• Open Access

Hedgehog Gli3 activator signal augments tumorigenicity of colorectal cancer via upregulation of adherence-related genes


To whom correspondence should be addressed.

E-mail: kenakano@med.kyushu-u.ac.jp


Hedgehog signal is re-activated in several cancers. In this study, we examined the role of Gli3 on malignant phenotype of tumorigenicity for colorectal cancer and its relationship with p53, WNT and ERK/AKT signals. Gli3 expression was detected in HT29 and SW480 (p53-mutant) cells, but not in DLD-1 (p53-mutant) or HCT116 (p53-wild type) cells by reverse transcription-polymerase chain reaction and immunocytochemistry. Full-length Gli3 transfection increased anchor-independent growth for all cells regardless of p53 status, with upregulation of adhesion-related genes. Exogenous Sonic-Hedgehog increased activator-type of Gli3 and colony formation in Gli3-positive HT29 and SW480 cells. After implantation of Gli3-FL or mock-transfectant DLD-1 cells into SCID mice, tumor formation was highly observed in only Gli3-FL-transfectant group. In clinical specimens, Gli3 expression was detected in subsets of colorectal cancer and related with poorly-differentiated histological type, while Sonic–Hedgehog was present with high incidence. In conclusion, activator Gli3 signal augments tumorigenicity of colorectal cancer irrespective of p53 status.

Hedgehog (Hh) signal is morphogenically important for embryonic patterning and controlling growth and cell fate during neonatal development.[1-3] Three types of Hh ligands homologus, Sonic-Hedgehog (Shh), Indian-Hedgehog and Desert-Hedgehog, have been identified in mammals.[4] In the absence of Hh ligands, Patched (Ptch), a 12-pass transmembrane receptor, suppresses Smoothened (Smo) activity, a G-protein-coupled receptor-like protein. Binding of Hh ligands to Ptch leads to the Ptch inactivation and consequent Smo activation. The signal transmits to downstream transcriptional factors: GLI proteins: Gli1, Gli2 and Gli3.[1, 5] Whereas Gli1 lacks a repressor domain,[6] Gli2 and Gli3 possess repressor and activator domains. In the presence of Hh ligands, Gli2 and Gli3 are transmitted to nuclei as full-length activator (Gli2-FL and Gli3-FL) and transcribe the target genes of Hh signal. In contrast, when there are no or less Hh ligand stimuli, they are cleaved by the ubiquitin ligase and generate transcriptional repressor isoforms (Gli2-R and Gli3-R).[7-9] Gli2 generally functions as an activator of Hh signal[10] but Gli3 is proposed to act as a repressor of Hh in embryonic development.[8] Therefore, Hh ligand signal accelerates Gli1 transactivation and inhibits the formation of repressor isoforms (Gli2-R and Gli3-R), which is reflected in Hh activation.[11]

A series of evidence has demonstrated Gli1 gene amplification/overexpression in glioma,[12] medulloblastoma[13] and rhabdomyosarcoma,[14] and Shh overexpression in gastric,[15] pancreas,[16] and breast cancers.[17] The Shh-Gli1 signal controls the phenotypes of invasiveness[18-21] and stemness[22-26] in various types of cancer. In colorectal cancer, Shh was reported to overexpress in tumor specimens and enhance the proliferation of colorectal cancer cells in vitro together with Gli1 mRNA upregulation.[27] On the other hand, Gli1 transfection inhibits the proliferation of colorectal cancer cells via inactivation of WNT signal.[28] Alternatively, Gli2 and Gli3 knockout mice exhibit the phenotype of abnormal development of intestine[29-31] and Gli3 knockdown inhibits anchor-dependent proliferation via p53 activation in colorectal cancer cells with wild type p53.[32] However, the majority (60–70%) of colorectal cancer has mutations in p53 gene,[33-35] and the influence of Hh pathway on the proliferation and tumorigenicity of colorectal cancer remains unclear.

