Hedgehog-EGFR cooperation response genes determine the oncogenic phenotype of basal cell carcinoma and tumour-initiating pancreatic cancer cells

Inhibition of Hedgehog (HH)/GLI signalling in cancer is a promising therapeutic approach. Interactions between HH/GLI and other oncogenic pathways affect the strength and tumourigenicity of HH/GLI. Cooperation of HH/GLI with epidermal growth factor receptor (EGFR) signalling promotes transformation and cancer cell proliferation in vitro. However, the in vivo relevance of HH-EGFR signal integration and the critical downstream mediators are largely undefined. In this report we show that genetic and pharmacologic inhibition of EGFR signalling reduces tumour growth in mouse models of HH/GLI driven basal cell carcinoma (BCC). We describe HH-EGFR cooperation response genes including SOX2, SOX9, JUN, CXCR4 and FGF19 that are synergistically activated by HH-EGFR signal integration and required for in vivo growth of BCC cells and tumour-initiating pancreatic cancer cells. The data validate EGFR signalling as drug target in HH/GLI driven cancers and shed light on the molecular processes controlled by HH-EGFR signal cooperation, providing new therapeutic strategies based on combined targeting of HH-EGFR signalling and selected downstream target genes.

activity (ALDH) (Kim et al, 2011;Li et al, 2007;Rasheed et al, 2010;Simeone, 2008). This obvious heterogeneity can be explained by the existence of distinct subpopulation of rare TICs within the same tumor sample and/or by the mutation status of the tumor cells and cell lines. The latter has been shown to be a critical determinant of the marker combination to be used in order to identify and isolate TICs (Curtis et al, 2010).
The lack of a universal pancreatic TIC marker and the obvious discrepancy of the TIC identity in pancreatic cancer cells prompted us to use in this study a TIC enrichment technique that is based on the growth characteristics of TICs rather than surface marker expression or enzyme activity, i.e. high clonogenic growth and (tumor)sphere formation under non-adherent conditions (Simeone, 2008). For this purpose, we seeded pancreatic cancer cells at low clonal density (10 4 cells per mL) into a 3D softagar matrix rather than on non-adherent tissue culture plates. This allowed us to readily, reproducibly and more precisely monitor, quantify and isolate highly clonogenic TICs cells from large spheres (macrospheres) in 3D culture. To demonstrate their tumor initiating capacity in xenograft assays, we isolated single macrospheres from the 3D soft-agar culture by aspiration with a small pipette. The isolation process was monitored under a dissecting microscope. Isolated macrospheres were dissociated into single cell suspensions, counted, re-suspended in 50µL 25 % Matrigel (BD Laboratories) and injected into the lower flanks of nude mice. The limiting dilution in vivo tumor growth assay that is generally considered the gold standard for the characterization of TICs, clearly identified the macrosphere forming pancreatic cancer cells as a tumor initiating subpopulation of human pancreatic cancer cells (see manuscript figure 6B). It is also noteworthy that our findings of high level expression of HH ligands (e.g. SHH) in the tumor-initiating macrospheres are consistent with a previous study showing elevated expression of SHH in highly tumorigenic CD24+/CD44+/ESA+ pancreatic cancer cells with TIC/cancer stem cell-like properties (Li et al, 2007). Furthermore, we cultured TICs from individually isolated macrospheres under 2D adherent conditions for one or more passages before re-seeding 10 4 TIC progeny into 3D matrix cultures or by repeated alternating cultures in 2D and 3D. Notably, the fraction of macrosphere-forming TICs remained constant (about 1-2%) even after several passages in 2D cultures or serial alternating 2D-3D passages.
This suggests that the majority of TIC progeny (98-99%) lose their sphere forming ability when cultured in 2D, but that the rare fraction of clonogenic TICs is maintained in 2D cultures (suppl. information, Figure S8A). Conversely, serial passages (n>2) of TICs in 3D non-adherent cultures led to a significant increase in the number of macrospheres providing additional evidence for their selfrenewal capacity (suppl. information, Figure S8B, C).

