An Orthotopic Murine Model of Sinonasal Malignancy Free

Malignant sinonasal tumors are clinically challenging because of their rarity, their proximity to vital structures, and their histologic variety within a complex anatomic region. These tumors comprise <3.6% of all malignancies seen in the head and neck (1). With an annual incidence of only 0.5 to 1 per 100,000 per population, these tumors have been difficult to study in large clinical trials, leaving treatments and outcomes to be based primarily on limited retrospective data.

The most common sinonasal malignancies are squamous cell carcinoma (SCC) followed by adenoid cystic carcinoma (ACC). Most (70-80%) of these tumors originate in the maxillary sinus, but only 25% remain confined to the sinus (2). Patterns of invasion include direct extension into the orbit or skull base via perineural spread and bony destruction, spread to the regional lymphatics, and distant metastasis. Recurrence generally reflects local tumor regrowth rather than distant metastasis (3, 4). Despite improved surgical approaches, refined radiotherapy techniques, and cytotoxic chemotherapies, 3-year disease-specific survival remains <35% in patients with T₃ and T₄ lesions (5). Future progress in disease-specific morbidity and mortality will require the identification of novel prognostic tumor markers and the implementation of molecular-based therapies.

One of the challenges to developing molecularly targeted approaches in sinonasal malignancies has been the lack of preclinical models facilitating laboratory-based investigations. In contrast to the use of traditional cytotoxic agents, which rely on differential killing of rapidly dividing cells, effective molecular approaches require a thorough understanding of tumor cell biology in the context of the native tumor microenvironment (6, 7). The vast majority of the information pertaining to the molecular pathways mediating vital cell processes and therapeutic response has been based on in vitro monolayer cultures (8, 9). The application of these models assumes that the growth and metastatic mechanisms of a tumor cell in vitro resemble those in situ. Although a valuable tool in the study of cellular functions, they may not reflect the natural host-tumor microenvironment interactions associated with tumor development and progression (10–13). Attempts to compensate for this inherent limitation of the in vitro condition have included implanting human tumor cells as subcutaneous xenografts in immunocompromised mice (14). Although subcutaneous growth occurs in vivo, it does not replicate the normal tumor-host interactions for these tumors and thus remains an inferior system for preclinical modeling (15, 16).

Current animal models of sinonasal malignancy consist of several purely intracranial murine brain tumor models (17) and isolated rabbit models of maxillary sinus cancer that do not show local spread or skull base invasion (18, 19). Heterotopic engraftment models have also been established from human sinus cancer cell lines (20, 21). However, to date, no animal model is established to study the clinical and biological behavior of these malignancies. Therefore, the aim in our current study was to create an orthotopic murine model of sinonasal malignancy that could reproduce the intracranial extension, orbital involvement, bony destruction, and spread along neural facial planes frequently seen in patients with aggressive sinonasal malignancies of various histologies.

Cell lines. To produce tumors, ACC (ACC-3; human salivary ACC cell line provided by Dr. Wantao Chen, Laboratory of Oral Tumor Biology, Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine; previously established from an ACC of the parotid gland; ref. 20) and SCC (DM14; derived in our laboratory from floor of mouth cancer cell line TU167) were harvested from subconfluent cultures maintained in frozen stocks. ACC-3 cells were maintained as monolayer cultures in RPMI 1640 containing penicillin-streptomycin (Flow Laboratories), nonessential amino acids, sodium pyruvate, l-glutamine, and 10% fetal bovine serum. DM14 cells were grown in DMEM supplemented with 10% fetal bovine serum, penicillin/streptomycin, sodium pyruvate, glutamate, and nonessential amino acids. We maintained adherent monolayer cultures at 37°C in 5% CO₂ and 95% air. The cultures were maintained free of Mycoplasma species and no longer than 12 weeks after recovery from frozen stocks. For injections, the cells were trypsinized and resuspended in serum-free HBSS at appropriate concentrations. Cells from both of the human tumor lines were prepared in 30 μL HBSS in the following concentrations: 5 × 10⁵, 2 × 10⁵, 5 × 10⁴, and 5 × 10³.

