Mammalian Target of Rapamycin–Dependent Acinar Cell Neoplasia after Inactivation of Apc and Pten in the Mouse Salivary Gland: Implications for Human Acinic Cell Carcinoma

Malignant tumors of the salivary glands are rare, accounting for ∼5% of all head and neck malignancies and <0.5% of all malignancies in the body (1). Although initially treatable with surgery, the majority of patients eventually succumb to their disease (2). In general, salivary gland neoplasms respond poorly to chemotherapy, and hence, few treatment options exist for metastatic disease (3). Very little is known about the molecular alterations associated with the development and progression of salivary gland malignancies. A major obstacle to identifying these events has been a lack of preclinical animal models that could provide a basis for understanding salivary gland neoplasia and for testing therapies. Several murine salivary gland tumors have been identified serendipitously in mouse models engineered to generate mammary gland tumors (4–6). Because the mammary gland tumors were the focus of these studies, the salivary gland tumors were often not fully characterized. Developing mouse models that recapitulate human salivary gland malignancies will be a critical step to increase our understanding of these tumors and to discover targets for therapeutic intervention to improve patient outcome.

The mammalian target of rapamycin (mTOR) is a protein kinase that controls cell growth and proliferation in response to multiple upstream factors including insulin, growth factors [such as insulin-like growth factor-I (IGF-I) and IGF-II], and mitogens (7, 8). Activated mTOR regulates protein synthesis by phosphorylating ribosomal S6 kinase 1 (S6K) and eukaryotic translation initiation factor 4E–binding protein 1 (4EBP1), both of which are crucial for cap-dependent translation (9). Consequently, activation of mTOR signaling leads to increased levels of many proteins required for cell cycle progression, proliferation, angiogenesis, and survival pathways. Not surprisingly, deregulation of mTOR signaling is frequently associated with tumor initiation, growth, invasion, and metastasis (10, 11).

The lipid phosphatase Pten is one of several tumor suppressors that negatively regulate the mTOR pathway, and it is now thought to be the second most commonly mutated tumor suppressor in all human tumors, after p53 (12–14). Because cancer cells often become dependent upon activated mTOR signaling, this pathway is a promising target for therapeutic intervention (15). Rapamycin is a specific inhibitor of mTOR signaling that binds directly to the mTOR complex (mTORC1) and suppresses mTOR-mediated phosphorylation of S6K and 4EBP1 (9). Rapamycin analogues (CCI-779, RAD001, AP23575, AZD8055, and OSI-027) have been developed as anticancer drugs and are currently being tested in clinical cancer trials [reviewed by Liu and colleagues (16)]. CCI-779 (temsirolimus) and RAD001 (everolimus) have already been approved for the treatment of metastatic renal cell carcinoma, and more applications are pending Food and Drug Administration (FDA) approval (17, 18).

As a central regulator of cell growth, the mTOR pathway interacts with multiple signaling pathways (8, 9). One of these is the canonical Wnt signaling pathway, which promotes proliferation by activating β-catenin–dependent transcription and is itself often deregulated in human cancer (19). In addition to activating transcription, Wnt signaling augments the activity of mTOR, leading to increased phosphorylation of mTOR target proteins and cap-dependent translation (20, 21). This previously unrecognized function of Wnt signaling likely contributes significantly to the oncogenic effects of Wnt pathway mutations in human cancer. In support of this concept, Fujishita and colleagues recently showed that suppression of mTOR using a rapamycin analogue abrogated intestinal polyp formation in the Apc^Δ716 mouse model of colorectal cancer (21). The APC tumor suppressor gene is a key negative regulator of Wnt signaling and mutation of APC seems to be the initiating event in colorectal cancer. Similarly, we contributed to work showing that rapamycin inhibits Wnt1-induced mammary tumor growth (20).

In this study, we show that conditional inactivation of Apc and Pten in the mouse salivary gland causes a synergistic activation of canonical Wnt and mTOR signaling that depends on β-catenin. Salivary gland tumors similar to human acinic cell carcinoma (AcCC) quickly develop following mTOR and β-catenin activation, with 100% penetrance, often bilaterally. Although activation of both pathways is required for tumor initiation in this model, tumors completely regressed following rapamycin therapy. Importantly, we show that activation of the mTOR pathway is a common feature of human AcCC, suggesting that this mouse model is useful for studying the human disease. Furthermore, therapies targeting the mTOR pathway may be effective for treating human AcCC.

