Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Human GLI2 and GLI1 are part of a positive feedback mechanism in Basal Cell Carcinoma


Transgenic mouse models have provided evidence that activation of the zinc-finger transcription factor GLI1 by Hedgehog (Hh)-signalling is a key step in the initiation of the tumorigenic programme leading to Basal Cell Carcinoma (BCC). However, the downstream events underlying Hh/GLI-induced BCC development are still obscure. Using in vitro model systems to analyse the effect of Hh/GLI-signalling in human keratinocytes, we identified a positive feedback mechanism involving the zinc finger transcription factors GLI1 and GLI2. Expression of GLI1 in human keratinocytes induced the transcriptional activator isoforms GLI2α and GLI2β. Both isoforms were also shown to be expressed at elevated levels in 21 BCCs compared to normal skin. Detailed time course experiments monitoring the transcriptional response of keratinocytes either to GLI1 or to GLI2 suggest that GLI1 is a direct target of GLI2, while activation of GLI2 by GLI1 is likely to be indirect. Furthermore, expression of either GLI2 or GLI1 led to an increase in DNA-synthesis in confluent human keratinocytes. Taken together, these results suggest an important role of the positive GLI1-GLI2 feedback loop in Hh-mediated epidermal cell proliferation.


Basal Cell Carcinoma (BCC) of the skin represents the most common tumour in the western world. Genetic studies of the basal cell nevus or Gorlin syndrome, which predisposes patients to the early development of multiple BCCs, have shown that mutational inactivation of the Hedgehog (Hh)-receptor protein Patched (PTCH1) is the primary event in the formation of BCC (Gailani et al., 1996; Hahn et al., 1996; Johnson et al., 1996). In the absence of HH ligand, PTCH inhibits Hh-target gene expression, while loss of PTCH activity leads to increased expression of the Hh target genes GLI1 and PTCH itself (for reviews see Goodrich and Scott, 1998; Hahn et al., 1999; Ingham, 1998; Toftgard, 2000; Villavicencio et al., 2000). Further evidence supporting the importance of activated Hh-signalling in BCC has come from studies of genetic mouse models and skin grafting experiments: (1) heterozygous ptc+/− mice develop BCC-like features upon UV-irradiation, though sporadic BCC formation does not normally occur in mice (Aszterbaum et al., 1999). (2) Human keratinocytes expressing Sonic hedgehog (SHH) form BCC-like structures when grafted onto the back of nude mice (Fan et al., 1997). (3) Overexpression of mediators of Hh-signalling such as SHH, GLI1, Gli2 and an oncogenic form of SMOH, in epidermal cells of transgenic mice leads to the induction of BCC-like tumours (Fan et al., 1997; Grachtchouk et al., 2000; Nilsson et al., 2000; Oro et al., 1997; Xie et al., 1998). Gli1 and Gli2 are members of the GLI family of zinc finger transcription factors (Ruppert et al., 1988). During embryonic development of vertebrates both genes, whose expression is partially overlapping, are transcriptionally activated in response to Hh-signalling and are able to mediate most of the effects caused by activation of the pathway (reviewed in Matise and Joyner, 1999; Ruiz i Altaba, 1999).

GLI1 was originally identified as an oncogene amplified in glioblastoma. It acts as a transcriptional activator and together with E1A is able to transform primary cells (Kinzler et al., 1987; Ruppert et al., 1991). Expression of human GLI1 in basal keratinocytes of transgenic mice results in the development of several types of skin tumours, some of which show BCC-like features (Nilsson et al., 2000). Further evidence for a critical role of GLI1 in tumorigenesis comes from overexpression studies in frogs, where GLI1 can also induce structures resembling epidermal tumours (Dahmane et al., 1997).

Most studies of Gli2 have focused on embryonic development, where it plays a crucial role in neural, lung, and skeletal development (Ding et al., 1998; Hardcastle et al., 1998; i Altaba, 1998; Matise et al., 1998; Mo et al., 1997; Motoyama et al., 1998; Park et al., 2000). Only recently, it was shown that overexpression of mouse Gli2 in the basal epidermal layer of transgenic mice induced the formation of BCC-like tumours (Grachtchouk et al., 2000).

Though much is known about Hh signalling in Drosophila and murine development, our understanding of the molecular mechanisms and tumorigenic programs that are activated in response to Hh-signalling and GLI-activity in human keratinocytes is still very limited. Most studies have been carried out using mouse as a model system to study specific aspects of BCC development. Given the considerable differences between human and murine skin and the relative resistance of wild type mice to BCC development, murine studies alone will only provide a limited insight into the mechanisms by which Hh activation leads to BCC. To better understand the downstream effects of Hh/GLI activation in human BCC, we used retrovirally infected primary human keratinocytes and a tetracycline-inducible human keratinocyte cell line as in vitro model systems for the analysis of the effects of Hh/GLI-signalling on the gene expression pattern and cellular phenotype of human epidermal cells.

We provide evidence for a positive feedback mechanism between the two putative oncogenes GLI1 and GLI2 in human epidermal cells. Expression of either transcription factor stimulates DNA-synthesis in confluent human keratinocytes, suggesting that both factors may induce epidermal cell proliferation. This would point to a role of the GLI1–GLI2 feedback mechanism in tumorigenesis, an interpretation that is further supported by the elevated levels of GLI2 as well as GLI1 mRNA in BCCs.


A highly efficient retroviral expression system (Deng et al., 1997) was used to generate GLI1-expressing and EGFP-expressing human primary keratinocytes. Using DNA arrays hybridized with cDNA from GLI1-expressing human primary keratinocytes and EGFP-expressing control cells (Regl et al., unpublished), we found that GLI2 transcription was strongly induced by GLI1. Overexpression of mouse Gli2 in epidermal cells of transgenic mice leads to the development of BCC-like structures (Grachtchouk et al., 2000) though partial functional redundancy of mouse Gli1 and Gli2 raises the possibility that Gli2 induces BCC development by mimicking the activity of Gli1. The observation that GLI1 activates expression of GLI2 in human epidermal cells, however, points to a combined activity of both transcriptional regulators in BCC development. We therefore analysed in detail the induction of GLI2 by GLI1, its expression in BCCs and its possible function in BCC development. Using retrovirally infected primary keratinocytes and tetracycline inducible keratinocyte lines as human model systems for Hh/GLI-signalling in BCC we addressed the following questions: (1) Is GLI2 an early transcriptional target of GLI1? (2) Which splice variants of GLI2 are induced by GLI1? (3) Is GLI2 expressed in BCC (4) Does GLI2 control the transcription of GLI1? (5) What is the effect of GLI1/GLI2 expression on the proliferation of epidermal cells?

