Induction of basal cell carcinomas and trichoepitheliomas in mice overexpressing GLI-1

Transgenic Mice.

A bluescript KS (−) plasmid (p5′BK5II) containing a 5.2-kb fragment from the bovine keratin 5 (K5) promoter, a 0.6-kb fragment containing the rabbit β-globin intron, and a 1.1-kb fragment containing two 3′ poly(A) signal sequences was obtained from Jose Jorcano (CIEMAT, Madrid). A 3.5-kb fragment containing the full-length human GLI-1 cDNA was amplified from pGLI-1 K12 (24) by using GLI-1 specific primers (fwd: 5′-GCGCTAGCATGTTCAACTCGATGACCC-3′, rev: 5′-GCTACGTATTAGGCACTAGAGTTGAGGAA-3′) containing restriction half sites for SnaBI and NheI, respectively. The amplified product was ligated into the SnaBI and NheI sites of the p5′BK5II vector. The construct was verified by sequencing. The transgene was isolated by cutting with NotI/SalI. Founder transgenic mice were made by microinjecting the transgene into the pronuclei of fertilized [C57BL/6J × CBA] F₂ oocytes and were genotyped by PCR amplification of sequences specific for the β-globin intron and the human GLI-1 cDNA respectively from tail genomic DNA. All experiments were performed with mice hemizygous for the transgene, and transgenic mice were always directly compared with nontransgenic siblings. All transgenic mice were generated within the transgenic core facility at Novum/Huddinge hospital.

Immunohistochemical Staining.

Tissues used for immunohistochemistry were formalin fixed and paraffin embedded. Before being incubated in 1% H₂O₂ in methanol for 30 min, the paraffin sections were dewaxed, rehydrated, and submerged in 10 mM citrate buffer while heated in a 97°C water bath, for 20 min. Sections were incubated overnight at +4°C with 1:500 diluted rabbit polyclonal antibodies directed against either mouse K1, K5, K6, or loricrin (all from Berkeley Antibody Company, Berkeley, CA) or 1:200 diluted goat polyclonal antibodies against Gli-1 or Ptch 1 (both from Santa Cruz Biotechnology) or 1:500 diluted rabbit polyclonal antibodies against phosphohistone H3 (Upstate Biotechnology, Lake Placid, NY) or 1:200 diluted mouse monoclonal antibodies against Bcl-2 (Transduction Laboratories, Lexington, KY) or 1:700 diluted rabbit polyclonal antibodies against wild-type p53 (NovoCastra, Newcastle, U.K.). All antibodies were diluted in PBS containing 0.1% BSA. Detection was carried out with the Vectastain Elite Kit (Vector Laboratories) by using rabbit or goat IgG. Sections were washed with PBS before biotinylated secondary antibodies were applied for 1 h at room temperature, followed by extensive rinses and incubation with avidin–biotin immunoperoxidase. Immunohistochemical staining was visualized with 3-amino-9-ethylcarbazole (AEC, Vector Laboratories), and counterstaining was performed with Mayer's hematoxylin.

RNA Isolation and Reverse Transcription (RT)–PCR.

Total cellular RNA was isolated from the back skin, containing dermis as well as epidermis, of transgenic and wild-type sibling control mice by using guanidium thiocyanate-phenol-chloroform extraction as described previously (25). RT-PCR was carried out by using the Access RT-PCR system kit (Promega). The RT reaction was performed at 48°C for 45 min by using 1 μg of total RNA. The following primer pairs were used: 5′-TGCTCTTTCTCTCCAGCACCTCGGATC-3′ (sense: corresponding to the vector sequence) and 5′-CAGTGCCCGCTTCTTGGTCAA-3′ (antisense: corresponding to the human GLI-1 cDNA sequence from 301–327). These transgene-specific GLI-1 primers yielded a 369-bp PCR product. In the same reaction, β-actin-specific primers giving a 553-bp PCR product were used as an internal control.

Analysis of p53 and Ha ras Mutations.

Microdissection of 10-μm-thick paraffin sections was performed, and normal cells were separated from tumor cells by using a scalpel. Microdissected tissue was incubated in digestion buffer (10 mM KCl/1.5 mM MgCl₂/10 mM Tris, pH 9.0/0.5% Tween 20) containing 200 μg/ml proteinase K at 55°C for 6 h. All microdissected tissue contained >80% tumor cells. Amplification of exons 4–8 of the mouse p53 gene or exons 1 and 2 of the mouse Ha ras gene was performed as follows: 10 μl aliquots of tumor lysate were used for PCR amplification in a 20-μl reaction volume containing 2 μl 10 × reaction buffer without MgCl₂ (Promega); 250 μM of each dNTP; 3.2 pmol of each primer; and 0.5 units Taq polymerase (Promega). The primers used for amplification of the murine p53 or Ha ras exons and the individual MgCl₂ concentrations have previously been described (26, 27). Thirty-five cycles of amplification were carried out, with denaturation at 94°C for 1 min, annealing at 54 or 58°C (p53, depending on the exon), or 60°C (Ha ras) for 2 min, and extension at 72°C for 2 min. PCR products were purified by using MicroSpin columns (Amersham Pharmacia Biotech) and subjected to sequencing.

