A role for Hedgehog (Hh) signalling in the development of colorectal cancer (CRC) has been proposed. In CRC and other solid tumours, Hh ligands are upregulated; however, a specific Hh antagonist provided no benefit in a clinical trial. Here we use Hh reporter mice to show that downstream Hh activity is unexpectedly diminished in a mouse model of colitis-associated colon cancer, and that downstream Hh signalling is restricted to the stroma. Functionally, stroma-specific Hh activation in mice markedly reduces the tumour load and blocks progression of advanced neoplasms, partly via the modulation of BMP signalling and restriction of the colonic stem cell signature. By contrast, attenuated Hh signalling accelerates colonic tumourigenesis. In human CRC, downstream Hh activity is similarly reduced and canonical Hh signalling remains predominantly paracrine. Our results suggest that diminished downstream Hh signalling enhances CRC development, and that stromal Hh activation can act as a colonic tumour suppressor.
The Hedgehog (Hh) signalling pathway is one of the major regulators of embryonic development. Later in life, it can drive tumorigenesis: mutations that lead to cell-autonomous Hh pathway activation cause nearly all basal cell carcinomas, as well as subsets of medulloblastomas and rhabdomyosarcomas1. By contrast, the majority of solid tumours, including colorectal cancer (CRC), rarely carry mutations in Hh genes2. Instead, in CRC, Hh ligand expression is upregulated and conflicting data suggest either a paracrine role for Hh ligands in shaping a tumour-supportive microenvironment3 or autocrine pathway activation that promotes metastasis4.
Under homeostatic conditions, the main intestinal ligand, Indian hedgehog (Ihh), is secreted by differentiated enterocytes, whereas downstream signalling is activated exclusively in the stroma5. In the ‘canonical’ Hh signalling cascade, binding of the ligand to the inhibitory receptor Patched 1 (Ptch1) leads to derepression of the activating receptor Smoothened (Smo), thereby initiating a signalling cascade that culminates in the stromal activation of the Glioma-associated oncogene (Gli) proteins, Gli1, Gli2 and Gli3 (ref. 5). Gli1 expression is considered the most reliable indicator of downstream pathway activity, whereas Hh interacting protein (Hhip), Ptch1 and its homologue Ptch2 are further common downstream targets6.
The SMO antagonist, vismodegib, has recently been approved for the treatment of basal cell carcinomas7. Motivated by the upregulation of Hh ligands in CRC, a clinical trial with vismodegib added to first-line therapy in metastatic CRC was recently completed, but yielded a negative result8. A trial with vismodegib in pancreatic cancer, in which Hh ligands are similarly overexpressed, had an equally discouraging outcome9, whereas a further pancreatic cancer study using another SMO antagonist was halted due to an inferior outcome in the inhibitor-treated group10. Collectively, these clinical data challenge the paradigm of a tumour-promoting stroma shaped by Hh signalling and contest a putative oncogenic role for cancer cell-autonomous Hh activation in these tumour types.
Functionally, the stromal response to the Hh ligand is part of a paracrine loop that controls differentiation of the intestinal epithelium11. Diminished Hh signalling evokes an expansion of the intestinal stem cell compartment and leads to impaired enterocyte differentiation, as well as activation of Wnt signalling, the central oncogenic driver pathway in CRC11,12,13,14.
Given this discrepancy between data suggesting a tumour-promoting role for Hh in CRC on one hand and its pro-differentiating role under homeostatic conditions, together with negative results from clinical trials, on the other hand, we sought to get a more precise picture of the role played by Hh in colorectal tumourigenesis. Using Hh reporter mice, we provide evidence that downstream Hh signalling activity is reduced in murine colon tumours. Functionally, diminished Hh signalling promotes colitis-associated colonic tumourigenesis in mice, whereas stroma-specific Hh activation markedly curtails tumour development. Similarly, human CRCs harbour diminished expression of Hh downstream targets despite upregulated expression of the ligand, Sonic Hh (SHH). Together, our data suggest that stromal activation of Hh signalling can function as a suppressor of colorectal carcinogenesis.
Reduced stromal Hh signalling in murine colonic tumours
To induce colonic tumours in mice, we employed a chemical model based on injection of the mutagenic agent azoxymethane (AOM), followed by repeated oral treatment with dextran sodium sulphate (DSS) to induce epithelial damage and subsequent colitis (Fig. 1a)15. Tumours in this model recapitulate central aspects of human CRCs as they consistently exhibit mutations in the genes for adenomatous polyposis coli (Apc) or catenin β1 (Ctnnb1)16, and, less frequently, in Kras17. We consider this model particularly well suited to our purposes, as it can be combined readily with Cre-LoxP models to modify epithelial and mesenchymal Hh signalling, and because tumours arise specifically in the colon but not in the small intestine.
To assess downstream Hh signalling activity, we first subjected Gli1 reporter mice harbouring a β-galactosidase knock-in to Gli1 (Gli1lacZ/+ mice18), to AOM/DSS treatment, and then visualized Gli1 expression with whole-mount X-gal staining. Although non-malignant mucosa stained strongly, X-gal staining was weak to absent in AOM/DSS-induced tumours (Fig. 1b). Histological analysis revealed Gli1 expression exclusively in the stroma and reduced X-gal staining in the tumours (Fig. 1c,d).
Real-time quantitative PCR confirmed reduced Gli1 expression in tumours from wild-type (wt) mice, with a congruent reduction of the Hh targets, Hhip and Gli2 (Fig. 1e). Remarkably, expression of the receptor, Ptch1, and its homologue, Ptch2, as well as the main intestinal ligand, Ihh, was dissociated from attenuated downstream Hh signalling (Fig. 1e), whereas all tumours demonstrated increased Wnt activity, as expression of the Wnt targets, Axin2 and Leucine-rich repeat-containing G-protein coupled receptor 5 (Lgr5) was upregulated (Fig. 1e). RNA in situ hybridization (ISH) confirmed that downregulation of stromal Gli1 can occur despite high Ihh expression in the adjacent tumour cells (Fig. 1f).
Given that Hh downstream signalling in the intestine is exclusively stromal5, the reduced expression of downstream Hh targets could be the result of a reduction in the number of stromal cells. To explore this possibility, we used immunofluorescence (IF) to stain for the intermediate filaments desmin, vimentin and α-smooth muscle actin simultaneously. This marker combination covers the majority of normal and cancer-associated non-inflammatory stromal cells, for which no specific single marker exists19. However, we saw no differences in the amount of stromal cells between normal mucosa and adjacent tumours, indicating that stromal cell loss is not the reason for diminished expression of Hh downstream targets (Supplementary Fig. 1a–c).
Immunohistochemistry (IHC) of β-catenin in tumours from Gli1lacZ/+ mice revealed its nuclear translocation in areas of low stromal Gli1 expression (Fig. 1g), substantiating a model in which the stromal downstream Hh signal is diminished in areas of high epithelial Wnt activity. In line with this, we found that the expression of p-Smads1/5—markers of epithelial differentiation driven by bone morphogenetic protein (BMP) signalling that are highly expressed in differentiated enterocytes20—correlated spatially with stromal Gli1 (Fig. 1h).
Hence, colonic tumour development in the AOM/DSS model is unexpectedly associated with a diminution in the stromal Hh response. To explore Hh expression in a different colon cancer model, we probed RNA-sequencing data from tumours in which Apc activity is regulated by a doxycycline-dependent short hairpin RNA (shRNA) (Gene Expression Omnibus (GEO) data set GSE67186 and ref. 21). We found that these non-inflammatory colon tumours recapitulated the expression pattern of the AOM/DSS model, in that expression of the downstream Hh targets, Gli1, Gli2, as well as Gli3, was reduced significantly and was dissociated from expression of the ligands, Ihh (unchanged) and Shh (increased), as well as the receptor Ptch1 (increased) (Supplementary Fig. 2a–c).