Apc^flox/flox (B6/129), Pten^flox/flox (FVB/SLJ), β-catenin^flox/flox (FVB/SLJ), and MMTV-Cre (FVB) transgenic mice have previously been described in detail (22–25). Mice with different genotypes of the Apc, Pten, β-catenin, and MMTV-Cre transgenes were generated by crossbreeding. The genotype of each transgene was confirmed by PCR using tail biopsy DNA and primer sequences (available upon request). Mice were monitored three times a week for tumor formation by visual inspection and/or palpation, and tumors were usually found when they were ∼500 mm³. Tumor growth was measured three times a week using a caliper, and all mice were euthanized before the tumors reached 2,000 mm³. An outline of the animal work is shown in Supplementary Table S1. All experiments done were approved in advance by the Van Andel Research Institute Institutional Animal Care and Use Committee.

All mice were examined at necropsy for gross organ abnormalities. Salivary gland tissue and tumors, as well as mammary glands, were collected, fixed in 10% (v/v) neutral buffered formalin, and embedded in paraffin. Five-micrometer-thick sections were cut from paraffin-embedded tissues and stained with H&E. Histopathologic evaluation of murine tumors and tissues was done by a pathologist with expertise in head and neck cancer (A.K. El-Naggar). Sections of paraffin-embedded human acinic cell carcinomas were obtained from the University of Michigan pathology department.

Immunohistochemistry was done on formalin-fixed, paraffin-embedded tissue sections (5 μm thick) using the Discovery XT System from Ventana Medical Systems according to the manufacturer's instructions. The following primary antibodies were used at specified dilutions: rabbit polyclonal antibody to β-catenin (1:200, Cell Signaling Technology), CD31 (1:50, Neomarkers), and cytokeratin 6 (1:100, Covance); rabbit monoclonal to PTEN (1:20, Abcam), phospho-mTOR (1:20, Cell Signaling Technology), and phospho-S6 (1:75, Cell Signaling Technology); rat monoclonal antibody to cytokeratin 8 (1:200, Troma1, Hybridoma Bank, University of Iowa) and Ki-67 (1:30, Dako).

Phospho-S6 levels in immunostained sections were quantified by spectral imagine using a Nuance camera system (Cri). Emissions were collected between 420 and 720 nm in 20-nm increments. The resulting image cubes were converted to absorbance units and mathematically unmixed into their individual components using spectra deduced from control specimens and saved in a spectral library. The resulting component images were then analyzed using the Nuance system's colocalization tool. This tool applies a standard threshold to all images in an experimental set and is able to evaluate nuclear or nonnuclear positivity. Values were compared using Student's t test.

Rapamycin (LC Laboratories) was given through i.p. injections at a dose of 1 mg/kg/day for 14 days. Tumor size was measured using a caliper daily for 28 days following the first dose of rapamycin. Rapamycin was reconstituted in absolute ethanol diluted in aqueous solution of 5.2% Tween-80 (U.S. Biochemical) and 5.2% polyethylene glycol (Polyscience Inc.) before injection. Control animals were given i.p. injections of the aqueous solution of 5.2% Tween-80 and 5.2% polyethylene glycol. The final volume of all injections was 100 μL.

5-Bromo-2′-deoxyuridine (BrdUrd) was injected (100 mg/kg) 2 hours before the mice were sacrificed. Salivary gland tumors were fixed and embedded in paraffin. BrdUrd-positive cells were detected in tissue sections using an anti-BrdUrd mouse monoclonal antibody (1:300, Sigma-Aldrich) and the ExtrAvidin peroxidase mouse staining kit (Sigma-Aldrich). A terminal deoxyribonucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) assay for apoptosis was conducted on tissue sections according to the manufacturer's protocol using the DeadEnd colorimetric TUNEL kit (Promega).

Single cells were isolated from murine salivary gland tumors by mechanical disaggregation using Medimachine (BD Biosciences). Cells were kept on ice and immediately processed according to the manufacturer's protocol using the Annexin V labeling kit (BD Biosciences). Cell sorting and analysis were done by using the BD FACSCalibur Flow Cytometer and Cellquest 5.2.1 software, respectively (BD Biosciences).

