Ablation of Cancer Stem Cells by Therapeutic Inhibition of the MDM2–p53 Interaction in Mucoepidermoid Carcinoma

Salivary gland mucoepidermoid carcinomas are the most common salivary malignancy, accounting for 30% to 35% of all malignant salivary gland tumors (1–8). Patients diagnosed with mucoepidermoid carcinomas are treated using surgical resection and radiotherapies (9). Although surgery is often sufficient for patients with low- to intermediate-grade tumors, patients with high-grade tumors often display recurrent disease (10). As mucoepidermoid carcinomas are resistant to chemotherapy, treatment of patients with locoregional recurrence involves radiotherapy, but the long-term prognosis of these patients is modest (10, 11). Improved understanding of the pathobiology of mucoepidermoid carcinoma is essential for the identification of more effective therapies that prevent locoregional recurrence and enhance the survival of patients with mucoepidermoid carcinoma.

A subpopulation of cells, termed cancer stem cells (CSC), is uniquely tumorigenic and capable of multipotency and self-renewal. CSCs are essential for both initiating and maintaining the growth of many cancer types (12). Importantly, CSCs are resistant to conventional chemotherapies and radiation treatments due to their slow proliferation, the presence of transporter proteins that facilitate cell detoxification, as well as microenvironmental influences that enhance their survival (13–15). It has been postulated that survival of CSCs after treatment allows for tumor regrowth and recurrence (16–17). We have recently identified functionally the existence of CSCs in salivary gland mucoepidermoid carcinomas (18). Mucoepidermoid carcinoma cells sorted for ALDH^highCD44^high exhibit self-renewal and multipotency, and are highly tumorigenic when compared with control cells. Selective targeting and ablation of ALDH^highCD44^high cells might be a novel approach to prevent locoregional recurrences in patients with mucoepidermoid carcinoma.

p53 functions as a transcription factor that plays an essential role in many cellular functions, including regulation of cell cycle and senescence, as well as induction of apoptosis upon the genomic damage (19–21). The function of p53 in adult and embryonic stem cell differentiation has been described more recently. In normal differentiated cells, deletion or repression of p53 function leads to greater efficiency in dedifferentiating into induced pluripotent stem cells (22–26). In the context of cancer, research has shown that loss of p53 also led to the expansion of malignant reprogrammed progenitor cells in liver cancer (27). Mouse double minute (MDM) 2 regulates p53 by binding to it and signaling its degradation by the proteasome (28). Indeed, MDM2 is an oncogene that is overexpressed in many cancers, including salivary gland tumors (29).

Over the past few years, small-molecule inhibitors of the MDM2 binding to p53 have been developed. One small molecule in particular, MI-773, has been shown to have significantly greater specificity and affinity to MDM2 when compared with p53 binding, Nutlin-3a binding, and previous analogue MI-219 binding (30). Importantly, MI-773 was able to activate p53 activity and induce apoptosis in cancer cells at a much greater efficiency when compared with the standard, nutlin-3a (30). Notably, this therapeutic effect was seen in cell lines with both wild-type and mutated p53 (31), as long as they retain an additional wild-type copy of the p53 gene.

Although the therapeutic effect of MI-773 in inducing tumor cell apoptosis and xenograft tumor regression has been characterized, the effect of MI-773 on CSCs has yet to be determined. The focus of our study is to investigate the therapeutic effect of MI-773 on the CSCs using mucoepidermoid carcinomas as a model of a tumor that follows the CSC hypothesis (18). Our findings demonstrate that therapeutic inhibition of the MDM2–p53 interaction reduces the fraction of ALDH^highCD44^high CSCs in mucoepidermoid carcinomas, in addition to inducing apoptosis and cell-cycle arrest. Together, these findings unveil inhibitors of MDM2 as a novel strategy for ablation of CSCs.

