Hedgehog is an early and late mediator of pancreatic cancer tumorigenesis

Sonic hedgehog (SHH) is misexpressed in human adenocarcinoma and its precursor lesions. SHH expression was determined using in situ hybridization to detect SHH messenger RNA and immunohistochemistry (IHC) to detect the protein with an antibody directed against SHH⁹. Pancreatic tissues were obtained from 20 specimens resected for pancreatic cancer. Control pancreatic tissues with no evidence of abnormality or autolysis upon histological evaluation were obtained from autopsy specimens or from pancreatic resections for trauma. In normal adult human pancreata, no SHH was detected in the islets, acini or ductal epithelium (Fig. 1a). However, evaluation of pancreata from patients with adenocarcinoma reveals that SHH is aberrantly expressed in 70% of specimens. Normal ductal epithelium does not express detectable levels of SHH (Fig. 1b); however, as the ductal epithelium shows increasing degrees of atypia, PanIN-1 to -3 (Fig. 1c–e), a higher expression of SHH is observed. SHH expression is also detected in the malignant epithelium of adenocarcinoma samples (Fig. 1f). This expression pattern was also confirmed by our in situ hybridization for SHH mRNA (Supplementary Fig. 1).

Loss of regulation in this pathway has been implicated in several human cancers^10,11. Thus in order to determine the potential role of SHH misexpression in the adult human pancreas, pancreata from transgenic mice (gift of H. Edlund) in which Shh misexpression was driven by the pancreatic-specific Pdx-1 promoter were histologically and immunohistochemically analysed.

A total of four pancreata from three-week-old Pdx–Shh mice were histologically evaluated by a gastrointestinal pathologist (G.Y.L.). All four expressed a significant intestinal phenotype of the pancreatic epithelium. All four also exhibited abnormal tubular complex formations of the epithelium resembling pancreatic cancer precursor lesions (PanINs). These abnormal epithelial changes were characterized and classified according to the pathological classification scheme for human ductal lesions of the pancreas¹².

The transgenic pancreata display a mixed phenotype characterized by the transformation of acini into glandular structures lined by tall intestinal-type epithelium embedded in cellular spindle stroma. Focally, the metaplastic epithelium is composed of columnar cells with basally located nuclei and abundant supranuclear mucin (Fig. 2a). This phenotype is similar to human PanIN-1 (Fig. 2b), an early lesion in the stepwise sequence of pancreatic carcinogenesis. Furthermore, some glands are characterized by a lesser degree of mucin production but a higher grade of atypia, with nuclear disarray and stratification, mitosis and apoptosis (Fig. 2c), representing a higher degree of dysplastic change, similar to human PanIN-2 (Fig. 2d). According to the tumour progression model of pancreatic carcinoma, these features are suggestive of an increasing neoplastic potential.

The histological progression model of pancreatic adenocarcinoma is also associated with progressive genetic alterations¹³. The transgenic mice pancreata, which histologically resemble PanIN-1 and -2, also express some of the early genetic alterations observed in the human progression model. Three pancreata overexpress HER-2/neu in the epithelium (Fig. 2g). In addition, preliminary data indicate that a codon 12 mutation in K-ras was present in two pancreata (Fig. 2i): changes noted to occur in human PanIN-1 and -2 lesions^14,15. Thus the Pdx–Shh mice develop abnormal pancreata with morphological and genetic changes that resemble human ductal PanIN.

Hedgehog (Hh) signalling is activated through binding of Shh to a membrane receptor, Ptc. This ligand–receptor interaction inhibits Ptc function, thereby allowing the de-repression of another transmembrane protein, smoothened (Smo), which ultimately leads to the activation of the Hh pathway and to the expression of Hh-specific genes, one of which is Ptc1 (ref. 16). Ptc1 expression (determined by IHC) is not detected in normal mouse or human pancreata (Fig. 3a, b). However, Ptc1 protein expression was detected throughout the abnormal epithelium in both Pdx–Shh mice (Fig. 3c) and human neoplastic pancreata (Fig. 3d). Furthermore, in human neoplastic pancreata, PTC1 was also detected in the reactive mesenchymal cells that surround the abnormal epithelium (Fig. 3d).