Increased tumour burden upon diminished Hh signalling
In the murine colon, Ihh messenger RNA expression exceeds that of Shh by several orders of magnitude (Supplementary Fig. 2a and refs 22, 23), and an Shh conditional knockout model revealed a limited phenotype strictly confined to the terminal ileum24. To assess whether reduced Hh signalling can promote malignant transformation in the colon, we used a genetic model in which the main intestinal ligand, Ihh, can be knocked out by Tamoxifen (Tam) administration to VillinCreER;Ihhflox/flox;R26-LSL-ZsGreen mice (hereafter IhhΔVil), simultaneously activating expression of a conditional fluorescent reporter. Although prolonged loss of Ihh in intestinal epithelial cells leads to intestinal stem cell accumulation, loss of differentiated cells and activation of the Wnt pathway, it does not, by itself, result in adenoma formation12. However, when we exposed IhhΔVil mice to AOM/DSS (Fig. 2a), we found a significant increase in the tumour burden compared with controls (Fig. 2b). The ZsGreen reporter (indicating successful Cre-mediated recombination) persisted throughout normal and tumourous epithelium, demonstrating a permanent loss of Ihh and strongly suggesting that tumour development is not dependent on Ihh expression (Fig. 2c). Indeed, Ihh mRNA was barely detectable in IhhΔVil mice, Shh did not increase significantly and stromal downstream Hh targets were reduced (Fig. 2d and Supplementary Fig. 3a,b). It is worth noting that IhhΔViI mice lost more weight than controls after DSS treatment and had a higher mortality, which may imply that enhanced inflammatory activity contributes, at least partly, to the increased tumour frequency in this model (Supplementary Fig. 3c,d).
In a subsequent experiment, we used vismodegib to antagonize canonical Hh signalling at the level of the receptor Smo (Fig. 2e). Twice daily administrations of vismodegib (25 mg per kg body weight) led to a significant reduction in downstream Hh target gene expression (Supplementary Fig. 3e). We treated C57BL/6 wt mice with either vismodegib or vehicle and employed high-frequency ultrasound (US) with intraluminal contrast to assess polypoid colonic lesions at high resolution (Supplementary Movies 1 and 2 and ref. 25). We found that inhibition of canonical Hh signalling with vismodegib had an effect similar to the Ihh knockout, in that it led to more frequent and larger tumours (Fig. 2f,g). However, we observed no significant differences in terms of weight loss or mortality due to colitis (Supplementary Fig. 3f,g).
Taken together, the data demonstrated that canonical downstream Hh signalling is attenuated in murine colorectal carcinogenesis, and that Hh inhibition can further accelerate tumour development.
Stroma-specific activation of Hh signalling
To visualize Hh target cells in a near-native setting, we crossed Gli1CreERT2 mice to Rosa26-LSL-tdTomato reporter mice, yielding inducible Gli1CreERT2;Rosa26-LSL-tdTomato mice. Seven days after administration of Tam, we found tdTomato+ cells exclusively residing in the stroma, including a mixed population of cells expressing the cytoskeletal proteins desmin, vimentin and α-sma (Fig. 3a). Gli1 expression remained exclusively stromal after DSS-induced epithelial damage, as indicated by analysis of both the Gli1lacZ/+ and Gli1CreERT2;tdTomato reporter mice (Supplementary Fig. 4a–d), indicating that downstream Hh signalling is not activated ectopically in damaged epithelium.
In Gli1CreERT2 mice, Cre is knocked in to the endogenous Gli1 locus18; thus, the loss of one functional Gli1 allele alters downstream Hh signalling. To circumvent this problem in further experiments aimed at activating Hh signalling specifically in stromal cells, we used mice expressing Cre from a transgenic collagen, type I, α2 (Col1a2) promoter26. Analysis of Tam-induced Col1a2CreER;R26-LSL-tdTomato reporter mice confirmed recombination in an exclusively stromal cell population (Fig. 3b) and collagen 1 protein expression overlapped with tdTomato expression in Gli1CreERT2;Rosa26-LSL-tdTomato mice (Fig. 3c). We found that Col1a2CreER targets a persistent cell population as indicated by abundant tdTomato+ fibroblasts 100 days after recombination (Fig. 3d). As the Col1a2 transgene is expressed in a wide range of organs26, we administered the active metabolite 4OH-Tam intraluminally, to minimize systemic effects in subsequent experiments, where we employed Col1a2CreER-expressing mice to activate Hh signalling specifically in the stroma (Fig. 3e and Supplementary Fig. 5).
Hh activation antagonizes colonic tumour development
To activate Hh signalling in Col1a2+ cells, we crossed Col1a2CreER mice to Ptch1neo[Δ]Ex2[Δ] mice with LoxP-sites flanking exon 2 of Ptch1 (ref. 27), creating Col1a2CreER;Ptch1flox/+ mice (hereafter Ptch1Col1(het)mice). In this model, Tam treatment leads to the knockout of one Ptch1 allele specifically in Col1a2-expressing stromal cells. Challenged with AOM/DSS (Fig. 4a), Ptch1Col1(het) mice developed significantly fewer and smaller tumours than controls, revealing a protective effect of stromal Hh activation on colorectal carcinogenesis (Fig. 4b,c). Following the loss of one Ptch1 allele, Gli1 expression increased significantly in tumour tissue, whereas the effect in normal mucosa was less pronounced, probably reflecting background Hh activity in homeostasis (Fig. 4d). In line with this, we found no phenotypical differences between wt littermates and Ptch1Col1(het) mice during a 6-month follow-up after recombination (data not shown). Moreover, we observed no differences between Ptch1Col1(het) mice and control animals with respect to their sensitivity to DSS treatment or mucosal immune cell composition at the endpoint of the AOM/DSS protocol (Supplementary Fig. 6a,b). In tumours from Ptch1Col1(het) mice, Gli1 transcripts were enriched in the stromal compartment as expected (Supplementary Fig. 6c,d), yet the amount of stroma did not differ significantly between tumours from control animals and those from Ptch1Col1(het) mice, as assessed by IF (Supplementary Fig. 6e). Importantly, treatment of Ptch1Col1(het) mice with vismodegib (Fig. 4e) overcame the protective effect of Hh activation, indicating that it is largely mediated in a Smo-dependent manner (Fig. 4f,g). Using Ptch1Col1(het) mice, treated with vismodegib, and harbouring an additional inducible tdTomato allele, we confirmed that Col1a2CreER-targeted cells remained exclusively stromal throughout the AOM/DSS protocol (Fig. 4h), showing that Hh activation remains confined to the stroma in this model.
The loss of one Ptch1 allele had a marked effect on tumour development, yet it did not detectably influence the course of the DSS-induced colitis. Nevertheless, a role for Hh in modulating the immune infiltrate in the colon cannot be excluded and previous studies have suggested that stromal Hh signals can act as modifiers of the intestinal immune response11,14. To investigate the effect of Hh activation in a sporadic carcinoma model, we induced tumours in Ptch1Col1(het) mice and controls with consecutive AOM injections in the absence of DSS15 (Fig. 4i). In accordance with AOM/DSS-treated mice, stromal Hh activation attenuated tumour development in this model (Fig. 4j,k).
Collectively, our results indicated that Hh-driven stromal signals can counteract the neoplastic transformation of overlying epithelial cells.
Diminished BMP inhibitor expression upon Hh activation
To investigate the functional consequences of stromal Hh activation in greater depth, we performed gene expression analysis on colonic tissue from Ptch1ΔCol1 mice, in which one pulse of Tam induces homozygous inactivation of Ptch1 in the stromal compartment, resulting in activation of Hh signalling in normal mucosa.
Gene expression data confirmed the upregulation of Hh targets (false discovery rate (fdr)=0.001), dominated by Gli1, Ptch1, Ptch2, and Hhip (Fig. 5a,b upper left panel and Supplementary Tables 1 and 2). Gene set enrichment analysis (GSEA) indicated a significant enrichment of transcripts associated with enterocyte differentiation (fdr=0.086), such as intestinal alkaline phosphatase (Alpi) or the brush border myosin Ia gene (Myo1a) (Fig. 5b upper right panel and Supplementary Table 3).
