Loss of WNT-TCF addiction and enhancement of HH-GLI1 signalling define the metastatic transition of human colon carcinomas

Abstract

Previous studies demonstrate the initiation of colon cancers through deregulation of WNT-TCF signalling. An accepted but untested extension of this finding is that incurable metastatic colon carcinomas (CCs) universally remain WNT-TCF-dependent, prompting the search for WNT-TCF inhibitors. CCs and their stem cells also require Hedgehog (HH)-GLI1 activity, but how these pathways interact is unclear. Here we define coincident high-to-low WNT-TCF and low-to-high HH-GLI transitions in patient CCs, most strikingly in their CD133(+) stem cells, that mark the development of metastases. We find that enhanced HH-GLI mimics this transition, driving also an embryonic stem (ES)-like stemness signature and that GLI1 can be regulated by multiple CC oncogenes. The data support a model in which the metastatic transition involves the acquisition or enhancement of a more primitive ES-like phenotype, and the downregulation of the early WNT-TCF programme, driven by oncogene-regulated high GLI1 activity. Consistently, TCF blockade does not generally inhibit tumour growth; instead, it, like enhanced HH-GLI, promotes metastatic growth in vivo. Treatments for metastatic disease should therefore block HH-GLI1 but not WNT-TCF activities.

Figure 1. Repressed WNT-TCF and enhanced HH-GLI characterize the metastatic transition of human CCs detected in fresh patient samples

1. Histograms of individual changes in gene expression determined by qRT-PCR in individual patient tumour samples, obtained from the operating room, for both CD133⁺ (red bars) and CD133⁻ (blue bars) cells of each sample for WNT-TCF, HH-GLI, ES-like stemness and WNT-inhibitors (additional genes are shown in Fig S1 and in Varnat et al, 2009). CCs are grouped by TNM stage (numerals) plus liver metastases (m). Normal colon (nc) and normal liver (nl), and subcutaneous xenografts (x; mCC17) are also included. Description of tumours is as in Varnat et al (2009). Values for GLI1 and HIP are from Varnat et al (2009) shown here for comparison. In all cases ct expression levels are normalized by the geometric mean of the ct values of the EEFIA1 and GAPDH, yielding relative expression levels. Statistics comparing TNM1,2 versus TNM3,4 plus liver metastases are shown by the brackets above each graph. Error bars of s.e.m.'s are not included to enhance clarity. However, asterisks show significance (p < 0.05) using the Student's t-test between CD133⁺ (red columns) values of TNM1,2 versus those of TNM3,4+ metastases. ns: not significant. In addition, we provide individual comparisons of TNM1,2 versus TNM3,4 and statistics: the CD133⁺/CD133⁻ ratios were: LGR5: 3.2 in TNM1,2 versus 0.5 in TNM3,4, p < 0.001; SOX4: 2.7 versus 0.4, p < 0.001; cMYC: 3.1 versus 0.8, p < 0.001; DKK1: 0.4 versus 2.3, p < 0.001; SFRP1: 0.4 versus 1.3, p < 0.01; SOX2: 1.4 versus 3.3, p = 0.03; AXIN2, a direct TCF target (e.g. Leung et al, 2002), displayed a decreasing trend in both populations but significantly only in CD133⁺ cells (19.9 vs. 6.5, p < 0.01); KLF4 showed increased expression in both CD133⁺ and CD133⁻ populations in metastatic versus non-metastatic CCs: 7.7-fold increase in CD133⁺ cells of TNM1,2 versus TNM3,4, p = 0.003; eightfold for CD133⁻ cells, p < 0.0001. Ls = LS174T, HT = HT29. Analyses of gene expression in the GEO database was uninformative since there is no data on CD133⁺ cells and CCs are not TNM sorted.

2. Graphic representation of the changes in HH-GLI, WNT-TCF, ES-like stemness and WNT inhibitor signatures in normal colon and during CC progression shown in (Figs 1A; S1). The signatures shown are the averages of the CD133⁺/CD133⁻ ratios of GLI1, GLI2, PTCH1, SHH, HIP and SNAIL1 for HH-GLI; of cMYC, CD44, LGR5, AXIN2 and SOX4 for WNT-TCF; of NANOG, OCT4, SOX2 and KLF4 for ES-like stemness and of DKK1 and SFRP1 for WNT inhibitors. Asterisks denote significant changes (p < 0.05) between the values of TNM1,2 versus those of liver metastases as indicated by the colour code of the asterisks (e.g. red for HH-GLI).