Animal care and implantation of tumor cells. Male nude mice (ages 6-8 weeks) were purchased from the Animal Production Area of the National Cancer Institute-Frederick Cancer Research and Development Center. The mice were maintained in a pathogen-free environment and fed irradiated mouse chow and autoclaved reverse osmosis-treated water at facilities in accordance with current regulations and standards. The mice were used in accordance with the Animal Care and Use Guidelines of The University of Texas M. D. Anderson Cancer Center. All of the animal procedures were done in accordance with a protocol approved by the Institutional Animal Care and Use Committee. We injected experimental animals with tumor cells from two human cancer lines (previously described) in either the maxillary sinus or the soft palate to determine the optimal site of orthotopic tumor implantation. All mice were anesthetized with intraperitoneal sodium pentobarbital (50 mg/kg body weight) before injection. We used groups of five mice for each concentration and each cell line. Mice were euthanized with CO₂ when they lost >20% of their body weight, had ulcerated tumors, or became moribund in accordance with Institutional Animal Care and Use Committee guidelines.

Maxillary sinus model. We sought to determine if human xenograft tumors could be established in the maxillary sinus of mice. Mice were anesthetized with an intraperitoneal phenobarbital injection. After confirming the degree of anesthesia, we placed the mice in a supine position. Tumor cells were implanted via transcutaneous injection underneath the infraorbital muscle groups, with the needle angled toward the lateral wall of the maxillary sinus. We used a 1 mL tuberculin syringe (Hamilton) with a 30-gauge hypodermic needle for the injections. The absence of a fluid leak from the nasal cavity confirmed a successful injection. The mice were observed until the effects of anesthesia had resolved.

Soft palate model. We sought to determine if human xenograft tumors could be established in the soft palate of mice. Mice were anesthetized with an intraperitoneal phenobarbital injection. After confirming the degree of anesthesia, we placed the mice in a supine position. The mandible was depressed, and the tongue was retracted to show the junction of the hard and soft palate. After identifying the soft palate, we implanted tumor cells via direct injection into the muscle of the soft palate using a 30-gauge hypodermic needle and a 1 mL tuberculin syringe (Hamilton). After injection, a well-localized wheal over the soft palate and the absence of a fluid leak from the nasal or oral cavity confirmed a successful injection. The mice were observed until the effects of anesthesia had resolved.

Survival assessment. Initial survival studies were done to determine the proper time to euthanize the animals for the tumor growth study. During the initial survival studies, we checked the animals daily for hypomotility and absence of grooming behavior. The body weights of animals were measured twice a week to detect weight loss. We euthanized mice with CO₂ when they became moribund or when their weight loss was >20%. At the conclusion of the study, survival curves were plotted for time of animal death according to the above-mentioned criteria. Kaplan-Meier estimation was used to graph the survival curve using moribundness and >20% weight loss as surrogates for survival.

Necropsy and tissue preparation. For tumor growth and invasion evaluation, we euthanized the animals with CO₂ 18 days after cell line implantation. At the time of death, the full heads of the mice were obtained, fixed in a periodate-lysine-paraformaldehyde solution for 24 h, and decalcified in Immunocal formic acid bone decalcified (Decal) for 10 to 12 days. Each head was divided into blocks by one median and two paramedian sagittal sections. All head samples were embedded in OCT compound (Tissue-Tek OCT Compound; Sakura Finetek) after going through successive passage in 10%, 15%, and 20% sucrose concentrations. H&E staining was done on histologic sections of the head to determine the extent of tumor growth and the degree of invasion into surrounding structures. Also at the time of euthanasia, tumors were measured and volume was estimated with the formula: volume of a sphere = 4 / 3πr³. Additionally, the cervical lymph nodes and lungs were removed and placed in 10% buffered formalin solution overnight for fixation. Each specimen was stained with H&E and evaluated under light microscopy for the presence of regional or distant metastasis.