By contrast, Hh activation led to reduced expression of colonic stem cell-associated genes (fdr=0.0313), including Lgr5, Cdca7 and Cdk6, all of which are restricted to the stem cell compartment at the crypt base (Fig. 5b lower left panel and Supplementary Table 4)28. ISH confirmed the reduced expression of Lgr5, whereas Gli1 was increased as expected (Supplementary Fig. 7a,b). These effects of augmented stromal Hh signalling were paralleled by a marked reduction in secreted BMP inhibitors such as Gremlin 1 (Grem1) and Noggin (Nog) (fdr=0.092) (Fig. 5b lower right panel and Supplementary Table 5). Interestingly, Gdf10 (also known as Bmp3b), a member of the BMP family exhibiting Hh-dependent expression, at least in the brain29, was among the top 50 upregulated genes (Fig. 5a).
Stroma-specific Hh activation entailed further changes, including alterations in known modifiers of intestinal tumourigenesis. For example, the mesenchymal factor, Foxl1, which was upregulated in Ptch1ΔCol1 mice (Fig. 5a), is directly controlled by Gli proteins30 and attenuates tumour development in a small intestinal adenoma model31. As a further example, the gene encoding plasma glutathione peroxidase (Gpx3), which has a protective effect in the AOM/DSS model32, was similarly upregulated (Fig. 5a).
To identify genes that show robust alterations in expression as a result of Hh activation specifically in stromal cells, we analysed two independent data sets: (i) human colonic fibroblasts (Fig. 5c,d) and (ii) murine intestinal mesenchymal cells (Fig. 5e,f), treated with Hh ligands in vitro (based on the GEO data sets GSE17840 and GSE29316 (refs 14, 33)). In both cases, activation of canonical Hh signalling was paralleled by significant upregulation of Foxl1/FOXL1 and significant downregulation of the BMP inhibitor Grem1/GREM1 (Fig. 5c–f). Other BMP inhibitors such as Nog/NOG, Chrdl1/CHRDL1 and Sostdc1/SOSTDC1 were significantly downregulated in one of the data sets, but not in both (Supplementary Table 6). Based on these data, we revisited tumours from Ptch1Col1(het) mice, in which Gli1 expression is increased (Fig. 4d), and analysed the expression of selected BMP inhibitors using real-time quantitative PCR. In line with the in vitro studies, we found that Grem1 expression is downregulated in tumours from Ptch1Col1(het) mice compared with controls (Fig. 5g).
Together, the data indicated that stromal Hh activation leads to complex transcriptional changes with a prominent, but not exclusive, role for the modulation of BMP signalling. These changes result in restriction of the colonic stem cell state signature and induction of epithelial differentiation markers.
Hh activation blocks the growth of established tumours
AOM/DSS treatment of Ptch1ΔCol1 mice led to severe morbidity and mortality, precluding their use for the study of tumour initiation (Supplementary Fig. 7c,d,e). Without AOM/DSS treatment, Ptch1ΔCol1 mice had to be killed several weeks after Tam treatment, as they developed large, cystic stromal tumours predominantly at the ileoceacal junction, which occasionally ruptured or led to gross abdominal distension (Supplementary Fig. 7f,g). Similar tumours have been described in a related model, in which Ptch1 was specifically inactivated in a more restricted stromal compartment34.
Nevertheless, we were able to induce tumours with AOM/DSS in Ptch1ΔCol1mice that had not received Tam previously and we used this model to activate Hh signalling in established tumours. It has recently been shown that high levels of Wnt signalling are necessary for tumour maintenance, and that restoration to homeostatic Wnt levels causes tumour regression21. This prompted us to ask whether activated stromal Hh signalling might restrain tumour growth in a similar manner. Once AOM/DSS-induced lesions of similar sizes were detected, we used three-dimensional (3D) μUS to monitor tumour volumes over time in Ptch1ΔCol1 mice and control mice after Tam administration (Fig. 6a,b). Using Col1a2CreER;R26-LSL-tdTomato mice, we confirmed that tdTomato+ cells were directly adjacent to the tumour cells (Fig. 6c).
One week after Tam treatment, delayed growth was evident in tumours from Ptch1ΔCol1 mice compared with littermate controls (Fig. 6d). Over the course of several weeks, tumours with Hh-activated stroma frequently entered permanent macroscopic growth arrest or, in the case of smaller lesions, showed regression, whereas tumours in control animals progressed (Fig. 6d and Supplementary Movies 1 and 2). IHC analysis of cleaved caspase-3 (Fig. 6e) and Ki67 (Fig. 6f) indicated increased tumour cell apoptosis and diminished proliferative activity, respectively, following stromal Hh activation.
To assess whether Grem1 is reduced as a consequence of stromal Hh activation in the tumours, we used RNA ISH and semi-quantitatively evaluated the staining intensity35. Although Grem1 was strongly expressed in the tumour stroma of control mice, expression was significantly reduced in Ptch1ΔCol1 mice (Fig. 6g). In addition, tumour-epithelial p-Smad1/5 staining was consistently more abundant in tumours from Ptch1ΔCol1 mice, indicating increased epithelial BMP downstream activity as a result of the augmented stromal Hh signal (Fig. 6h).
Diminished downstream Hh response in human colon cancer
Based on the unexpected expression patterns of Hh pathway members in murine tumours, we decided to re-assess Hh expression in human CRC in more detail, including the pathways’ downstream targets. To this end, we first analysed microarray gene expression data from The Cancer Genome Atlas (TCGA), representing 155 colon cancer patients2. Although SHH was upregulated in carcinomas as previously reported3, IHH was moderately downregulated (Fig. 7a). As in the mouse model, the downstream targets, GLI1 and HHIP, were downregulated (Fig. 7b) and, strikingly, no correlation between GLI1 and Hh ligand expression was observed (Fig. 7c). In addition, we found that the BMP targets, ID1 and ID2, were significantly downregulated in the carcinomas, whereas GREM1 and NOG expression increased, albeit not significantly in the case of GREM1 (Fig. 7d). Moreover, expression of the BMP ligands, GDF10/BMP3b and BMP5, was reduced significantly (Fig. 7d).
In some colon cancer cell lines with mutations in KRAS or BRAF, cancer cell-autonomous GLI1 activation has been described36. However, we found no correlation between the presence of mutated KRAS or BRAF and GLI1 mRNA levels (Supplementary Fig. 8a). Instead, there was a weak but significant correlation between mutations in the APC gene and low GLI1 levels (Supplementary Fig. 8a).
GLI1 mRNA levels correlated well with stromal gene expression signatures in the TCGA data set (Supplementary Fig. 8b), indicating that activation of downstream Hh signalling is largely confined to the stromal compartment. To confirm this further, we probed gene expression data from CRC patient-derived xenografts, in which murine stromal cells replace the human stroma after transplantation (based on GEO GSE56710 and GSE35144 data sets)37. In the xenografts, expression of the ligands, SHH and IHH, was exclusively epithelial, whereas downstream signalling was largely restricted to the stroma, as transcripts of GLI1 (>75%) and GLI2 (95%), as well as HHIP (100%) were predominantly or exclusively murine (Fig. 7e).
To assess gene expression in the different cellular compartments in a more quantitative manner, we analysed Hh and BMP pathway components in CRCs sorted for epithelial cells (CD45−, Epcam+), leukocytes (CD45+, Epcam−) and stromal cells (CD45−, Epcam−; based on the GEO39395 data set)38. We found a similar pattern to that in the xenografts, such that downstream Hh targets were enriched in the stroma to a highly significant degree, whereas ligand expression was epithelial (Fig. 7f,g). GREM1 and the BMP ligand BMP5 were enriched significantly in the stromal fraction, in keeping with the results from the mouse model, whereas tumour cells expressed BMP4 in several cases. Of note, the Hh co-receptor CDON and PTCH1 were expressed in the tumour cells at a level similar to that seen in the stroma, although this did not lead to increased expression of Hh downstream targets in the epithelial compartment (Fig. 7f).
