Development and Radiation Response Assessment in A Novel Syngeneic Mouse Model of Tongue Cancer: 2D Culture, 3D Organoids and Orthotopic Allografts

Oral squamous cell carcinoma (OSCC) are aggressive cancers that contribute to significant morbidity and mortality in humans. Although numerous human xenograft models of OSCC have been developed, only a few syngeneic models of OSCC exist. Here, we report on a novel murine model of OSCC, RP-MOC1, derived from a tongue tumor in a C57Bl/6 mouse exposed to the carcinogen 4-nitroquinoline-1-oxide. Phenotypic characterization and credentialing (STR profiling, exome sequencing) of RP-MOC1 cells was performed in vitro. Radiosensitivity was evaluated in 2D culture, 3D organoids, and in vivo using orthotopic allografts. RP-MOC1 cells exhibited a stable epithelial phenotype with proliferative, migratory and invasive properties. Exome sequencing identified several mutations commonly found in OSCC patients. The LD50 for RP-MOC1 cells in 2D culture and 3D organoids was found to be 2.4 Gy and 12.6 Gy, respectively. Orthotopic RP-MOC1 tumors were pan-cytokeratin+ and Ki-67+. Magnetic resonance imaging of orthotopic RP-MOC1 tumors established in immunocompetent mice revealed marked growth inhibition following 10 Gy and 15 Gy fractionated radiation regimens. This radiation response was completely abolished in tumors established in immunodeficient mice. This novel syngeneic model of OSCC can serve as a valuable platform for the evaluation of combination strategies to enhance radiation response against this deadly disease.


Introduction
Oral cancers are a major cause of morbidity and mortality worldwide with more than 90% of these cancers diagnosed histologically as oral squamous cell carcinoma (OSCC) [1]. Globally, approximately 350,000 of patients are diagnosed with OSCC resulting in more than 150,000 deaths every year [2]. Chronic exposure of the oral epithelium to carcinogens such as alcohol and tobacco are the major risk factors for development of OSCC [1][2][3]. Radiation therapy is an integral part of the management of patients with OSCC [4]. Significant advancements in radiation delivery methods have led to improved outcomes in patients but resistance to radiation remains a clinical problem [5,6].
The development of clinically relevant animal models of OSCC is an essential first step in the examination of novel strategies that could enhance radiotherapeutic efficacy against this debilitating disease. In this regard, numerous human tumor xenograft models of OSCC have been employed to examine the effects of radiotherapy in vivo [7][8][9]. However, the need for an immunodeficient host

Generation and Credentialing of RP-MOC1
The workflow for generation of RP-MOC1 cells is shown schematically in Figure 1A. A clinically visible exophytic lesion on the tongue of a C57Bl/6 mouse exposed to the carcinogen 4NQO in drinking water for 16 weeks was transplanted subcutaneously into a recipient mouse. The established tumor was excised, and a portion of the tumor was also placed in formalin fixative for histopathologic evaluation. The remaining tissue was digested and seeded as a monolayer. Cultured cells were sorted using flow cytometry based on expression of the stem cell marker (CD44) and epithelial cell adhesion molecule, EpCAM (CD326). Post sort analyses showed co-expression of the two markers in 98.2% of the cells ( Figure 1B).
Microscopic evaluation of these cells (named as RP-MOC1) revealed polygonal shaped cells with a cobblestone morphology, characteristic of OSCC ( Figure 1C). Species specific PCR evaluation confirmed the murine origin of the cells with no mammalian interspecies contamination (Table S1). Short tandem repeat (STR) profiling of RP-MOC1 cells confirmed a genetic profile identical to C57Bl/6 mouse strain consistent with the origin of the cells (Table S2). The growth curve of RP-MOC1 cells over a five-day period showed a typical "S" shape with a calculated doubling time of 37.6 ± 2.4 h ( Figure 1D).
The in vitro behavior of RP-MOC1 cells was studied using wound healing, migration, and invasion assays. The wound healing assay showed the ability of RP-MOC1 cells to migrate with a near complete wound closure seen by 48 h ( Figure 1E). Quantitative analyses using the transwell migration assay showed~7-fold increase (p = 0.006) in the motility of these cells from 11.42 ± 1.46 cells (at 12 h) Cancers 2020, 12, 579 3 of 16 to 75.67 ± 11.68 cells (at 24 h) ( Figure 1F). A three-fold increase (p = 0.049) in the number of cells invading through the matrigel coated transwell chamber by 24 h (189.9 ± 44.19 cells) compared to 12 h (60.21 ± 14.82 cells) ( Figure 1G).  Donor tongue tissue from a C57BL/6NCr mouse exposed to 4NQO in the drinking water for 16 weeks was initially transplanted subcutaneously to establish an allograft. The established tumor was subsequently excised, digested with collagenase and seeded in a petri dish. (B) Flow cytometry was used to sort cells co-expressing the stem cell marker (CD44) and epithelial cell adhesion molecule, EpCAM (CD326