GLI1 expression in human keratinocytes leads to GLI2 transcription and GLI1 auto-activation

We used real-time PCR analysis to quantitatively and qualitatively analyse the induction of GLI2 in response to exogenous GLI1. We also addressed the issue of GLI1 auto-activation which has been described in amphibians (Dahmane et al., 1997). The primers used for these analyses are shown in Figure 1b and Table 1. Figure 1a shows a graphical display of a representative real-time PCR experiment demonstrating activation of GLI2 transcription in response to GLI1 expression in primary human keratinocytes (HEK). As shown in Figure 1c, GLI1 expression led to a 24.5-fold increase of Gli2 mRNA levels at 24 h post infection (p.i.). After 48 h and 144 h p.i. GLI2 mRNA expression was induced 28.6-fold and 210.9-fold, respectively. The increase of GLI2 was paralleled by an increase of endogenous GLI1 transcript levels (24.8-fold at 24 h p.i, 26.7-fold at 48 h p.i. and 29.9-fold at 144 h p.i.), suggesting auto-activation of GLI1 transcription. PTCH mRNA expression, a direct GLI target used as positive control, was also gradually induced by GLI1 (22.9-fold at 24 h p.i., 23.6-fold at 24 h p.i. and 24.8-fold at 144 h p.i.).

Figure 1

Real-time analysis of human keratinocytes expressing GLI1. (a) Representative real-time PCR result showing that GLI2 is strongly induced in GLI1-expressing HEK cells. (b) Location of primer pairs used to distinguish between GLI2α/β and GLI2γ/δ splice variants as well as endogenous GLI1 (GLI1(endo)) and retrovirally expressed GLI1 (GLI1(rv). (c) mRNA expression of PTCH, endogenous GLI1 (GLI1(endo)), GLI2 and splice variants GLI2α/β in cells overexpressing GLI1(rv) for 24, 48 and 144 h, respectively. (d) Luciferase reporter gene assay showing that GLI2β encodes a potent transcriptional activator. HeLa cells were transfected with a GLI2β, GLI1 or mutant GLI1 expression plasmid (pGLI1mut, control) together with a luciferase reporter plasmid containing 6 Gli binding sites and a lacZ expression plasmid for normalization. Activation of reporter gene expression is expressed as fold induction of luciferase activity compared to control transfected cells. Data represent results from three independent experiments each carried out in duplicate. (e) Analysis of the time course of GLI1 transgene, GLI2α/β and PTCH transcription in HaCaT cells expressing GLI1 under the control of the tetracycline repressor. Upper figure illustrates induction of GLI1 protein expression in response to tetracycline, which was added for the times indicated. Equal amounts of total cellular protein were loaded on each lane. RNA was harvested after 3, 6, 24 and 48 h of tetracycline treatment and subjected to real-time PCR analysis. Note that in (c) and (e) fold induction values are plotted on a log2 scale due to very high induction values. Mean values of six measurements in three independent real-time PCR experiments are shown. The standard deviation for each experiment was below 15%. To exclude amplification of genomic DNA, all samples were also tested without reverse transcriptase treatment prior to PCR amplification. The specificity and quality of the PCR reactions were controlled by direct sequencing of the PCR- amplicons, melting curve analysis and gel electrophoresis. ‘Primer only’ controls, without template were done to ensure the absence of primer dimers. Large ribosomal protein P0 (RPLP0) was used as a reference standard for all analyses to control for the amount of sample material (Martin et al., 2001)

Table 1 Primer sequences for real-time PCR analysis

GLI1 induces the GLI2α/β splice variants in human keratinocytes

Alternative splicing of GLI2 mRNA can result in short isoforms which – based on functional analysis of protein domains of GLI proteins – lack the putative transactivation domain of human GLI2 raising the possibility that these isoforms act as transcriptional repressors rather than activators (Sasaki et al., 1999). We therefore characterized the isoform specific pattern of GLI2 expression in both GLI1 infected human keratinocytes (Figure 1c) and in BCCs (Table 2). We distinguished between the long GLI2α/β and the short GLI2γ/δ splice variants using the primer pairs shown in Figure 1b and Table 1 for real-time PCR analysis. At 24 h p.i. GLI1 expression led to a 24.9-fold increase of Gli2α/β mRNA. After 48 h and 144 h GLI2α/β mRNA expression was induced 28.6-fold and 210.9-fold, respectively (Figure 1c). At all time points the short GLI2 variants γ and δ were expressed only at trace amounts (Ct value >38) in both EGFP and GLI1 expressing cells, indicating that GLI1 induces exclusively the putative transcriptional activator forms. Using primers to distinguish between the GLI2 α and β isoforms we found a ratio of β to α of about 3 : 1 (data not shown). GLI2β is more closely related to mouse GLI2 than the other isoforms (Tanimura et al., 1998).

Table 2 Gene expression in BCC compared to normal human skin

To investigate whether the predominant isoform GLI2β indeed acts as a transcriptional activator we co-transfected HeLa cells with GLI2β expression plasmid and a GLI-reporter construct containing six Gli-binding sites upstream of the firefly luciferase reporter gene (Dai et al., 1999). GLI2β expression resulted in a 9.8-fold increase in reporter gene activity, suggesting that human GLI2β encodes a strong transcriptional activator. Positive control transfection with GLI1 expression plasmid led to a 4.2-fold increase in luciferase activity (Figure 1d). The stronger induction of reporter gene expression by GLI2β compared to GLI1 contrasts results obtained with mouse Gli2, which, when compared to Gli1, has been shown to be only a weak activator unless the putative N-terminal repressor domain is removed (Sasaki et al., 1999). The different efficiency of human and mouse Gli2 genes to activate transcription underlines the necessity of species-specific model systems to study GLI-gene function.