In summary, the results obtained from different CRC cohorts demonstrated that human colon cancer exhibits a marked decrease in downstream Hh target gene expression, which is predominantly, if not exclusively, of a stromal nature.
Under homeostatic conditions, the mode of intestinal Hh signalling is exclusively paracrine, from the ligand-secreting epithelium to the receptive underlying stroma11,39. However, a role for cell-autonomous downstream Hh activity in colon cancer cells has been suggested, in particular at later tumour stages, in which Hh activation drives metastasis4,40.
In the current study, our analyses of stromal and epithelial gene expression strongly suggest that there is little, if any, cell-autonomous Hh downstream activity in colon cancer cells (Fig. 7e,f). Together with the marked overall downregulation of GLI1 and HHIP (Fig. 7b), these results provide consistent evidence that stromal downstream Hh activity is diminished in colon cancer. However, as some CRC cell lines are dependent on GLI1 (refs 36, 40), it is possible that downstream Hh signalling might play a role in a tumour cell subpopulation. On the other hand, most CRC cell lines do not express a complete set of Hh downstream targets41. Aside from a role for canonical Hh signalling, it has previously been suggested that Hh ligands can prevent colon cancer cells from entering apoptosis by binding to the dependence receptor, CDON, in an autocrine manner (independent of the role of CDON as a Hh co-receptor)42. As we find that CDON is expressed in the tumour cell compartment (Fig. 7f), this leaves a possible tumour-promoting role for a non-canonical SHH/CDON/PTCH1 axis, acting independently of downstream Hh signalling (Fig. 7h).
In contrast to carcinomas, human colonic adenomas from patients with familial adenomatous polyposis as well as small intestinal adenomas from Apclox15 mice exhibit increased levels of IHH/Ihh and GLI1/Gli1 (ref. 12). In Apclox15 mice the presence of Ihh is required for adenoma development12, whereas in the AOM/DSS model Ihh loss fuels tumourigenesis (Fig. 2a). Importantly, treatment of later stage small intestinal murine adenomas with Hh inhibitors promoted, rather than restricted, tumour growth in the Apcmin model12. Almost all functional data on intestinal Hh signalling from in vivo models have thus far been acquired in the small intestine11,12,13,14,43. However, the tissue architecture and cellular composition of the small intestine differ considerably from those of the colon, in which the vast majority of human intestinal cancers arise. It is therefore noteworthy that colonic tumours in mice, whether they are induced chemically with AOM/DSS (Fig. 1), or genetically, as in the ShApc model21 (Supplementary Fig. 2), recapitulate the Hh expression pattern seen in human CRCs12. Hence, tissue-specific properties of the small intestine and colon may explain the different requirements in the two locations with regard to dependence on Hh signalling.
Across various colon cancer mouse models and in human CRC, downstream Hh activity can be lost despite increased ligand expression (Figs 1 and 7, and Supplementary Fig. 2). This might be the result of alterations in stromal gene expression programmes that render stromal cells indifferent to the epithelial ligand44. An alternative explanation has recently been proposed by Shyer et al.45, who found that during development Hh ligands are concentrated at the villus tips merely as a result of the physical folding of the epithelium. This leads to Hh activation in the stroma and differentiation of the epithelium at the villus tips45. It is tempting to speculate that, as a result of perturbations in the tissue architecture during tumourigenesis, ligand concentrations are insufficient to activate downstream Hh signalling in the tumour stroma. Interestingly, in the Apc-shRNA model, the expression of Gli1, Gli2, and Hhip is consistently reduced further in tumours carrying an additional mutation in Kras (Supplementary Fig. 2c), which show higher levels of dysplasia than tumours in Apc-shRNA animals, as well as invasive behaviour, supporting a model in which disrupted tissue architecture interferes with ligand availability and induction of the stromal Hh response21. Moreover, in our hands, AOM/DSS-induced tumours occasionally displayed signs of invasion through the muscularis mucosae (Supplementary Fig. 9 and refs 16, 46).
Although the AOM/DSS model recapitulates the Hh expression pattern of human CRC as well as parts of the mutagenic landscape of sporadic human colon cancers16, tumour development is accelerated by colitis, which is not a self-evident feature of most human carcinomas. Nevertheless, sporadic colorectal carcinogenesis is characterized by, and is dependent on, a complex inflammatory cell infiltrate and immunocompromised mice are protected from both inflammatory-associated and sporadic colon tumours47,48. On a clinical note, our data indicate that caution is warranted particularly when using Hh antagonists in patients with inflammatory bowel disease, who are at an increased risk of developing CRC49. However, together with data from pancreatic and bladder cancer, in which Hh antagonism similarly accelerates carcinogenesis50,51, our results point to the possibility that treatment with Hh antagonists could increase the risk for several solid malignancies, including CRC.
Functionally, the data suggest a dominant, but not exclusive role for BMP modulation as a mediator of stromal–epithelial crosstalk. BMP signalling is active in adenomas, but is downregulated in CRCs52 (Fig. 7d) and thus mirrors the expression pattern of downstream Hh effectors. In sporadic CRC, high GREM1 expression is associated with treatment resistance and inferior survival53. Moreover, ectopic epithelial expression of GREM1 causes hereditary mixed polyposis syndrome, which is characterized by the appearance of multiple colonic polyps that frequently progress to carcinomas54. This can be modelled in mice, where ectopic Grem1 expression leads to the reacquisition of stem cell properties in differentiated cells53, which are consequently able to initiate intestinal neoplasia. We show that, as a consequence of Hh activation, attenuated BMP inhibitor expression is conversely associated with diminution of the intestinal stem cell signature (Fig. 5a,b). Thus, it seems plausible that loss of stromal BMP inhibitors due to the activation of stromal Hh signalling (Fig. 5b) may have the potential to restrain colonic tumour initiation and progression (Fig. 7h).
The importance of paracrine Hh signalling in tumour–stroma cross-talk has been recognized for more than a decade—yet accumulating data now indicate that Hh-driven signals might represent exploitable targets that are not limited to colon cancer therapy, but extend to other cancer types.
C57BL/6 mice (referred to as wt) were obtained from Scanbur, Sweden, and used at >8 weeks of age. All mice were on a C57BL/6 background. Gli1lacZ/+ and GliCreERT2 mice were a kind gift from Fritz Aberger, whereas R26-LSL-tdTomato (B6.Cg-Gt[ROSA]26Sortm14[CAG-tdTomato]Hze/J) and Col1a2CreER (B6.Cg-Tg[Col1a2-cre/ERT]7Cpd/J) mice were obtained from the Jackson Laboratory (Bar Harbor, USA). Ptch1fl/+ mice have been described previously27. For IhhΔVil mice, VillinCreERT2 mice were crossed with Ihhflox/flox mice and Rosa26-ZSGreen mice, to generate VillinCreERT2-ZsGreen-Ihhflox/flox animals.
Mice receiving DSS were assessed daily in a cumulative score until recovery (weight, movement and body posture, piloerection, skin condition, food consumption, excretions and breathing). Littermates were used as controls and experimental groups were housed in one cage wherever possible. In the case of Cre/Lox-p models, the control animals lacked either the Cre or floxed alleles or both.
Chemical models of CRC
AOM (Sigma-Aldrich, A5486) at 12.5 mg kg−1 bodyweight was administered intraperitoneally (i.p.) on day 0 and DSS (DB001; Tdb Consultancy, Sweden) was initiated at a concentration of 2% on day 5 and was administered for 5 days15. DSS cycles were repeated twice with 14 days of normal drinking water in between. All DSS preparations were renewed after 2 days. For the sporadic model, AOM was administered i.p. once a week at 12.5 mg kg−1 body weight for 10 consecutive weeks.