RP-MOC1 Cells Harbor Genetic Alterations Similar to Human Head and Neck Cancer
We performed exome sequencing at a depth of 20× and 30×, with 88% and 74% of targeted regions covered, respectively. The data were mapped to mm10 genome (reference strain C57BL/6NCr) and GENCODE GRCm38.75 annotation (gene set). Following filtration and alignment with GATK 3.5 against the reference genome, we found that the coding regions of RP-MOC1 genomes contained of 3404 single nucleotide variations (SNVs) and 146 small insertions and deletions (INDEL) (Supplementary File 1). Out of 3404 SNVs, 1727 (50%) missense variants were observed. The majority of the SNP variants found were G>T transversion (30.8%) followed by C>A transversion (28.6%). We compared the filtered SNVs against the mutation landscape of human head and neck SCC Donor tongue tissue from a C57BL/6NCr mouse exposed to 4NQO in the drinking water for 16 weeks was initially transplanted subcutaneously to establish an allograft. The established tumor was subsequently excised, digested with collagenase and seeded in a petri dish. (B) Flow cytometry was used to sort cells co-expressing the stem cell marker (CD44) and epithelial cell adhesion molecule, EpCAM (CD326). (C) Microscopic image of RP-MOC1 showing characteristic cobblestone morphology with anisocytosis and cellular pleomorphism. Image was acquired at 20× magnification. (D) Growth curve of RP-MOC1 cells over a 5-day period. The doubling time for this cell line based on MTT was determined to be 37.6 ± 2.4 h. (E) Wound healing assay. Images of RP-MOC1 cells at 0 h and at 48 h are shown. The image at 48 h illustrates near complete wound closure due to migration of tumor cells. Images were acquired at 4× magnification. Scale bar: 500 µm. Bar graphs showing increase in the number of migrated cells (F) and invasive cells (G) between 12 h to 24 h. Data is reported as mean ± SEM and from two independent experiments. (ns, no significance; * p < 0.05; ** p < 0.01).

RP-MOC1 Cells Harbor Genetic Alterations Similar to Human Head and Neck Cancer
We performed exome sequencing at a depth of 20× and 30×, with 88% and 74% of targeted regions covered, respectively. The data were mapped to mm10 genome (reference strain C57BL/6NCr) and GENCODE GRCm38.75 annotation (gene set). Following filtration and alignment with GATK 3.5 against the reference genome, we found that the coding regions of RP-MOC1 genomes contained of 3404 single nucleotide variations (SNVs) and 146 small insertions and deletions (INDEL) (Supplementary File 1). Out of 3404 SNVs, 1727 (50%) missense variants were observed. The majority of the SNP variants found were G>T transversion (30.8%) followed by C>A transversion (28.6%). We compared the filtered SNVs

Radiation Response of RP-MOC1 Cells in 2D Culture and as 3D Organoids
Next, we examined the response of RP-MOC1 cells to radiation in vitro. We exposed RP-MOC1 cells in 2D culture to increasing doses of radiation (0-10 Gy) and evaluated the response using the colony forming assay. At fifteen days post radiation, a 20%-100% reduction in colony formation was observed with increasing radiation dose from 1 to 10 Gy compared to controls (Figure 2A). The radiation dose to achieve 50% of cell death was 2.4 ± 0.48 Gy ( Figure 2B). We also examined the response RP-MOC1 organoids to radiation (dose range 1-20 Gy). The morphology of the organoids was evaluated at 7 days post radiation treatment. Size (diameter measured in pixels) of the organoids was measured from bright field microscopic images ( Figure 2C). The radiation response curves of RP-MOC1 organoid culture showed the lethal dose (LD 50 ) mean value of 12.61 ± 1.22 Gy ( Figure 2D). Compared to control (102.3 ± 6.37), 20 Gy radiation resulted in a significant reduction in size (52.88 ± 7.25) reflective of growth inhibition (p = 0.004). The difference in diameter between control and 10 Gy radiation (70 ± 15.18) was not statistically significant (p = 0.05).