In this study, the transfection of full-length Gli3 (Gli3-FL), but not Gli1 or Gli2, upregulated adherence-related genes and increased anchor-independent growth and tumorigenicity for colorectal cancer cells, regardless of p53 status. Shh signal augmented Gli3-FL isoform and the growth potential in endogenous Gli3-positive colorectal cancer cells. In clinical specimens, Gli3 expression was detected in subsets of colorectal cancer together with Shh overexpression. It is concluded that Shh-enhancing Gli3 activator signal may be involved in the tumorigenicity for colorectal cancer.

Materials and Methods

Plasmids and siRNA

Plasmids pFN21A-Gli1, -Gli2 and -Gli3 were purchased from Promega (Madison, WI, USA). Open reading frame sequences of Gli3-FL and EmGFP, which were amplified polymerase chain reaction (PCR), were integrated into the multicloning site of pLVSIN vector (Takara-bio, Shiga, Japan). Gli3-R (Gli3 repressor-form) expressing vector, which contains the sequence of the repressor domain at the N-terminal site but deletes that activator domain at the C-terminal site, was constructed by PCR using the following primers: forward, 5′-GCTACGACATCATGGAGGCCCAGTCCCACAGC-3′ and reverse, 5′-GCGGCCGCTTACGGTCGGCCAGGCGACCTGGACTG-3′. The plasmids for reporter assay of WNT activity, pTOPFLASH and pFOPFLASH (mutant control) were kindly provided by Dr Akihide Ryo (Yokohama City University). Gli3- and Smo-siRNA were purchased from Dharmacon RNA Technologies.

Cell lines

Human colorectal cancer cell lines p53-wild type HCT116 and p53-mutant HT29, SW480 and DLD-1 were obtained from American Type Culture Collection (Manassas, VA, USA) and cultured in RPMI 1640 mediated with 10% FBS (Sigma–Aldrich, Hokkaido, Japan) at 37°C in 5% CO2 humidified incubator. HT29 and DLD-1 cells stably expressing Gli3-FL and EmGFP were isolated with puromycin after transfection of the lentivirus vectors. These cells are designated as HT29/Gli3, HT29/Mock, DLD-1/Gli3 and DLD-1/Mock, respectively.

Proliferation assay

The indicated Gli expression plasmids or Gli3 siRNA was transfected into cells by Lipofectamine 2000 or Lipofectamine RNAiMAX (Life Technologies, Carlsbad, CA, USA), respectively. At 48 h after transfection, the cells were re-seeded in 12-well plates and we examined the anchor-dependent growth of colorectal cancer cells in vitro.

Soft agar colony formation assay

Soft agar colony formation was performed as previously described.[28] In brief, the cells transfected with the indicated plasmids or siRNA were mixed into a top layer of RPMI1640 medium containing 0.3% agar and 10% FBS. In separate experiments, each well was then further covered with supplemented-RPMI1640 containing 1.0 μg/mL of recombinant human Sonic-Hedgehog (rhSHH; R&D Systems, Minneapolis, MN, USA), or 10 μM of cyclopamine, as a Smo inhibitor (Toronto Research Chemicals, North York, Canada). Two weeks later, the colonies were stained with crystal violet (Sigma–Aldrich).

Luciferase reporter assay

At twenty-four hours after transfection with pTOPFLASH or pFOPFLASH reporter plasmid, pFN21A-Gli3 or mock plasmid, and pRL-SV40 as transfection control, luciferase activities were determined using Dual-Luciferase Reporter Assay System (Promega).

Isolation of RNA, and conventional and real-time RT-PCR

The procedures of total RNA isolation and cDNA synthesis were performed as previously reported.[36] The conventional and real-time RT-PCR reactions were performed using the sets of primers (Table S1) with MasterMix Kit (Qiagen, Hilden, Germany) and iQSYBER Green Supermix (Bio-Rad Laboratories, Philadelphia, PA, USA), respectively.


Cultured cells were immunostained with primary antibodies: anti-Gli1, anti-Gli3, anti-Shh, anti-Ptch, anti-Smo (1:100, Santa-Cruz Biotechnology, Santa Cruz, CA, USA) or anti-Gli2 (1:200, Rockland, Gilbertsville, PA, USA), followed by Alexa 488-labeled secondary antibodies (Life Technologies), as previously described.[37]

Western blot analysis

Protein samples were extracted using M-PER (Thermo Fisher Scientific, Chicago, IL, USA). Immunoblot was carried on with primary antibodies against phospho-ERK, total-ERK (1:1000, Cell Signaling Technology, Danvers, MA, USA), phospho-AKT, total-AKT, Gli3 (1:250, Santa-Cruz Biotechnology) or α-tubulin (1:1000, Sigma–Aldrich). Blots were developed with ECL plus Western Blotting Detection System (Amersham Biosciences, Piscataway, NJ, USA).


