Rapamycin Prevents Early Onset of Tumorigenesis in an Oral-Specific K-ras and p53 Two-Hit Carcinogenesis Model

Head and neck squamous cell carcinomas (HNSCC), the majority of which occur in the oral cavity, remain a significant cause of morbidity and mortality worldwide. A major limitation in HNSCC research has been the paucity of animal models to test the validity of current genetic paradigms of tumorigenesis and to explore the effectiveness of new treatment modalities and chemopreventive strategies. Here, we have developed an inducible oral-specific animal tumor model system, which consists in the expression of a tamoxifen-inducible Cre recombinase (CreER^tam) under the control of the cytokeratin 14 (K14) promoter (K14-CreER^tam) and mice in which the endogenous K-ras locus is targeted (LSL-K-ras^G12D), thereby causing the expression of endogenous levels of oncogenic K-ras^G12D following removal of a stop element. Surprisingly, whereas K14-CreER^tam can also target the skin, K14-CreER^tam/LSL-K-ras^G12D mice developed papillomas exclusively in the oral mucosa within 1 month after tamoxifen treatment. These lesions were highly proliferative but never progressed to carcinoma. However, when crossed with p53 conditional knockout (p53^flox/flox) mice, mice developed SCCs exclusively on the tongue as early as 2 weeks after tamoxifen induction, concomitant with a remarkable activation of the mammalian target of rapamycin (mTOR) signaling pathway. The availability of this ras and p53 two-hit animal model system recapitulating HNSCC progression may provide a suitable platform for exploring novel molecular targeted approaches for the treatment of this devastating disease. Indeed, we show here that mTOR inhibition by the use of rapamycin is sufficient to halt tumor progression in this genetically defined oral cancer model system, thereby prolonging animal survival. [Cancer Res 2009;69(10):4159–66]

Head and neck squamous cell carcinomas (HNSCC), the vast majority of which occur in the oral cavity, remain a significant cause of morbidity and mortality worldwide (1), resulting in more than 10,000 deaths each year in the United States alone (2). Like most cancers, HNSCC progression involves the sequential acquisition of genetic and epigenetic alterations in tumor suppressor genes and oncogenes (reviewed in ref. 3). The most frequent alterations include loss of heterozygosity and promoter silencing of the p16 and inactivating mutations in the p53 tumor suppressor genes (3). HNSCC often overexpress the epidermal growth factor receptor and some of its active variants (4) and harbor activating mutations in the ras oncogene with a rate ranging from 3% to 5% in Western countries and up to 50% in India and Southeast Asia (5–7).

Advances in the understanding of the molecular mechanisms involved in HNSCC have been hampered by the limited availability of appropriate animal models for oral malignancies. In this regard, genetically engineered mouse models are powerful tools to investigate specific cause and effect relationships between molecular changes and cancer development and progression (8). However, the use of genetically defined approaches for HNSCC has been limited due to the lack of specific systems targeting the oral stratified epithelium. For example, expression of the K-ras oncogene in the oral epithelium under the control of a cytokeratin 5 (K5) promoter-driven tet-on inducible system leads to SCC in the oral mucosa and the salivary glands, skin, and other squamous epithelia (9, 10). A K5 promoter-driven Cre recombinase system that is activated by the local administration of a synthetic progesterone derivative has been used to express a mutant K-ras allele in the oral epithelium of mice, leading to the development of oral benign papillomas (11). These mice develop massive oral carcinomas when crossed with animals lacking transforming growth factor receptor II (TGFβRII); therefore, they are no longer responsive to the growth-suppressive activity of TGFβ (7). In these conditional TGFβRII^−/− mice (7), progression to SCC was achieved by the use of chemical carcinogens, likely promoting ras mutations. Thus, we have explored additional approaches to achieve the development of malignant lesions in a genetically defined cancer cell autonomous-driven model, which may therefore facilitate recapitulating oral tumorigenesis.