Radiation Response of RP-MOC1 Cells in 2D Culture and as 3D Organoids
Next, we examined the response of RP-MOC1 cells to radiation in vitro. We exposed RP-MOC1 cells in 2D culture to increasing doses of radiation (0-10 Gy) and evaluated the response using the colony forming assay. At fifteen days post radiation, a 20%-100% reduction in colony formation was observed with increasing radiation dose from 1 to 10 Gy compared to controls (Figure 2A). The radiation dose to achieve 50% of cell death was 2.4 ± 0.48 Gy ( Figure 2B). We also examined the response RP-MOC1 organoids to radiation (dose range 1-20 Gy). The morphology of the organoids was evaluated at 7 days post radiation treatment. Size (diameter measured in pixels) of the organoids was measured from bright field microscopic images ( Figure 2C). The radiation response curves of RP-MOC1 organoid culture showed the lethal dose (LD50) mean value of 12.61 ± 1.22 Gy ( Figure 2D). Compared to control (102.3 ± 6.37), 20 Gy radiation resulted in a significant reduction in size (52.88 ± 7.25) reflective of growth inhibition (p = 0.004). The difference in diameter between control and 10 Gy radiation (70 ± 15.18) was not statistically significant (p = 0.05).

Histopathologic Credentialing of RP-MOC1 Tumors
We performed histopathologic evaluation of the initial subcutaneous tumor and subsequent orthotopic allografts. In addition to hematoxylin and eosin (H&E) staining, tumor sections were immunostained for cytokeratin (Pan-CK), the proliferation marker, Ki-67, the epithelial marker, Ecadherin, and the mesenchymal marker, vimentin.
Representative photomicrographs of H&E and immunostained sections of the initial subcutaneous tumor and subsequent orthotopic RP-MOC1 tumors are shown in Figure 3. Hematoxylin and eosin (H&E) stained sections showed invasive keratinizing moderately differentiated squamous cell carcinoma. The tumor cells showed vesicular to hyperchromatic nuclei with abundant eosinophilic cytoplasm and indistinct cellular outlines. Quantification of Ki-67 staining based on H-score did not reveal a significant difference (p > 0.05) in the proliferation index between subcutaneous (223 ± 56) and orthotopic (301 ± 49) tumors. The connective tissues expressed

Histopathologic Credentialing of RP-MOC1 Tumors
We performed histopathologic evaluation of the initial subcutaneous tumor and subsequent orthotopic allografts. In addition to hematoxylin and eosin (H&E) staining, tumor sections were immunostained for cytokeratin (Pan-CK), the proliferation marker, Ki-67, the epithelial marker, E-cadherin, and the mesenchymal marker, vimentin.
Representative photomicrographs of H&E and immunostained sections of the initial subcutaneous tumor and subsequent orthotopic RP-MOC1 tumors are shown in Figure 3. Hematoxylin and eosin (H&E) stained sections showed invasive keratinizing moderately differentiated squamous cell carcinoma. The tumor cells showed vesicular to hyperchromatic nuclei with abundant eosinophilic cytoplasm and indistinct cellular outlines. Quantification of Ki-67 staining based on H-score did not reveal a significant difference (p > 0.05) in the proliferation index between subcutaneous (223 ± 56) and orthotopic (301 ± 49) tumors. The connective tissues expressed positive staining of vimentin compared to epithelial SCC cells in the initial subcutaneous tumor and orthotopic tumor grafts.

Radiation Response of Orthotopic RP-MOC1 Tumors in Immunocompetent and Immunodeficient Hosts
Building on the in vitro observations, we examined the response of orthotopic RP-MOC1 tumors to radiation in vivo. Non-invasive magnetic resonance imaging (MRI) was performed once every 3-4 days post implantation to assess tumor growth and response to radiation. Images were acquired at 20× magnification. Scale bar: 100 µm.