GLI2 induction is delayed compared to the putative direct target PTCH

To analyse in more detail the time course of GLI2 induction in response to GLI1, we genetically modified the human keratinocyte cell line HaCaT (Boukamp et al., 1988) such that human GLI1 protein expression can be regulated in a tetracycline-dependent manner (Eichberger, unpublished). This system allows the analysis of early transcriptional responses to GLI1, which should facilitate the identification of candidate direct GLI1 target genes. Western blot analysis of tetracycline-treated cells showed that GLI1 protein was induced in modified HaCaT cells within 3 h of tetracyclin treatment and gradually accumulated over a time range of 48 h (Figure 1e, upper image) demonstrating precisely controlled induction of GLI1 expression. Real-time PCR analysis showed that GLI2 mRNA levels were unchanged after 3 h of tetracycline treatment, weakly elevated after 6 h (1.5-fold) but strongly induced after 24 h (23.1-fold ) and 48 h (27.8-fold) (Figure 1e). The kinetics of PTCH mRNA accumulation closely paralleled that of GLI1 protein (Figure 1e). PTCH mRNA levels increased 2.3-fold after 3 h, 23.1-fold after 6 h, 24.4-fold after 24 h and 26.3-fold after 48 h of tetracyclin treatment. The delayed response of GLI2 to GLI1 expression suggests that, in contrast to PTCH, GLI2 is unlikely to be a direct target of GLI1. In line with this finding we did not detect any Gli consensus binding sequences in the putative GLI2 promoter region (data not shown).

GLI2 is overexpressed in BCCs

Since GLI1 is highly expressed in BCCs (Bonifas et al., 2001; Dahmane et al., 1997; Ghali et al., 1999; Green et al., 1998) and is able to activate GLI2 transcription, we analysed GLI2 levels in BCCs and normal skin by real-time PCR. We tested a panel of 21 tumour samples for GLI2 mRNA. Since tumour and skin samples may contain variable proportions of dermal components we also measured the expression of the dermal marker vimentin. Vimentin expression was measured in 10 individual normal skin samples and six preparations with low vimentin expression were pooled and used for analysis. The mean ratio of vimentin expression in BCC samples and skin was 1 : 1.3, demonstrating comparable amounts of mesoderm in BCC and normal skin preparations. Pooling of normal skin was done to reduce site-specific and individual variations. Normal skin samples with a large proportion of adipose tissue were excluded from the analysis.

As shown in Table 2, all 21 BCC samples had elevated levels of GLI2 mRNA compared to normal skin. The increase of GLI2 transcripts in BCCs varied from twofold (BCC5) to 67-fold (BCC25) with an average increase of 10.5-fold as calculated from mean ΔCt values (see below for details). Expression of GLI1, PTCH and MYC was analysed as controls. Consistent with previous reports (Bonifas et al., 2001; Dahmane et al., 1997; Ghali et al., 1999; Green et al., 1998), GLI1 transcription was strongly up-regulated in all BCC samples ranging from 18-fold (BCC35) to more than 4000-fold (BCC20) (on average 270-fold induced) as was PTCH (2.3-fold (BCC5) to 52.4-fold (BCC3)). MYC was consistently downregulated in all tumours as has been described previously by Bonifas et al. (2001).

The average mRNA concentration of GLI2 in BCCs was about 14-fold higher than the level of GLI1 (mean ΔCtGLI2=2.9 versus mean ΔCtGLI1=6.7, for ΔCt calculation see Materials and methods section). In normal skin GLI2 levels were about 360-fold higher than GLI1 levels (mean ΔCtGLI2=6.3 versus mean Δ CtGLI1=14.8). Almost exclusively the long splice variants GLI2α/β were expressed in tumour samples at a ratio of about 1 : 1 (data not shown). In a recent study of Hh-target gene expression in BCCs, GLI2 was not found to be consistently expressed in BCCs at levels significantly different from normal skin. Detailed results were not shown, but the different findings may be due to a different method used (RNase protection), and the small number of tumour samples analysed (n=7) (Bonifas et al., 2001).

GLI2β activates expression of GLI1 and induces osteoblast differentiation of mouse embryonic fibroblasts

Genetic studies in mice have shown that Gli2 is required for normal Gli1 expression levels in the central nervous system (Ding et al., 1998). Combined with our data on GLI1-mediated induction of GLI2 this points to a possible feedback loop between GLI1 and GLI2 expression. To test this hypothesis, we transiently transfected HaCaT cells with either a GLI1, GLI2β or control expression plasmid (pGLI1mut) and analysed transcription of GLI2, PTCH and GLI1 by real-time PCR. Consistent with our previous findings in primary keratinocytes and inducible HaCaT lines, overexpression of GLI1 resulted in a 1.6-fold increase in PTCH mRNA levels and a 2.7-fold induction of GLI2 mRNA. Transfection of HaCaT cells with GLI2β resulted in a fivefold increase in GLI1 mRNA levels, suggesting a positive feedback mechanism between GLI1 and GLI2 in human keratinocytes (Figure 2a). Since it is conceivable that this mechanism applies to human cells only, we turned to mouse C3H10T1/2 fibroblasts, which are frequently used to assay Hh-signalling (Kinto et al., 1997). We transfected C3H10T1/2 cells with either human GLI1, GLI2β or control expression plasmids and measured mouse Gli1 or mouse Gli2 transcripts by real-time PCR. As shown in Figure 2b expression of human GLI1 in mouse fibroblasts led to a 6.5-fold increase of mouse Gli1 mRNA levels and to a 2.1-fold increase of mouse Gli2 mRNA. Overexpression of human GLI2β resulted in a 12.3-fold activation of mouse Gli1 transcription and a 2.6-fold increase of mouse Gli2 demonstrating conservation of the feedback loop between man and mouse.