A panel of University of Michigan Human Mucoepidermoid Carcinoma cells (UM-HMC-1, UM-HMC-3A, and UM-HMC3B), generated and characterized in our laboratory (32), were cultured using DMEM (Gibco) supplemented with 10% FBS (Gibco), l-glutamine (Gibco), penicillin and streptomycin (Gibco), 20 ng/mL EGF (Sigma), 400 ng/mL hydrocortisone (Sigma), and 5 μg/mL insulin (Sigma). Primary human microvascular endothelial cells (HDMEC; Lonza) were grown using the EBM2-MV cell culture medium (Lonza).

Tissue section slides for IHC and immunofluorescence staining were deparaffinized using xylene and ethanol, and then incubated with 0.1% Triton X-100 (Sigma) for 10 minutes, 3% hydrogen peroxide for 10 minutes, and Background Sniper (Biocare Medical) for 10 to 30 minutes. IHC slides were incubated with 1/100 monoclonal anti-human p53 antibody (Santa Cruz Biotechnology, catalog no. #sc-126) or 1/200 monoclonal anti-human MDM2 antibody (Santa Cruz Biotechnology, catalog no. #sc-965) overnight at 4°C. Following incubation with primary antibody, sections were washed and then incubated with MACHI 3 Probe (Biocare Medical) for 20 minutes. After two 10-minute washes, sections were incubated with MACHI 3 HRP polymer for 20 minutes and again washed twice for 10 minutes. Sections were then incubated with DAB for 3 minutes and quenched in water. Sections were finally incubated with hematoxylin for 45 seconds, dehydrated in ethanol, and mounted. Immunofluorescence sections were incubated with 1/50 monoclonal anti-human ALDH1 antibody (Abcam, catalog no. #ab52492) overnight at room temperature. The following day, sections were washed and incubated with Alexa Fluor 488 (anti-rabbit; Invitrogen). Then, sections were incubated with 3% hydrogen peroxide for 30 minutes followed by 1-hour incubation with ready-to-use monoclonal anti-human CD44 antibody (Thermo Fisher Scientific, catalog no. #MA5-13890) and a 20-minute incubation with Alexa Fluor 594 (anti-mouse; Invitrogen). Nuclei were stained with DAPI. TUNEL staining was performed using in situ TUNEL staining Kit (Roche). Seven 200× image fields were taken per tissue section and pixel density was quantitated using ImageJ software.

UM-HMC cells and xenograft mucoepidermoid carcinoma tissue lysates were prepared using an NP40-based lysis buffer. Lysates were run using PAGE gels and probed using 1/1,000 monoclonal anti-human p53 (Santa Cruz Biotechnology, catalog no. #sc-126), 1/500 monoclonal anti-human MDM2 (Santa Cruz Biotechnology, catalog no. #sc-965), 1/500 monoclonal anti-human p21 (Cell Signaling Technology, catalog no. #2947), 1/1,000 monoclonal anti-human Bmi-1 (Cell Signaling Technology, catalog no. #5856), 1/1,000 monoclonal anti-human Cyclin-A (Santa Cruz Biotechnology, catalog no. # sc-751), 1/500 monoclonal anti-human Cyclin-D (Santa Cruz Biotechnology, catalog no. # sc-753), 1/1,000 monoclonal anti-human Cyclin-E (Cell Signaling Technology, catalog no. #4129), 1/1,000 monoclonal anti-human CDK2 (Santa Cruz Biotechnology, catalog no. #sc-6248), 1/500 monoclonal anti-human CDK4 (Santa Cruz Biotechnology, catalog no. # sc-260), 1/1,000 monoclonal anti-human CDK6 (Santa Cruz Biotechnology, catalog no. # sc-177), 1/1,000 monoclonal anti-human Notch-1 (Cell Signaling Technology, catalog no. #3608s), 1/1,000 monoclonal anti-human Notch-2 (Cell Signaling Technology, catalog no. #4530s), 1/1,000 monoclonal anti-human Notch-3 (Cell Signaling Technology, catalog no. #5276s), 1/1,000 monoclonal anti-human Oct-4 (Cell Signaling Technology, catalog no. #2750s), 1/1,000 monoclonal anti-human Nanog (Santa Cruz Biotechnology, catalog no. # sc-293121), and 1/1,000 monoclonal anti-human Bcl-x_L (BD Transduction, catalog no. # 610747).