Radiation Response of Orthotopic RP-MOC1 Tumors in Immunocompetent and Immunodeficient Hosts
Building on the in vitro observations, we examined the response of orthotopic RP-MOC1 tumors to radiation in vivo. Non-invasive magnetic resonance imaging (MRI) was performed once every 3-4 days post implantation to assess tumor growth and response to radiation.
Tumor-bearing albino C57Bl/6 mice were assigned to control or radiation groups~10 days post tumor inoculation. Animals in the radiation arm received 10 Gy (5 daily fractions of 2 Gy) or 15 Gy (5 daily fractions of 3 Gy). The panel of images shown in Figure 4A represent axial T2-weighted images tumor bearing mice over a 30-day period. Tumor volumes were calculated from multi-slice T2-weighted images ( Figure 4B,C). As evident from the images and the quantitative assessment, significant inhibition of tumor growth was seen with radiation at both doses compared to untreated controls. No significant difference in tumor volume was observed between 10 Gy and 15 Gy radiation exposure. Individual tumor growth curves illustrate the heterogeneity in response to radiation ( Figure 4C). A gradual reduction in body weight associated with tumor growth was seen during the study period ( Figure 4D). Tumor-bearing albino C57Bl/6 mice were assigned to control or radiation groups ~10 days post tumor inoculation. Animals in the radiation arm received 10 Gy (5 daily fractions of 2 Gy) or 15 Gy (5 daily fractions of 3 Gy). The panel of images shown in Figure 4A represent axial T2-weighted images tumor bearing mice over a 30-day period. Tumor volumes were calculated from multi-slice T2-weighted images ( Figure 4B,C). As evident from the images and the quantitative assessment, significant inhibition of tumor growth was seen with radiation at both doses compared to untreated controls. No significant difference in tumor volume was observed between 10 Gy and 15 Gy radiation exposure. Individual tumor growth curves illustrate the heterogeneity in response to radiation ( Figure 4C). A gradual reduction in body weight associated with tumor growth was seen during the study period ( Figure 4D).  Comparative evaluation of the radiation response of RP-MOC1 tumors in immunodeficient mice ( Figure 5) revealed no evidence of tumor growth inhibition following radiation treatment at both doses. The panel of images shown in Figure 5A represent T2-weighted images of SCID mice bearing orthotopic RP-MOC1 tumors over a two-week monitoring period. Grouped ( Figure 5B) and individual ( Figure 5C) MR-based tumor volume measurements for animals in the control and radiation groups are also shown. No difference in tumor volume was observed between control and radiated animals, highlighting the role of the immune system in tumor response to radiation.
Comparative evaluation of the radiation response of RP-MOC1 tumors in immunodeficient mice ( Figure 5) revealed no evidence of tumor growth inhibition following radiation treatment at both doses. The panel of images shown in Figure 5A represent T2-weighted images of SCID mice bearing orthotopic RP-MOC1 tumors over a two-week monitoring period. Grouped ( Figure 5B) and individual ( Figure 5C) MR-based tumor volume measurements for animals in the control and radiation groups are also shown. No difference in tumor volume was observed between control and radiated animals, highlighting the role of the immune system in tumor response to radiation.

Discussion
The translation of basic research into improved therapies for head and neck cancer patients will require the development of clinically relevant animal models that can capture the complex interactions between the tumor and the host. In the present study, we report on the establishment, cellular, genetic, and histopathologic credentialing of a new immunocompetent mouse model of OSCC. Our results illustrate the potential translational utility of this novel model as a valuable platform for the conduct of imaging-guided preclinical radiation trials against OSCC.