3. Heat map representation of individual gene expression levels determined by qRT-PCR shown in (Figs 1A; S1) in normal and cancer samples shown as CD133⁺/CD133⁻ ratios. White boxes denote values within the 0.7–1.4 ratio range (see Fig S2). Other colours follow the code given below the table. All values are given in Fig S2. The pathway level switch at the metastatic transition from a high-to-low crypt/adenoma WNT-TCF signature in TNM1,2 CCs, to a low-to-high HH-GLI signature in TNM3,4 and liver metastases is highlighted by bold boxes. HH-GLI data is derived from (Varnat et al, 2009). Note that the signatures of xenografts mimic those of advanced TNM3,4 CCs. The percentage of CD133⁺ cells in each sample is also given at the bottom of the table, as is the normalized CD133⁺/CD133⁻ expression ratio of CD133 mRNA.

Figure 2. Effects of modulating WNT-TCF and HH-GLI levels in CC cells

1. Images of Ls174T-dnTCF4^ERT2 colonies on plates after staining with crystal violet. Cells were transduced with lentivectors as noted. Addition of TAM (+TAM) activated dnTCF4^ERT2 and is compared with mock-treated cells (−TAM). Control cells (c) were transduced with GFP-only parental lentivectors. Control assays were also performed with the ERT2 part only ±TAM confirming that ERT2 or TAM per se had no adverse effects (not shown). Quantification is shown in Fig S4A.

2. Quantification of the rescue of the anti-proliferative effects, shown as BrdU+ nuclei/total DAPI-labelled nuclei counted per field, of shSMOH by concomitant inhibition of the GLI inhibitor SUFUH through expression of shSUFUH. Results in two primary CCs are shown as indicated. >10 fields were counted per condition. Asterisks denote significant (p < 0.05) changes in Student's t-tests. ns: not significant. Error bars represent s.e.m.

3. Heat map of early (16 h) changes of selected key genes for HH-GLI, WNT-TCF and ES-like stemness signatures in mCC11 cells determined by qRT-PCR after expression of cDNAs or shRNAs as indicated. Values are ratios of individual experimental over control (GFP-only transfected cells) values after normalization. Results for CC14 and Ls174T are shown in Fig S4B. White boxes represent ratio values within the 0.7–1.4 range. Blue boxes represent ratio values equal to or smaller than 0.5. Red boxes represent values equal to or greater than 1.5.

4. Positive regulatory loop between GLI1 and cMYC. mRNA expression levels determined by qRT-PCR are shown as ratios over control transfected (GFP only) cells. Heat map colours and values are as in (C). Enhanced cMYC levels were obtained by its overexpression while blockade of endogenous cMYC activity was obtained through the expression of dnMYC (also known as OMOMYC; Soucek et al, 2002). Note the general differential regulation of GLI1 and PTCH1 versus GLI3 and p21, the latter being a proven cMYC target. GLI2 levels were unaffected. The lack of upregulation of PTCH1, a GLI1 target (Agren et al, 2004), when GLI1 is upregulated by cMYC in Ls174T cells may be due to different target response kinetics (see Zbinden et al, 2010).

Figure 3. Regulatory HH-GLI and WNT-TCF interactions and modulation of GLI1 activity by CC oncogenes and tumour suppressors

A,B. GLI-dependent luciferase assays in purified CD133⁺ mCC11 (A) and Ls174T (B) cells as indicated with wt or mutant (mGBS) reporters. Assays with CC14 and its CD133⁺ population are shown in Fig S6. Triangles in this and other panels show dose-dependent effects with increasing concentrations of the second plasmid as written in the figures (e.g. GLI1 + βCAT under the left triangle in A have increasing amounts of βCAT with constant amounts of GLI1 plasmids). C. GLI-dependent luciferase assay showing βCAT (bCAT) action on chimeric GLI1 proteins. See text for details. D. TCF-dependent luciferase reporter assays using wt (TOP) or mutant (FOP) reporters as indicated in CD133⁺ mCC11 cells. Similar results with CC14, CD133⁺ CC14 and Ls174T cells are shown in Fig S7A–C. E. GLI-dependent luciferase reporters testing for the regulation of GLI1 activity by oncogenic KRAS (KRAS^V12G), MEK1 (p45MAPKK^S222D) and AKT1 (N-myristoylated AKT1). Similar results in CD133⁺ mCC11 cells are shown in Fig S8. F. Changes in the expression of key HH-GLI and WNT-TCF pathway genes as determined by qRT-PCR in CC14 and mCC11 after inhibition of endogenous MEK1 activity with U0126 or endogenous AKT activity with SH6 (10 µM, 48 h). Expression values were normalized with housekeeping genes as in Fig 1 and the ratios of the experimental over control mock-treated cells are shown. White boxes denote values within the 0.7–1.4 range. Blue boxes with values equal to or smaller than 0.5. Red boxes with values equal to or greater than 1.5. G. GLI-dependent luciferase reporters testing for the modulation of GLI1 by increased exogenous p53 and by inhibition of endogenous p53 activity with shp53 as noted, using also mutant (mt) reporters. H. Scheme of the proposed interactions between WNT-TCF, HH-GLI, oncogenes and tumour suppressors in CC cells as determined in vitro. Plasmids were nucleofected at similar concentrations unless noted by triangles, where , and 1/1 ratios were used for the plasmid in second position. Percent changes in relation to the GLI1 value (equated to 100%, all after normalization with the renilla internal control) are shown (A–C, E and G). Percent changes in relation to the βCAT value are shown in (B and D). ns: difference not significant between the samples at the end of the line. Asterisks denote significant (p < 0.05) changes in Student's t-tests. Error bars represent s.e.m.