Spontaneous Tumor Formation in K5 GLI-1 Transgenic Mice.

To examine the effect of deregulated GLI-1 expression in mammalian skin, we generated transgenic mice expressing the human GLI-1 gene under the control of the bovine K5 promoter (Fig. 1a). This K5 promoter fragment has previously been shown to direct transgene expression to the basal cell compartment of stratified squamous epithelia as well as the outer root sheet of hair follicles (28). Three independent K5 GLI-1 transgenic founders were generated, all of which developed tumors within 1–13 wk. In addition to tumor formation, the GLI-1 transgenic mice exhibited failure to thrive and died prematurely at 1 to 6 mo of age. Founder 1 developed a hyperkeratotic tumor on the back within the first week and died at 4 wk of age. Founder 2 showed the first signs of tumor development at 4 wk of age and developed a 1-cm tumor-ulcer at 10 wk of age, and subsequently tumors developed on the back (Fig. 1b). The third founder showed the weakest phenotype, because 13 wk passed before tumors developed on the back. RT-PCR analysis of total RNA from dorsal back skin showed K5 GLI-1 mRNA in all three founders, demonstrating expression of the transgene in these mice (Fig. 1c). A summary of the macroscopically detected tumors on each founder is described in Table 1.

Four Different Types of Adnexal Skin Tumors in GLI-1 Overexpressing Mice.

Histologic examination of the observed nodules on founder 2 identified two different skin tumors separated by nontumorigenic skin. The first nodule showed a large primitive tumor (Fig. 2a). The peripheral layer of these tumor formations showed palisading similar to human BCCs, and the cells had a uniform nonanaplastic appearance (Fig. 2 a and b). The dermis was completely filled with tumor masses also lined with palisading basaloid cells but with large necrotic areas in the center of the tumor nests. Inside the same tumor islands, abundant amyloid was identified, indicating the presence of phagocytosis and apoptosis of tumor cells. Colloid bodies and scattered calcinosis were also observed in this tumor. The stroma was arranged in parallel bundles around the tumor masses with a retraction zone resulting in peritumoral lacuna also seen in human BCCs (Fig. 2b). Such retraction zones were observed in some but not all tumor nests. In addition to the nodular BCC observed on the back, several BCCs were found on the nose, including a superficial BCC showing strong resemblance to its human counterpart (Fig. 2 c and d).

A second type of tumor observed was macroscopically similar to the BCC-like tumor described above but showed different features under the microscope. This tumor consisted of numerous tumor islands of epithelial cells appearing like pieces in a jigsaw puzzle (Fig. 2e). The thick hyaline sheath that surrounded the tumor islands like a cylinder was periodic acid/Schiff reagent positive (Fig. 2f). All the above characteristics are typical for a cylindroma, and this tumor was consequently classified as a cutaneous cylindroma (29).

A third type of tumor developed on both founder 1 and 2, consisting of large well-circumscribed tumors located in the dermis reaching down to the muscle layer without direct connection with the epidermis. The tumors consisted of multiple tumor nests of epithelial origin all showing horn cysts in the center (Fig. 2g). The centers were fully keratinized and surrounded by basophilic tumor cells, similar to the hair matrix cells. These tumors are very similar to human TE, a benign human skin tumor believed to develop from pluripotential epithelial cells in the hair follicle (29).

Six different tumors were examined from founder 3. Histologically, all six tumors were shown to be benign well-circumscribed noduli with a silhouette placed deep in the dermis (Fig. 2h). No connection with epidermis was observed, even though the overlying epidermis seemed thickened and hyperplastic. The numerous aggregations consisted mostly of follicular germinative cells arranged in a palisade pattern in the periphery. No keratinizing cysts were observed. These tumors were classified as counterparts to the benign human skin tumor, trichoblastoma (29).

In summary, four different types of skin tumors were found on the K5 GLI-1 mice: BCCs, TE, cylindromas, and trichoblastomas. The tumors described above were macroscopically detected; however, more tumors of similar types were detected when examining additional (n = 5) skin samples from these mice (data not shown).