Discussion
The translation of basic research into improved therapies for head and neck cancer patients will require the development of clinically relevant animal models that can capture the complex interactions between the tumor and the host. In the present study, we report on the establishment, cellular, genetic, and histopathologic credentialing of a new immunocompetent mouse model of OSCC. Our results illustrate the potential translational utility of this novel model as a valuable platform for the conduct of imaging-guided preclinical radiation trials against OSCC.
To generate the OSCC cell line, donor tongue tissue from a 4NQO exposed mouse was stained for the cancer stem cell marker, CD44 and the membrane glycoprotein, EpCAM and subsequently sorted using flow cytometry. These markers are highly expressed in OSCC and have been associated with an invasive and proliferative phenotype [28,29]. Mutational profiling of the cell line revealed a characteristic tobacco-associated signature (C:G > A:T transversions) that has been reported in smokers [30,31]. Exome sequencing revealed that RP-MOC1 cells harbor common somatic mutations of TP53, NFE2L2, CSMD3, STEAP4, UNC13C and NOTCH2 similar to human head and neck cancer. In addition to mutation of the bonafide tumor suppressor, TP53, we identified SNP missense mutation in NFE2L2, the loss of which reflects dysregulated oxidative stress [32,33]. STEAP4 is a metalloreductase known for its role in cell proliferation and metabolism that has been shown to be overexpressed in OSCC [34]. While UNC13C and STEAP4 mutations have been found human head and neck cancers, their role in oral carcinogenesis remains unclear. Similarly, deletion of CSMD1 gene has been associated with lymph node metastasis and poor prognosis in several cancers and is seen commonly in smokers [35]. We detected the less common NOTCH2 mutation in RP-MOC1 cells. Hijoka et al. have previously shown NOTCH2 overexpression in OSCC compared to normal oral mucosa, highlighting its potential role in oral tumorigenesis [36]. GSEA pathway analysis of SNP variant genes revealed association with epithelial-to-mesenchymal transition (EMT) consistent with the proliferative, migratory and invasive phenotype of RP-MOC1 cells in vitro.
Radiation therapy remains an integral part of standard of care for head and neck cancer patients [4,37]. We therefore examined the radiation response of RP-MOC1 cells in 2D and 3D organoid cultures. The LD50 for RP-MOC1 cells in 2D culture was 2 Gy (single dose) while 3D organoid cultures exhibited a higher LD50. Our observations are consistent with a previous study by Storch et al. [38] in which 3D cultures of head and neck and lung cancers exhibited reduced DNA double strand breaks and increased survival as a result of cell-cell and cell-matrix interactions [39]. Building on the in vitro observations, we investigated the histology, growth kinetics and response of our model in vivo. Consistent with the proliferative and invasive phenotype observed in vitro, RP-MOC1 tumors grew orthotopically in the floor of the mouth and demonstrated invasive squamous histology with positive immunostaining for pan-cytokeratin and Ki67. In vivo radiation response studies performed in orthotopic RP-MOC1 tumors established in immunocompetent C57Bl/6 showed significant tumor growth inhibition on MRI examination following 10 Gy and 15 Gy fractionated radiation regimens. Given the well-recognized immune-mediated effects of radiation, we examined the radiation response of RP-MOC1 tumors in immunodeficient mice. Non-invasive MRI showed no evidence of antitumor response with the two regimens in SCID mice. This observation is consistent with a previous work by Liang et al. on the role of host immune responses in tumor response to radiation [40]. In the study, Liang and colleagues demonstrated prolonged stable responses or partial responses as a result of both tumor cell proliferation and immune cell infiltration resulting in 'radiation-induced tumor equilibrium'. Consistent with our observations, the study demonstrated a complete lack of radiotherapeutic efficacy in SCID mice compared to immunocompetent BALB/c mice.
Our observations should be interpreted in the context of the limitations of our study. While preclinical models are essential tools for cancer research, no experimental tumor model can fully recapitulate the disease characteristics seen in OSCC patients. Syngeneic models enable conduct of studies in immunocompetent hosts. However, the murine origin of the tumors needs to be considered. The observed radiation response in our model system may not directly reflect the sensitivity of human OSCC to radiation. In this regard, we have previously examined the radiation response of patient derived xenograft (PDX) models of OSCC [9]. However, human tumor xenograft models necessitate the use of immunodeficient hosts. While our orthotopic tumor model recapitulates the aggressive growth characteristics of the disease, the stromal and vascular components in syngeneic models and human tumor xenograft models are of murine origin. Indeed, no animal model can fully recapitulate human tumors in terms of their biology or therapeutic response. Oral cancers represent a heterogeneous group of neoplasms with varying biological behavior. In our study, we successfully established and credentialed the response of one sub type of these cancers, namely, tongue cancer. We therefore submit that preclinical investigation into strategies aimed at sensitizing tumors to radiation should include systematic evaluation in syngeneic murine models and human OSCC xenograft models that account for the biological heterogeneity of this patient population. In this regard, our 3D organoid platform could be used as a first step for in vitro screening of potential 'radiosensitizers' prior to in vivo evaluation. Our in vivo orthotopic model could also serve as a valuable tool to examine the activity of novel immunotherapeutic strategies against this disease. While immune checkpoint inhibitors have recently received approval for clinical use in head and neck cancer patients [41], the optimal dose, schedule and sequence of combining checkpoint inhibitors with standard of care chemoradiation regimens is still unclear [42,43]. As such, several ongoing trials are investigating the combination of these immune-oncology agents with chemotherapy, radiation and targeted therapies [43]. Given the costs associated with large scale randomized trials, our model could be useful in the conduct of imaging-guided preclinical trials of these immunomodulatory agents with radiation. The model could also be used to understand mechanisms of resistance to radiation and immune checkpoint inhibitors. To address some of these questions, we have begun developing additional murine OSCC lines and will report on our findings in the future.