Statistical analysis. Statistical analysis was done with SPSS software. In the tumor volume studies, an ANOVA was done to assess whether an overall difference existed between the mean volumes of the tumors growing from cell suspensions of different concentrations. If the test showed that a significant difference existed, two-tailed Student's t tests were done. The t test comparisons were used to compare the differences between tumor volume and the initial tumor cell concentration. All tests were done separately for the maxillary sinus and soft palate data. The level of significance was set at 5% (P < 0.05). In survival studies, the log-rank test was used to determine if differences in survival times were significant (P < 0.05).

Orthotopic implantation is technically feasible and can be used with a spectrum of human tumor xenografts. Maxillary sinus is seen on computed tomography of normal mouse (images provided courtesy of VoxPort); red shaded regions denote maxillary sinus (Fig. 1A). Normal soft palate is seen on illustration (Color Atlas of Anatomy of Small Laboratory Animals, Volume II. Popesko, Rajtova, and Horak. Elsevier; 2002. p. 123); sagittal computed tomography section of normal mouse (images provided courtesy of VoxPort) and H&E-stained section from normal mouse (Fig. 1B).

Human SCC line DM14 forms large bulky tumors when injected into either the maxillary sinus or the soft palate of nude mice. Successful implantation and tumor formation was achieved with 5 × 10⁵ cells of the human SSC cell line DM14 implanted into either the maxillary sinus (Fig. 2A-C) or the soft palate (Fig. 2D-F) of nude mice. High rates of local invasion were seen within 30 days in mice with tumor implanted in either the maxillary sinus or the soft palate, showing the feasibility of these orthotopic injection techniques for xenograft implantation and growth.

To further characterize the model and to establish optimal time points for euthanasia and histologic evaluation, we performed a longitudinal observational study in mice injected with various concentrations of DM14 cells implanted in either the soft palate (Fig. 2G) or the maxillary sinus (Fig. 2H). In the maxillary sinus model, the Kaplan-Meier survival analysis revealed a statistically significant difference in survival when mice injected with 5 × 10⁵ cells were compared with mice injected with 5 × 10⁴ (P = 0.002) or 5 × 10³ (P = 0.002) cells. The difference between mice bearing 5 × 10⁵ cells and mice bearing 2 × 10⁵ cells was not statistically significant (P = 0.926). For the group of mice injected in the soft palate, the Kaplan-Meier survival analysis revealed a statistically significant difference in survival between mice injected with 5 × 10⁵ cells and all other groups (versus 2 × 10⁵, P = 0.004; versus 5 × 10⁴, P = 0.0018; and versus 5 × 10³, P = 0.0018). Using the Kaplan-Meier survival analysis, we determined the median survival time for mice that received 5 × 10⁵ cells in the maxillary sinus was 25.50 days. The median survival time for mice that received 5 × 10⁵ cells implanted in the soft palate was 18.5 days. To maximize the number of surviving mice bearing tumor for histologic analysis, we elected to euthanize all mice at 18 days in the next experiment.

Human SCC line DM14 shows local tissue invasion and regional and distant metastasis when injected into either the maxillary sinus or the soft palate of nude mice. Mice that received 5 × 10⁵ DM14 cells in the maxillary sinus developed tumors that invaded through the lamina papyracea into the orbital cavity (Fig. 3A). Tumors were not restricted to local invasion and also showed regional metastasis to the cervical lymphatics (Fig. 3B). When 5 × 10⁵ DM14 cells were implanted in the soft palate, mice developed tumors that invaded superiorly into the cranial vault (Fig. 3C). Tumors in the soft palate showed metastasis to the lungs (Fig. 3D and E). Additionally, tumors implanted in the maxillary sinus displayed perineural invasion as evidenced on histologic sections from the lateral maxillary wall (Fig. 3F).

