Human colon cancer epithelial cells harbour active HEDGEHOG-GLI signalling that is essential for tumour growth, recurrence, metastasis and stem cell survival and expansion

Abstract

Human colon cancers often start as benign adenomas through loss of APC, leading to enhanced beta CATENIN (beta CAT)/TCF function. These early lesions are efficiently managed but often progress to invasive carcinomas and incurable metastases through additional changes, the nature of which is unclear. We find that epithelial cells of human colon carcinomas (CCs) and their stem cells of all stages harbour an active HH-GLI pathway. Unexpectedly, they acquire a high HEDGEHOG-GLI (HH-GLI) signature coincident with the development of metastases. We show that the growth of CC xenografts, their recurrence and metastases require HH-GLI function, which induces a robust epithelial-to-mesenchymal transition (EMT). Moreover, using a novel tumour cell competition assay we show that the self-renewal of CC stem cells in vivo relies on HH-GLI activity. Our results indicate a key and essential role of the HH-GLI1 pathway in promoting CC growth, stem cell self-renewal and metastatic behavior in advanced cancers. Targeting HH-GLI1, directly or indirectly, is thus predicted to decrease tumour bulk and eradicate CC stem cells and metastases.

Figure 1. HH-GLI pathway expression and localization in human CCs

1. Analysis of CCs localized in the colon. Hematoxylin and eosin (H&E) staining, in situ hybridization for /GLI1/ and /PTCH1/ mRNAs (blue; CC7) and SHH (red; CC14) in (A) and GLI1 protein (green; CC6) in (B) in CC epithelial cells as indicated.

2. Analysis of metastatic CCs localized in the liver. H&E staining, in situ hybridization for /GLI1/ and /PTCH1/ mRNAs (blue; mCC2), SHH (red; mCC11) and cytokeratin (red; mCC11) in (C) and GLI1 protein (green; mCC1) in (D) in CC epithelial cells as indicated.

3. The top panels of (A, C) show wider views of the local CC6 and metastatic mCC2 tumors, depicting local colon invasion and high-grade dysplastic morphology (A) and liver invasion (C). Nuclei (blue) are stained with DAPI immunofluorescence images for SHH, cytokeratin and GLI1. l: liver cells, s: stroma; t: tumor. Scale bar = 50 µm for all panels.

Figure 2. Gene expression changes in human CC samples

1. Heat map of gene expression determined by RT-qPCR shown as CD133⁺/CD133⁻ expression ratios. CC are TNM staged. Numerical values are given in Fig S2 of Supporting Information. Here and in all figures expression levels were normalized with the geometric mean of the ct values of EEFIa and GAPDH. Ls = LS174T, HT = HT29, m = metastatic tumour in the liver.

2. Histograms of individual gene expression changes in CD133⁺ (red bars) and CD133⁻ (blue bars) cells. The graphs use the same samples as in (A). x = mCC17 xenograft. Numerals refer to TNM stages. nc: normal colon; nl: normal liver; m = metastases.

Figure 3. Requirement of HH-GLI signalling for CC growth in vitro

A. Representative CC primary cultures labelled with pan-Cytokeratin antibodies (red) with more than 90% Cytokeratin⁺ cells (see Fig S1 of Supporting Information), demonstrating their epithelial nature. B, C. Effects of siRNAs (B) and lentivectors (C) as noted on proliferation (BrdU index: number of BrdU⁺ cells over total number of cells), and apoptosis (cleaved Caspase-3 index: number of cleaved Caspase-3⁺ cells over total number of cells). Asterisks in these and all panels denote significant changes (p < 0.05) as compared with controls. mRNA target destruction efficiencies for siGLI1 was 85% on average in seven primary cultures tested; 83.3% for siGLI2; and 83.5% for siGLI3; for shRNAs: 86.9% for shSMOH and 83.9% for shPTCH1 on average in four primary cultures analyzed, measured by qPCR. Primary cultures of normal liver or normal colon did not grow under the conditions used for CCs. D. Effects of cyclopamine compared with the inactive compound tomatidine on proliferation (right) shown as BrdU index, and apoptosis (left), shown as cleaved Caspase-3 index. Cyclopamine decreases proliferation and increases apoptosis in a concentration-dependent manner. Samples are shown grouped by TNM stage and liver metastases are shown to the right. CC14 proliferation was similarly inhibited by shSMOH, GLI3 and cyclopamine treatment, and enhanced by shPTCH1 and GLI1 (not shown). E. Graph of the variation in proliferation index (BrdU incorporation/cell number) compared with total cell number (number of DAPI⁺ cells) of five CC cells as indicated, responding to expression of GLI1, GLI3R or both simultaneously through lentivector transduction. Control cells were transduced with a parental empty lentivector. All cells respond similarly. Scale bar = 150 µm (A).