RNA was isolated from sorted and unsorted UM-HMC cells and reverse transcribed. cDNA was PCR-amplified using sense and anti-sense primers targeting full-length p53, as well as residues 54-716, 460-1179, and 876-1412 (33). PCR products were run on a 1.5% agarose gel; fragments were excised and purified for Sanger sequencing performed by the University of Michigan Sequencing Core. In addition, we isolated genomic DNA using the Wizard genomic DNA Purification Kit (Promega) and performed Sanger sequencing to evaluate the status of p53 in the UM-HMC cell lines.

HEK293T cells were transiently cotransfected with the lentiviral packaging vectors psPAX2, pMD2G, and shRNA-p53 (seq#1: TACACATGTAGTTGTAGTG, seq#2: TAACTGCAAGAACATTTCT) or scramble sequence control (shRNA-C; Vector Core, University of Michigan, Ann Arbor, MI) by the calcium phosphate method. UM-HMC-3A cells were infected with supernatants containing lentivirus and selected with 1 μg/mL of puromycin (Sigma-Aldrich) for at least 1 week. Knockdown of p53 was verified by Western blot analysis.

UM-HMC cells were plated at a density of 500 cells/well in a 96-well plate and allowed to attached overnight. MI-773 was solubilized in DMSO and added at increasing concentrations (0.01–100 μmol/L) for 24 to 72 hours (100 μmol/L DMSO served as vehicle control). WST-1 reagent (Roche) was added to cells and incubated at 37°C for 4 hours, and plates were analyzed in a microplate reader (GENious; Tecan).

UM-HMC cells were exposed to 1 μmol/L MI-773 or 1 μmol/L DMSO (vehicle control), for 48 to 96 hours. For ALDH/CD44 staining, single-cell suspensions of 2 × 10⁶ cells/tube were incubated in ALDH substrate (Cayman Chemicals), or ALDH inhibitor diethylaminobenzaldehyde (DEAB; Cayman Chemicals), for 40 minutes at 37°C, as described previously (16). Then, cells were washed and exposed to 1:20 anti-CD44 (APC; BD Pharmingen) for 30 minutes at 4°C. At least 10,000 events were analyzed per cell type and experimental condition at the University of Michigan Flow Cytometry Core. Results were analyzed using FlowJo software (FlowJo, LLC). For propidium iodine staining, cells were fixed in 70% ethanol overnight at −20°C and stained in 1 mg/mL propidium iodide (BD Biosciences), 1% sodium citrate (Thermo Fisher Scientific), 1 mg/mL RNase A (Sigma), 10% Triton X-100 (Thermo Fisher Scientific) for 20 minutes. For Annexin V staining, cells were washed in PBS then suspended in 1× Binding Buffer (BD Biosciences). Annexin V (APC; BD Biosciences) was added to 1 × 10⁵ cells and incubated for 20 minutes at room temperature. In all flow cytometry experiments, 7AAD (BD Biosciences) was used as a live/dead control. Data were obtained from triplicate wells and represent at least 3 independent experiments.

The Human Apoptosis Array (catalog no. #ARY009; R&D Systems) was used to assess the expression of apoptosis-associated proteins in vehicle versus MI-773–treated cells. UM-HMC-3A cells (2 × 10⁵) were seated and treated with 5 μmol/L MI-773 or 5 μmol/L DMSO (vehicle control) for 24 hours. Following treatment, the cells were lysed and analyzed according the company protocol. Array blots were quantitated using ImageJ software.

Highly porous poly-l-lactic acid scaffolds were seeded with 6 × 10⁵ UM-HMC-3A or UM-HMC-3B cells together with 4 × 10⁵ primary human HDMECs (Lonza) in a 1:1 mix of growth factor–reduced (Matrigel; Corning) and EGM2-MV (Lonza), as described previously (18, 32, 34). Scaffolds were implanted in the subcutaneous space of the dorsal region of CB-17 SCID mice (Charles River Laboratories). Scaffolds were measured weekly until tumors reached an average volume of 500 mm³. Mice were treated daily with 50 to 200 mg/kg MI-773 or vehicle (polyethylene glycol-200 and D-α-tocopherol polyethylene glycol 100 succinate; Sigma-Aldrich) via oral gavage. After 7 days, mice were euthanized, tumors were removed, and the tissues were digested using the Tumor Dissociation Kit (Miltenyi Biotec, 130-095-929) and the GentleMACS Dissociator (Miltenyi Biotec) following the manufacturer's instructions. Single-cell tumor tissue suspensions were stained for flow cytometry, as described above.