Germline deletion of either Apc or Pten in mice results in embryonic lethality. Mice heterozygous for the Pten gene have a higher incidence of tumors in many organs, whereas mice heterozygous for the Apc gene develop intestinal adenomas (26–30). To study the effects of Apc and Pten deficiency in the salivary gland and to overcome the embryonic death caused by conventional knockout strategies, we interbred homozygous Apc and Pten flox mice (Apc^flox/floxPten^flox/flox) mice with the MMTV-Cre transgenic mice developed by W. Muller and colleagues (McGill University). On a FVB purebred genetic background, this MMTV-Cre transgene is highly expressed in the mammary epithelium (25, 31). However, in a cross to Rosa26 (129 background), we found low penetrance (<10%) of MMTV-Cre expression in the mammary epithelium, but very high expression in the salivary gland (data not shown). Hence, on a mixed genetic background, this MMTV-Cre transgene can be used to effectively delete floxed alleles in the salivary gland. We confirmed that recombination occurred at the Apc and Pten locus in the salivary gland of MMTV-Cre/Apc^flox/flox/Pten^flox/flox (hereafter referred to as Apc^−/−Pten^−/−) by allele-specific PCR (Fig. 1A). To determine the extent of recombination, we also carried out Pten immunohistochemistry on salivary glands from 6-month-old MMTV-Cre/Pten^flox/flox (Pten^−/−) and MMTV-Cre/Pten^+/+ (Pten^+/+) mice. In Pten^+/+ mice (n = 4), Pten was ubiquitously expressed in all tissues from the three major salivary glands, with prominent staining in the ducts (Fig. 1B). In Pten^−/− salivary glands (n = 10), we observed a variable number of both ductal and acinar cells that lacked Pten expression (Fig. 1B). In 2 of 10 salivary glands analyzed, Pten expression was completely lost in cells from all three salivary glands, reflecting highly efficient Cre recombinase activity (Fig. 1C). Pten was ubiquitously expressed in the mammary epithelium of Pten^−/− female mice, and these mice exhibited normal mammary gland development (Supplementary Fig. S1A and B). Taken together, these results show that the MMTV-Cre transgene was preferentially expressed in the salivary gland in our model system.

Salivary gland tumors developed in the Apc^−/−Pten^−/− mice (n = 25) with short latency and 100% penetrance (Fig. 2A). Tumors were palpable as early as 6 weeks after birth, and within 25 weeks, all animals developed at least one large tumor mass in the ventral neck region. Both female and male mice developed tumors at a similar rate, with a median age at tumor onset of 10 weeks. Necropsy revealed well-demarcated and encapsulated tumors that seemed to always involve the parotid gland (Fig. 2B). The tumors grew very rapidly, with size doubling roughly every 5 days (Supplementary Fig. S2A). Sixty percent of the mice developed two or three tumors before being euthanized per guidelines for the humane treatment of animals. In addition, 23% of Apc^+/−Pten^−/− mice (n = 30) developed parotid gland tumors; these tumors appeared within the same time frame of 25 weeks that Apc^−/−Pten^−/− mice developed tumors. Apc^+/−Pten^−/− mice that had not developed tumors by 25 weeks were resistant to tumor development later in life. Once established, Apc^+/−Pten^−/− tumors grew as rapidly as Apc^−/−Pten^−/− tumors. In contrast to Apc^−/−Pten^−/− mice, only 6% of Apc^+/−Pten^−/− mice developed more than one salivary gland tumor. Mice that retained one copy of Pten were resistant to tumor development: none of the Apc^−/−Pten^+/− mice (n = 26) developed tumors (Fig. 2A). In addition, none of the Apc^+/−Pten^+/− or Apc^+/+Pten^+/+ control mice developed tumors. None of the mice, regardless of genotype, developed mammary gland tumors.