Establishment of the RP-MOC1 Cell Line and Culture Conditions
To induce oral lesions, female C57BL/6NCr mice (Charles River, Wilmington, MA, USA) were exposed to the carcinogen 4NQO (Sigma-Aldrich, St. Louis, MO, USA) in drinking water for 16 weeks. An exophytic tongue lesion from a donor mouse was transplanted subcutaneously into a naïve recipient animal. When the resultant tumor reached a volume of~200 mm 3 , the tissue was excised, enzymatically dissociated and digested in DMEM 1× media containing 1% of penicillin-streptomycin (Life Technologies, Carlsbad, CA, USA), and 5 mg/mL of collagenase type IV (Life Technologies) for 3 h. After digestion, the cells were trypsinized and washed with cell culture media consisting of DMEM 1× with 10% fetal bovine serum and 1% penicillin-streptomycin (Life Technologies). Finally, the cells were filtered through a 40 µm cell strainer and the cell suspension was seeded in culture flask with cell culture media at 37 • C, 5% CO 2 for 7 days prior to cell sorting using flow cytometry.

Fluorescence Activated Cell Sorting (FACS)
One million cells were labeled with a cocktail of antibodies containing CD44 BV421 (clone IM7, BD Biosciences, San Jose, CA, USA) and CD326 PE (clone G8.8, BD Biosciences, San Jose, CA, USA) for 30 min at 4 • C. Isotype BV421 (Rat IgG2b, κ; BD Biosciences) and isotype PE (IgG2a, κ; BD Biosciences, San Jose, CA, USA) were used as controls to determine the separation between positive and negative cells. Compensation was performed using single cells labeled with appropriate CD44 BV421 (clone IM7, BD Biosciences) and CD326 PE (clone G8.8, BD Biosciences, San Jose, CA, USA), respectively. Epithelial cells, defined by their positive epithelial cell adhesion molecule (CD326; EpCAM) and CD44 expression were sorted (FACS ARIA, BD Biosciences, San Jose, CA, USA) and a minimum of 500,000 cells were collected. Sorted cells were cultured immediately in cell culture media at 37 • C, 5% CO 2 . Flow data were analyzed using FCS Express 6.0 (De Novo Software, Pasadena, CA, USA).

Authentication and IMPACT Testing of RP-MOC1
To verify the species origin and mycoplasma contamination of RP-MOC1, the cells were sent to IDEXX Laboratories, Minnesota, USA for STR DNA fingerprinting and PCR species evaluation performed using the CellCheckTM Mouse kit.

Exome Sequencing
Genomic DNA was prepared from RP-MOC1 cells in culture (Qiagen DNeasy Blood & Tissue Kit, Hilden, Germany) and subjected into Illumina libraries according to the manufacturer's protocol (Illumina Inc, San Diego, CA, USA). Mutation reads were normalized to the reference C57BL/6NCr genome (mm10 genome) and a genomic database (GENCODE GRCm38.75) of 18 commonly used strains of inbred laboratory mice. Mutational data on human OSCC was obtained from publicly available data in TCGA; cBioPortal (http://www.cbioportal.org/public-portal/). The identified SNPs were subjected to Gene Set Enrichment Analysis (GSEA) pathway analysis [44].

Wound Healing Assay
Wound healing migration assay was performed using a 35-mm µ-Dish (ibidi GmbH, Munich, Germany). A total of 70 µL of 35,000 cells were seeded into each chamber of the cell culture insert for overnight. The next day, cells were treated with 5 µg/mL of Mitomycin-C (Calbiochem, San Diego, CA, USA) for 2 h at 37 • C before the cell culture insert was gently removed with sterile forceps. The cultures were washed twice with PBS. Then, 2 mL of cell culture media was added into the 35 mm µ-Dish. The cells that migrated into the denuded area were captured with microscope at 4× magnification. Two independent experiments were performed for this assay.

Transwell Migration Assay
Cells were tested in transwell migration assay using cell culture inserts with PET membrane of 8 µm pore sizes, according to the manufacturer's protocol (Corning, Tewksbury, MA, USA). Briefly, cells were cultured and serum starved for 24 h. Cells were then harvested and suspended in serum free cell culture media at a concentration of 3.0 × 10 5 cells/mL. In the 24 well plate, inserts were placed into wells containing 500 µL of cell culture media that acted as a chemoattractant, and followed by 200 µL of serum starved cells were seeded onto the insert for 12 and 24 h, respectively. Post 12-and 24-h incubation, the non-migrated cells were first eliminated by scraping with wet cotton swabs, while the bottom of the membrane was fixed and stained with 0.1% crystal violet (Sigma-Aldrich, St. Louis, MO, USA) in 20% methanol for 2 h at room temperature. Membranes were washed with water to remove excessive stained and air dried. The membrane was then viewed and captured under the microscope at 20× magnification. The number of stained cells were counted in 4 randomly chosen microscopic fields and averaged. Three independent experiments were performed with 2 replicates.