To investigate the patterns of expression of MDM2 and p53 in human mucoepidermoid carcinoma, we performed IHC analyses that revealed strong MDM2 and relatively weaker p53 expression (Fig. 1A). To validate these results in vitro, we performed PCR and Western blot analyses on a panel of human mucoepidermoid carcinoma cell lines (UM-HMC) generated and characterized in our laboratory (32) as follows: UM-HMC-1, primary mucoepidermoid carcinoma from a minor salivary gland of buccal mucosa; UM-HMC-3A, local mucoepidermoid carcinoma recurrence, left hard palate; and UM-HMC-3B, originated from the lymph node metastases of the same patient who donated tissues for UM-HMC-3A. We observed lower p53 mRNA expression in the UM-HMC-3A cells, as compared with UM-HMC-1 and UM-HMC-3B (Fig. 1B). At the protein level, both UM-HMC-1 and UM-HMC-3A cell lines showed strong MDM2 expression and low p53 expression (Fig. 1C). Interestingly, we noticed the opposite pattern in the metastatic UM-HMC-3B cells that showed minimal baseline MDM2 expression and elevated p53 expression (Fig. 1C). We next compared MDM2 and p53 expression in CSCs (ALDH^highCD44^high) and non-CSCs, that is, the pool of remaining cells (control cells). Higher expression of MDM2 was observed in ALDH^highCD44^high cells when compared with control cells, particularly in cells sorted from sorted from the UM-HMC-3B cell line (Fig. 1D). Expression of p53 remained consistent between ALDH^highCD44^high and non-CSCs in both cell lines, but p53 was generally stronger in UM-HMC-3B than in UM-HMC-3A (Fig. 1D). Sanger sequencing suggested the presence of A278P polymorphism in UM-HMC-1, UM-HMC-3A, and UM-HMC-3B cells as well as V157F in UM-HMC-3A and UM-HMC-3B cells with retention of wild-type alleles. Next, we performed exome sequencing to confirm the presence/absence of mutations that would affect the function of p53 in the UM-HMC cell lines. This analysis revealed that all UM-HMC cell lines studied here have wild-type, functional, TP53 gene.

To evaluate the effect of MI-773 on the fraction of mucoepidermoid carcinoma CSCs, we treated UM-HMC-1 cells with 1 μmol/L MI-773 for 48 to 96 hours and performed FACS analysis for ALDH^highCD44^high cells. We observed a significant and consistent decrease in the ALDH^highCD44^high cell population at each time period evaluated (Fig. 2A). To determine whether this decrease was due to selective death (or decrease in proliferation) of the CSC population or differentiation into a non-CSC cell type, we performed a WST-1 analysis. After 72-hour treatment with increasing concentrations of MI-773, we observed no difference in the cytotoxicity curves for ALDH^highCD44^high cells when compared with control non-CSC cells, suggesting that MI-773–mediated decrease in CSCs (Fig. 2A) is not mediated by preferential killing these cells (Fig. 2B). To verify specificity of these results, we repeated this analysis in UM-HMC-3A and UM-HMC-3B cells and observed a significant decrease in the fraction of ALDH^high CD44^high cells upon treatment with MI-773 (Fig. 2C and E). As described for UM-HMC-1 above, we did not observe significant differences in the overall cell density of CSCs and control non-CSC cells (Fig. 2D and F). Interestingly, UM-HMC-3B cells were more resistant to MI-773, with IC_50s about 10-fold higher than the other 2 mucoepidermoid carcinoma cell lines examined here. Together, these results demonstrate that MI-773 reduces the fraction of ALDH^highCD44^high cells in mucoepidermoid carcinoma cells in vitro, and that this effect on CSCs does not correlate with the overall sensitivity of bulk cells to the drug.

