LETTERS A paracrine requirement for hedgehog signalling in cancer Robert L. Yauch 1 *, Stephen E. Gould 1 *, Suzie J. Scales 1 , Tracy Tang 1 , Hua Tian 1 , Christina P. Ahn 1 , Derek Marshall 1 , Ling Fu 1 , Thomas Januario 1 , Dara Kallop 1 , Michelle Nannini-Pepe 1 , Karen Kotkow 2 {, James C. Marsters Jr 1 , Lee L. Rubin 2 { & Frederic J. de Sauvage 1 Ligand-dependent activation of the hedgehog (Hh) signalling pathway has been associated with tumorigenesis in a number of human tissues 1–7 . Here we show that, although previous reports have described a cell-autonomous role for Hh signalling in these tumours 1–7 , Hh ligands fail to activate signalling in tumour epi- thelial cells. In contrast, our data support ligand-dependent activation of the Hh pathway in the stromal microenvironment. Specific inhibition of Hh signalling using small molecule inhibi- tors, a neutralizing anti-Hh antibody or genetic deletion of smoothened (Smo) in the mouse stroma results in growth inhibi- tion in xenograft tumour models. Taken together, these studies demonstrate a paracrine requirement for Hh ligand signalling in the tumorigenesis of Hh-expressing cancers and have important implications for the development of Hh pathway antagonists in cancer. An increasing number of solid tumours have been reported to be reliant on ligand-dependent Hh pathway signalling within the tumour epithelium 1–7 . Because the typical role for Hh in develop- ment is mediated by paracrine effects on mesenchymal cells 8,9 , we wanted to explore further the mechanisms behind Hh-dependent tumorigenesis. We have evaluated the effect of a specific and potent small molecule SMO-acting antagonist of the Hh pathway (HhAntag) on cell viability across a large panel of cancer cell lines (Fig. 1a) 10,11 . HhAntag has been demonstrated to be ,10-times more potent than the natural product SMO antagonist, cyclopamine, at inhibiting Hh pathway activity 10 . We observed a range of cellular sensitivities to HhAntag with half-maximal inhibitory concentration (IC 50 ) values for growth inhibition ranging from ,2 mM to .30 mM. In contrast to previous reports 6,7 , no tissue specificity of in vitro sensitivity to HhAntag was observed. The extent of basal Hh pathway activity in these cell lines, as measured by the expression of GLI1,a well characterized endogenous direct transcriptional target gene, did not correlate with cellular sensitivity to HhAntag (Fig. 1a), nor could we detect a positive correlation between cellular sensitivity and other Hh pathway genes (Supplementary Fig. 1 and Supplementary Table 1). Finally, these differences were not a unique attribute of the synthetic small molecule HhAntag, as a similar discordance of all variables was observed with cyclopamine (Supplementary Fig. 2) 12 . The requirement for high concentrations of HhAntag or cyclopa- mine to elicit growth inhibition and the lack of correlation with Hh pathway activation suggested that the observed in vitro growth repression might be due to non-specific effects of these molecules. This is supported by the observation that only 100 nM HhAntag is needed to completely inhibit Hh signalling in a Hh-responsive human mesenchymal cell line (HEPM) expressing a GLI luciferase reporter construct (HEPM-rep, Fig. 1b) and the IC 50 of 5 nM is ,400-times lower than that required to inhibit cell growth by 50% in the most sensitive cancer cell line (1.9 mM). To address specifically whether the in vitro growth inhibitory effect of high concentrations of Hh antagonists is due to off-target activity, we focused on two pan- creatic cell lines described recently: CFPAC-1, a SMO-positive and cyclopamine-sensitive cell line, and BxPC-3, a SMO-negative cell line with differentially reported sensitivity to cyclopamine 4,6,13 . Despite differences in SMO expression (Supplementary Table 1), we found that these cell lines exhibit similar sensitivities to the growth inhibi- tory effects of high concentrations of cyclopamine and HhAntag (Fig. 1a, b). Conversely, no growth inhibition was observed using two completely different mechanisms of Hh pathway inhibition: PKA activation using the agonist forskolin 14 (Fig. 1c) or the anti- Hh ligand blocking antibody 5E1 (Fig. 1d). Finally, neither Hh path- way activation with recombinant Sonic hedgehog homologue (rSHH) nor pathway inhibition with high concentrations (10 mM) of Hh antagonists had any effect on endogenous GLI1 messenger RNA levels in these cells (Fig. 1e), despite regulating pathway activity in HEPM cells. Previous reports relied on transfected GLI reporter constructs to monitor endogenous Hh pathway activity. rSHH was unable to activate GLI reporter activity when transfected into either CFPAC-1 or BxPC-3 cells (Supplementary Fig. 3). Although high concentrations of HhAntag and cyclopamine inhibited GLI reporter activity in CFPAC-1 and BxPC-3 cells, comparable repressions of unrelated transcriptional reporters were also observed at these high concentrations (Supplementary Fig. 4). Similar effects of Hh antago- nists on viability and signalling were observed across a variety of cell lines that have been used previously (Supplementary Fig. 5) 5–7 . Taken together, the data strongly argue against Hh signalling in these epi- thelial cancer cell lines and suggest that the previously observed effects of Hh antagonists on growth and reporter activity are due to off-target effects of these molecules when used at high concentrations. Despite our failure to detect Hh pathway activity in epithelial cells in vitro, we determined that subsets of human tumours and cell lines do indeed express Hh ligands. Microarray expression analysis of human tissue specimens revealed that subsets of colorectal, endomet- rial, ovarian and pancreatic cancers overexpressed Hh ligand mRNA (data not shown). Quantitative polymerase chain reaction with reverse transcription (RT–PCR) profiling of an independent set of human tissue specimens confirmed that the transcript levels of SHH and/or Indian hedgehog homologue (IHH) ligands were significantly upregulated in subsets of these cancers (Fig. 2a–c). In contrast, we did not observe increased SHH or IHH expression in a smaller panel of lung cancer specimens (Fig. 2d). Differential expression of SHH and 1 Genentech Inc., 1 DNA Way, South San Francisco, California 94080, USA. 2 Curis Inc., 45 Moulton Street, Cambridge, Massachusetts 02138, USA. {Present address: Harvard Stem Cell Institute, Harvard University, Biolabs Room 1065, 16 Divinity Avenue, Cambridge, Massachusetts 02138, USA. *These authors contributed equally to this work. Vol 455 | 18 September 2008 | doi:10.1038/nature07275 406 ©2008 Macmillan Publishers Limited. All rights reserved