Tissue samples were obtained from patients with colorectal cancer who underwent the operation at Kyushu University and affiliated Hospitals from 1991 to 2010. The approval for the use of tissues was obtained from patients in accordance with the Ethical Committees of Clinical Study at Kyushu University. Immunohistochemistry was performed, as previously described,[17] using primary antibodies: anti-Gli1, anti-Gli3, anti-Shh (1:100, Santa-Cruz Biotechnology) or anti-Gli2 (1:250, Rockland). The level of protein expression was categorized to three grades: score 0, no detectable staining; score 1, staining less than 30%, and score 2, staining more than 30% of tumor cells.

Colon cancer xenograft model

Animal procedures were conducted in accordance with the Guide for the Care and the Use of Laboratory Animals approved by the Kyushu University. HT29/Mock, HT29/Gli3, DLD-1/Mock and DLD-1/Gli3 (1.0 × 103 or 1.0 × 104 cells in PBS) were implanted subcutaneously at the flank regions of Severe Combined ImmunoDeficiency (SCID) mice (Charles River Laboratories, Kanagawa, Japan). The tumorigenicity was monitored for 12 weeks after cell implantation.

Gene expression array

The cRNA was amplified, labeled, and hybridized to a 44K Agilent 60-mer oligomicroarray. Relative hybridization intensities and background hybridization values were calculated using Agilent Feature Extraction Software ( We calculated Z-scores and ratios from the normalized signal intensities of each probe for comparison between Gli3-manipulated and control samples. Then, we established the criteria for regulated genes: (upregulated genes) Z-score ≥2.0 and ratio ≥1.5-fold, (downregulated genes) Z-score ≤−2.0 and ratio ≥0.66.

Statistical analysis

The data are presented as the means ± SD. Student's t-tests were used to compare the continuous variable between two groups. Associations between Gli3 staining and clinicopathological features listed were analyzed by χ2 test. < 0.05 was considered as significant.


mRNA and protein expression of Hedgehog signal components

In conventional RT-PCR (Fig.S1A), Gli1 mRNA was detected in HCT116, HT29 and SW480. The mRNA expressions of Shh, Ptch, and Gli2 were detected in all cell lines. Gli3 mRNA was expressed in HT29 and SW480; however, very faint expression was detected in HCT116 and DLD-1. Smo mRNA was little expressed in DLD-1. In immunocytochemistry (Fig. S1B), Gli1 protein was detected in all cells except DLD-1, and Gli3 protein was expressed in HT29 and SW480. Smo protein was not detected in DLD-1 cells, whereas it was observed in other cells.

Gli1 or Gli2 transfection does not increase the colony formation and proliferation

Gli1 transfection downregulated the colony formation by 61% and 56% of the control for HT29 and DLD-1 cells (= 0.03 and 0.02), and also suppressed the proliferation (Fig. S2). Gli2 decreased the number of colonies on soft agar gels to 61%, 68% and 67% of the mock control for HT29, SW480 and DLD-1 cells, respectively (Fig. S2C). Gli2 did not affect the proliferation for all cell lines (Fig. S2D), despite Gli1 mRNA being significantly upregulated to 140%, 144% and 176% of the mock control in HCT116, HT29 and SW480 (= 0.02, 0.002 and 0.002, respectively).

Gli3-FL, but not Gli3-R, transfection increases colony formation and proliferation

Gli3-FL transfection upregulated the colony formation ability by 1.6–3.6 folds compared with the mock control in HCT116, and HT29, SW480 and DLD-1 cells (= 0.004, 0.019, 0.043 and 0.012, respectively; Fig. 1A). On the other hand, the increase in anchor-dependent growth of cells transfected with Gli3-FL was limited to 1.2 to 1.4 folds in HT29, SW480 and DLD-1 (< 0.05), and not significantly detected in HCT116 cells (Fig. 1B). Unlike the Gli3-FL transfection, Gli3-R did not significantly change the colony formation ability compared with the mock control in all cells (Fig. 1C). Furthermore, Gli3-R transfection did not significantly change the proliferation for all cell lines (Fig. 1D).