To assess the impact of “off-target” effects, we used shRNA vectors to knock down p53 expression in mucoepidermoid carcinoma cells and exposed the cells to MI-773 (Fig. 3A). Whereas shRNA-p53 sequence #1 was effective in reducing p53 expression, shRNA-p53 sequence #2 showed little effect and was used as an additional control (Fig. 3B). Interestingly, we observed a significant reduction in MDM2 and p21 expression in p53-silenced cells (Fig. 3B), suggesting that indeed these two proteins are downstream targets of p53. We observed that the IC₅₀ for MI-773 in p53-silenced cells is approximately 15-fold higher than in vector control cells (Fig. 3A). These results show that p53 silencing enhances resistance of UM-HMC-3A cells to MI-773 treatment and suggests that the therapeutic effect of MI-773 is primarily through the activation of functional p53, and not simply off-target effects.

To determine the overall effect of MI-773 on mucoepidermoid carcinoma cell density and expression of proteins downstream of p53, we performed WST-1 and Western blot analyses on unsorted UM-HMC-1, UM-HMC-3A, and UM-HMC-3B cells treated with MI-773. We observed a consistent and dose-dependent decrease in total cell number in all mucoepidermoid carcinoma cell lines used here (Fig. 3C, E, and G). However, it became clear again that the UM-HMC-3B cells were more resistant to MI-773, with an IC₅₀ at 72-hour treatment that was approximately 72-fold higher than the IC₅₀ of UM-HMC-1 cells, and about 10-fold higher than the IC₅₀ of UM-HMC-3A cells. We performed analysis of covariance on the normalized WST-1% to determine whether cell density differences were based on treatment time controlling for concentration of MI-773 using the R 3.1.0 software (35). For cell lines UM-HMC-1, UM-HMC-3A, and UM-HMC-3B, the MI-773 treatment time was significant, controlling for log-concentration (P < 10⁻⁶; P < 10⁻⁶; P = 0.0114, respectively). We next treated the cells with 0 to 20 μmol/L MI-773 and observed a robust dose-dependent increase in MDM2, p53, and p21 expression, suggesting that p53 signaling is indeed activated in UM-HMC-1 and UM-HMC-3A cells (Fig. 3D and F). In contrast, expression of MDM2, p53, and p21 was not significantly affected by treatment with MI-773 in UM-HMC-3B cells (Fig. 3H). This may begin to explain the resistance that these cells exhibit against inhibition of the MDM2–p53 interaction with MI-773. Surprisingly, we observed a significant reduction of the key regulator of stem cell self-renewal Bmi-1 in the 3 mucoepidermoid carcinoma cell lines evaluated here (Fig. 3D, F, and H). This observation provides a putative mechanism to explain the reduction in CSC fraction upon treatment with MI-773 in all cell lines (Fig. 2A, C, and E).