Figure 2

Analysis of positive feedback mechanisms between GLI1 and GLI2. (a) Real-time PCR analysis of HaCaT cells transfected with human GLI1 (phGLI1) or human GLI2β (phGLI2). Transcripts analysed are shown in brackets. (b) Real-time PCR analysis of mouse C3H10T1/2 fibroblasts transfected with human GLI1 (phGLI1) or human GLI2β (phGLI2) and analysed for the expression of mouse Gli1 (mGli1) and mouse Gli2 (mGli2). (c) C3H10T1/2 cells transfected with human GLI1 (upper left) or human GLI2β (lower left). Dark staining of cells indicates alkaline phosphatase expression due to osteoblast differentiation. Cells transfected with control DNA do not show any staining (upper and lower right image). (d) Quantitative analysis of osteoblast differentiation induced by GLI1 and GLI2β, respectively, expressed as -fold induction of alkaline phosphatase (AP) activity compared to control transfected cells. Results shown in (a), (b) and (d) were calculated from three independent experiments each carried out in duplicate

GLI1 has previously been shown to trigger osteoblast development in C3H10T1/2 cells (Altaba, 1999; Murone et al., 1999). Since GLI2β expression induces the expression of Gli1 in these cells, we asked whether GLI2β is also able to induce osteoblast differentiation in C3H10T1/2 cells. Figure 2c shows that transfection of C3H10T1/2 cells either with GLI1 or with human GLI2β induced the expression of alkaline phosphatase, which specifically marks cells undergoing osteoblast differentiation. Quantitative analysis of alkaline phosphatase induction revealed that GLI2β was more effective than GLI1 in the induction of alkaline phosphatase activity (14.9- versus 2.3-fold activation) (Figure 2d). Since we found that osteoblast differentiation by SHH-Np treatment is specifically linked to activation of Gli1 but not of Gli2 expression (data not shown), the increased activity of alkaline phosphatase in GLI2β transfected cells is probably due to the stronger induction of endogenous mouse Gli1 by GLI2β compared to cells transfected with GLI1 (see Figure 2b). This is also consistent with our observation that GLI2β is a stronger transcriptional activator than GLI1 (see Figure 1d).

GLI1 is a putative direct target of GLI2

Analysis of the mouse Gli1 promoter revealed the presence of putative Gli binding sites which can be bound by the zinc finger domain of human GLI3 protein (Dai et al., 1999), though a biological relevance of this interaction has not yet been shown. Given the high sequence similarity of the zinc finger DNA-binding domain of GLI-proteins and the high conservation of the putative Gli-binding sites in the murine and human GLI1 promoter (G Regl and F Aberger, unpublished) (Liu et al., 1998) we hypothesized that GLI1 expression may be directly regulated by GLI2.

If GLI1 were a direct target gene of GLI2, induction of GLI2 protein expression should be paralleled by a simultaneous increase in GLI1 mRNA levels. We therefore constructed a HaCaT cell line expressing His-tagged GLI2β in a tetracycline-regulated manner. The transcriptional response of GLI1 to GLI2β was measured by real-time PCR. Increased levels of GLI2β protein were first detectable by Western blot analysis after 2 h of tetracycline treatment (data not shown). Immediately following detectable GLI2β protein expression, GLI1 mRNA levels started to rise (1.8-fold after 2 h; 4.8-fold after 2.5 h; 6.2-fold after 3 h; 6.4-fold after 6 h and 9.5-fold after 12 h (Figure 3)). Transcriptional activation of the direct GLI-target PTCH was first detectable after 2.5 h of treatment (1.8-fold induction). Unlike GLI1 and PTCH, levels of endogenous GLI2, as assayed by using primers specific to the 3 UTR of the endogenous GLI2 mRNA, were unchanged at early time points. A clear increase of GLI2 mRNA levels was only observed after 6 h of tetracycline treatment (Figure 3), suggesting that GLI2 activation does not involve direct autoactivation. In contrast, the rapid transcriptional response of GLI1 to GLI2 suggests that GLI1 is likely to be directly regulated by GLI2.

Figure 3

Real-time PCR analysis of tetracycline-regulated GLI2β HaCaT cells. GLI2β (His-tagged) expression was induced by tetracycline treatment for the time indicated. Induction of GLI1, PTCH and endogenous GLI2 (GLI2endo) transcription in response to GLI2 transgene expression was measured by real-time PCR. Data represent results of three independent experiments, each analysed in duplicate PCR runs. Standard deviation was below 20 per cent between all replicate experiments. Large ribosomal protein P0 (RPLP0) was used as a reference standard for all analyses to control for the amount of sample material (Martin et al., 2001)

GLI1 and GLI2 stimulate DNA-synthesis of confluent human keratinocytes

Expression of GLI1 and mouse Gli2 in the epidermis of transgenic mice has been shown to induce BCC development. Furthermore, in murine neural cells stimulation of cell proliferation by Hh-signalling is associated with an increase in Gli1 expression, suggesting that Gli1 mediates the proliferative effect of Hh-signalling (Dahmane et al., 2001; Grachtchouk et al., 2000; Nilsson et al., 2000). To analyse whether human GLI proteins play a role in the control of human keratinocyte proliferation, we expressed either GLI1 or GLI2 in tetracycline-inducible HaCaT cells and determined the incorporation of BrdU as an indicator of cell proliferation. Expression of GLI1 and GLI2, respectively, was induced for 65 h by tetracycline addition and DNA-synthesis was measured either at 60% confluence or at 48 h post confluence. BrdU-positive cells in tetracycline-treated and untreated samples were quantified by flow-cytometry (Figure 4e,f) or analysed by fluorescence microscopy (Figure 4a–d). While at sub-confluence we did not find any significant differences in BrdU-incorporation between GLI1 or GLI2 expressing cells and controls (data not shown), expression of either GLI-gene induced a twofold increase in BrdU-incorporation in post confluent keratinocytes as determined by flow cytometry (Figure 4e,f). Together with the positive feedback mechanism described, these results suggest that the activity of GLI1 and/or GLI2 increases proliferation of epidermal cells, probably by opposing epithelial cell cycle arrest of keratinocytes, an effect that has also been shown to be caused by SHH-expression in confluent primary keratinocytes (Fan and Khavari, 1999).