We show here that mice expressing a tamoxifen-inducible Cre recombinase under the control of the cytokeratin 14 (K14) promoter (K14-CreER^tam) and the ras oncogene from its own promoter after Cre excision of a stop signal (LSL-K-ras^G12D mice) develop large papillomas exclusively in the oral cavity and hyperplasia in the tongue within 1 month of a tamoxifen treatment. Furthermore, if these mice are crossed with floxed p53 conditional knockout (p53^flox/flox) mice, the compound mice develop carcinomas exclusively on the tongue as early as 2 weeks after tamoxifen induction. The use of this K14-CreER^tam/LSL-K-ras^G12D/+/p53^flox/flox animal model may help recapitulate oral cancer progression in a genetically defined two-hit system involving ras activation and p53 inactivation, thus facilitating the study of the molecular mechanisms underlying HNSCC progression. In this regard, <1?show [nonbspace]>we show that the activation of the mammalian target of rapamycin (mTOR) signaling pathway is an early event in both oral benign <1?show [nonbspace]>and malignant lesions and that targeting mTOR by the use of rapamycin halts tumor progression in this genetically defined oral cancer model, thereby prolonging animal survival.

Generation of mice with K-ras^G12D activation and p53 deletion. Mouse strains, K14-CreER^tam (The Jackson Laboratory), LSL-K-ras^G12D, and p53^flox/flox, have been described (12–14). Their genetic backgrounds were as follows: K14-CreER^tam and p53^flox/flox, FVB/N; LSL-K-ras^G12D, 129Sv. K14-CreER^tam mice were crossed with LSL-K-ras^G12D/+ mice to generate K14-CreER^tam/LSL-K-ras^G12D/+ mice and subsequently bred with p53^flox/flox mice to generate K14-CreER^tam/LSL-K-ras^G12D/p53^flox/+ mice. These mice were further crossed with p53^flox/flox to generate K14-CreER^tam/LSL-K-ras^G12D/p53^flox/flox line. K14-CreER^tam mice were also crossed to Rosa26-LacZ homozygous reporter mice to generate K14-CreER^tam/Rosa26-LacZ mice. K14-CreER^tam mice were used as heterozygous in all the lines established. Tamoxifen was administered to 1-mo-old animals, 1 mg/mouse/d orally, for 5 consecutive days. Control mice received only the tamoxifen solvent (sunflower oil) with the same schedule. Genotyping was performed on tail biopsies by PCR using specific primers (12–14). Excision of the stop cassette from the LSL-K-ras^G12D allele was determined in DNA purified from tissues of wild-type and the indicated compound mice treated with or without tamoxifen using the following primers: 5′-GGGTAGGTGTTGGGATAGCTG-′1 and 3′-TCCGAATTCAGTGACTACAGATGTACAGAG-3′ (sequence obtained from Dr. Tyler Jacks laboratory, Koch Institute at MIT, Cambridge, MA). Cre-mediated recombination of the p53 floxed allele was confirmed as previously described (14). All experiments involving mice were carried out according to NIH-approved protocols, in compliance with the Guide for the Care and Use of Laboratory Animals.

Rapamycin administration. Rapamycin (LC Laboratories) was diluted in aqueous solution of 5.2% Tween 80 (Sigma) and 5.2% polyethyleneglycol (PEG-400; Hampton Research), as previously described (15), and injected i.p. at a final dose of 10 mg/kg every other day.