Tumors Show Differentiation Marker Expression Similar to Human BCCs.

Terminal differentiation of keratinocytes involves the sequential expression of specific keratins in the basal and spinous layers and loricrin, a component of the cornified envelope, in the granular layer. To determine whether the histologic features seen in the K5 GLI-1 mice are accompanied by an altered pattern of differentiation marker expression, immunohistological analysis was performed. Staining of K5 GLI-1 BCC with antibodies against the basal K5 revealed high expression in the tumor, consistent with the pattern seen in human BCC (Fig. 3a) (30). K1, which is expressed only in cells committed to terminal differentiation, was found to be abundantly expressed in the suprabasal layer of the transgenic skin but not in tumor cells (Fig. 3b). In addition, loricrin, a major component of the epidermal cornified envelope, showed normal expression in the transgenic stratified epithelia (Fig. 3c). Taken together, these findings suggest that the interfollicular epidermal keratinocytes present in the K5 GLI-1 transgenic mice differentiate normally. Analysis of K6 and histone 3, both associated with hyperproliferation, showed abundant expression in tumor cells and in overlaying hyperproliferative epidermis but not in histologically normal epidermis (Fig. 3 d and e). This observation indicates that, whereas tumor cells and reactive epidermal keratinocytes are actively proliferating, uninvolved epidermal cells proliferate at normal rate.

Increased Expression of GLI-1 Target Genes in Tumor Cells.

Human BCCs and TEs consistently overexpress PTCH1 and GLI-1 (5, 17, 18, 31). Furthermore, earlier studies in frogs and mice have shown that ectopic expression of GLI-1 causes increased expression of endogenous Gli-1 and Ptch1 (15, 16, 18). We used immunohistochemical analysis to examine the expression of these genes in transgenic tumors as well as in wild-type skin. Gli-1 was strongly expressed in tumor cells, and expression was also observed in adjacent reactive epidermis, although at a lower level (Fig. 4a). Even though the antibody used does not distinguish between the human and murine Gli-1, we interpret the observed difference in signal as reflecting increased expression of the endogenous murine gene because no positive signal was observed in histologically normal epidermis. No Gli-1 expression could be detected in wild-type skin (Fig. 4b). Consistent with the Gli-1 expression, increased expression of murine Ptch1 was observed in transgenic skin compared with wild type (Fig. 4 c and d).

Human BCCs have demonstrated a marked increase in expression of Bcl-2, in contrast to the other common epidermal malignancy, squamous cell carcinoma (30). In the BCC from a K5 GLI-1 transgenic mouse Bcl-2 expression was markedly elevated with no detectable expression in wild-type skin (Fig. 4 e and f), suggesting a potential role in GLI-1-induced BCC formation.

Wild-Type p53 and Ha ras in Transgenic Skin Tumors.

The tumor-suppressor gene p53 is commonly mutated in most types of human epithelial malignancies, including BCCs (6, 32). The level of p53 protein is low in normal cells; however, increased expression because of stabilization is observed in cells harboring a mutated p53. When analyzing a tumor sample from a K5 GLI-1 mouse for p53 protein expression, a moderately intense dispersed pattern was observed primarily in peripheral cells, suggestive of a possible involvement of p53 in the formation of this tumor (Fig. 4g). Similar results were obtained when analyzing additional tumors (n = 3; data not shown). However, direct sequencing of p53 exons 4–8 after microdissection of this particular BCC did not reveal any mutations (Table 1). Moreover, when the p53 mutation analysis was extended to involve all macroscopically observed tumors (n = 10), no p53 mutation was found in any tumor. This finding suggests that GLI-1 can drive tumor formation independently of a mutated p53 protein.

Other genes commonly mutated in nonmelanoma skin cancer are members of the ras oncogene family, comprised of the N, Ki, and Ha ras genes. Mutations in all three ras genes have been found in BCCs, with Ha ras showing the highest mutation frequency (33). In addition, in patients with nonmelanoma skin cancer, there is a high incidence of Ha ras loss of heterozygosity (36% for patients with BCC) in normal-appearing skin (34). These findings suggest a potential involvement of the Ras signaling pathway in BCC development. However, examination of the transgenic mouse tumors (n = 10) for Ha ras mutations revealed no mutations in exons 1 and 2, including the hot spot regions codons 12, 13, and 61, suggesting a mechanism whereby GLI-1 induce tumor growth independently of a mutated Ha ras gene (Table 1). However, we cannot exclude presence of mutations in other ras genes, and further analysis is ongoing to address this issue.