For IhhΔVil mice specifically, 10-week-old animals were injected i.p. with 1 mg Tam (T5648; Sigma-Aldrich) for 5 consecutive days. Two weeks after Tam initiation, mice were injected with AOM (10 mg kg−1) and received drinking water supplemented with 1.75% DSS for 4 days, which was then changed to untreated drinking water. During weeks 5 and 8, the mice received 1.25% DSS for 4 and 3 days, respectively. The DSS concentration and treatment duration were adjusted due to augmented weight loss. After 12 weeks, the mice were killed.
Vismodegib and Tam treatment
Vismodegib (V-4050, LC Laboratories) was administered i.p.55,56 in dimethyl sulfoxide twice daily at 25 mg kg−1 body weight for 6 days per week.
For intracolonic administration, 4OH-Tam (H6278; Sigma-Aldrich) was dissolved in 99.5% ethanol, vortexed, further diluted in corn oil (S5007, Sigma-Aldrich) to reach the working concentration and sonicated for 30 min at 37 °C in a water bath. The solution was kept at 37 °C in darkness until it was instilled intraluminally at 1 mg 25 g−1 bodyweight. Instillation was performed with US guidance under isoflurane anaesthesia (described below). For visualization of the instilled substance with microcomputed tomography (μCT) imaging, 0.1 mg ml−1 barium sulphate (Barium Sulphate (#11381499) 97%, Alfa Aesar) was added to the Tam/oil suspension. The μCT was performed under isoflurane anaesthesia (see below) using a Quantum FX μCT imaging system (Caliper Life Sciences, Perkin Elmer, Inc.) a few minutes after administration of the Tam/oil/barium mixture.
For systemic labelling experiments in Col1a2CreER;R26-LSL-tdTomato and Gli1CreERT2;R26-LSL-tdTomato mice, 5 mg of Tam was dissolved in corn oil using sonication at 37 °C at 20 mg ml−1 and injected once i.p.
The Vevo 2100 system (Visualsonics, Canada) was used for high-frequency US57. Mice were anaesthetized with 1–2% v/v isoflurane (Baxter, KDG9621). Hair over the abdomen was removed using a depilatory cream. The mice were scanned in a supine position on a heating plate (40 °C) with constant cardiovascular monitoring and protective eye ointment. Scans were performed at 40 MHz (probe MS550D; axial resolution 40 μm) using a 3D-motor for z axis resolution57. For administration of intraluminal contrast during colonic imaging, a plastic gavage tube was advanced carefully via the anus and small amounts of US gel were injected to inflate the colonic lumen gently under US surveillance. The largest tumour cross-sectional area was determined visually. For 3D scanning, the z resolution was set to 76 μm and the whole tumour length was scanned. Tumour outlines were drawn manually for each slice and 3D reconstruction was performed using Visualsonics Vevo software (v1-5.0).
For IF, the colon was removed, faeces were gently washed out with cold PBS and the tissue was immediately frozen on dry ice in Tissue-Tek O.C.T. compound and stored at −80 °C. Sections were cut using a cryotome at 20–100 μm and incubated for 30 min in PBS with 0.5% Triton X-100 (Sigma-Aldrich; T8787) and 0.1% TOPRO-3 (Invitrogen; T3605). Slides were mounted between two coverslips (Menzel, Nr. 1.5 (0.16–0.19 mm)) and scanned using a Zeiss LSM710 confocal microscope in single-photon mode. For antibody staining, 20–100 μm sections were cut in a cryotome and incubated overnight in antibody diluted in PBS plus 0.5% Triton-X-100, 0.5% BSA and 0.5% dimethyl sulfoxide at 4 °C. A species-appropriate secondary antibody (1:250 dilution) was chosen from Alexa Fluor dyes (Invitrogen) using an appropriate wavelength depending on the primary antibody and the presence of endogenous fluorophores. The antibodies and dilutions were as follows: anti-vimentin (Santa Cruz, sc7557, 1:500); anti-desmin (Abcam, ab8592, 1:500); anti-α-sma (Abcam, ab5694, 1:500); anti-collagen1 (Abcam, ab34710, 1:200); anti-β-catenin (Cell Signalling, #9582, 1:100); anti-Ita6 (CD49f, BD Pharmigen, # 555734, 1:100).
IHC and X-gal staining
For IHC, the colon was incised longitudinally and fixed in 4% paraformaldehyde (PFA) in PBS overnight at 4 °C before routine processing and paraffin embedding. Next, 4 μm sections were baked at 60 °C for 1 h and rehydrated.
The sections were subjected either to a heat-induced epitope retrieval step or to proteinase-induced epitope retrieval, depending on the primary antibody used. Heat-induced epitope retrieval was performed in citrate buffer (pH 6) or DIVA reagent (Biocare Medical, DV2004) using a pressure cooker. Protein-induced epitope retrieval was performed in a 37 °C water bath using 0.05% protease (Sigma). Sections were incubated with antibodies against B220 (eBioscience; clone RA3-6B2; 1:200), CD3 (Dako; #N1580; 1:20), F4/80 (Invitrogen; clone BM8; 1:100), MPO (Dako #A0398; 1:500), β-catenin (Cell Signalling; #9582; 1:100), phospho-smad1/5 (Ser463/465; Cell Signalling; 9516S; 1:50), cleaved caspase-3 (Asp175; Cell Signalling; 9661S; 1:200) or Ki67 (Novocastra; NCL-Ki67p; 1:200) for 1 h at room temperature (RT). After rinsing, the sections were incubated with a biotinylated secondary antibody: donkey anti-rabbit, rabbit anti-rat antibody (Dianova) or rabbit anti-goat (Vector, BA-5000), followed by incubation with alkaline phosphatase-labelled streptavidin (Dako; CD3, B220, F4/80, MPO, β-catenin) for 30 min at RT or streptavidin peroxidase (Invitrogen; for all others). Alkaline phosphatase was revealed with Fast Red (Dako; AR17911-2) and peroxidase was revealed with DAB (Invitrogen; DAB plus reagent kit; #002020). Sections were counterstained with haematoxylin or eosin (for X-gal stained tissue) and slides were mounted with water-based mounting medium (Polyscience; #18606). Negative controls omitted the primary antibodies. Triple-IF was performed with antibodies against desmin (AF3844, R&D, 1:200) and vimentin (#5741, Cell Signalling, 1:100), followed by secondary antibodies as described above together with a Cy3-coupled anti-sma antibody (C6198, Sigma Aldrich, 1:200).
X-gal staining was performed as previously described with slight modifications27. Briefly, colon tissue was fixed in 2% PFA/PBS with 0.2% v/v glutaraldehyde for 30 min at RT. Tissues were washed once for 15 min in 2 mM MgCl in PBS plus 0.01 % Nonidet P-40 (BDH Laboratory Supplies; #56009). Tissues were incubated for 15 h at 37 °C in X-gal substrate solution (stock solution: 100 ml of washing buffer, 1 ml of 5 mM potassium ferricyanide, 1 ml of 5 mM potassium ferrocyanide and 2.5 ml of 40 mg ml−1 X-gal (Sigma, B9285) in N,N-dimethylformamide). Specimens were washed once in PBS, fixed in 4% PFA/PBS for 4 h at RT and paraffin embedded.
Microphotographs were taken using a Leica DMLB microscope with a Leica DC300F camera or with a Panoramic MIDI scanner (3DHISTECH), imported with PanoramicViewer software (v.1.15.4). Adobe Photoshop (CS5, v12.1) was used to adjust brightness and contrast. Changes were always applied to the whole image.
The colour deconvolution plug-in for ImageJ was used to quantify DAB and X-gal intensities, and to manually define regions of interest. ‘Periphery’ was defined as the outermost area with the strongest X-gal staining. Regions were drawn manually and the pixel intensities for DAB (brown) and X-gal (blue) were quantified on an 8-bit scale.