Figure 4

Analysis of BrdU-incorporation as indicator of cellular proliferation of confluent human keratinocytes in response to GLI1 and GLI2β, respectively. (ad) Fluorescence microscopy analysis of tet-inducible GLI1 (a-b) and GLI2β (cd) HaCaT cell lines either untreated (a, c) or treated with tetracycline (b,d). BrdU-incorporation was detected with anti-BrdU-FITC labelled antibodies. Cells were counter-stained with DAPI. (ef) Flow-cytometry analysis of tet-inducible GLI1 (e) and GLI2β (f) HaCaT cells. DNA-synthesis is expressed as percentage of BrdU positive cells. In total, 10 000 cells were counted in each experiment. Data represent results of three independent experiments each carried out in duplicate. tet: tetracycline; bar: 20 μm


The Hedgehog (Hh) signal transduction pathway plays a pivotal role in a number of vertebrate and invertebrate developmental processes such as cell type specification, pattern formation and regulation of cell proliferation. Inappropriate activation of this pathway can lead to developmental defects as well as to the formation of tumours such as medulloblastomas, rhabdomyosarcomas and BCCs (Goodrich and Scott, 1998; Ruiz i Altaba, 1999; Toftgard, 2000).

Numerous studies in Drosophila and vertebrates have provided convincing evidence that GLI genes act as key mediators of Hh-signalling in the nucleus. Studies in murine systems have provided evidence that GLI1 acts as a major tumour-inducing factor in BCC development (Nilsson et al., 2000), though major differences between murine and human skin tissue may limit the relevance of murine model systems to human neoplasia. To fully understand the role of GLI1 in the development of BCCs, additional functional studies are necessary in a purely human system.

Positive feedback mechanism of GLI1 and GLI2 in human epidermal cells

Using primary human keratinocytes expressing GLI1 as a model system to identify Hh-target genes with relevance to BCC formation, we found that GLI2 expression was highly elevated in response to GLI1. In addition, we showed that GLI2 mRNA is consistently expressed in BCCs at elevated levels compared to normal skin. We found, however, no clear correlation between GLI2 and GLI1 mRNA levels, nor did we find a correlation between GLI1 and the mRNA levels of the direct GLI target PTCH. GLI1 protein has previously been shown to be localized mainly in the cytoplasm of BCC tumour cells (Ghali et al., 1999; Dahmane et al., 1997) raising the possibility that the biological activity of GLI1 is also regulated by mechanisms controlling its subcellular localization. The disparity between GLI1, GLI2 and PTCH mRNA levels in tumours may thus be due to variable amounts of nuclear/cytoplasmic GLI1 protein in the tumour samples analysed.

Induction of GLI2 expression by GLI1 is of particular interest since mouse Gli2 has recently been shown to induce BCC development when expressed under the control of the bovine Keratin 5 promoter (Grachtchouk et al., 2000). The tumorigenic effect of Gli2, however, may be explained by considerable functional redundancy of Gli1 and Gli2. Mice lacking functional Gli2 protein display severe developmental malformations, which can be rescued by knock-in of Gli1 into the Gli2 locus, demonstrating that Gli1 can compensate for the absence of Gli2 protein (Bai and Joyner, 2001). Our results demonstrating the existence of a positive feedback mechanism between GLI1 and GLI2 argue for a role of both transcription factors in mediating the oncogenic effect of activated Hh-signalling in human epidermis. Thus, a possible scenario in the formation of BCC is the activation of GLI1 by inactivating mutations in PTCH followed by the induction of GLI2. Alternatively, activation of Hh-signalling may first lead to the induction of GLI2 expression which then activates GLI1 expression. This hypothesis is consistent with our observation that human GLI2 is a potent activator of GLI1 transcription in both the human keratinocyte cell line HaCaT and in mouse C3H10T1/2 fibroblasts. Irrespective of the sequence of GLI gene activation, our data provide evidence for a positive feedback mechanism between GLI1 and GLI2 in BCC development. This feedback loop may be required for the activation and maintenance of an oncogenic gene expression pattern in human epidermis, which is not initiated in normal skin due to the absence of GLI1 expression. The inability of GLI2 to activate GLI1 in normal skin may be due to the requirement for (an) additional factor(s) present in BCC only. Alternatively, post-translational processing of GLI2 in normal skin may result in the generation of inactive GLI2 activator or even repressor forms, a hypothesis that is supported by expression studies of murine and amphibian Gli2 function in mouse and Drosophila (Aza-Blanc et al., 2000; Sasaki et al., 1999). Once high-affinity antibodies against GLI2 become available, in vivo analysis of GLI2 protein expression will allow to address the role of protein processing in the context of skin tumorigenesis.

GLI1 is a putative direct target gene of GLI2

Analysis of the mouse Gli1 promoter revealed the presence of eight putative Gli-binding sites and it was shown that in vitro GLI3 is able to bind to and stimulate gene expression from these cis-regulatory sequences (Dai et al., 1999). However, the in vivo relevance of this interaction remains to be established. Based on the high conservation of the zinc finger DNA binding domain of GLI proteins, it is also conceivable that GLI2 can bind to the human GLI1 promoter, rendering GLI1 a direct target of GLI2. Using a precisely regulatable human keratinocyte expression system we provided evidence that GLI1 is likely to represent a direct target of GLI2. This is consistent with the presence of highly conserved putative Gli-binding sites in the human GLI1 promoter. In contrast, GLI2 is unlikely to be a direct GLI target, since activation of GLI2 expression by GLI1 as well as GLI2 itself is significantly slower than activation of the direct GLI-target PTCH (Shin et al., 1999). This is further supported by in silico analysis of the putative human GLI2 promoter, which failed to identify any consensus Gli-binding sites (G Regl and F Aberger, unpublished data).

Activation of GLI2 by GLI1 is likely to be context dependent. Mice ectopically expressing Gli1 in the neural tube do not show any alterations of the Gli2 expression pattern (Hynes et al., 1997). Similarily, overexpression of Gli1 does not induce the expression of Gli2 in animal cap explants of Xenopus. In contrast, injection of human GLI1 into frog embryos results in the formation of skin tumours, which express Gli2 mRNA (Mullor et al., 2001). Interestingly, homozygous Gli1 knock-out mice are phenotypically normal, suggesting that Gli2 expression does not strictly depend on functional Gli1 protein. More likely, Gli2 is able to compensate for the lack of active Gli1 protein (Park et al., 2000). Our experiments with mouse C3H10T1/2 cells show that human GLI1 expression leads to an increase in Gli2 mRNA levels but up to now, there is no clear evidence for a role of mouse Gli1 in the activation of Gli2 in mouse. Analysis of Gli2 expression in Gli1-induced BCCs from mouse would be important to clarify the role of mouse Gli1 in the regulation of Gli2.