Figure 1.

Effects of Gli3-FL, Gli3-R and Gli3-siRNA on anchor-independent and -dependent growth. In colony formation assay, the anchor-independent growth was increased for Gli3-FL transfection (A), not changed for Gli3-R (C) and decreased for Gli3-siRNA (E) compared with the controls. The anchor-dependent proliferation was higher for Gli3-FL (B), not different for Gli3-R (D) and decreased for Gli3-siRNA (F) compared with the controls. *P < 0.05, **P < 0.01.

Gli3 RNA silencing decreases colony formation and proliferation

In a preliminary experiment with real-time RT-PCR, the knockdown efficiency of Gli3 siRNA was 70% and 66% in HT29 and SW480 cells. Gli3 siRNA decreased the number of colonies by 46% and 78% of the controls for HT29 and SW480 cells (< 0.05, P = 0.07, respectively; Fig. 1E). The proliferations were decreased by Gli3 siRNA, but the degree of inhibition was less than 25% in HT29 (= 0.10) and SW480 (< 0.05) cells (Fig. 1F).

rhSHH increases colony formation ability and proportion of activator Gli3 isoform

The number of colonies on soft agar gels was increased by 1.0 μg/mL of rhSHH to 1.4- and 1.9-fold compared with the vehicle control for endogenous Gli3-positive HT29 and SW480 cells (= 0.006 and 0.024, Fig. 2A); however, it was not significantly changed for HCT116 and DLD-1 cells without endogenous Gli3 expression. On the other hand, the proliferation was not affected by rhSHH for all cell lines (Fig. 2B), of which the dose (1.0 μg/mL) upregulated Gli1 expression in HCT116, HT29 and SW480 cells expressing endogenous Smo (< 0.01; Fig. S3). Then, we analyzed the proportional change between activator and repressor isoform of Gli3 protein in the presence of rhSHH by Western blots (Fig. 2C). rhSHH retained Gli3 protein as Gli3-FL, such that the intensities of Gli3-FL protein in HT29 and SW480 cells treated with rhSHH were 2.25 and 1.56 times higher than those with vehicle (= 0.02 and 0.002, respectively). On the contrary, the expression of Gli3-R protein was not significantly changed by the addition of rhSHH in both cell lines. Moreover, the colony numbers were increased by co-transfection of Gli1 and Gli3-FL to 1.5–2.6-fold compared with the mock control (= 0.002 for HCT116, 0.0016 for HT29, 0.022 for SW480 and 0.0049 for DLD-1; Fig. S4).

Figure 2.

Effects of rHSHH, cyclopamine and Smo-siRNA on anchor-independent and -dependent growth. In colony formation assay, the anchor-independent growth was increased by rhSHH (A), whereas it was decreased by cyclopamine (D) or Smo-siRNA (E) in Gli3-positive HT29 and SW480 cells. Smo-siRNA deteriorated the rhSHH-induced colony growth (E). The anchor-dependent pro-liferation was not different between the groups with rhSHH or vehicle (B). Western blots (C) show the increased Gli3-FL but not Gli3-R protein by rhSHH. *P < 0.05, **P < 0.01.

Smo-targeting cyclopamine and siRNA decrease colony formation ability

Cyclopamine decreased the colony formation ability for HT29 and SW480 to 37% and 18% of the vehicle (Fig. 2D). For Smo-positive HCT116 and Smo-negative DLD-1 cells both of which did not detect endogenous Gli3 expression, cyclopamine did not affect the colony formation significantly. In addition, Smo-siRNA decreased the number of colonies to 51% and 56% of the control-siRNA (= 0.04 and 0.001) in the presence of rhSHH and 56% and 61% (= 0.0005 and 0.01) in the absence of rhSHH in HT29 and SW480 cells (Fig. 2E). The enhancing effect of rhSHH on colony formation ability was diminished when transfected with Smo-siRNA.

