 24, 36–40). Increased HH pathway activity either from mutational activation or autocrine signaling may induce the expression of genes affecting proliferation (Cyclin D1, Cyclin D2, N-Myc, Igf2, Hes1), cell survival (Bcl2), angiogenesis (Vegf), and genetic instability (p53; refs. 41–51). Moreover, HH ligand blocking antibodies (in the case of ligand-dependent signaling), pharmacologic SMO antagonists, or small interfering RNA (siRNA) directed against SMO or GLI1 have been shown to inhibit tumor growth.
Hh signaling may act in a paracrine fashion during embryonic development in which cells secreting Hh ligands are distinct from those responding with pathway activation, and several recent studies have shown that pathologic Hh signaling may also occur in a similar manner that involves the tumor microenvironment (52–54). In a recent report examining B-cell lymphomas and multiple myeloma, HH ligand was found to be primarily produced by stromal cells derived from the bone marrow, spleen, or lymph nodes, rather than tumor cells (Fig. 2C; ref. 52). However, the activation of HH signaling within tumor cells was found to act as an important survival factor because treatment with SMO antagonists induced cell death and decreased BCL2 expression in vitro as well as inhibited tumor growth in vivo. This effect was specifically mediated by HH signaling as the expression of GLI1 or a constitutively activated form of SMO abrogated these effects.
An alternative mode of paracrine Hh signaling has been recently described in which tumor cells secrete HH ligands that induce pathway activity within stromal cells (Fig. 2D; refs. 53–55). In a study examining human pancreatic and colon primary tumors and cell lines grown as xenografts in mice, the expression of HH ligand by tumor cells did not correlate with their expression of target genes, but rather was associated with canonical pathway activity by tumor infiltrating stromal cells derived from the murine host (53). Decreased tumor growth was observed following administration of a pharmacologic SMO antagonist or HH-neutralizing monoclonal antibody and correlated with decreased expression of mouse Gli1 expression, whereas human GLI1 expression remained unchanged. Further studies using a transgenic mouse model of pancreatic cancer similarly failed to identify canonical Hh pathway activity within epithelial tumor cells despite expression of a constitutively activated form of Smo in these cells (54). These results strongly suggest that tumor derived HH ligand primarily induces pathway activity within infiltrating stromal cells that in turn influences tumor growth. Although the precise factors generated by stromal cells that modulate tumor growth are unknown, HH pathway activation may induce the secretion of soluble factors, such as insulin growth factor 2 (IGF2) or vascular endothelial growth factor (VEGF), that influence tumor cell proliferation and survival (43, 48, 49).
Conflicting data about the precise mode of Hh signaling in specific tumors, such as pancreatic and colon carcinoma (see below), may be due to differences in experimental systems. Given the spatial restriction of Hh signaling during normal development, the study of human tumors may require three-dimensional models to precisely determine the role HH plays in cancer biology and the antitumor effects of pathway inhibitors. It is also likely that the exact role of Hh signaling will be dependent on the specific tumor type as well as distinct clinical and biological factors, such as disease stage or genetic lesions. For example, intact Hh signaling, specifically Smo expression, is required for myeloid leukemia arising from the BCR-ABL, but not the MLL-AF9, fusion gene (56–58). Although delineating disease specific information will likely require extensive studies, it is likely required to optimize the clinical use of Hh pathway inhibitors.