All tumors developing in Apc^−/−Pten^−/− and Apc^+/−Pten^−/− mice showed marked morphologic resemblance to human AcCC, with formation of acinar structures composed of cells with abundant basophilic granular cytoplasm with and without squamous metaplasia (Fig. 2C). Both Apc^−/−Pten^−/− and Apc^+/−Pten^−/− tumors showed nuclear and cytoplasmic staining for β-catenin (Fig. 2D and Supplementary Fig. S2B). All tumors exhibited strong activation of the mTOR pathway (shown by the large amounts of phosphorylated mTOR and p-S6) and lacked detectable expression of Pten (Fig. 2D and Supplementary Fig. S2B). Similar to human AcCC, immunohistochemical staining for cytokeratin CK-8 was positive for a significant fraction of tumor cells (Fig. 2D). In contrast to human AcCC, 35% of murine tumors also contained areas of CK-6–positive squamous metaplasia (Fig. 2D). The rapid tumor growth was evident from the large number of tumor cells that expressed proliferation marker Ki-67 (Fig. 2D).

Because Apc^−/−Pten^−/− and Apc^+/−Pten^−/− (but not Apc^−/−Pten^+/− mice) developed tumors, we established a Pten^−/− mouse line to test whether loss of Pten alone could induce acinic cell tumorigenesis. However, neither Pten^−/− nor Pten^+/− mice developed tumors within 78 weeks (n = 16 of each genotype), and histologic evaluation of salivary gland tissue was devoid of neoplastic lesions. During necropsy, we noticed that the salivary glands of Pten^−/− mice were often enlarged relative to those of littermate controls (Supplementary Fig. S3A). Histologic analysis showed that loss of Pten was associated with an increase in cell size and a corresponding increase in p-S6 levels (Fig. 3A and B). No significant difference in the number of Ki-67–positive cells was observed between Pten^−/− or Pten^+/+ salivary glands (Fig. 3C). Hence, the salivary gland enlargement occurred without evidence of increased cell proliferation. Similarly, loss of Apc alone was insufficient to induce tumorigenesis. In fact, we did not observe any histologic changes in the salivary glands of Apc^−/− mice (data not shown). In sharp contrast, loss of both Apc and Pten resulted in rapid proliferation of cells with strong activation of mTOR and Wnt/β-catenin signaling (Fig. 3D and Supplementary Fig. S3B). Taken together, these results suggest that loss of Apc and Pten synergistically activated Wnt/β-catenin and mTOR signaling to induce tumorigenesis.

Because APC has been implicated in regulating a number of cellular functions (32), it was important to test whether the observed Pten synergy was mediated through β-catenin. Therefore, we established an Apc^−/−Pten^−/−β-catenin^−/− mouse line. None of these mice (n = 9) developed salivary tumors within 52 weeks, implying that stabilization of β-catenin was indeed required for tumorigenesis. Similar to Pten^−/− mice, Apc^−/−Pten^−/−β-catenin^−/− salivary gland cells were enlarged and contained phosphorylated S6 (Supplementary Fig. S3C). Moreover, loss of one copy of β-catenin (Apc^−/−Pten^−/−β-catenin^+/−) was sufficient to inhibit tumor formation in mice (n = 5) with this genotype for 52 weeks.

To determine if the salivary gland tumor growth in Apc^−/−Pten^−/− and Apc^+/−Pten^−/− mice was dependent on activated mTOR and to provide a rationale for treatment of similar tumors in human patients, we treated tumor-bearing mice with the mTOR inhibitor rapamycin. Rapamycin directly binds and inhibits the mTORC1 complex. Low-dose rapamycin treatment was initiated on animals having tumors of different sizes and was continued for 2 weeks. Tumor growth was measured during rapamycin therapy and for at least an additional 2 weeks. Intraperitoneal injection of 1 mg/kg rapamycin effectively inhibited mTOR activity, as made evident by a loss in p-S6 4 hours after treatment (Supplementary Fig. S4A). Nuclear accumulation of β-catenin was still detectable in tumor cells following rapamycin therapy (Supplementary Fig. S4B). Compared with the rapid salivary tumor growth of untreated control animals (Fig. 4A), rapamycin therapy dramatically inhibited tumor growth (Fig. 4B). Apc^−/−Pten^−/− and Apc^+/−Pten^−/− tumor-bearing mice responded similarly to rapamycin treatment (data not shown). Analysis of cell proliferation on day 4 (after three doses of rapamycin) showed that the number of tumor cells with BrdUrd incorporation was reduced by at least 80% in rapamycin-treated mice (P = 2.6 × 10⁻¹⁰; Fig. 4C). Importantly, rapamycin not only inhibited tumor growth but also induced tumor regression (Fig. 4B). To determine if rapamycin induced tumor cell death, we analyzed tumor tissues and cells for evidence of necrosis and apoptosis. Flow cytometric analysis of tumor cells on day 4 (after three doses) showed that on average, 68% (SD 13.7%) of tumor cells from rapamycin-treated mice were Annexin V–positive (Fig. 4D). TUNEL assays similarly showed large numbers of apoptotic cells following rapamycin treatment (Fig. 4D). These results show that rapamycin quickly and effectively inhibited tumor cell proliferation and induced cell death. The gene expression profiles of Apc^−/−Pten^−/− tumors following rapamycin or vehicle therapy can be downloaded from the Gene Expression Omnibus database (accession no. GSE23435).