Influences of Gli3 to Hh, WNT, MAPK and PI3K signals

Gli3-FL transfection did not change Gli1 mRNA expression in all cell lines (Fig. 3A), but increased Gli2 mRNA expression in HT29, SW480 and DLD-1 (< 0.05; Fig. 3B). The effect of Gli3-FL on Shh expression varied in each cells (Fig. 3C). In contrast, Gli3-R transfection significantly decreased Gli1 mRNA in HCT116, HT29 and SW480 cells (< 0.05; Fig. 3D), but Gli3-siRNA did not change Gli1 expression in HT29 and SW480 cells (Fig. 3F). On the other hand, Gli2 expression was inhibited in HCT116 and SW480 cells by Gli3-R transfection (Fig. 3E) and in HT29 cells by Gli3-siRNA (Fig. 3F). Gli3-FL did not change WNT signal activity (Fig. S5). Moreover, Gi3-FL transfection did not affect the activities in MAPK (phosphorylated-ERK expression) and PI3K (phosphorylated-AKT expression) pathways (Fig. S6).

Figure 3.

Interaction between Gli3 and other Hh components. RNA samples after transfection with the indicated plasmids or siRNAs were subjected to qRT–PCR. Gli3-FL transfection did not affect Gli1 (A) and increased Gli2 expression (B) for three of four cell lines, while the changes in Shh expression (C) were varied. Gli3-R transfection inhibited Gli1 (D) and Gli2 (E) expressions for the indicated cells. Gli3-siRNA did not change Gli1 or Gli2 (F) except Gli2 expression in HT29 cells. *P < 0.05, **P < 0.01.

Gene expression array

The transcripts of 600 genes were upregulated in DLD-1/Gli3 cells compared with Gli3-negative DLD-1/Mock. Upon gene ontology (GO) classification using DAVID software, many annotation terms were found to be associated with genes in the extracellular region, biological adhesion, cell adhesion, and extracellular matrix (Table 1). Among the genes in these GOs, integrin 4, growth differentiation factor 15 (GDF15), neurexophilin 4 and semaphilin 3B were upregulated in both HCT116 and DLD-1 cells. Extra cellular matrix (ECM) receptor interaction and focal adhesion signals were identified by KEGG classification for the pathway modulated by Gli3 transfection into Gli3-negative HCT116 and DLD-1cells. In addition, the upregulated 10 and 8 (4) genes in the categories of tumorigenicity and stem (cancer stem) cell for DLD-1 were depicted in Figure S7. In HCT116, nine and two genes in the categories of tumorigenicity and cancer stem cell were upregulated by Gli3-FL transfection. In these experiments, only the ASS1 (argininosuccinate synthase 1) gene was upregulated in both cells. The changes in some picked-up gene expressions in qRT-PCR paralleled those in microarray analyses (data not shown), validating the accuracy for the gene profiling in microarray.

Table 1. Lists of GO and pathway identified by KEGG classification associated with the genes which were upregulated by Gli3-FL transfection
GO biological processDLD-1/Gli3HCT116/Gli3
Gene numberP-valueGene numberP-value
Extracellular region (GO:0005576)448.2 E-71013.3 E-4
Biological adhesion (GO:0022610)344.3 E-4355.6 E-2
Cell adhesion (GO:0007155)344.3 E-4355.7 E-2
Proteinaceous extracellular matrix (GO:0005578)201.2 E-3202.3 E-2
 Term of pathwayCountP-value
  1. GO, gene ontology; KEGG, kyoto Encyclopedia of genes and genomes; DLD-1, p53-mutant and Gli3-undetectable colorectal cancer cells; HCT116, p53 wild-type and Gli3-undetectable colorectal cancer cells

  2. a

    Other pathways: Autoimmuno thyroid disease; systemic lupus erythematosus; athema; glycine, serin and threonine metabolism; alanine, aspartate and glutamate metabolism; natural killer cell mediated cytotoxicity; antigen processing and presentation; small cell lung cancer (P < 0.05).