Hedgehog Signaling in Cancer Stem Cells
Emerging data from many human tumors including glioblastoma, breast cancer, pancreatic adenocarcinoma, multiple myeloma, and chronic myeloid leukemia (CML) have suggested that Hh signaling regulates cancer stem cells (CSC; refs. 39, 40, 56, 59–62). CSCs have been functionally defined by their capacity to undergo self-renewal and give rise to differentiated progeny that recapitulates the original tumor in an ectopic setting (63). Self-renewal is required for maintenance of the malignant clone, and reports examining mouse models of CML have provided evidence that Hh signaling regulates this property (56, 62). In these studies, the loss of Hh signaling by genetically disrupting Smo resulted in the inhibition of BCR-ABL expressing leukemic stem cells and prolonged survival. Active Hh signaling pathway has also been identified in glioblastoma CSCs, and pathway inhibition with cyclopamine or siRNA directed against pathway components results in the loss of tumorigenic potential (39, 40). In breast cancer, pathway activation in CSCs using Hh ligand and GLI1 or GLI2 expression or inhibition with cyclopamine or siRNA directed against GLI1 or GLI2 alters the expression of BMI-1, a central regulator of self-renewal in normal stem cells, and tumorigenic potential in vitro and in vivo (59). In multiple myeloma, CSCs that phenotypically resemble normal memory B cells have been found to display relatively higher levels of Hh signaling than the mature plasma cells constituting the tumor bulk (61). Pathway activation with SHH ligand resulted in CSC self-renewal and expansion, whereas the SMO antagonist cyclopamine or the ligand-neutralizing antibody 5E1 induced terminal differentiation and loss of clonogenic growth potential (Fig. 3A and B). Thus, HH signaling may dictate CSC fate decisions that include self-renewal and differentiation possibly by generation of a malignant niche (64). Data from multiple myeloma demonstrate that Hh signaling can act through multiple signaling modes within the same cancer and can mediate interactions between CSCs, differentiated tumor cells and the microenvironment (Fig. 3C).
In addition to tumor formation, CSCs have been implicated in disease progression and the development of metastasis in solid tumors (Fig. 4; refs. 65, 66), and Hh signaling may play a critical role in this process similar to the Notch and Wnt pathways in cancer (67, 68). In colon carcinomas derived from primary clinical specimens, Hh signaling has been found to be preferentially activated within CSCs as evidenced by relatively higher expression of GLI1, GLI2, PTCH1, and HIP within this cellular compartment (37). Moreover, the relative expression of these pathway components as well as the target gene SNAIL1, which is associated with the epithelial-to-mesenchymal transition (EMT) and implicated in metastasis, increases in CSCs with disease stage. Inhibition of Hh pathway activity with cyclopamine or siRNA against SMO, GLI1, and GLI2, or expression of the repressive form of GLI3 reduced tumor cell proliferation and induced apoptosis. Moreover, cyclopamine reduced tumor regrowth in vivo and small hairpin RNA (shRNA) directed against SMO eliminated the formation of metastatic disease. The relationship between EMT and clonogenic growth potential has also been examined in pancreatic CSCs, and cyclopamine has been found to inhibit each of these functional properties and the formation of metastatic disease (60).
Clinically Targeting Aberrant Hedgehog Signaling
Initial evidence that Hh signaling could be pharmacologically inhibited arose from the identification of cyclopamine, a steroidal alkyloid derived from Veratrum californicum, as an active compound capable of producing congenital defects in sheep (69). Recognition that SHh loss-of-function mutations recapitulate the neural defects seen with Veratrum teratogenicity led to findings that cyclopamine inhibits pathway activity by directly binding and inactivating Smo (70–72). Several pathway antagonists have been subsequently identified in compound screens that inhibit Gli-dependent reporter activity (73–77). Despite the unbiased nature of these screens against any one component of the Hh signaling pathway, the vast majority of compounds have been found to bind to and inhibit Smo despite being structurally distinct from one another and cyclopamine (reviewed in ref. 78). It is possible that these whole cell screens have primarily identified SMO antagonists because of its location on the cell surface, the central role it plays in Hh signal transduction, and the relative ease with which inhibitors of other 7-pass transmembrane proteins have been identified. An alternative approach to identifying novel Hh inhibitors has involved chemically modifying cyclopamine to improve its specificity, bioavailability, and pharmacokinetics (79).