Another member of the Hh pathway that is similarly overexpressed in both human neoplastic and mouse transgenic pancreata is Smo. In normal human and mouse pancreata Smo is not generally detected by IHC in the ducts or acini. Very infrequently Smo can be detected in an occasional duct cell or scattered acinar cell within the normal pancreas (Fig. 3f). However, Smo was overexpressed in the abnormal pancreatic epithelium found in Pdx–Shh mice (Fig. 3g), as well as in the epithelium of human adenocarcinoma and precursor lesions (Fig. 3h). The identification of Hh signalling members overexpressed in both mesenchyme and abnormal epithelium suggests that misexpression of Shh may be acting through both a paracrine and an autocrine mechanism to cause many of the morphological changes that characterize pancreatic cancer. To determine whether the HH pathway has an equally important role in the malignant biological behaviour of this tumour, we undertook work with pancreatic cancer cell lines.

Twenty-six human adenocarcinoma lines were screened for expression of HH signalling components. These cell lines were derived either from primary tumours (for example, the Panc series Panc 01.28 to 10.05 (ref. 17)) or from liver, lymph node and spleen metastases (CFPAC1, Hs 766T and SW 1990, respectively). All lines tested expressed two or more components of HH signalling, including PTCH1, SMO, HIP and GLI1 (Supplementary Fig. 2). Thus, sustained HH signalling activity is detected in pancreatic adenocarcinoma cell lines isolated from both primary and metastatic tumours.

Cyclopamine, a steroidal alkaloid that inhibits Hh signalling through direct interaction with Smo^18–20, was used to test whether pancreatic adenocarcinoma cells require HH signalling for proliferation and survival. Initially five pancreatic cancer cell lines—BxPC3, Panc 01.28, CFPAC, L3.6sl and Panc 05.04—were qualitatively assessed. After 7 days of treatment, cyclopamine produced no change in cell density or morphology in control samples that included either untreated cells or cells treated with tomatidine, an inactive cyclopamine analogue. By contrast, a marked reduction in cell density was noted in cell lines CFPAC, L3.6sl and Panc 05.04. Morphologically, the majority of these cells detached from tissue culture plates during the treatment period, and the few remaining cells appeared aplastic (Fig. 4a, and data not shown).

Three representative pancreatic cancer cell lines—BxPC3-SMO^low, L3.6sl-SMO^high and Panc 05.04-SMO^high—were then used to quantitatively determine the effects of cyclopamine on proliferation and apoptosis by means of FACS analysis after 5-bromodeoxyuridine (BrdU) labelling. Cyclopamine did not induce apoptosis and only marginally reduced proliferation in the BxPC3-SMO^low control cell line. By contrast, apoptosis increased 2.5- to 3.5-fold accompanied by a 75–80% reduction in proliferation in the SMO^high cyclopamine-responsive cell lines L3.6sl and Panc 05.04 (Fig. 4b–d).

Our initial studies with five cell lines showed that only a subset of pancreatic cancer cell lines are susceptible to cyclopamine treatment. This could be the result of activating mutations of the HH pathway downstream of SMO, which should not be inhibited by this antagonist. To determine the percentage of cancer cell lines susceptible to cyclopamine treatment, we expanded our analysis to a total of 13 lines. Twelve of thirteen lines analysed expressed SMO; half of these (50%) responded to cyclopamine treatment. As predicted, growth of SMO-negative Panc 01.28 cells was not affected (Supplementary Fig. 3). Thus our data suggest that mutations in the HH pathway downstream of SMO might contribute to the aberrant proliferation of these cells. However, we cannot exclude the possibility that mutations in other, HH-independent, pathways regulate proliferation in cyclopamine-resistant pancreatic cancer cell lines.