DLD-1/Gli3ECM-receptor interaction85.2 E-3
Axon guidance106.0 E-3
Hypertrophic cardiomyopathy86.2 E-3
Dilated cardiomyopathy89.5 E-3
Focal adhesion113.5 E-2
Arrythmogenic right ventricular cardiomyopathy64.7 E-2
HCT116/Gli3Type I diabetes mellitus91.9 E-4
Complement and coagulation cascades113.3 E-4
Viral myocarditis114.2 E-4
ECM-receptor interaction111.6 E-3
Allograft rejection72.6 E-3
Graft-versus host disease73.9 E-3
Focal adhesion175.6 E-3
Cell adhesion molecules135.6 E-3
Regulation of actin cytoskelton171.0 E-2

Gli3-FL transfectants exhibit tumorigenicity in vivo

After implantation of DLD-1/Gli3 and DLD-1/Mock (Gli3 negative) cells in SCID mice, subcutaneous tumors were formed in four of six SCID mice with 1.0 × 104 DLD-1/Gli3 cells, and in none of four mice with 1.0 × 103 DLD-1/Gli3 cells. On the contrary, DLD-1/Mock (1.0 × 103 and 1.0 × 104) implanted mice did not develop any tumor (Fig. 4A). Furthermore, we implanted HT29/Gli3 cells and compared the tumorigenicity with parental HT29/Mock cells with endogenous Gli3 expression. In contrast to DLD-1 cells, three of four mice developed subcutaneous tumors after implantation of 1.0 × 104 cells for both HT29 transfectants.

Figure 4.

In vivo tumorigenicity of Gli3-FL-transfectants and immunohistochemistry for Gli and Shh expres-sions in patient colorectal cancer tissues. (A) HT29/Gli3, HT29/Mock, DLD-1/Gli3 or DLD-1/Mock cells were subcutaneously injected into SCID mice. No tumor is shown in the mice receiving DLD-1/Mock cells without endogenous Gli3 or Smo. (B) Colorectal cancer tissues of patients were immunostained with antibodies for the indicated Gli and Shh. Three grades of expression level for Gli3 are shown as score-0, 1 and 2.

GLIs and Shh protein expressions are detected with various patterns in colorectal cancer tissues of patients

In fifty clinical specimens, we analyzed Gli3 protein staining and diagnosed as Gli3-positive for 12 out of all samples (24%), of which seven (14%) and five (10%) tumors were given scores 1 and 2, respectively (Table 2). There was no relation between the scores of Gli3 expression and clinicopathological features: age, location and stage of tumors. Positive Gli3 staining was more frequently detected in poorly differentiated adenocarcinoma, and a significant relationship was noted between the two parameters (= 0.02, Gli3 score 1 + 2 vs. poorly differentiated type). The Gli3 staining was observed in the area of adenocarcinoma cells with heterogeneity pattern, but not detected in normal mucosa (Fig. 4B). The localization of Gli3 expression was dominant in nuclei rather than cytoplasm of adenocarcinoma cells. Gli1 expression was observed in 26 of 50 specimens. Gli2 expression was also detected in 28 of 50 samples. On the other hand, Shh expression was strongly detected in 30 of 50 specimens, and the localization of Shh protein was mainly in cytoplasm of adenocarcinoma cells and little in stromal cells (Fig. 4B). In the Gli3 positive specimens, Shh expression was detected at high incidence of 67% (8 of 12 Gli3 positive specimens).

Table 2. Clinicopathological features and Gli3 immunostaining of 50 colorectal carcinoma specimens
Clinicopathological features n Gli3P-value
0 (= 38)1 (= 7)2 (= 5)
  1. Colorectal adenocarcinoma specimens were immunostained with anti-Gli3 antibody. Score 0, no detectable staining. Score 1, staining less than 30% of adenocarcinoma cells. Score 2, more than 30% of adenocarcinoma cells. Differentiations of colorectal cancers were classified according to the WHO classification. UICC, Union for International Cancer Control.