At sacrifice, the tumor volume measured in external cross-section was directly correlated with the implanted cell concentration (Fig. 3G and H). For mice injected in the soft palate, differences between groups were statistically significant in two comparisons (5 × 10⁵ versus 5 × 10⁴, P = 0.020; 5 × 10⁵ versus 5 × 10³, P = 0.001). Similar significant differences in tumor volumes were seen between groups of mice injected in the maxillary sinus (5 × 10⁵ versus 5 × 10⁴, P = 0.008; 5 × 10⁵ versus 5 × 10³, P = 0.001). Tumors growing in the maxillary sinus were of larger volume for all cell densities when compared with tumors growing in the soft palate. Further investigation is needed on whether this difference was related to the physical restrictions imposed by the anatomy of the soft palate or microenvironmental cues.

For the human SCC cell line DM14, we observed a spectrum of tumorigenicity related to initial concentration and site of injection. These trends are described in Table 1.

Human adenoid cystic tumor cell line ACC-3 forms tumors when implanted orthotopically in either the maxillary sinus or the soft palate. In initial survival studies and Kaplan-Meier analysis of mice bearing cell line ACC-3 tumors in either the maxillary sinus or soft palate, the median survival time was 24 days (data not shown). When injected into the maxillary sinus, these mice developed bulky, large tumors (seen in coronal view; Fig. 4A) that also invaded the orbital tissues (Fig. 4B) and showed infiltration and destruction of the bony paranasal sinus suprastructure (Fig. 4C). Similarly, when mice were injected in the soft palate, they developed bulky, large tumors (seen in sagittal view; Fig. 4D), which showed bone and intracranial invasion (Fig. 4E). When the mice were euthanized on day 24, tumor volume measured in external cross-section was found to be directly correlated to the cell concentration in the tumor cell inoculum (Fig. 4F and G). When comparing the different volumes of the tumors in the maxillary sinus, we found significant differences in mean tumor volume between groups receiving different tumor cell concentrations: (5 × 10⁵ versus 5 × 10⁴, P = 0.006; 5 × 10⁵ versus 5 × 10³, P = 0.008; 2 × 10⁵ versus 5 × 10⁴, P = 0.004; 2 × 10⁵ versus 5 × 10³, P < 0.0001). There was no evidence of a significant difference between the mean volume of the 5 × 10⁵ and 2 × 10⁵ densities (P = 0.16) or between the 5 × 10⁴ and 5 × 10³ cell concentration inocula (P = 0.071). When comparing volumes of the tumors in the soft palate, the difference in mean tumor volume between groups of mice receiving 5 × 10⁵ versus 5 × 10³ cell density inocula was not significant (P = 0.056). However, a significant difference was found between the groups injected with cells at the 2 × 10⁵ and 5 × 10³ densities (P = 0.023).

Human adenoid cystic tumor cell line ACC-3 shows local tissue invasion and regional and distant metastasis when injected into either the maxillary sinus or the soft palate of nude mice. To determine the reliability of our model in replicating the phenotype of human ACC, we examined the histologic characteristics of nude mice with the human tumor line ACC-3 implanted in their maxillary sinus and soft palate. After being sacrificed on day 24, mice that were injected with 5 × 10⁵ ACC-3 cells in the soft palate developed tumors that locally invaded the soft palate and paranasal sinuses (Fig. 5A); they also showed perineural invasion and invasion inside the cranial vault (Fig. 5B). Histologic examination showed that these tumors compressed the medulla (Fig. 5C). Intracranial involvement was not restricted to tumors originating from cells injected into the soft palate. Mice bearing 5 × 10⁴ ACC-3 cells implanted in the maxillary sinus also showed intracranial extension of tumor and brainstem compression in sagittal sections (Fig. 5D). When these cells were implanted in the soft palate, tumors showed upward extension and bony destruction though the cribiform plate (Fig. 5E). When 5 × 10⁵ ACC-3 cells were implanted in the maxillary sinus, tumors extended laterally into the bony structures of the nasal and paranasal sinuses (Fig. 5F). Tumors showed capacity for distant spread regardless of the injection site; mice with 5 × 10⁵ cells implanted in the soft palate developed cervical lymph node metastasis (Fig. 5G) and mice with 5 × 10⁵ cells implanted in the maxillary sinus developed pulmonary metastasis (Fig. 5H).