Figure 4. Differential expression of WNT-TCF and HH-GLI pathway components in CC cells in vitro versus in vivo in xenografts

A–C. Histograms of individual gene expression changes determined by qRT-PCR for CD133⁺ (red bars) and CD133⁻ (blue bars) cells, after normalization as in Fig 1, shown for selected WNT-TCF (A), HH-GLI (B) and ES-like stemness genes (C) in HT: HT29; Ls: Ls174T; mCC: mCC11 cells. v: in vitro culture; x; in vivo xenografts. Analyses of additional genes are shown in Fig S9. Error bars of s.e.m.'s are not included to enhance clarity. However, asterisks denote significant (p < 0.05) changes in Student's t-tests using PCR triplicates in each case. ns: not significant. D. Logarithmic scale representation of averaged gene expression changes in CD133⁺ cells (cell lines Ls174T and HT29, and primary mCC11) cultured in vitro versus in vivo in xenografts. Primary data derived from (A) and Fig S9. The values for HH-GLI, βCAT/TCF and ES-like stemness signatures represent the averages, after normalization of the values of GLI1, GLI2, HIP and PTCH1 for HH-GLI, of LGR5, SOX4, AXIN2, cMYC and CD44 for WNT-TCF, and of NANOG, SOX2 and OCT4 for ES-like stemness, shown in panels A-C and Fig S9. In vitro levels were equated to 1 to normalize the graph, and in vivo levels were correspondingly adjusted. Asterisks denote significant (p < 0.05) changes between the same cell/signature averages in vitro versus in vivo. ns: difference not significant. Note the clear and strong differences in CD133⁺ cells.

Figure 5. Effects of dox-inducible TCF4 inhibition on CC Ls174T and DLD1 cell line xenograft growth in mice

A,B,D,E. Xenograft growth curves (A and E) and representative images (B and D) control (Ctrl) and dnTCF4^dox (dnTCF4) expressing Ls174T (A and B) or DLD1 (D and E) cells treated with doxycycline (DOX) in vivo. Arrows pointing down in (A and E) denote the time of initiation of dox treatment in vivo after tumour appearance. Arrow pointing up in (E) indicates the time of termination of dox treatment. +dox->-dox in (D lower panel) denotes the growth of the tumour after removal of dox. Tumours in (D) are outlined. n = 8 tumours for each condition. Asterisk denotes significant (p < 0.05) change in Student's t-tests. Error bars represent s.e.m. C. Gene expression changes induced by dox-inducible dnTCF4 in Ls174T and DLD1 cells in vitro and in vivo shown as the ratios of experimental over control cells without dox, all after normalization as in Fig 1. White boxes denote values within the 0.7–1.4 range. Blue boxes with values equal to or smaller than 0.5. Red boxes with values equal to or greater than 1.5. F. Growth curves of individual, recurrent DLD1 xenograft tumours from (E) starting at day 49. One tumour was left untreated (blue line), continued to grow aggressively and the animal was euthanized. Three other mice treated again with dox showed a block of tumour growth and reduction of tumour volume. To simplify comparisons, tumour volumes were each equated to 1 at day 49 before restarting dox treatment in order to have all tumours share the same starting point. Actual averaged tumour volumes at day 49 are shown in (E). G. CD133⁺/CD133⁻ gene expression ratios in DLD1 cells in vivo versus in vitro, determined by qRT-PCR and after normalization as in Fig 1. As we had found for other CC cells (Varnat et al, 2009), DLD-1 CD133⁺ cells had enriched CD133 mRNA levels as compared to CD133⁻ cells (6.7-fold in vivo and 7.6-fold in vitro). CD133⁺ cells were more abundant in vivo than in vitro (16% vs. 5%, respectively). Colours of boxes are as in (C).