RNA in situ hybridization
The RNAScope 2.0 High Definition Kit (DAB) (#310035) and the Multiplex Fluorescent Kit (#320850) were used35. The following probes were all from Advanced Cell Diagnostics: Mus musculus leucine-rich repeat containing G protein coupled receptor 5 (Lgr5), #312171; mRNA M. musculus GLI-Krüppel family member GLI1 (Gli1), mRNA, #311001; M. musculus Ihh mRNA, #413096; M. musculus Grem1 mRNA, #314741; M. musculus Ptch1 mRNA, #402811. Hybridization was performed according to the manufacturer’s protocol with slight modifications (pre-treatment 2 was prolonged to 20 min and the developing time was extended to 15 min). Expression was assessed semi-quantitatively as suggested by the manufacturer from 0 (absent), 1 (sporadic positive cells with single spots), 2 (frequent spots with occasional clusters of signal) to 3 (abundant spots and clusters).
Real-time quantitative PCR
A 20–30 mg sample of snap-frozen tissue was disrupted using a VDI 12-homogenizator (VWR International) and mRNA was extracted according to the manufacturer’s protocol using an RNeasy midi kit (Qiagen, #75142). Real-time PCR was performed using the 7500 Fast Real-Time PCR System (Applied Biosystems). Power Green PCR Master Mix (Applied Biosystems; #4368708) comprised SYBR Green I Dye, AmpliTaq Gold DNA Polymerase, dNTPs, buffer components and ROX as the passive reference dye. The following primers were used: glyceraldehyde-3-phosphate dehydrogenase (Gapdh), forward: 5′- GGTGTGAACGGATTTGGCCGTATTG -3′, reverse: 5′- CCGTTGAATTTGCCGTGAGTGGAGT -3′; Rplp0/ARP, ribosomal protein, large, P0, forward: 5′- TGCACTCTCGCTTTCTGGAGGGTGT -3′, reverse: 5′- AATGCAGATGGATCAGCCAGGAAGG -3′; GLI family zinc finger 1 (Gli1), forward: 5′- CGTTTAGCAATGCCAGTGACC -3′, reverse: 5′- GAGCGAGCTGGGATCTGTGTAG -3′; Ihh, forward: 5′- GGCTTCGACTGGGTGTATTA -3′, reverse: 5′- CGGTCCAGGAAAATAAGCAC -3′; Ptch1 forward: 5′- TTGGGATCAAGCTGAGTGCTG -3′, reverse: 5′- CGAGCATAGCCCTGTGGTTCT -3′; Shh forward: 5′- TGGAAGCAGGTTTCGACTGG -3′, reverse: 5′- GGAAGGTGAGGAAGTCGCTGT -3′. Gli2 forward: 5′- TGAGGAGAGTGTGGAGGCCAGTAGCA -3′, reverse: 5′- CCGGGGCTGGACTGACAAAGC -3′; Hhip forward: 5′- TAACGGCCCTTTGGTTGGTGGATTT -3′, reverse: 5′- AGCAAAGCCCAGTGACCAAGCAATG -3′; Lgr5 forward: 5′- CCAATGGAATAAAGACGACGGCAACA -3′, reverse: 5′- GGGCCTTCAGGTCTTCCTCAAAGTCA -3′; Axin2 forward: 5′- CAGGAGGATGCTGAAGGCTCAAAGC -3′, reverse: 5′- CTCAAAAACTGCTCCGCAGGCAAAT -3′; Ihh forward: 5′- GGCTTCGACTGGGTGTATTA -3′, reverse: 5′- CGGTCCAGGAAAATAAGCAC -3′; Ptch2 forward: 5′- CCCGTGGTAATCCTCGTGGCCTCTAT -3′, reverse: 5′- TCCATCAGTCACAGGGGCAAAGGTC -3′; Grem1 forward: 5′- AGACCTGGAGACCCAGAGTA -3′, reverse: 5′- GTGTATGCGGTGCGATTCAT -3′; Nog forward: 5′- AAGGATCTGAACGAGACGCT -3′, reverse: 5′- GCGAAGTAGCCATAAAGCCC -3′; Chrd forward: 5′- GATGCTGTTCCCACTGCAC -3′, reverse: GGCCCATCCTCTTGGTCATA -3′; and Bambi forward: 5′- CGCCACTCCAGCTACTTCTT -3′, reverse: 5′- TGAGCAGCATCACAGTAGCA -3′.
Gene expression microarrays
RNA was isolated from 1 cm of fresh-frozen whole colonic tissue from Col1a2CreER;Ptch1fl/fl mice (n=4) and controls (n=4) starting 1.5 cm proximal to the anus. RNA quality was assessed using the Agilent 2200 Tape Station system; RNA integrity numbers >8 were considered sufficient. Affymetrix Mouse Gene ST 2.0 arrays were used. Robust multi-array average median polish procedures were applied for normalization. For GSEA analyses58, the GSEA Java plug-in v2.1.0 was used to probe non-log-transformed normalized expression data with standard settings (with the exception that the permutation type was set to ‘gene_set’ and gene sets <15 genes were permitted). Gene lists were derived either from publications as indicated, from the Molecular Signature Database (www.broadinstitute.org/gsea/msigdb), or from gene ontology searches via the AmiGO2 database v 2.1.4 (www.geneontology.org), curated to include ‘manual assertion’ and ‘experimental evidence’, and excluding ‘sequence similarity evidence’ (see also Supplementary Tables 2–5 for the gene sets used).
Processed microarray data for colon adenocarcinoma (AgilentG4502A_07_3) were analysed using gene expression data obtained from TCGA2 accessed via the UCSC Cancer Browser59. In total, the data set obtained included expression data for 155 primary tumours and 19 normal tissue samples. All analyses were performed in R (version 3.1.1). Associations between GLI1 expression and the selected genes were evaluated using Spearman’s rank correlation coefficient. The correlations were derived using the rcorr function from the Hmisc package. Differences in expression levels between tumours and normal tissue samples were analysed using the Wilcoxon rank-sum test from the R stats package. Pre-calculated ESTIMATE scores for the TCGA colon adenocarcinoma data were retrieved from the ESTIMATE web repository (http://ibl.mdanderson.org/estimate/)60.
Analysis of GEO data sets
Data for the GEO data sets GSE67186, GSE39395, GSE17840 and GSE29316 were retrieved from the GEO website (http://www.ncbi.nlm.nih.gov/geo). For RNA-sequencing data (GSE67186), DESeq61 was used to normalize the raw count data and to test for differential expression comparing both the control (shRenilla) to each treatment group (shAPC and shAPC/Kras) and the treatment groups to each other. Microarray data from Affymetrix Human Genome U133 Plus 2.0 (GSE39395 and GSE29316) and Affymetrix Mouse Genome 430 2.0 arrays (GSE17840) were robust multi-array average normalized using the affy package62 and differential expression between treatment groups was analysed using limma63. For GSE39395, gene expression in stromal cells was compared with both epithelial cells and leukocytes, whereas in the cases of GSE29316 and GSE17840, Hh-ligand treated cells were compared with untreated or vehicle-treated controls. The expression pattern of a priori selected genes was subsequently analysed and differential expression was called based on an fdr< 0.05.
All data (except for microarray and sequencing data) were analysed using GraphPad Prism v6.0e and are presented as the mean and s.e.m. unless otherwise specified in the figure legends. Statistical tests are indicated in the figure legends. All t-tests were two-sided. The type 1-error rate was set to 5%, except for GSEA, for which results with an fdr <0.25 were considered significant as previously suggested58.
All animal experiments were performed in accordance with local ethical guidelines and with approval from the local ethics committee (the Swedish Board of Agriculture and the Animal Ethics Committee of the Amsterdam Medical Center, The Netherlands).
The gene expression data that support the findings of this study have been deposited at GEO (with the accession code GSE67172); expression data referenced in this study are available in GEO (GSE67186 (ref. 21), GSE39395 (ref. 38), GSE17840 (ref. 14) and GSE29316 (ref. 33)). All other data that support the findings of this study are available from the corresponding author (R.T.) upon request.
How to cite this article: Gerling, M. et al. Stromal Hedgehog signalling is downregulated in colon cancer and its restoration restrains tumour growth. Nat. Commun. 7:12321 doi: 10.1038/ncomms12321 (2016).