Our study is the first description of an orthotopic preclinical model of sinonasal malignancy that closely mimics the behavior of human disease. Our model shows extension into the cranial vault, orbital invasion, perineural spread, and distant metastasis. Using the human SCC line DM14 and the ACC cell line ACC-3, our model manifests marked resemblance to the phenotype and the malignant behavior of two of the most aggressive human tumor types that are seen in this complex anatomic area.

Numerous retrospective studies have shown that tumor type, stage, and extent of direct extension are significant predictors of patient outcome. Multivariate analysis of 220 patients treated between 1975 and 1994 showed that factors associated with worse 5-year carcinoma specific actuarial survival were tumor histology, extension into the pterygomaxillary fossa, and invasion of the dura (21). Another retrospective review of 100 patients with sinonasal malignancies showed factors associated with worse 5-year survival to be recurrence following craniofacial tumor resection, involvement of the orbital soft tissues, and invasion of the sphenoid sinus (22). A multivariate study of 334 patients from 17 institutions identified local invasion into surrounding structures with worse outcome. Additionally, surgical margins, histology, and the intracranial extent of tumor were independent predictors of overall, disease-specific, and recurrence-free survival (23). Clinical experience with human disease strongly suggests an intrinsic relationship between survival and extent of local spread, bone invasion, and perineural spread. For any animal model to accurately recapitulate this disease, it must reliably and reproducibly show these characteristics.

A hallmark of orthotopic cancer models is the ability to model the disseminated metastasis seen in human cancer. Subcutaneous models are a valuable tool, but they are limited by a lymphatic drainage and vascular supply that are vastly different than those of the orthotopic site (24). Thus, subcutaneous xenografts cannot reproduce the patterns of regional and distant metastasis that are characteristic of sinonasal cancers. Additionally, the differences in the stromal cellular composition and the extracellular matrix between the orthotopic and subcutaneous microenvironments contribute to the extremely low metastatic rates of subcutaneously established tumors. Therefore, orthotopic models of highly metastatic cancers have been established in nude mice for carcinomas of the colon, stomach, pancreas, breast, bladder, lung, thyroid, and tongue (7, 24–29). Similarly, our orthotopic model of sinonasal malignancy provides an experimental system for exploring the events associated with metastasis within both the regional draining lymphatic basin and in distant sites.

Orthotopic models of cancer are also able to reproduce the site-specific spectrum of clinical findings and thereby permit analysis of the effects these findings have on survival. As seen in patients with malignant sinonasal tumors, our orthotopically generated sinonasal tumors in nude mice reproducibly penetrated the cranial vault, invaded the orbit, and spread through bone and perineurium within 18 to 30 days after the tumor cells were implanted. This model can now be used to investigate the genetic and molecular aberrations that may serve as a novel substrate for targeted therapeutic strategies. These therapies must be able to specifically address the characteristics that render these malignancies clinically aggressive.

Rapid progress has been made in our understanding of the molecular alterations contributing to the development of SCC arising outside the sinonasal tract (30). As evidenced by the discovery of epidermal growth factor receptor signaling in head and neck cancer, targeting aberrant molecular pathways has been directly translated into tangible patient benefit (31, 32). However, our understanding of the molecular biology of sinonasal malignancies and their potential for molecularly targeted therapy has not yet been able to reap the full benefit of current research efforts centered on disease outside the sinonasal cavity. Our model offers a mechanism to apply several of the well-established molecular therapeutics to a malignancy with precious few therapeutic options.