Activation of GLI1 in response to GLI2 expression is consistent with the finding that Gli2 deficient mice express Gli1 mRNA at reduced levels (Ding et al., 1998). Also, Gli2 expression in mouse epidermis induces the development of BCCs expressing Gli1 (Grachtchouk et al., 2000). These data suggest a conserved function of Gli2 in the regulation of Gli1 expression. This is, however, likely to be a cell-type specific or context dependent effect, since the expression domains of Gli1 and Gli2 are only partially overlapping. Our data also suggest that the function of vertebrate Gli2 has not been completely conserved, since, unlike human GLI2β, Xenopus Gli2 is unable to induce osteoblast differentiation in C3H10T1/2 cells (Altaba, 1999). In addition, the potent activator function of GLI2β in our Gli-reporter assays contrasts with previous results on mouse Gli2, which was shown to be only a weak transcriptional activator, unless the putative N-terminal repressor domain is deleted (Sasaki et al., 1999). No evidence for processing of GLI2β was found by immunological detection of N-terminally His-tagged GLI2β. This suggests that full-length GLI2β–the isoform most closely related to mouse Gli2 (Tanimura et al., 1998)–functions as an activator (unpublished data). The stronger transcriptional activation by human GLI2β is further supported by our transfection experiments using HaCaT or mouse embryonic fibroblasts. In all cases, expression of GLI2β had a more pronounced effect than GLI1 either on target gene activation or induction of osteoblast differentiation as determined by quantitative analysis of alkaline phophatase induction in mouse embryonic fibroblasts. These species-specific differences of GLI-gene activity suggest the use of well defined non-heterologous assay systems for future functional analysis of GLI-genes in a given biological context.

GLI1 and GLI2β have a stimulatory effect on epidermal cell proliferation

Stratified epithelium displays a balance between proliferation and cell cycle arrest. Distortion of such a balance can be associated with tumorigenesis, including BCC development. Activation of Hh-signalling in epidermal cells causes hyperplasia, though the molecular mechanisms of Hh-mediated tumour formation are still poorly defined. Recently, it was shown that expression of SHH in primary keratinocytes can overcome the epithelial growth arrest of cells. The opposing effect on cell cycle arrest by SHH expression was characterized by an increase of cell proliferation of confluent cells treated with calcium to induce differentiation (Fan and Khavari, 1999). Whether the increase of proliferation of SHH-expressing cells is mediated by GLI-genes has not yet been shown. Our results suggest that the opposing effect of Hh-signalling on epithelial growth arrest may result from the activation of the positive feedback mechanism between GLI1 and GLI2, since expression of either transcriptional regulator in human keratinocytes significantly increased the number of cells in S phase in confluent cultures.

Taken together, these findings suggest a model of BCC development where inactivation of PTCH in human epidermis leads to activation of GLI1-GLI2 feedback signalling, which causes or promotes epithelial neoplasia by perturbing the balance of cell proliferation and growth arrest. Phenotypic analysis of cells lacking either GLI1, GLI2 or a combination of both as well as identification and functional analysis of GLI1 and GLI2 target genes in human keratinocytes will be required to reveal the detailed molecular mechanisms underlying tumorigenic conversion of epidermal cells triggered by inappropriate Hh/GLI-signalling.

Materials and methods

Retroviral infection of keratinocytes, cell culture and cell transfections

To generate a retroviral bicistronic GLI1-EGFP expression construct, the pIRES2-EGFP plasmid (Clontech) was modified by cloning an adapter containing a SalI site into the vector AseI site to create pI2E-A. GLI1 coding cDNA was amplified by PCR with the sense primer 5′-gacagagtgtcgacacaccct-3′ and antisense primer 5′-gattccctactcttttaggca-3′ and after digestion with SalI was cloned into pI2E-A at the XhoI/SmaI sites creating pI2E-A-GLI1. The construct was verified by DNA sequencing. The retroviral vector SIN-IP-GFP (gift from P Khavari) was digested with XhoI/NotI to excise the CMV-GFP sequence (SIN-IP). pI2E-A-Gli1 was digested with SalI/NotI to isolate the CMV-Gli1-IRES-EGFP sequence and cloned into SIN-IP creating SIN-Gli1-EGFP. Amphotropic retrovirus was produced as previously described (Deng et al., 1997) except that the more efficient Phoenix packaging cell line was used. Primary human keratinocytes were isolated and infected as previously described (Deng et al., 1997; Rheinwald and Green, 1975) except that cells were plated at a density of 0.5–1.0×106 cells per 10 cm dish in Defined Keratinocyte-SFM (Invitrogen) 16–18 h prior to infection. Cells were harvested at the time points indicated, washed once in low calcium PBS, and the pellets snap frozen in liquid nitrogen.

The T-REx system (Invitrogen) was used to generate double-stable HaCaT lines expressing GLI1 or GLI2β under the control of the tetracycline repressor. Transgene expression was induced by culturing the cells in Dulbecco's Modified Medium (DMEM, high glucose, Life Technologies) containing 10% fetal calf serum (FCS) and 1 mg/l tetracycline (Invitrogen). HaCaT and C3H10T1/2 were grown in HEPES-buffered DMEM (pH 7.2), containing 10% FCS, 100 mg/l streptomycin and 62.5 mg/l penicillin at 37°C. Cells were transfected using SuperFect (Qiagen), according to the manufacturer's instructions.

For transfection studies full-length human GLI1 or GLI2β was cloned into pcDNA3.1zeo(+) expression vector (Invitrogen). The control plasmid (pGLI1mut (control)) was constructed by deleting an N-terminal 1658 bp fragment including the DNA binding domain with HindIII-KpnI followed by blunt-ending and religation. Cells were harvested for RNA isolation 24–48 h post transfection.

Western blot analysis, alkaline phosphatase staining and luciferase reporter assays

Western blot analysis of GLI1 protein expression in tetracyclin-regulated HaCaT cells was done according to standard procedures using a polyclonal goat anti-GLI1 antibody (C-18, Santa Cruz Biotechnology) and a secondary rabbit anti-goat IgG coupled with horse-radish peroxidase (Amersham Pharmacia Biotech) followed by ECL detection (Amersham Pharmacia Biotech).