The first clinical data reported using a SMO antagonist have involved GDC-0449 that was originally identified in a chemical compound screen and further modified and developed through a collaborative agreement by Curis and Genentech (74). The initial phase I trial examining the safety and tolerability of GDC-0449 enrolled a total of 68 solid tumor patients (80). Early data suggested efficacy in patients with advanced BCC, presumably because of the high frequency of aberrant Hh signaling, and the trial was expanded to further examine this group of patients. Of a total of 33 patients with locally advanced or metastatic BCC, 55% showed clinical responses that included two complete remissions. Although efficacy data have not yet been reported from the other 35 patients without BCC enrolled in the trial, a published abstract indicated that a single patient with adenocystic carcinoma experienced stable disease (81). Reported toxicities were mild, and the most common events were mild loss of taste, hair loss, weight loss, and hyponatremia. Importantly, biopsies of noninvolved skin showed in vivo inhibition of HH signaling evidenced by downregulation of GLI1 expression following drug administration. A transient subjective tumor response was also reported in a single medulloblastoma patient treated with GDC-0449 under compassionate use (82). Early results from another SMO antagonist, BMS-833923 (XL139), codeveloped by Exelixis and Bristol-Myers Squibb have been recently reported and shown clinical benefit in patients with medulloblastoma and basal cell nevus syndrome (83). On the basis of the safety and efficacy data generated in the early phase clinical trial, multiple phase II trials examining GDC-0449 in combination with standard agents in solid tumors have been initiated (Table 2). Moreover, early phase clinical testing of alternative SMO antagonists have been undertaken by several other companies including Exelixis/Bristol-Myers Squibb, Novartis, Infinity, and Pfizer (Table 2). Although all these agents target SMO, they are chemically distinct and no comparative studies have been carried out. It is possible that differences in SMO binding may lead to efficacy against some mutations that may arise during treatment (82), or that differences in pharmacokinetic properties may influence tissue distribution that provides more favorable inhibition of specific tumors, such as those in the central nervous system, or attenuation of side effects.
Theses early results provide critical proof-of-concept that inhibiting HH pathway activity has antitumor activity in humans, but they also generate several questions. For example, the majority of BCC tumor specimens from patients treated with GDC-0449 displayed increased levels of GLI1 expression consistent with the nature of the disease, but slightly more than half of the patients experienced clinical responses despite circulating levels of drug that were well above the concentration required to inhibit Hh signaling in vitro (74, 80). Additionally, the examination of pretreatment specimens for the presence of primary cilia or HH pathway activity within tumor stromal cells may have correlated with responses and validated the importance of these factors in the clinical setting. Similarly, serial tumor biopsies were not required, but it would have been interesting to correlate the degree of intratumoral pathway inhibition with clinical responses to draw a conclusive relationship between pathway activity and human cancers. Finally, the rapid loss of response and finding of an acquired mutation in SMO in the medulloblastoma patient treated with GCD-0449 suggest that loss of response in BCC patients may have occurred in a similar manner (82, 84), and isolation of tumor tissue at disease progression could have explored this possibility.
Alternative Strategies to Inhibit Hedgehog Signaling
Initial results with SMO antagonists are encouraging, but the development of alternative means to inhibit Hh signaling may be desirable for several reasons. First, several genetic lesions acting downstream of SMO have been detected in human cancers, such as mutations or loss of heterozygosity of the negative pathway regulators SUFU or RENKCTD11 or amplifications in GLIs (33, 35, 85). Moreover, GLI1 activity may be induced in a SMO-independent manner by transforming events, such as EWS-FLI in Ewing sarcoma or mutant KRAS in pancreatic cancer (21, 22, 86, 87). Because Hh signal transduction ultimately activates the GLI transcription factors, inhibitors modulating GLI may be useful, similar to therapeutic strategies being developed to target Wnt/β-catenin signaling in cancer (68). A screen for compounds capable of inhibiting Gli-mediated transcription identified Gli-antagonist (GANT)-58 and GANT-61 (88). Both of these compounds act downstream of Smo and SuFu and inhibit tumor growth both in vitro and in vivo. The precise inhibitory mechanism of these compounds is unclear, but GANT-61 limits Gli1 DNA binding. A screen designed to identify molecules capable of inhibiting Gli1 transcriptional activity without targeting Smo identified four distinct Hh pathway inhibitors (HPI; ref. 89). Each of these compounds inhibits Gli at a level epistatic to SuFu, but their precise mechanisms are unknown. HPI-1 inhibits both Gli1 and Gli2, perhaps by altering their post-translational modification, HPI-4 seems to partially inhibit ciliogenesis, and HPI-2 and 3 may inhibit the conversion of Gli2 from a repressive to an activated form. Although less specific, GLI activity may also be inhibited by modulating the activity of other pathways known to interact with Hh signaling, such as phosphoinositide 3-kinase (PI3K)/AKT, mitogen-activated protein kinase (MAPK), and transforming growth factor β (TGF-β; refs. 23, 24, 90–92).