For 7/9 rapamycin-treated mice, tumor regression continued after the therapy was discontinued (Fig. 4B). Histologic sectioning of salivary tissue on day 28, 2 weeks after the completion of rapamycin therapy, showed fibrotic tissue in place of the regressed tumor (Fig. 5A). Very few Ki-67–positive proliferating cells were detected in this tissue, indicating that the malignancy had effectively been abolished by the rapamycin therapy (Fig. 5B). The mTOR pathway is proangiogenic (33–35), and we noticed that tumors collected from mice on day 15, after the completion of 14 doses of rapamycin, were pale relative to tumors from vehicle-treated animals (Fig. 5C). To determine the effect of rapamycin on angiogenesis in our AcCC model, we counted microvessel number and size on day 15 using the endothelial marker CD31. The microvessel number and size were significantly decreased in tumors from rapamycin-treated mice (P = 2.0 × 10⁻⁹; Fig. 5D). Taken together, these results show that rapamycin inhibited tumor angiogenesis, which likely contributed to the complete regression of tumors.

To determine if the canonical Wnt and/or mTOR signaling pathways are deregulated in human AcCC, we did immunohistochemistry for β-catenin and p-S6 on paraffin-embedded tumor sections from 11 patients diagnosed at the University of Michigan. Nuclear staining for β-catenin was only found in 1/11 tumor samples (Supplementary Fig. S5A). In contrast, activated mTOR signaling was evident in all tumor samples: the tumor tissues contained high levels of p-S6, whereas the adjacent normal parotid gland tissues exhibited low levels (Fig. 6A). The staining pattern of p-S6 correlated with an increased immunostaining for phosphorylated mTOR (Fig. 6A). We quantified the level of p-S6 in tumor and adjacent normal tissues using the Nuance spectral imaging system and found a significant difference in all samples (P < 1 × 10⁻¹⁰, Student's t test comparing p-S6–positive pixel numbers; Fig. 6B). We also did immunohistochemistry for Pten but did not detect a loss of Pten expression in any of the 11 human AcCC samples (Supplementary Fig. S5A).

In this report, we describe a novel mouse model for salivary gland tumorigenesis. This model is based on the inactivation of the tumor suppressor genes Apc and Pten using floxed alleles of these genes together with MMTV-Cre transgene expression. To achieve targeted deletion of the floxed Apc and Pten alleles, we used an MMTV-Cre mouse line in which Cre recombinase is preferentially expressed in the salivary gland when the mice are kept on a mixed genetic background. Although Pten loss in other MMTV-Cre mouse strains causes mammary gland hyperplasia and tumors (31, 36), we observed infrequent loss of this gene in the mammary epithelium of our mouse line. Instead, the expression of the MMTV-Cre transgene caused the inactivation of the Apc and Pten floxed alleles in cells of all three major salivary glands, and 100% of female and male Apc^−/− Pten^−/− (MMTV-Cre/Apc^flox/flox/Pten^flox/flox) mice developed salivary gland tumors. This MMTV-Cre line should be very useful for the development of further salivary tumor models, which has been difficult until now because of strong Cre expression in the mammary gland and the resulting mammary tumor phenotypes.