For alkaline phosphatase staining transfected C3H10T1/2 cells were grown for 24 h in DMEM containing 10% FCS, washed twice in phosphate buffered saline and fixed in methanol/acetone (50%/50%). Alkaline phosphatase activity was visualised by incubating cells in BCIP/NBT (Boehringer Mannheim/Roche). The staining reaction was stopped by rinsing the cells with phosphate buffered saline. For quantitative measurement of alkaline phosphatase induction C3H10T1/2 cells were seeded into 6-well plates at 20% confluence 24 h before transfection. Cells were transfected with GLI-expression constructs together with lacZ expression plasmid for data normalization. Cells were harvested 24 h post transfection in 100 μl of lysis buffer (50 mM Tris-HCl, pH 7.5; 1 mM MgCl2; 0.15% NP-40). Alkaline phosphatase activity of each cleared lysate was analysed in 96-well microtiter plates using an alkaline phosphatase detection kit (Sigma). Absorbance was measured at 405 nm with a Spectra platereader (Tecan). Samples were normalized for beta-galactosidase activity, which was determined by absorbance measurement at 405 nm following a 1 h incubation of lysed samples in staining buffer containing 0.65 mg/ml o-nitrophenyl-β-D-galacto-pyranoside.

GLI1/GLI2 induced luciferase activity was measured with a Lucy I luminometer (Anthos) using a commercial luciferase assay system (Promega). Results were normalized for expression of lacZ, which was co-transfected with the reporter plasmid, GLI- and control constructs. All constructs (GLI1, GLI1mut (control), GLI2β and LacZ) were cloned into pcDNA3.1zeo(+) expression vector (Invitrogen).

RNA isolation and Real-Time PCR analysis

BCC and normal skin were snap frozen after surgical removal and stored at −70°C until used. Total RNA was isolated with TRI-Reagent (Molecular Research Center, Inc.) followed by further purification using the High Pure RNA Isolation Kit (Boehringer Mannheim), which includes a DNase treatment to remove genomic DNA. Total RNA from primary keratinocytes, HaCaT cells and mouse C3H10T1/2 cells was isolated using the High Pure RNA Isolation Kit. RNA quality and quantity were assayed on a Bioanalyzer 2100 (Agilent). cDNA was synthesized with SuperScript II (Rnase H) reverse transcriptase (GibcoBRL) according to the manufacturer's instructions. Real-time PCR analysis was done on a Rotorgene 2000 (Corbett Research) using SYBR Green PCR Mix (PE-Biosystems). Primer sequences used for real-time analysis are shown in Table 1. Induction values (x) were calculated using the following formula: x=2−ΔΔCt, where Ct represents the mean threshold cycle of all replicate analyses of a given gene and ΔCt represents the difference between the Ct values of the gene in question (target) and the Ct value of the reference gene (RPLP0 or Cyclophilin E). ΔΔCt is the difference between the ΔCt values of the samples for each target (e.g. ΔCt for GLI2 in tumour) and the mean ΔCt of the calibrator (e.g. ΔCt for GLI2 in normal skin).

BrdU labelling and flow cytometry

To study indicators of cellular proliferation tetracycline-regulated GLI1 and GLI2 HaCaT cells, respectively, were grown for 65 h either in the presence or absence of 1 μg/ml tetracycline (Invitrogen). Bromodeoxyuridine (BrdU)-incorporation was either determined at 60% confluence or 48 h post confluence using the FLUOS in situ cell proliferation kit (Boehringer Mannheim). Cells were incubated for 90 min in the presence of BrdU before detection with Fluorescein-labelled anti-BrdU antibody. Flow-cytometry was performed on a FACSCalibur (Becton Dickenson) and data were analysed with CellQuest software (Becton Dickenson). Microscopic imaging was done on an Olympus IX 70 microscope equipped with a SPOT CCD-camera (Diagnostic Instruments Inc.).


  1. Altaba AR . 1999 Development 126: 3205–3216

  2. Aszterbaum M, Epstein J, Oro A, Douglas V, LeBoit PE, Scott MP, Epstein Jr EH . 1999 Nat. Med. 5: 1285–1291

  3. Aza-Blanc P, Lin HY, Ruiz i Altaba A, Kornberg TB . 2000 Development 127: 4293–4301

  4. Bai CB, Joyner AL . 2001 Development 128: 5161–5172

  5. Bonifas JM, Pennypacker S, Chuang PT, McMahon AP, Williams M, Rosenthal A, De Sauvage FJ, Epstein Jr EH . 2001 J. Invest. Dermatol. 116: 739–742

  6. Boukamp P, Petrussevska RT, Breitkreutz D, Hornung J, Markham A, Fusenig NE . 1988 J. Cell. Biol. 106: 761–771

  7. Dahmane N, Lee J, Robins P, Heller P, Ruiz i Altaba A . 1997 Nature 389: 876–881

  8. Dahmane N, Sanchez P, Gitton Y, Palma V, Sun T, Beyna M, Weiner H, Ruiz i Altaba A . 2001 Development 128: 5201–5212

  9. Dai P, Akimaru H, Tanaka Y, Maekawa T, Nakafuku M, Ishii S . 1999 J. Biol. Chem. 274: 8143–8152

  10. Deng H, Lin Q, Khavari PA . 1997 Nat. Biotechnol. 15: 1388–1391

  11. Ding Q, Motoyama J, Gasca S, Mo R, Sasaki H, Rossant J, Hui CC . 1998 Development 125: 2533–2543

  12. Fan H, Khavari PA . 1999 J. Cell. Biol. 147: 71–76

  13. Fan H, Oro AE, Scott MP, Khavari PA . 1997 Nat. Med. 3: 788–792

  14. Gailani MR, Stahle-Backdahl M, Leffell DJ, Glynn M, Zaphiropoulos PG, Pressman C, Unden AB, Dean M, Brash DE, Bale AE, Toftgard R . 1996 Nat. Genet. 14: 78–81