To demonstrate that cyclopamine-induced inhibition of cancer cell proliferation is secondary to inactivation of the HH pathway, a Gli-luciferase reporter construct containing Gli-binding sites upstream of a TK minimal promoter²¹ was transfected into control and cyclopamine-responsive cell lines. GLI1 is a downstream transcription factor and transcriptional target of the Hh pathway²². All cell lines tested expressed GLI1 (Supplementary Fig. 2), and transfection of the Gli-reporter construct resulted in significant elevation of luciferase activity (Supplementary Fig. 4a), further indicating that Hh signalling is active in adenocarcinoma cells. Treatment of transfected cells with cyclopamine abolishes luciferase activity in responsive cell lines, whereas control cells display only an insignificant reduction in luciferase activity. Active HH signalling thus appears to be essential for cell survival and proliferation in a subset of cultured human pancreatic cancer cell lines.

As a more definitive test of the ability of cyclopamine to inhibit tumour cell proliferation, control (BxPC3) and cyclopamine-responsive cells (L3.6sl and Panc 05.04) were injected subcutaneously into immunocompromised nude mice (Fig. 5a). Two treatment models were used—concurrent or delayed treatments—in which cyclopamine was given either at the time of tumour implantation (concurrent) or when the subcutaneous tumours were palpable (delayed). Cyclopamine or carrier control was given for 7 days, at which time the tumours were excised and weighed. In the delayed treatment model, no difference in weight was noted between control and cyclopamine-treated BxPC3-SMO^low tumours (Fig. 5c). By contrast, a 50–60% decrease in tumour mass was observed in Panc 05.04- and L3.6sl-derived tumours, respectively (Fig. 5b, c)—an even more marked effect was noted in the concurrent treatment model, which revealed an 84% reduction in tumour mass of L3.6sl-derived tumours (Fig. 5b, c).

Histological analysis of L3.6sl-derived tumours (delayed treatment) revealed a compact mass of epithelial cells in untreated controls, whereas treated tumours appeared as loose epithelial cell aggregates with a larger amount of interspersed mesenchymal cells (Fig. 5d, e). TdT-mediated dUTP nick end labelling (TUNEL) assays demonstrated that cyclopamine treatment resulted in a sixfold increase in apoptotic cells compared with untreated controls (Fig. 5f–h). Therefore both in vitro and in vivo experiments reveal that HH signalling is important for proliferation and survival and thus may be an important mediator of the malignant biological behaviour of pancreatic adenocarcinomas in humans.

The observation that HH signalling has a critical and central role in pancreatic development and results in malignant transformation when mutated prompted us to ask about its potential role in pancreatic cancer^16,23–25. Our experiments lead us to believe that misregulation of HH signalling has a critical early role in the initiation and a later role in the maintenance of human pancreatic ductal adenocarcinoma, a hypothesis that is supported by findings in the accompanying paper³¹. Our evidence is based on comparative analysis of pancreata from Pdx1–Shh transgenic mice and of human adenocarcinoma. We find that HH signalling components are undetectable in normal human ductal epithelium but are strongly expressed in pancreatic precursor (PanIN) and invasive lesions. When Shh is misexpressed in Pdx–Shh mice, we detected the presence of abnormal pancreata with features that histologically, immunohistochemically and genetically resemble human PanIN lesions. Moreover, our analysis of human pancreatic cancer cell lines indicates that the HH signalling pathway is required for both cell division and the suppression of apoptosis. An important but unresolved question concerns the identity of the pancreatic cells that become transformed in response to increased HH signalling. Although pancreatic stem cells have not yet been isolated, results from studies in other tissues suggest that increased HH signalling in progenitor cells can elicit cancer formation. Future studies will be needed to elucidate the role of HH activity in stem-cell-mediated tissue regeneration and pancreatic adenocarcinomas.

Pancreatic cancer, the fourth highest cause of death from cancer in the United States, remains an incurable and rapidly lethal disease, with a 5-yr survival rate of less than 3%. Many factors contribute to this poor prognosis, but ultimately it is our lack of knowledge of the molecular determinants involved in the pathogenesis and progression of this tumour that has limited our ability to design effective treatments. Identification of a role for the HH pathway in the initiation and maintenance of pancreatic cancer suggests that this pathway may hold promise for new diagnostic and therapeutic approaches.