Differentiation (WHO)


(0 vs1 + 2:<0.05)

TMN stage (UICC)
Lymph node (N)


Hedgehog (Hh) signal plays a crucial role for organ development and homeostasis in embryonic and postnatal phases, and is re-activated in several types of cancers.[1, 2, 12-17] In this study, we investigated the biological significance of Hh signal in tumorigenicity for colorectal cancer. Surprisingly, we found that the Gli3 activator signal, which was emerged by Gli3-FL transfection or the increase in non-cleaved Gli3-FL isoform by Shh stimuli, enhanced the anchor-independent growth of colorectal cancer irrespective of p53 status. Furthermore, the tumorigenicity in vivo was gained by Gli3 overexpression, and Gli3 and Shh expressions were observed in subsets of clinical specimens of colorectal cancer. The current study has firstly demonstrated the significance of Gli3 activator signal on the malignant phenotype of tumorigenicity for colorectal cancer.

It has been shown that Shh increases the growth of colorectal cancer[27] and the Shh-Gli1 signal contributes to the enhancement in stemness[38] and xenograft tumor growth.[39] In our previous study[28] and Figure S2, Gli1 transfection deteriorated the WNT activity and growth of colorectal cancer cells, while Shh increased the proliferation and Gli1 expression.[27] These could be explained by the idea that Gli1 may augment the in vivo tumor formation of colorectal cancer with stem-like phenotype, while Shh may stimulate the growth via other Hedgehog pathways. Thus, we elucidated the effect of Gli2 and Gli3 transfection (overexpression) on the colony formation and proliferation for colorectal cancer cells. Gli2 transfection reduced the anchor-independent growth in some cell lines, but it did not affect the anchor-dependent growth. The degree of Gli1 upregulation by Gli2 transfection was only a few folds (data not shown), and the inhibitory effect of Gli1 on growth might be obscured by a Gli2 overexpression induced effect. In contrast, the growth phenotype of colorectal cancer was enhanced by Gli3-FL transfection in all cells. The effect of Gli3-FL on proliferation was weaker than that on colony formation, suggesting that Gli3-FL may mainly affect anchor-independent growth for colorectal cancer cells. Gli3-FL transfection might override the inhibitory effect of Gli2 on anchor-independent growth although it also upregulated Gli2, because the degree of Gli2 upregulation by Gli3-FL transfection was only several folds, which was much less than that by Gli2 transfection (data not shown). These results suggest that the enhancement in anchor-independent and -dependent growth is mediated through the Gli3-specific and Gli1/Gli2 independent signal pathway.

Gli3-FL protein is eventually cleaved to Gli3-R by proteolytic reaction,[8] and we further examined whether Gli3-R contributed to the growth phenotype by transfection of Gli3-R expression plasmid. Despite the inhibition in Gli1 expression, Gli3-R overexpression did not affect either anchor-independent or anchor-dependent growth in these cells. It may be because the degree of Gli1 inhibition by Gli3-R transfection was insufficient to change the growth phenotype biologically. We then examined the effect of Gli3 silencing on the both types of growth in endogenous Gli3-positive cells harboring p53-mutation. The proliferation was downregulated by Gli3 knockdown for these cells, which is consistent with a previous report using colorectal cancer cells with p53-wild type.[32] The anchor-independent growth was also impaired by Gli3 knockdown. These results indicate that the activator type of Gli3-FL protein may be important for the growth of colorectal cancer through p53-independent mechanisms.

The colony formation ability was accelerated with rhSHH in endogenous Gli3-positive cells, where the increase in activator type of Gli3-FL isoform was confirmed by Western blot analysis. Even though rhSHH increased Gli1 expression, rhSHH induced Gli3 activator isoform might override the inhibitory effect of Gli1 against colony formation based on our data that co-transfection of Gli1 and Gli3-FL increased anchor-independent growth. In contrast, the colony formation of Gli3-negative cells was not changed with rhSHH. To verify the linkage of Shh/Smo signal to Gli3-induced anchor-independent growth of colorectal cancer, we used cyclopamine, an inhibitor of Smo. The administration of cyclopamine suppressed the colony formation for endogenous Gli3-positive cells, which is consistent with a previous study,[40] but not for Gli3-negative cells. To exclude the possibility for non-specific effect of chemical inhibitor, we also used Smo-targeting siRNA and confirmed the specific linkage of Smo to the Gli3-mediated growth signal. These results indicate that the Shh/Smo activation could retain Gli3 protein as a full-length activator, which may in turn accelerate the colony formation ability for colorectal cancer.