The inactivation of Apc and Pten in the salivary gland led to rapid development of tumors that manifested remarkable resemblance to human AcCC. In humans, AcCC is a rare malignant neoplasm of the parotid gland (reviewed in refs. 37, 38), where its histogenesis, genetic characteristics, and biology remain undefined. The development of our model will shed more light on AcCC development and progression and allow for comparative analysis with other salivary carcinomas. As in humans, the murine tumors seemed almost exclusively in the parotid gland.

Tumor formation required inactivation of both Apc and Pten because mice with inactivation of either gene alone failed to develop salivary tumors. As expected, the murine tumors exhibited strong activation of β-catenin/Wnt and mTOR signaling. However, the Apc^−/−Pten^−/− salivary gland tumors that developed were completely dependent on continuous mTOR signaling. Mice treated over 14 days with a low dose of the mTORC1 inhibitor rapamycin (1 mg/kg) showed pronounced and continued tumor regression. Even after the discontinuation of rapamycin therapy, we found only fibrotic tissue replacing the tumor and there was no recurrence 14 days posttreatment. Interestingly, reports of mouse acinic cell tumors developing in MMTV-v-Ha-ras, MMTV-CD8-IGF1R and MMTV-HER2 transgenic mice have been published (39–41). Because these growth factors can activate mTOR signaling, mTOR may play a tumorigenic role in salivary precursor cells leading to the development of these tumors.

To validate our model, we screened 11 human AcCC tumors for evidence of Wnt or mTOR signaling and observed that all tumors had significantly elevated mTOR signaling (Fig. 6A and B). In several samples, strong p-S6 staining was present in virtually all tumor cells suggesting that an mTOR pathway mutation occurred at an early stage in tumorigenesis. We screened for the loss of Pten by immunohistochemistry but found no evidence for Pten inactivation in any of the 11 tumors. Because mTOR activation can result from mutations in multiple pathway components (8), high-throughput sequencing of human AcCC tumors to define mutations in the mTOR pathway is an important future goal.

Although Wnt pathway activation was required for tumorigenesis in the mouse model, its role in human AcCC is not clear. In mice, the tumor phenotype caused by mTOR and Wnt pathway activation required β-catenin because the Apc^−/−Pten^−/−β-catenin^−/− mice failed to develop tumors. Thus, tumor formation required β-catenin and was not a consequence of other functions related to Apc loss. However, only 1 of 11 human tumors we surveyed showed evidence of β-catenin stabilization. One possible role for Apc inactivation in the mouse model is to promote cell proliferation through activating β-catenin–dependent transcription. It is also possible that Apc inactivation increased the level of mTOR signaling as occurs in mouse models for colorectal cancer (21, 42). That all of the human AcCC tumors had activated mTOR signaling but not Wnt signaling suggests that human tumors activate other signaling pathway(s) that cooperate with mTOR activation in AcCC pathogenesis.

It is worth noting that inactivation of Pten alone resulted in salivary gland hypertrophy and hypercellularity without increased cell proliferation. This is consistent with the results of several published mouse models in which inactivation of Pten alone did not lead to tumor development (42–44), but can cause cellular senescence (45, 46). Similar to our findings, others have shown that Pten deletion in mouse brain caused increased cell size but no effect on cell proliferation (47, 48). It may be that terminally differentiated cells are refractory to the growth-promoting effects of activated mTOR. However, when Pten is inactivated in the presence of other tumor-associated mutations, for example, in p53 or Apc, this can be highly tumorigenic (42–44). In summary, our study provides a relevant preclinical model for investigating therapies for AcCC of the salivary gland. Furthermore, the data suggest that rapamycin treatment might be useful for treating patients with AcCC, which is important to evaluate because limited treatment options exist beyond surgical excision of the tumor.

No potential conflicts of interest were disclosed.

We thank William Muller, Tak Mak, Tetsuo Noda, and Rolf Kemler for providing access to the mouse lines used in this project; Silivo Gutkind and Alfredo Molinolo for sharing their expertise; Nicole Evans, Kristin Feenstra, Juraj Zahatnansky, Kyle Furge, Karl Dykema, and members of the Williams' lab for technical assistance; and the Van Andel Research Institute's vivarium staff for outstanding animal husbandry.

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.