Matrigel Invasion Assay
Experiments were carried out as described in the transwell migration assay except that the inserts used were pre-coated with Matrigel basement matrix from Biocoat Matrigel 24 well invasion chamber (Corning, Tewksbury, MA, USA). Briefly, cells were cultured and serum starved for 24 h. Cells were then harvested and suspended in serum free cell culture media at a concentration of 5.0 × 10 5 cells/mL. In the 24 well plate, inserts were rehydrated with serum free cell culture media for 2 h before they were placed into wells containing 750 µL of cell culture media that acted as chemoattractant. This was followed by seeding 500 µL of serum starved cells onto the insert for 12 and 24 h. Post 12-and 24-h incubation, non-migrated cells were scraped by wet cotton swab, while the bottom of the membrane was fixed and stained with 0.1% crystal violet (Sigma-Aldrich, St. Louis, MO, USA) in 20% methanol for 2 h at room temperature. Membranes were washed with water, air dried and microscopic images captured at 20× magnification. The number of stained cells were counted in 4 randomly chosen microscopic fields and averaged. Three independent experiments were performed with 2 replicates.

Clonogenic Assay
Five hundred thousand cells were plated to five different T-25 flasks. Next day, the T-25 flasks were radiated by the irradiator (Shepherd Mark I Model 68, 4000 Ci Cesium 137 source irradiator, J.L. Shepherd and Associates), with 1, 2, 4, 10 Gy at position 2. Post 24 h, 120 cells were plate to 60 mm petri dish. Cell culture media was refreshed every 3 days. After 15 days of culture, the colonies were fixed and stained with 0.1% crystal violet (Sigma-Aldrich, St. Louis, MO, USA) in 20% methanol for 2 h at room temperature. Colonies were counted manually. Three independent experiments were performed in triplicate. . The difference in luminescent signal post radiation normalized to control (0 Gy) was used as the measurement of cell viability. Cell death was calculated as the difference in viable cells from 100% viable control. The radiation dose that able to achieve 50% cell death (EC50) was determined from absorbance signal versus concentration curve using GraphPad software by applying the nonlinear regression and the equation log (agonist) vs. normalized response. All experiments were carried out in triplicate and performed in three independent experiments. Bright field microscopy images of the organoid were captured and the diameter of the control (0 Gy), 10 Gy and 20 Gy were measured using cellSens Standard 1.6 software (Olympus, Center Valley, PA, USA).

In Vivo Radiation Studies
Experimental studies were carried out using eight-to-twelve week old female C.B 17 severe combined immunodeficient (SCID) (C.B-Igh-1 b /IcrTac-Prkd scid ; Laboratory Animal Shared Resource, Roswell Park) and C57BL/6NCr mice with an average body weight~20 g. Mice were kept in sterile micro-isolator cages (4-5 mice per cage) in a pathogen-free environment and provided with standard chow/water and maintained on 12 h light/dark cycles in a HEPA-filtered environment. Animals were injected with 1 × 10 6 RP-MOC1 cells orthotopically in the floor of mouth. Seven to ten days post injection, non-invasive magnetic resonance imaging (MRI) was performed to confirm tumor growth. Tumor-bearing mice were randomized into control (n = 5, untreated) or one of two radiation arms (10 Gy; 5 doses of 2 Gy radiation delivered on consecutive days (n = 5); 15 Gy; 5 doses of 3 Gy radiation delivered on consecutive days (n = 5). Irradiation was performed using the Philips RT 250 Orthovoltage Cancers 2020, 12, 579 12 of 16 X-ray unit (Philips Medical Systems, Andover, MA, USA). Radiation was delivered through an axial beam directed to the tumor in the floor of the mouth. A lead shield was utilized to protect the thoracic cavity from exposure to radiation. Body weights of the animals were measured once every 3 days throughout the duration of the study as a measure of toxicity. Experimental procedures were performed under aseptic conditions and in accordance with protocols approved by the Institutional Animal Care and Use Committee (Protocol #1183M; Original approval: July 2017; Renewal; September 2019; Animal welfare assurance number: A-3143-01.