Histology, histochemistry, immunohistochemistry, and Western blotting. One hour before euthanasia, mice were injected i.p. with 5′-bromo-2′-deoxyuridine (BrdUrd; 100 μg/g body weight) for cell proliferation assays. All tumor lesions and control tissues were dissected, fixed overnight in buffered 4% paraformaldehyde at room temperature, dehydrated, and embedded in paraffin. H&E-stained sections were used for diagnostic purposes, and unstained serial sections for immunohistochemical studies. Immunohistochemistry was performed as previously described (10). Western blots were performed on lysates from frozen tissues. Protein concentration was determined, and 50 μg of total proteins were separated on SDS-PAGE, transferred to nitrocellulose membranes, and blocked with 5% dry milk. Antibodies used include cyclin D1 (92G2), phospho-S6 ribosomal protein Ser^240/244 (pS6), and phospho-p44/42 mitogen-activated protein kinase (MAPK) Thr²⁰²/Tyr²⁰⁴ (pMAPK 9101) from Cell Signaling Technology; and cytokeratin 1 (MK1; Covance); antibodies to proliferating cell nuclear antigen (PCNA; Zymed), BrdUrd (Biogenex), tubulin, MAPK (extracellular signal-regulated kinase 1/2), and DUSP-14/MKP-6 (Santa Cruz Biotechnology) were also used. The antibodies for K14 and filaggrin were kindly provided by Dr. J. Segre (National Human Genome Research Institute, NIH, Bethesda, MD). β-Galactosidase histochemistry was performed on cryostat section from snap-frozen OCT-embedded tissue samples (10). Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) assay was performed using the In Situ Cell Death Detection kit, AP (Roche) according to the manufacturer's instructions. The evaluation of the immunohistochemistry was conducted blindly, without knowledge of the origin and genotype. Tissues were classified based on the staining intensity (0, no staining; 1, weak staining; 2, moderate staining; and 3, strong staining) and the percentage of positive cells (0, between 0% and 25% of stained cells; 1, between 25% and 50%; 2, between 50% and 75%; 3, between 75% and 100% of cells stained). Results were scored by multiplying the percentage of positive cells by the intensity, as previously described (16).

Hierarchical clustering and data visualization. The staining scores that resulted from immunohistochemistry were converted into scaled values centered on zero, and hierarchical analysis was performed using Cluster program¹

with average linkage based on Pearson's correlation coefficient as the selection parameter on the unsupervised approach. The results were visualized using the Java TreeView software. The clustered data were arranged with markers on the horizontal axis and tissue samples on the vertical axis as recently described (16). Two biomarkers with a close relationship are located next to each other.

Statistical analysis. Kaplan-Meier survival curves and its statistical analysis, as well as log-rank test followed by a post hoc comparison (Holm-Sidak test), were performed using the SigmaStat software package.

Development of tumoral oral lesions on tamoxifen-induced ras activation in K14-CreER^tam/LSL-K-ras^G12D/+ mice. We explored the consequences of targeting ras to the basal layer of the oral mucosa and tongue by examining whether mice expressing a tamoxifen-inducible form of Cre recombinase under the control of the K14 promoter (17) could induce Cre/LoxP-mediated recombination when crossed with the Rosa26 reporter animal line. The administration of tamoxifen to K14-CreER^tam/Rosa26-LacZ mice resulted in the expression of β-galactosidase in epithelial cells within the interfollicular skin, hair follicle, and bulge region of the hair follicle (Fig. 1), as previously reported (12). Of interest, these mice also expressed β-galactosidase in scattered cell populations in the epithelium lining the tongue (12) and oral mucosa (Fig. 1) but not when treated with the vehicle (Fig. 1). Thus, this system enabled the inducible Cre-mediated recombination of sporadic cells within the oral cavity, likely involving the epithelial stem cell population, as groups or columns of β-galactosidase–positive cells initiated at the basal layer were visible beyond 1 month after tamoxifen treatment in both the tongue and oral mucosa, in spite of the rapid turnover of the oral epithelial cells (Fig. 1).