  15. Ghali L, Wong ST, Green J, Tidman N, Quinn AG . 1999 J. Invest. Dermatol. 113: 595–599

  16. Goodrich LV, Scott MP . 1998 Neuron 21: 1243–1257

  17. Grachtchouk M, Mo R, Yu S, Zhang X, Sasaki H, Hui CC, Dlugosz AA . 2000 Nat. Genet. 24: 216–217

  18. Green J, Leigh IM, Poulsom R, Quinn AG . 1998 Br. J. Dermatol. 139: 911–915

  19. Hahn H, Christiansen J, Wicking C, Zaphiropoulos PG, Chidambaram A, Gerrard B, Vorechovsky I, Bale AE, Toftgard R, Dean M, Wainwright B . 1996 J. Biol. Chem. 271: 12125–12128

  20. Hahn H, Wojnowski L, Miller G, Zimmer A . 1999 J. Mol. Med. 77: 459–468

  21. Hardcastle Z, Mo R, Hui CC, Sharpe PT . 1998 Development 125: 2803–2811

  22. Hynes M, Stone DM, Dowd M, Pitts-Meek S, Goddard A, Gurney A, Rosenthal A . 1997 Neuron 19: 15–26

  23. i Altaba AR . 1998 Development 125: 2203–2212

  24. Ingham PW . 1998 EMBO J. 17: 3505–3511

  25. Johnson RL, Rothman AL, Xie J, Goodrich LV, Bare JW, Bonifas JM, Quinn AG, Myers RM, Cox DR, Epstein Jr EH, Scott MP . 1996 Science 272: 1668–1671

  26. Kinto N, Iwamoto M, Enomoto-Iwamoto M, Noji S, Ohuchi H, Yoshioka H, Kataoka H, Wada Y, Yuhao G, Takahashi HE, Yoshiki S, Yamaguchi A . 1997 FEBS. Lett. 404: 319–323

  27. Kinzler KW, Bigner SH, Bigner DD, Trent JM, Law ML, O'Brien SJ, Wong AJ, Vogelstein B . 1987 Science 236: 70–73

  28. Liu CZ, Yang JT, Yoon JW, Villavicencio E, Pfendler K, Walterhouse D, Iannaccone P . 1998 Gene 209: 1–11

  29. Martin KJ, Graner E, Li Y, Price LM, Kritzman BM, Fournier MV, Rhei E, Pardee AB . 2001 Proc. Natl. Acad. Sci. USA 98: 2646–2651

  30. Matise MP, Epstein DJ, Park HL, Platt KA, Joyner AL . 1998 Development 125: 2759–2770

  31. Matise MP, Joyner AL . 1999 Oncogene 18: 7852–7859

  32. Mo R, Freer AM, Zinyk DL, Crackower MA, Michaud J, Heng HH, Chik KW, Shi XM, Tsui LC, Cheng SH, Joyner AL, Hui C . 1997 Development 124: 113–123

  33. Motoyama J, Liu J, Mo R, Ding Q, Post M, Hui CC . 1998 Nat. Genet. 20: 54–57

  34. Mullor JL, Dahmane N, Sun T, Ruiz i Altaba A . 2001 Curr. Biol. 11: 769–773

  35. Murone M, Rosenthal A, de Sauvage FJ . 1999 Curr. Biol. 9: 76–84

  36. Nilsson M, Unden AB, Krause D, Malmqwist U, Raza K, Zaphiropoulos PG, Toftgard R . 2000 Proc. Natl. Acad. Sci. USA 97: 3438–3443

  37. Oro AE, Higgins KM, Hu Z, Bonifas JM, Epstein Jr EH, Scott MP . 1997 Science 276: 817–821

  38. Park HL, Bai C, Platt KA, Matise MP, Beeghly A, Hui CC, Nakashima M, Joyner AL . 2000 Development 127: 1593–1605

  39. Rheinwald JG, Green H . 1975 Cell 6: 331–343

  40. Ruiz i Altaba A . 1999 Trends Genet. 15: 418–425

  41. Ruppert JM, Kinzler KW, Wong AJ, Bigner SH, Kao FT, Law ML, Seuanez HN, O'Brien SJ, Vogelstein B . 1988 Mol. Cell. Biol. 8: 3104–3113

  42. Ruppert JM, Vogelstein B, Kinzler KW . 1991 Mol. Cell. Biol. 11: 1724–1728

  43. Sasaki H, Nishizaki Y, Hui C, Nakafuku M, Kondoh H . 1999 Development 126: 3915–3924

  44. Shin SH, Kogerman P, Lindstrom E, Toftgard R, Biesecker LG . 1999 Proc. Natl. Acad. Sci. USA 96: 2880–2884

  45. Tanimura A, Dan S, Yoshida M . 1998 J. Virol. 72: 3958–3964

  46. Toftgard R . 2000 Cell Mol. Life Sci. 57: 1720–1731

  47. Villavicencio EH, Walterhouse DO, Iannaccone PM . 2000 Am. J. Hum. Genet. 67: 1047–1054

  48. Xie J, Murone M, Luoh SM, Ryan A, Gu Q, Zhang C, Bonifas JM, Lam CW, Hynes M, Goddard A, Rosenthal A, Epstein Jr EH, de Sauvage FJ . 1998 Nature 391: 90–92

Download references


We thank Dr Ken Kinzler for the gift of GLI1 plasmid, Dr Paul Khavari for Sin-IP-GFP plasmid, Dr David Markovitz for providing human GLI2β expression plasmid, Dr Gary Nolan for the permission to use the Phoenix packaging cell line and Dr Norbert Fusenig for providing the HaCaT cell line. We are particularly grateful to Dr Harald Esterbauer for his advice and discussions about real-time PCR analysis and to Dr Emberger for histopathological analysis of BCC samples. This work was supported by FWF grant P14227-MOB (Austria), and the Medical Research Council (UK).

Author information



Corresponding author

Correspondence to Fritz Aberger.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Regl, G., Neill, G., Eichberger, T. et al. Human GLI2 and GLI1 are part of a positive feedback mechanism in Basal Cell Carcinoma. Oncogene 21, 5529–5539 (2002).

Download citation


  • Basal Cell Carcinoma
  • Hedgehog signalling
  • GLI genes
  • cell proliferation

Further reading


Quick links