Our model offers a validated preclinical tool to test if SCC and ACC arising in the sinonasal cavity can be successfully treated with reagents targeting specific molecular pathways. These include drugs targeting growth factors and their receptors, signal transduction molecules, oncogenes, hormones, apoptosis-related molecules, angiogenesis-related factors, as well as inhibitors of cell motility, invasion, and proteolysis. Decreased tumor size coupled with reductions in perineural invasion and lymphatic and pulmonary metastasis following therapy can serve as quantitative endpoints for efficacy evaluation.

Beyond the established reagents for epidermal growth factor receptor (33), small-molecule inhibitors of nuclear factor-κB (34), farnesyl transferase inhibitors that inhibit the ras oncogene and its downstream mediators (35), small molecules targeting signal transducers and activator of transcription 3 (36), antisense oligonucleotides targeting human telomerase reverse transcriptase mRNA (37), and angiogenesis inhibitors (38, 39) can all be rapidly tested with the model described in this work.

Additionally, given that the degree of direct extension is a significant predictor of patient outcomes in sinonasal malignancy, reagents designed to limit this process may influence patient survival. Studies of epithelial malignancies have consistently shown that loss of cell adhesion and the acquisition of mesenchymal features, a process termed epithelial-mesenchymal transition, precede their invasion and progression. Epithelial-mesenchymal transition is a complex process induced by modifications of multiple pathways leading to a spectrum of epithelial cellular changes including loss of polarity and cohesion, increased motility, and acquisition of mesenchymal phenotype. Of these pathways, the protein tyrosine kinase Src is a common target of different growth factor receptor activation in epithelial-mesenchymal transition (40). Preclinical evaluations have shown a strong rationale for targeting Src in head and neck SCC (41, 42). Other targets involved in extracellular matrix proteolysis, including matrix metalloproteinases and urokinase-type plasminogen activator and its receptor may prove fruitful lines of research as well.

Beyond the application of established therapies, our model may also allow for selective targeting of novel pathways associated with tumorigenesis. One recent example is the role of β-catenin signaling in the cancer stem cells postulated to be fundamental to SCC growth, invasion, and metastasis (43). In this exciting new work, the authors show multiple common molecular characteristics of both cancer stem cells and nonmalignant skin stem cells. Interestingly however, cancer stem cells were critically dependent on highly active β-catenin signaling and unique in their ability to initiate tumors when injected, at a very low number, into mice. Silencing β-catenin signaling led to the regression of skin tumors while leaving normal skin homeostasis apparently untouched, thus providing a strong rationale for therapeutic targeting of β-catenin in SCC arising within the sinonasal tract.

Orthotopic nude mouse models allow dissection of the molecular and cellular mechanisms of tumor growth and metastasis (44). They serve a critical role in identifying safe and effective therapies. Yet, nude mouse models are not without limitations. Insights gleaned from these models need additional substantiation from complementary studies in immunocompetent models coupled with analysis of archival human tumor specimens. Together, these studies may generate insight into the pathogenesis and molecular biology of sinonasal cancer. They also may provide the information critical to preclinical assessment of new drugs designed to combat a diverse and aggressive subset of malignancies.

We have developed an orthotopic model of sinonasal cancer in nude mice that reproduces the clinical and pathologic features of sinonasal malignancy in humans. Our model is technically feasible and can be used with multiple human tumor xenografts. The model should allow further studies that enhance our understanding of the molecular and cellular mechanisms underlying the pathogenesis of sinonasal malignancy. We believe our model will facilitate the development and evaluation of new therapies for a disease subset whose rarity leads to difficulty in accruing sufficient patients to adequately power large human trials.

No potential conflicts of interest were disclosed.

We thank Kristi M. Speights for excellent editorial assistance.