Teglund, S. & Toftgård, R. Hedgehog beyond medulloblastoma and basal cell carcinoma. Biochim. Biophys. Acta Rev. Cancer 1805, 181–208 (2010).
Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature 487, 330–337 (2012).
Yauch, R. L. et al. A paracrine requirement for hedgehog signalling in cancer. Nature 455, 406–410 (2008).
Varnat, F. et al. Human colon cancer epithelial cells harbour active HEDGEHOG-GLI signalling that is essential for tumour growth, recurrence, metastasis and stem cell survival and expansion. EMBO Mol. Med. 1, 338–351 (2009).
Büller, N. V. J. A., Rosekrans, S. L., Westerlund, J. & van den Brink, G. R. Hedgehog signaling and maintenance of homeostasis in the intestinal epithelium. Physiology 27, 148–155 (2012).
Scales, S. J. & de Sauvage, F. J. Mechanisms of Hedgehog pathway activation in cancer and implications for therapy. Trends Pharmacol. Sci. 30, 303–312 (2009).
Sekulic, A. et al. Efficacy and safety of vismodegib in advanced basal-cell carcinoma. N. Engl. J. Med. 366, 2171–2179 (2012).
Berlin, J. et al. A randomized phase II trial of vismodegib versus placebo with FOLFOX or FOLFIRI and bevacizumab in patients with previously untreated metastatic colorectal cancer. Clin. Cancer Res. 19, 258–267 (2013).
Catenacci, D. V. T. et al. Randomized phase Ib/II study of gemcitabine plus placebo or vismodegib, a Hedgehog pathway inhibitor, in patients with metastatic pancreatic cancer. J. Clin. Oncol. 33, 4284–4292 (2015).
Herter-Sprie, G. S., Kung, A. L. & Wong, K. -K. New cast for a new era: preclinical cancer drug development revisited. J. Clin. Invest. 123, 3639–3645 (2013).
van Dop, W. A. et al. Loss of Indian Hedgehog activates multiple aspects of a wound healing response in the mouse intestine. Gastroenterology 139, 1665–1676.e10 (2010).
Büller, N. V. J. A. et al. Stromal Indian hedgehog signaling is required for intestinal adenoma formation in mice. Gastroenterology 148, 170–180.e6 (2015).
Kosinski, C. et al. Indian Hedgehog regulates intestinal stem cell fate through epithelial−mesenchymal interactions during development. Gastroenterology 139, 893–903 (2010).
Zacharias, W. J. et al. Hedgehog is an anti-inflammatory epithelial signal for the intestinal lamina propria. Gastroenterology 138, 2368–2377.e4 (2010).
Neufert, C., Becker, C. & Neurath, M. F. An inducible mouse model of colon carcinogenesis for the analysis of sporadic and inflammation-driven tumor progression. Nat. Protoc. 2, 1998–2004 (2007).
Fazio, V. et al. The AOM/DSS murine model for the study of colon carcinogenesis: from pathways to diagnosis and therapy studies. J. Carcinog 10, 9 (2011).
Hu, Y., Le Leu, R. K. & Young, G. P. Detection of K-ras mutations in azoxymethane-induced aberrant crypt foci in mice using LNA-mediated real-time PCR clamping and mutant-specific probes. Mutat. Res. Toxicol. Environ. Mutagen 677, 27–32 (2009).
Bai, C. B., Auerbach, W., Lee, J. S., Stephen, D. & Joyner, A. L. Gli2, but not Gli1, is required for initial Shh signaling and ectopic activation of the Shh pathway. Development 129, 4753–4761 (2002).
Öhlund, D., Elyada, E. & Tuveson, D. Fibroblast heterogeneity in the cancer wound. J. Exp. Med. 211, 1503–1523 (2014).
Zeuner, A., Todaro, M., Stassi, G. & De Maria, R. Colorectal cancer stem cells: from the crypt to the clinic. Cell Stem Cell 15, 692–705 (2014).
Dow, L. E. et al. Apc restoration promotes cellular differentiation and reestablishes crypt homeostasis in colorectal cancer. Cell 161, 1539–1552 (2015).
van den Brink, G. R. et al. Sonic hedgehog expression correlates with fundic gland differentiation in the adult gastrointestinal tract. Gut 51, 628–633 (2002).
van den Brink, G. R. et al. Indian Hedgehog is an antagonist of Wnt signaling in colonic epithelial cell differentiation. Nat. Genet. 36, 277–282 (2004).
Gagné-Sansfaçon, J., Allaire, J. M., Jones, C., Boudreau, F. & Perreault, N. Loss of sonic Hedgehog leads to alterations in intestinal secretory cell maturation and autophagy. PLoS ONE 9, e98751 (2014).
Barbieri, A. et al. Intraluminal gel ultrasound and eco-color doppler: new tools for the study of colorectal cancer in mice. Vivo Athens Greece 27, 443–450 (2013).
Zheng, B., Zhang, Z., Black, C. M., de Crombrugghe, B. & Denton, C. P. Ligand-dependent genetic recombination in fibroblasts: a potentially powerful technique for investigating gene function in fibrosis. Am. J. Pathol. 160, 1609–1617 (2002).
Kasper, M. et al. Wounding enhances epidermal tumorigenesis by recruiting hair follicle keratinocytes. Proc. Natl Acad. Sci. USA 108, 4099–4104 (2011).
Muñoz, J. et al. The Lgr5 intestinal stem cell signature: robust expression of proposed quiescent ‘+4’ cell markers: transcriptomic and proteomic signature of Lgr5 + stem cells. EMBO J. 31, 3079–3091 (2012).
Mecklenburg, N. et al. Growth and differentiation factor 10 (Gdf10) is involved in Bergmann glial cell development under Shh regulation. Glia 62, 1713–1723 (2014).
Madison, B. B., McKenna, L. B., Dolson, D., Epstein, D. J. & Kaestner, K. H. FoxF1 and FoxL1 link Hedgehog signaling and the control of epithelial proliferation in the developing stomach and intestine. J. Biol. Chem. 284, 5936–5944 (2009).
Perreault, N., Sackett, S. D., Katz, J. P., Furth, E. E. & Kaestner, K. H. Foxl1 is a mesenchymal modifier of Min in carcinogenesis of stomach and colon. Genes Dev. 19, 311–315 (2005).
Barrett, C. W. et al. Tumor suppressor function of the plasma glutathione peroxidase Gpx3 in colitis-associated carcinoma. Cancer Res. 73, 1245–1255 (2013).
Chen, W. et al. Canonical hedgehog signaling augments tumor angiogenesis by induction of VEGF-A in stromal perivascular cells. Proc. Natl Acad. Sci. USA 108, 9589–9594 (2011).
Pelczar, P. et al. Inactivation of Patched1 in mice leads to development of gastrointestinal stromal-like tumors that express Pdgfrα but not kit. Gastroenterology 144, 134–144.e6 (2013).
Wang, F. et al. RNAscope: a novel in situ RNA analysis platform for formalin-fixed, paraffin-embedded tissues. J. Mol. Diagn. 14, 22–29 (2012).
Mazumdar, T., DeVecchio, J., Agyeman, A., Shi, T. & Houghton, J. A. The GLI genes as the molecular switch in disrupting Hedgehog signaling in colon cancer. Oncotarget 2, 638–645 (2011).
Isella, C. et al. Stromal contribution to the colorectal cancer transcriptome. Nat. Genet. 47, 312–319 (2015).
Calon, A. et al. Dependency of colorectal cancer on a TGF-β-driven program in stromal cells for metastasis initiation. Cancer Cell 22, 571–584 (2012).
Walton, K. D. et al. Hedgehog-responsive mesenchymal clusters direct patterning and emergence of intestinal villi. Proc. Natl Acad. Sci. USA 109, 15817–15822 (2012).
Mazumdar, T. et al. Hedgehog signaling drives cellular survival in human colon carcinoma cells. Cancer Res. 71, 1092–1102 (2011).