Members of the ras oncogene family (H-ras, K-ras, and N-ras) are often mutated in human cancer (18, 19). A high incidence of ras mutations occurs in Southeast Asia, where it is associated with areca nut chewing, but much less frequently mutated in the Western countries (5, 20, 21). The use of the LSL-K-ras^G12D strain that enables achieving endogenous levels of oncogenic K-Ras^G12D protein on the removal of the stop element (13) by the K14-CreER^tam–mediated recombination led to the developed large exophytic tumors exclusively in the oral mucosa in all mice studied as early as 1 month after the tamoxifen induction (n = 18; Fig. 2). In contrast, a group of K14-CreER^tam and LSL-K-ras^G12D/+ mice treated with tamoxifen and control K14-CreER^tam/LSL-K-ras^G12D/+ animals treated with solvent alone did not develop any lesions or exhibit any differences in survival rate (Fig. 3 and data not shown). Tumors arising in tamoxifen-treated K14-CreER^tam/LSL-K-ras^G12D/+ mice were hyperkeratotic squamous cell papillomas (Fig. 2). Increased cell proliferation, restricted to the basal layers (Fig. 2), was coupled with cell differentiation, as judged by the positive immunostaining for filaggrin (Fig. 2). We observed, however, changes in the pattern of expression of cytokeratins. For example, K14 was expressed poorly in the basal layer but strongly in the upper layers of the papillomas, different from its distribution in normal epithelium (Fig. 2) but similar to that observed in chemically induced skin papillomas (22). On the other hand, K1 was expressed in the upper layers of the hyperplastic epithelium but not in the control normal oral mucosa (Fig. 2). The majority of the tongues showed a marked hyperplasia with significant acanthosis, papillomatosis, and hyperkeratosis (Fig. 2). No animals developed carcinomas, but they had to be sacrificed 2 to 3 months after tamoxifen administration, as these lesions exceeded acceptable size and thus compromised food intake (Fig. 3).

We confirmed that the treatment with tamoxifen leads to the CreER^tam-dependent recombination of the LSL-K-ras^G12D allele by PCR analysis using oligonucleotides flanking the stop cassette (Fig. 2 and Supplementary Fig. S1). No recombination was observed in the absence of CreER^tam expression or tamoxifen induction, and LSL-K-ras^G12D allele recombination was readily demonstrable in K14-CreER^tam/LSL-K-ras^G12D/+ mice in the oral mucosa, tongue, and the papillomas arising in tamoxifen-treated mice (Fig. 2). LSL-K-ras^G12D allele recombination effectively occurred in the skin of K14-CreER^tam/LSL-K-ras^G12D/+ mice treated with tamoxifen (Supplementary Fig. S1), although these mice did not develop skin lesions (Supplementary Fig. S1). This is aligned with recent studies using a different keratinocyte-specific inducible Cre system in which mice did not develop skin malignancies unless persistently treated with phorbol esters as tumor promoter (23). Thus, in spite of ras being likely activated in many K14-expressing squamous epithelia in our animal system, these animals developed benign tumors only in the oral mucosa, suggesting that the oral epithelial cells might be particularly sensitive to ras-induced aberrant cell proliferation.

Excision of p53 cooperates with ras to induce tongue carcinomas. p53 is one of the most frequently mutated tumor suppressor gene in human malignancies (24). In HNSCC, mutations in p53 occur in >50% of the oral cavity cancers and the presence of mutations that render p53 functionally inactive is associated with tumor progression and decreased overall survival (25). Loss of heterozygosity of p53 and inactivating mutations in its coding sequence or the accelerated destruction of its protein product, p53, by viral oncoproteins, such as by human papillomavirus E6, represents a common molecular alteration in HNSCC (3). Thus, to examine whether alterations in p53 cooperate with ras during oral carcinogenesis, we bred the K14-CreER^tam/LSL-K-ras^G12D/+ mice with mice harboring a floxed allele of p53 (14). Remarkably, these mice developed carcinomas on the tongue as early as 2 weeks after tamoxifen administration (Fig. 3A) and progressed into clearly visible carcinomas in only 3 weeks (Fig. 3) The concomitant activation of LSL-K-ras^G12D and deletion of the floxed p53 alleles were confirmed by PCR analysis using DNA from these tumors (Supplementary Fig. S1).