Histology, immunohistochemistry and Western blot analysis
Skin specimen of mice were fixed over night at 4°C in 4% paraformaldehyde and subsequently embedded in paraffin. Sections were cut on a Leica 250M microtome. Specimen were deparaffinized, stained with hematoxylin/eosin and embedded in HistoFix (Carl Roth Laboratory Equipment).
Primary anti-JUN (#9165, Cell Signaling Technology), and anti-human SOX9 (#sc-20095, Santa Cruz Biotechnology) were used at 1:100 dilutions and incubated for 10min at room temperature. Heat induced antigen retrieval was done in EDTA buffer (pH 9.0) (Dako, Denmark). Anti-JUN, anti-SOX9 and anti-CXCR4 antibodies were detected with polymer-based envision reagent (Dako, Denmark) and visualized with DAB. Staining was performed on an automated Autostainer Plus platform (Dako, Denmark). Staining of mouse tissues for Keratin 17 (K17) and Sox9 was done as described previously (Yang et al, 2008).
Western blot analyses were done with the following antibodies:   Table S2). Hh/Gli mediated regulation of cell cycle genes with putative GLI binding sites was validated also in mouse cerebellar granular precursor cultures treated with Sonic Hh (data not shown).   The GLI binding site sequences and their location relative to the transcriptional start site are shown. Cell cycle genes with GLI binding sites labeled with an asterisk were selected for further analysis by chromatin immunoprecipitation analysis. Selection criteria were sequence, number and clustering of putative GLI binding sites.  IL1R2  CCATGAAGGCCAGCAATACCACATCAC  CGGGATTGTCAGTCTTGACCCCAGA  136bp  PTCH  TCCTCGTGTGCGCTGTCTTCCTTC  CGTCAGAAAGGCCAAAGCAACGTGA  200bp  CD133  AAAGTGGCATCGTGCAAACCTGTGG  TCATCGTACACGTCCTCCGAATCCA  200bp  NANOG  GGGGTTTCACTGTGTTAGCCAGGATGG  GCCAGAGACGGCAGCCAAGGTTATT  150bp  OCT4  AAACGACCATCTGCCGCTTTGAGG  TCCAGGTTGCCTCTCACTCGGTTCT  196bp  GLI1  TCTGGACATACCCCACCTCCCTCTG  ACTGCAGCTCCCCCAATTTTTCTGG  191bp  SHH  TGATGAACCAGTGGCCAGGAGTGA  ACCGAGCAGTGGATATGTGCCTTG  217bp  IHH  GCGCCGACCGCCTCATGACC  TCTGATGTGGTGATGTCCACCG  178bp  DHH  GGAGAGGGAGGGGGAGGGAGAAAAT  TTAGCCTCTCCCCCAGTGCTTCAGC  150bp  JUN  AGTGCGATGTTTCAGGAGGCTGGAG  TTCTCAAAGCAGGAATTGGTGGCAGA  177bp  SOX9  CCTCCTGCCTTTGCTTGTTCACTGC  CTCGGGCACTTATTGGCTGCTGAAA  106bp  FGF19  AGGAGATCCGCCCAGATGGCTACAA  GCAGCATGGGCAGGAAATGAGAGAG  138bp  TGFA  CGCCCTGTTCGCTCTGGGTATTG  CGGTGATGGCCTGCTTCTTCTGG  281bp  SPP1  GCCAGCAACCGAAGTTTTCACTCCA  GCACCATTCAACTCCTCGCTTTCCA  183bp  CXCR4 GGTGGTCTATGTTGGCGTCT     Table S2). B) Luciferase reporter assays showing induction of FOXM1, TOPBP1 and KNTC1 promoter activity in response to GLI1 and GLI2 expression. The PTCH promoter served as positive control. Reporter assays were done in 293 cells. C) qPCR analysis showing that the expression of direct cell cycle GLI target genes are not affected by concomitant EGFR signaling. IL1R2 served as positive control for cooperative HH/GLI-EGFR signaling. PTCH induction is shown as EGFR-independent GLI target to demonstrate comparable GLI1 activity in the different samples.
Only the EGFR/GLI target IL1R2 (Kasper et al, 2006)   A) The proportion of rare macrosphere-forming cells to cells with low clonogenicity is constant when tumor-initiating macrosphere cells isolated from 3D cultures were passaged in 2D cultures before being re-seeded and grown in 3D cultures. B) By contrast, the fraction of tumorinitiating macrosphere-forming cells is highly enriched if isolated from 3D cultures and directly reseeded in 3D cultures without intervening growth in 2D cultures. C) Quantification of macrosphereforming cell experiments shown in B after two serial passages).

MAC 3D>3D MAC 2D>3D
suppl. Figure S8 *** ** ** Figure S10: RNAi-mediated knockdown efficiency of human and mouse lentiviral shRNA constructs. For details of the constructs see materials and methods. shRNA mediated knockdown of target mRNAs was determined by quantitative real-time PCR. Target mRNA levels in cells transduced with nontarget control shRNA (shCont) were set to 100 and served as reference value for cells transduced with target shRNA (shTarget). Mouse shRNAs were validated in ASZ001 BCC cells, human shRNA contructs in the pancreatic cancer cell lines L3.6sl and Panc-1.
human mouse