Chatel, G. et al. Hedgehog signaling pathway is inactive in colorectal cancer cell lines. Int. J. Cancer 121, 2622–2627 (2007).
Delloye-Bourgeois, C. et al. Sonic Hedgehog promotes tumor cell survival by inhibiting CDON pro-apoptotic activity. PLOS Biol. 11, e1001623 (2013).
Varnat, F., Zacchetti, G. & Ruiz i Altaba, A. Hedgehog pathway activity is required for the lethality and intestinal phenotypes of mice with hyperactive Wnt signaling. Mech. Dev. 127, 73–81 (2010).
Cohen, M. et al. Ptch1 and Gli regulate Shh signalling dynamics via multiple mechanisms. Nat. Commun. 6, 6709 (2015).
Shyer, A. E., Huycke, T. R., Lee, C., Mahadevan, L. & Tabin, C. J. Bending gradients: how the intestinal stem cell gets its home. Cell 161, 569–580 (2015).
Tanaka, T. et al. A novel inflammation-related mouse colon carcinogenesis model induced by azoxymethane and dextran sodium sulfate. Cancer Sci. 94, 965–973 (2003).
Quante, M., Varga, J., Wang, T. C. & Greten, F. R. The gastrointestinal tumor microenvironment. Gastroenterology 145, 63–78 (2013).
Whetstone, R. D., Wittel, U. A., Michels, N. M., Gulizia, J. M. & Gold, B. Colon carcinogenesis in wild type and immune compromised mice after treatment with azoxymethane, and azoxymethane with dextran sodium sulfate. Mol. Carcinog. 55, 1187–1195 (2015).
Jess, T., Rungoe, C. & Peyrin-Biroulet, L. Risk of colorectal cancer in patients with ulcerative colitis: a meta-analysis of population-based cohort studies. Clin. Gastroenterol. Hepatol. 10, 639–645 (2012).
Rhim, A. D. et al. Stromal elements act to restrain, rather than support, pancreatic ductal adenocarcinoma. Cancer Cell 25, 735–747 (2014).
Shin, K. et al. Hedgehog signaling restrains bladder cancer progression by eliciting stromal production of urothelial differentiation factors. Cancer Cell 26, 521–533 (2014).
Kodach, L. L. et al. The bone morphogenetic protein pathway is active in human colon adenomas and inactivated in colorectal cancer. Cancer 112, 300–306 (2008).
Davis, H. et al. Aberrant epithelial GREM1 expression initiates colonic tumorigenesis from cells outside the stem cell niche. Nat. Med. 21, 62–70 (2015).
Jaeger, E. et al. Hereditary mixed polyposis syndrome is caused by a 40-kb upstream duplication that leads to increased and ectopic expression of the BMP antagonist GREM1. Nat. Genet. 44, 699–703 (2012).
Arzumanyan, A. et al. Hedgehog signaling blockade delays hepatocarcinogenesis induced by hepatitis B virus X protein. Cancer Res. 72, 5912–5920 (2012).
Nagao-Kitamoto, H. et al. GLI2 is a novel therapeutic target for metastasis of osteosarcoma. Int. J. Cancer 136, 1276–1284 (2015).
Gerling, M. et al. Real-time assessment of tissue hypoxia in vivo with combined photoacoustics and high-frequency ultrasound. Theranostics 4, 604–613 (2014).
Subramanian, A. et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).
Cerami, E. et al. The cBio Cancer Genomics Portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov 2, 401–404 (2012).
Yoshihara, K. et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat. Commun. 4, 2612 (2013).
Anders, S. & Huber, W. Differential expression analysis for sequence count data. Genome Biol. 11, R106 (2010).
Gautier, L., Cope, L., Bolstad, B. M. & Irizarry, R. A. affy—analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 20, 307–315 (2004).
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43, e47–e47 (2015).
This work was supported by the Center for Innovative Medicine (CIMED) at Karolinska Institutet, the Swedish Research Council (K2013-67X-20434-07-4) and the Swedish Cancer Society (#150665, all to R.T.) as well as the Ruth and Richard Julin Foundation (2015juli44263, to M.G.), Bengt Ihre-Fonden (2012-SLS 254491, to S.A.) and Karolinska Institutet’s forskningsfonder (2014fobi42063 to S.A.). M.G. received a postdoctoral stipend from the German Research Foundation (DFG Ge 2386/1-1) and the Swedish Cancer Society (2014/1376). B.E. and L.K. were supported by MD fellowships from the Boehringer Ingelheim Fonds, Germany. L.B. gratefully acknowledges a Marie Curie postdoctoral fellowship (#297639). We are grateful to Elin Tüksammel and Maryam Saghafian for technical assistance in the animal facility; Tarja Schröder, Contanze Cieluch and Simone Spieckermann for help with IHC; and Agneta Andersson for performing RNA ISH. We thank Maria Kasper for valuable discussions, Anja Füllgrabe for assistance with microarray data processing and Åsa Kolterud for advice on Col1a2CreER mice. Parts of this study were performed at the Live Cell Imaging Unit, Department of Biosciences and Nutrition, Karolinska Institutet, supported by grants from the Knut and Alice Wallenberg Foundation, the Swedish Research Council, CIMED, and the Jonasson donation to the School of Technology and Health, Kungliga Tekniska Högskolan, Sweden. We thank the Bioinformatics and Expression Analysis (BEA) core facility, which is supported by the Board of Research at Karolinska Institutet and the Research Committee at the Karolinska Hospital.
The authors declare no competing financial interests.
Supplementary Figures 1-9 and Supplementary Tables 1-6. (PDF 2338 kb)
Supplementary Movie 1
μUS of a Ptch1ΔCol1 mouse prior to Tamoxifen administration Ultrasound (US) of a murine abdomen scanned from the anus rostral to the level of the liver. Note that this scan was performed by moving the US probe manually over the abdomen rather than using a 3D motor. The colonic lumen was gently inflated with US gel (peristaltic waves lead to variation in colon diameter). The colon is the dark (hypoechoic) ventral structure. From frame 209 to frame 231 (indicated at the bottom), a tumour appears on the lower left part of the colonic lumen (corresponding to the dorsal right side of the mouse). A very small mucosal irregularity occurring ventrally in the same area can be assumed. Frames 272 to approximately 330 depict a larger tumour. Here, the kidney is clearly visible as a landmark in the lower right part of the image (dorsal left side of the animal). From frames 359 onwards, a large tumour can be seen, that in additional directed scans did not appear connected to the aforementioned tumour. In the vicinity of this third tumour, liver tissue is visible (from frame 435 onwards). At this time point, Tam was administered (week 0, Fig. 6c). (AVI 11802 kb)
Supplementary Movie 2
μUS of a Ptch1Col1 mouse after Tamoxifen administration The same mouse seen in Supplementary video 1, scanned 3 weeks after the first scan. Note that the tumours are not discernable in this scan, using the kidney (from approximately frame 200) or liver tissue (from frame 535 onwards) as landmarks. Further scans revealed that the third tumour was still visible. (AVI 12628 kb)
Rights and permissions
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
About this article
Cite this article
Gerling, M., Büller, N., Kirn, L. et al. Stromal Hedgehog signalling is downregulated in colon cancer and its restoration restrains tumour growth. Nat Commun 7, 12321 (2016). https://doi.org/10.1038/ncomms12321
This article is cited by
Cellular and molecular mechanisms of Hedgehog signalling
Nature Reviews Molecular Cell Biology (2023)
Cancer-associated fibroblasts and resistance to anticancer therapies: status, mechanisms, and countermeasures
Cancer Cell International (2022)
Crosstalk between cancer-associated fibroblasts and immune cells in the tumor microenvironment: new findings and future perspectives
Molecular Cancer (2021)
PGE2-JNK signaling axis non-canonically promotes Gli activation by protecting Gli2 from ubiquitin-proteasomal degradation
Cell Death & Disease (2021)
Emerging roles of the Hedgehog signalling pathway in inflammatory bowel disease
Cell Death Discovery (2021)
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.