The tumors were well-differentiated SCCs, with moderate cell atypia and focal dyskeratosis (Fig. 3). Dysplasias and microcarcinomas were also observed. Tongue carcinomas presented increased cell proliferation, with BrdUrd-positive cells distributed at all levels of the epithelium (Fig. 3). Unlike in the normal squamous epithelium, K14 expression was stronger in the upper layers in the carcinomatous epithelium (Fig. 3), similarly to that observed in ras-induced oral papillomas. K1 and filaggrin were expressed only in the upper layers of the malignant epithelium (Fig. 3). Interestingly, we observed a distinct susceptibility of tumor progression between the tongue and the oral mucosa. In contrast to the large SCC lesions in the tongue, the oral mucosa of the K14-CreER^tam/LSL-K-ras^G12D/+/p53^flox/flox mice developed only squamous papillomas, with no signs of malignant transformation. Thus, the squamous epithelium of the tongue seems to be highly susceptible to ras-induced transformation but only when relieved from the tumor-suppressive activity of p53.

Molecular mechanisms contributing to ras-induced HNSCC progression. As expected, the carcinomas exhibited an elevated proliferative capacity, as judged by the increase in CCDN1 and PCNA staining (Fig. 4). The normal oral mucosa and tongue and their tumoral lesions all expressed K14, albeit with a distinct distribution pattern (see above). All papillomas expressed high levels of K1. This suggested that in all benign lesions, cell differentiation might restrain the oncogenic potential of ras. In contrast, in the absence of a functional p53, epithelial cells from the tongue may lose their ability to couple proliferation with differentiation, thereby resulting in malignant growth.

We next focused on the status of activation of the MAPK and Akt/mTOR pathway, two key signaling pathways activated by Ras (26, 27), by examining the levels of pMAPK and pS6, the most downstream target of the Akt/mTOR signaling route. We did not observe an increase in the level of pMAPK in ras-induced tumors, which is aligned with the observation that MAPK activation is not often observed in primary human HNSCC (28). This was confirmed by Western blot analysis of lysates from papillomas and carcinomas (Supplementary Fig. S2). By systematic analysis of potential events leading to MAPK inactivation in these cancer lesions, we found that ras-induced tumors overexpress a MAPK phosphatase, DUSP-14 (refs. 29, 30; Supplementary Fig. S2). Although the potential role of DUSP-14 in preventing MAPK activation is under current evaluation, the most remarkable finding in these tumors was a clear increase in the level of pS6 in carcinomas and oral papillomas, which contrasts with the low level of pS6 in the normal tissues (Fig. 4).

The relationship between proliferative and differentiation markers and the activation of MAPK and the Akt/mTOR pathway was nicely reflected by the nonsupervised hierarchical class comparison of the immunostaining results (16), showing the existence of three well-defined clusters (Fig. 4). Control tissues grouped together, whereas hyperplasias and papillomas formed a separate group regardless of their genotype. All carcinomas defined a clearly distinct group, characterized by the increased expression of proliferative markers, PCNA and CCDN1, and basal keratins, K14, a reduced expression of suprabasal keratins, K1, and a remarkable accumulation of pS6. Indeed, the activation of the mTOR pathway correlated with the malignant conversion from papilloma to carcinoma, thus suggesting that the mTOR signaling route may contribute to carcinoma progression caused by the concomitant activation of ras and the inactivation of p53 in the oral cavity.