Magnetic Resonance Imaging
Longitudinal MRI examinations were performed using a 4.7T/33-cm horizontal bore magnet (GE NMR Instruments, Fremont, CA, USA) incorporating AVANCE digital electronics (Bruker Biospec with Paravision 3.0.2; Bruker Medical Inc., Billerica, MA, USA). Animals were anesthetized using 2.5% Isoflurane (Benson Medical Industries, Markham, ON, Canada) prior to and during imaging. Tumor volumes were calculated from multi-slice, axial T2-weighted spin echo images incorporating RARE (rapid acquisition with relaxation enhancement) [45].

Histology and Immunohistochemistry
Mice were humanely euthanized according to recommendations of the Panel on Euthanasia of the American Veterinary Medical Association and tumor tissues resected for further processing. Tumor tissues were fixed in 10% neutral buffered formalin (Sigma-Aldrich, St. Louis, MA, USA) and embedded in paraffin. Tissues were sectioned at a thickness of 4 µm, mounted on positively charged slides, and stained for hematoxylin and eosin (H&E) stain and immunohistochemistry. Immunohistochemistry was performed on FFPE sections of the tissues using the Envision technique, Dako Real EnVision Detection System and Peroxidase/DAB+ (Dako Corporation, Carpinteria, CA, USA) according to the manufacturer's protocol (Table S4). Briefly, FFPE sections were deparaffinized in xylene and rehydrated in ethanol series. Antigen retrieval was carried out using pre-heat steamer for 30 min. The sections were immersed in blocking solution (Dako Corporation, Carpinteria, CA, USA) for 10 min at room temperature followed for blocking the endogenous peroxidase activity. The sections were then incubated with serum free blocker (Dako Corporation, Carpinteria, CA, USA) for an hour at room temperature followed by 30 min to an hour incubation of primary antibody at room temperature (Pan cytokeratin, Abcam, Cambridge, UK, Catalog #ab9377; E-cadherin, Cell Signaling Technology, Danvers, MA, USA, Catalog #3195; Vimentin, Boster Biological Technology, Pleasanton, CA, USA, Catalog # PB9359) Ki-67, R&D Systems Inc., Minneapolis, MN, USA Catalog #MAB7617). After washing with TBS (pH 7.4) plus 0.1% Tween 20 (Bio-Rad Laboratories, Hercules, CA, USA), sections were incubated with the peroxidase labeled secondary antibody from the Envision kit for an hour for the immunoreactivity performances. Finally, sections were stained with 3 3 diaminobenzidine substrate chromogen, (Dako Corporation, Carpinteria, CA, USA) counterstained with Harris hematoxylin, dehydrated and mounted. Images were captured at 20× magnification. Ki-67 stained specimens were captured at 40× magnification and eight random areas were analyzed using NIH Image J software (NIH, Bethesda, MD, USA). Intensity of the stain was classified into 0 = negative/no stain, 1 = weak, 2 = moderate and 3 = strong. For each image, an observer who was blinded to the identity of the samples assessed 200 cell nuclei for staining intensity and assigned a value from 0 to 3. The H-scores for Ki-67 were calculated based on the sum of scores from each sample between subcutaneous and orthotopic tumor samples.

Sample Sizes and Statistics
All statistical analysis was performed using GraphPad version 7.00 for Windows (GraphPad Software, San Diego, CA, USA). The RP-MOC1 in vitro doubling time was determined by least-squares fitting of single-exponential curves to the cell viability against time (day). The comparison between the 12-h and 24-h migrated and invaded cells from transwell migration and invasion assays were compared using unpaired Student's t-test. H-scores from Ki-67 immunostained sections of subcutaneous and orthotopic tumors were also analyzed using unpaired Student's t-test. Non-invasive imaging was performed using 4-5 animals per cohort. Comparisons of organoid for 0 Gy, 10 Gy and 20 Gy and animal tumor volume across cohorts were analyzed using one-way ANOVA test. All results are expressed as mean values ± SEM of at least three independent experiments, except when otherwise indicated. p-values of <0.05 were considered statistically significant.

Conclusions
We have successfully established and credentialed a novel immunocompetent model of OSCC that exhibits a mutational and histologic profile similar to human disease. The model can serve as a valuable platform for evaluation of combination strategies to enhance radiotherapeutic efficacy against this deadly disease.