Figure 4. Requirement of HH-GLI signalling for CC growth in vivo

A-C. Representative CC subcutaneous xenografts in nude mice taken at the same time for each group (10–20 days) as indicated with cells expressing shRNAs or cDNAs noted. Control cells were transduced with a GFP-only parental lentivector. D-G. CC xenograft growth curves transduced or treated as shown. In (G), the graph starts at the beginning of treatment with cyclopamine rather than at the time of cell grafting (D–F). For each curve at least six independent tumours were scored. All graphs show tumour volume over days. Scale bar = 1 cm for (A–C).

Figure 5. Interference with HH-GLI prevents CC recurrence

Suppression of HT29 xenografts by cyclopamine treatment and their subsequent recurrence after additional treatments of 5 (yellow) or 10 (orange) days, but not after an additional 20 day treatment (red). n = 5 for each cohort. Insets show the morphological appearance of a recurrence (left) and its histology (right) showing the presence of XGal⁺ LacZ-HT29 cells (blue) surrounded by host stroma (pink). Scale bar = 0.5 cm for (left inset), 50µm (right inset).

Figure 6. Interference with HH-GLI prevents CC metastatic growth and GLI1 induces EMT in CC Cells

A, B. Wholemount (A) and histological sections (B) of LacZ-HT29 lung metastases after HH-GLI pathway modulation. The development of large metastases is contrary to local ethical rules. Sections were counterstained with eosin (pink). C. Quantification of metastastic colony number from injected cells expressing shSMOH, GLI1 or both. ns = not significative. n = 8 mice each for control (c), shSMOH and GLI1; n = 5 for shSMOH + GLI1. D, E. Whole-mount image (D) and quantification (E) of the metastatic behavior of LacZ-Ls174T and LacZ-CC14 and the modulation by shSMOH or shPTCH1. F. Images of transduced CC14 cells with control (left) or GLI1 (right) GFP⁺ lentivectors showing EMT induction by GLI1 (right) with elongated dispersing cells (right) that disintegrate the compact epithelial islands characteristic of CC cultures (left). G. RT-qPCR analyses of CC14 cells as in (F) showing the regulation of EMT genes by enhanced HH-GLI activity through shPTCH1 or GLI1 in relation to control (set at 1). Note the increased expression of EMT genes and the repression of E-CADHERIN. Scale bar = 5 mm for (A), 80 µm for (B), 30 µm for (D) and 60 µm for (F).

Figure 7. Requirement of HH-GLI function for CC stem cell survival and modulation of self-renewal in vivo

A. Diagram of the novel red/green in vivo protocol described in the text. B, C. FACS plots of CC14 (B) or mCC11 (C) cell populations with untransduced, TomatoRed+ (RFP⁺) and GFP⁺ cells. The GFP/RFP ratio is given for each case.

Figure 8. GLI1 drives in vitro expansion of CC cells and HH-GLI function is required for the growth of the bulk of CC tumours

A, B. In vitro (A) and in vivo (B) red/green assay, showing the dominance of GLI1-expressing cells (A) and the requirement of SMOH through the use of shSMOH (B) in CC14 cells. Assays were with untransduced CC14 cells (black in FACS plots), or CC14 cells transduced with RFP⁺ control (red) or GFP⁺ GLI1-expressing (green) lentivectors, or RFP plus GFP-only control lentivectors as indicated. FACS plots of cells at passage 1 and 4 are shown in all cases. Control GFP-only transduced CC14 cells are maintained and the green/red cell ratio around 1 is preserved. 10⁶ cells were injected in (B).