Targeting mTOR halts the progression of K-ras–induced and p53-induced oral tumors. These observations, prior studies supporting that the activation of the Akt-mTOR pathway is a frequent event in HNSCC (15, 31–33), and the emerging use of mTOR inhibitors in cancer treatment (34) prompted us to explore whether interfering with mTOR signaling could prevent tumor development in this model. The treatment of mice harboring ras-induced tumors with rapamycin, a potent inhibitor of mTOR function (35), led to a dramatic decrease in pS6 levels in the tumoral lesions, as judged by immunohistochemical and Western blot analysis (Fig. 5). We next explored whether the development of these lesions was dependent on mTOR activation. For these studies, we induced Cre-mediated ras activation in animals harboring wild-type p53 or its floxed allele by tamoxifen administration and, 1 week later, initiated the rapamycin treatment. The treatment of K14-CreER^tam/LSL-K-ras^G12D/+ mice with rapamycin prevented the development of oral tumors and increased their life span (P < 0.0001; Fig. 5). This was also clearly observed in mice in which ras activation and p53 deletion led to squamous carcinomas. Indeed, rapamycin treatment prevented tumor progression and extended the life expectancy of these mice (Fig. 5).

The combined inactivation of p53 and activation of ras in a cellular compartment that includes the oral epithelial stem cells by the administration of tamoxifen to K14-CreER^tam/LSL-K-ras^G12D/+/p53^flox/flox mice may provide a relevant two-hit animal model system resulting in the rapid development of oral squamous carcinomas in the tongue. This finding is consistent with the emerging notion that oral carcinomas arise from genetic alterations in the epithelial stem cell compartment (36). K14-CreER^tam also targets gene expression to self-renewing cells in other epithelial tissues, including the hair follicles and the interfollicular epithelium in the skin (12). However, activation of a single allele of ras and alterations in p53 were unable to promote tumorigenesis in the skin, but sufficient for oral carcinogenesis, suggesting that different epithelial stem cells may harbor distinct transforming potential. There were even differences between the oral mucosa and the epithelium of the tongue. Whereas ras activation with or without p53 inactivation caused the development of papillomas in the oral mucosa, ras alone caused only hyperproliferation of the epithelium of the tongue. However, ras activation and p53 deletion resulted in the rapid development of squamous carcinomas exclusively in the tongue. This increased susceptibility to malignant conversion of the squamous epithelium of the tongue may explain the higher incidence of tongue HNSCC with respect to all other anatomic locations in the head and neck region (37) even if the exposure to carcinogens, such as those in tobacco, results in p53 mutations in the epithelium lining the entire oropharynx (3).

This genetically defined HNSCC model system that does not require the use of tumor promoters affecting the stroma and connective tissue may now enable exploring the nature of the tumor cell autonomous molecular events driving SCC progression. Among them, the ability to uncouple epithelial cell proliferation with differentiation and the activation of the Akt-mTOR signaling pathway resulting in pS6 accumulation seem to represent early events in HNSCC progression (15, 16, 31–33). On the other hand, this model system enabled us to explore whether interfering with the activity of mTOR may represent a suitable approach to prevent or treat oral cancer. Indeed, we observed that decreasing mTOR activity by the administration of rapamycin after genetic recombination leading to p53 deletion and ras activation was sufficient to prevent tumor progression. These findings, together with the recent observation that rapamycin causes the regression of SCC lesions caused by the classic two-step skin chemical carcinogenesis model (38) resulting from the accumulation of ras and p53 mutations (38, 39), support the potential clinical benefit of targeting mTOR for the prevention and treatment of HNSCC, particularly in clinical cases that present premalignant and cancerous lesions harboring activated mutations in ras. Ultimately, the K14-CreER^tam/LSL-K-ras^G12D/+/p53^flox/flox two-hit animal model system may represent a suitable platform for exploring the underlying molecular mechanism and genetic and epigenetic event determining the susceptibility to malignant progression of tumoral lesions arising from the distinct stratified epithelia of the oral cavity and may facilitate the development of new targeted approaches for the treatment and prevention of HNSCC.

No potential conflicts of interest were disclosed.

Grant support: Intramural Research Program of NIH, National Institute of Dental and Craniofacial Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. Maria L. Paparella and Kantima Leelahavanichkul for valuable advice and help.