Mutations in Hh pathway components have not been described in most cancers displaying aberrant signaling. Therefore, the inhibition of ligand binding may inhibit pathway activation, and the ligand neutralizing monoclonal antibody 5E1 has been shown to have clear antitumor activity in several diseases (see above). More recently, robotnikinin, was identified as a compound that binds to SHh and blocks its ability induce pathway activity at the level of Ptch (93). It is also possible that specific on-target side effects attributable to the inhibition of Hh pathway activity in some cells, such as defects in bone growth seen in young mice treated with a Smo antagonist (94), may be avoided by blocking the activity of a specific Hh ligand.
Clinical Translation of Hedgehog Inhibitors
It is likely that the effective clinical use of Hh inhibitors will require a precise understanding of the role that the pathway plays in a specific tumor. In cancers with mutational pathway activation affecting all tumor cells, pathway inhibition may impact the growth and survival of all tumor cells evident as reduction in disease burden as observed in BCC patients receiving GDC-0449 (80). Alternatively, Hh may be required early in carcinogenesis or in the formation of precancerous lesions, but once tumors are established, it may not be needed for tumor maintenance or progression. In these cases, Hh inhibitors may be active chemo-preventative agents in high-risk patients, but would be ineffective in treating established disease. Alternatively, the Hh pathway may be activated in response to chemotherapy or during regrowth of minimal residual disease, perhaps by regulating CSCs (95). In this case Hh inhibitors would be most effective as maintenance therapy following tumor debulking with surgery or cytotoxic chemotherapy. Hh signaling has also been implicated in mediating chemoresistance (96, 97), therefore, inhibitors may act as potent chemo-sensitizers and require the simultaneous administration of cytotoxic chemotherapy.
Given the diverse biological effects of Hh activity, accurately measuring the activity of Hh inhibitors may also require context-dependent indicators of clinical response. The Hh pathway may regulate EMT and the formation of metastatic disease (37, 38, 60). In this case, treatment with Hh inhibitors may not result in clinically significant inhibition of the primary tumor, but could prevent the formation of new metastasis and prolong progression free survival. Similarly, if Hh signaling is primarily active in CSCs, clinical responses to Hh inhibitors may be delayed in patients with extensive disease. Effective inhibition of CSCs should lead to improved overall survival, but most early phase trials are not designed or powered to detect differences in long-term outcomes (98). Therefore, reliable biomarkers that accurately quantify CSCs may be useful endpoints within early clinical testing. Finally, patient selection is likely to be a critical factor in establishing the clinical efficacy of Hh pathway antagonists similar to most targeted therapies. Because mutations seem to drive only a minority of Hh-related cancers, selection of patients will not be possible by simply genotyping tumor specimens, but better understanding of the specific role the Hh signaling pathway plays in each tumor type is likely to be valuable during the development of these novel agents. It is clear that several issues in the precise role of the Hh signaling pathway in human cancer remain unresolved, including the precise mechanisms of signal transduction, the exact mode of signaling between tumor cells and the microenvironment, and the role of signaling in the regulation of CSCs. However, it is clear that the Hh pathway can play an important role in tumor cell growth and survival, and these functions are likely to be broadly applicable across a variety of malignancies. As several novel inhibitors have entered clinical testing, precise correlative studies may allow many of these conflicts to be resolved within the most relevant system, the patients themselves.
Disclosure of Potential Conflicts of Interest
W. Matsui, commercial research grant, Merck; patent interest, Infinity Pharmaceuticals; consultant, Bristol-Myers Squibb, Pfizer.
We thank Craig Peacock, Karen McGovern, and Sarah Brennan for helpful comments and critical review of the manuscript.
Grant Support: NIH R01CA127574, P01CA15396, K23CA107040; the Gabrielle's Angel Foundation for Cancer Research; the Sidney Kimmel Foundation for Cancer Research; the Multiple Myeloma Research Foundation; and the Leukemia and Lymphoma Society.
- Received December 31, 2009.
- Revision received March 31, 2010.
- Accepted April 9, 2010.