Figure 6. Effects of TAM-inducible TCF4 blockade in Ls174T and primary human CCs in xenografts in mice

A. Growth curves of xenografts with Ls174T, CC14, mCC11 or CC36 cells carrying dnTCF4^ERT2 or ERT2 alone with (+) or without (−) TAM treatment as indicated. Plots show logarithmic Y-axis scales. Arrows point to the beginning of TAM addition once the tumours were visible. Growth curves are overlapping except for those of CC36. ns: not significant in Student's t-tests. Error bars represent s.e.m. B. Representative images of xenograft tumours of Ctrl (ERT2, +TAM) or dnTCF4 (dnTCF4^ERT2, +TAM) taken at the same time. C,D. Histological sections of xenografts shown in (B) displaying BrdU incorporation (brown in nuclei in top two rows, C) or morphology after H&E staining (bottom row, C), and quantification of cell proliferation (D). Asterisks denote significant (p < 0.05) changes in Student's t-tests. ns: not significant. Error bars represent s.e.m. E. RT-qPCR analyses of human CC xenografts harvested 8-12d after continuous treatment with TAM (+) or vehicle (−) similar to those shown in (A and B). The values are ratios of human-specific gene expression levels after normalization between tumours expressing dnTCF4^ERT2 over those expressing the control ERT2 part only with or without TAM as indicated. In the absence of induction by TAM the ratios are ∼1. White boxes denote values within the 0.7–1.4 range. Blue boxes with values equal to or smaller than 0.5. Red boxes with values equal to or greater than 1.5. Scale bar = 5 mm (B), 100 µm (C).

Figure 7. In vivo competition reveals enhancement of the CD133⁺ CC stem cell population after TCF blockade

CC36 cells transduced with RFP⁺ (red) lentivectors, or CC36-dnTCF4^ERT2 transduced with GFP⁺ (green) lentivectors were mixed in similar amounts (FACS green to red ratio of 0.8) in vitro. Untransduced cells were also present (black cells in the FACS plots). This mixture was injected subcutaneously into immunocompromised mice. Their treatment with TAM (+TAM) or corn oil vehicle only (−TAM) once a day for 8 days resulted in a 3.5-fold increase in the green/red ratio in MACS sorted CD133⁺ cells in the tumours treated with TAM as compared with control tumours (−TAM). The increase of green over red cells in tumours without TAM maybe due to a slight leakiness of the system leading to a small amount of dnTCF4 activity in the absence of TAM. Nevertheless, +TAM tumours were greener than −TAM tumours, which were yellow due to near equivalent green and red cell populations as exemplified in the micrographs in the bottom panels. n = 2 per sample. Average ratios are shown inside the FACS plots. Scale bar = 4 mm.

Figure 8. Enhanced metastatic behavior by blockade of TCF activity in vivo

A,B. Images of dissected lungs after X-Gal staining (blue) showing metastatic growth of Ls174T-LacZ⁺-dnTCF4^ERT2 (A) or mCC11-LacZ⁺-dnTCF4^ERT2 (B) cells after systemic treatment with TAM (dnTCF4) or corn oil vehicle as control (ctrl). C. Quantification of lung metastases per animal in LacZ⁺-dnTCF4^ERT2 injected animals treated with TAM as compared with those identical injected but treated only with vehicle (corn oil). Results for CC36, CC14 and mCC11 are shown. Asterisks denote significant (p < 0.05) changes in Student's t-tests. Error bars represent s.e.m. Scale bar = 1.5 mm (A), 0.1 mm (A inset), 150 µm (B).

Figure 9. Model of molecular mechanisms underlying the metastatic transition and the proposed evolution of cancer stem cells by reprogramming

1. Diagram of the changes in pathway use in non-metastatic versus metastatic CCs, highlighting the metastatic transition and the recapitulation of non-metastatic CCs by in vitro conditions and of metastatic CCs by in vivo xenograft conditions. WNT-TCF and HH-GLI interactions are highlighted.

2. Model for the proposed role of GLI1 driving the metastatic transition when its activity levels, enhanced by multiple oncogene and loss of tumour suppressor events (the oncogene load), reach a threshold. This model recapitulates morphogenetic gradient interpretation resulting in distinct cell fates during development. Cancer stem cell reprogramming is proposed to be GLI1 driven at the metastatic transition.