To examine whether the Gli3-FL activator signal reflects for in vivo tumorigenicity of colorectal cancer, we implanted the Gli3-FL transfectant colorectal cancer cells into SCID mice. In HT29 cells, which have endogenous Gli3 and Shh, the tumorigenicity was not different between Gli3-FL and mock transfectants. However, in DLD-1, which has no endogenous Gli3 or Smo, a small number of Gli3-FL-transfectants formed tumors but not the mock. These results suggest that the tumorigenic potential of colorectal cancer may be gained if a certain level of full-length Gli3 isoform is present.

With respect to the responsible factors for the tumorigenicity, we postulated the WNT signal as a candidate because of frequent mutations in WNT components for colorectal cancer, which participate in cell growth, survival and migration.[41] Actually, the HCT116 cell has gain-of-function mutation of β-catenin,[42] and HT29, SW480 and DLD-1 have loss-of-functional mutations in adenomatous polyposis coli (APC) genes.[43] However, Gli3-FL transfection had no effect on WNT activity in colorectal cancer cells (Fig. S5). Alternatively, mutations in Ras are also reported in about 40–50% of colorectal cancers.[44, 45] We thus analyzed the related MAPK/ERK and PI3K/AKT activities, but the increase in both activities was not detected in Gli3-FL transfectants (Fig. S6). These results suggest that the Gli3-induced tumorigenic phenotype may not be mediated by WNT, MAPK/ERK or PI3K/AKT pathway in colorectal cancer. Therefore, we tried to seek the possible factors using gene microarray analysis. In Gli3-FL transfectant HCT116 and DLD-1 cells, adherence-related genes were picked up as the candidates. Moreover, Gli3 expression upregulated several genes in “tumorigenicity” and “cancer stem cell” categories, although the common gene in the categories was only ASS1 in both cells, suggesting that these genes might synergistically contribute to the Gli3-induced tumorigenicity. Further investigation in future should be of great interest to determine which molecules are direct mediators for Gli3/Smo-induced tumorigenicity.

Relevant to clinical significance of Gli3/Shh/Smo signal of colorectal cancer specimens from patients, the positivity of Gli3 protein was 24% of total samples and the pattern of staining exhibited heterogeneity in adenocarcinoma cells. 58% of Gli3 positive samples were diagnosed as poorly differentiated type and the relationship between the two parameters was shown. Though a poorly differentiated cancer is usually recognized to lose adhesive activity, Mallin et al. have shown the upregulated expression of the adhesion-related gene: GDF15, in poorly differentiated colorectal cancer.[46] Taken together, Gli3 expression was detected with a heterogenic pattern in the tumor region in the present study, suggesting that the expression of adhesion-related genes may not directly conflict with the loss of differentiation of colorectal cancer. Gli3 expression did not significantly correlate with TNM stage and lymph node metastasis, indicating that other factors, in addition to Gli3, are involved in malignant phenotypes for tumor progression. On the other hand, Shh was stained at the whole area of adenocarcinoma cells in the almost specimens, but not normal mucosa, as previously reported,[27] and almost Gli3 positive specimens were Shh positive. These results suggest that the isoform of Gli3 we detected seems to be Gli3-FL type dominant, although the anti-Gli3 antibody we used reacts with both Gli3-FL and Gli3-R. The Hh target therapy may improve the outcome for patients with Gli3/Smo activated colorectal cancer. A randomized controlled study with a large number of patients should be addressed in future.

In conclusion, Hedgehog Gli3 activator signal, but not Gli1 or Gli2, is involved in anchor-independent growth and tumorigenicity for colorectal cancer through upregulation of adherence-related genes irrespective of p53 status. Taken together, histological examination of patient specimens indicates that Gli3 and Shh are expressed in subsets of colorectal cancer and may be therapeutic targets for colorectal cancer.


We thank Drs Akinori Iwashita and Masayuki Sada (Fukuoka University) for helpful discussion on histological study and Ms Kaori Nomiyama for general assistance.

Disclosure Statement

The authors have no conflict of interest.