To further investigate the mechanisms underlying MI-773–mediated effects on mucoepidermoid carcinoma cell density, we treated UM-HMC cells with low-dose MI-773 (1 μmol/L), stained with a hypotonic solution of propidium iodide, and performed flow cytometry. We observed a G₁ cell cycle and a correspondent decrease in the fraction of UM-HMC-3A cells in both S and G₂–M, suggesting that MI-773 did indeed induce G₁ cell-cycle arrest (Fig. 4A). We repeated this experiment with UM-HMC-1 cells and observed a significant decrease in the fraction of UM-HMC-1 cells in S and G₂–M (Supplementary Fig. S1). Indeed, we consistently observed similar trends in cell-cycle experiments performed with UM-HMC-1 and UM-HMC-3A. We next treated UM-HMC-3A cells with increasing concentrations of MI-773 to evaluate the effect of MI-773 on apoptosis via staining with both 7AAD and Annexin-V. We observed an increase of cells positive for Annexin-V alone (early apoptosis) as well as cells positive for both Annexin-V and 7AAD (late apoptosis), suggesting that MI-773 induces apoptosis of UM-HMC-3A cells (Fig. 4B). Then, we repeated these experiments with the resistant, UM-HMC-3B cells. MI-773 did not cause changes in the cell-cycle distribution of UM-HMC-3B cells (Fig. 4C), and did not induce apoptosis (Fig. 4D). To confirm that indeed MI-773 was mediating UM-HMC-3A cell apoptosis (and not nonspecific cell death), we treated UM-HMC-3A cells with MI-773 for 24 hours and ran an Apoptosis Antibody Array (R&D Systems). We observed a significant upregulation (that is more than 2-fold) of several known inducers of apoptosis, such as Bax, cleaved caspase-3, cytochrome c, and cIAPs (Fig. 4E and F). This assay also confirmed the increase in expression of p53 and p21 (Fig. 4E and F) that was observed in the Western blots (Fig. 3). To evaluate the responses with our resistant cell line (i.e., UM-HMC-3B), we treated these cells with MI-773 and ran an Apoptosis Antibody Array (R&D Systems). As expected, we observed very minor changes (i.e., less than 2-fold) in expression of the apoptosis-inducing proteins in UM-HMC-3B cells exposed to MI-773 (Fig. 4G and H). Collectively, these results demonstrate that treatment with MI-773 induces cell-cycle arrest and apoptosis in UM-HMC-3A, but not in the resistant UM-HMC-3B cells.

To determine the effect of therapeutic inhibition of the MDM2–p53 interaction with MI-773 on tumor cell apoptosis and fraction of CSCs in vivo, we transplanted mucoepidermoid carcinoma cells (UM-HMC-3A or UM-HMC-3B) with HDMECs on a biodegradable scaffold in mice, as described previously (18, 34). Once tumors reached an average volume of 500 mm³, mice were treated daily with 100 mg/kg MI-773 by oral gavage. Treatment was performed for a week only, as these experiments were focused on the effects of MI-773 in tumor cell apoptosis and fraction of CSCs (not its effects on tumor size). We observed modest and nonstatistically significant (P > 0.05) tumor regression in xenografts generated with UM-HMC-3A cells (Fig. 5A) and in the UM-HMC-3B tumors (Fig. 5B) in mice treated with MI-773. To determine the systemic cytotoxicity of MI-773, we monitored mouse weight during treatment and observed that mice in the treatment group had modest weight loss (Fig. 5C and D).

As an initial readout for MI-773 activity in vivo, we examined the effect of this small-molecule inhibitor on the expression of MDM2 and p53. IHC revealed a qualitative increase in expression of MDM2 and p53 upon treatment with MI-773 (Fig. 5E). In attempt to quantify these differences, we performed Western blots from tumor tissues retrieved from mice that received MI-773 or vehicle. We observed a substantial increase in the expression of MDM2, p53, and p21 in UM-HMC-3A tumors treated with MI-773. In contrast, UM-HMC-3B tumor showed modest upregulation of MDM2 and p53, whereas no changes in p21 expression were observed (Fig. 5F).

Tissue sections were stained for in situ TUNEL to determine the effect of MI-773 on apoptosis in vivo. Seven standard images were taken from each tumor (n = 3 tumors/experimental condition) and pixel density of apoptotic cells was quantitated using ImageJ software. We observed a significant increase in TUNEL-positive cells in the MI-773–treated UM-HMC-3A tumors, when compared with the vehicle-treated tumors (Fig. 5G). In contrast, MI-773 did not mediate a significant increase in the number of apoptotic cells in the UM-HMC-3B tumors treated with MI-773 (Fig. 5H).

To determine whether a higher dose of MI-773 would be more effective in UM-HMC-3B tumors, we repeated these experiments treating 5 mice (10 tumors) with a higher dose of MI-773 (i.e., 200 mg/kg) as well as 5 mice (10 tumors) with a vehicle control. Again, we did not observe a significant reduction in tumor volume (Supplementary Fig. S2A). But, in this case, we had to discontinue treatment after 5 days as average mouse weight dropped to about 80% of pretreatment average (Supplementary Fig. S2B). As before, we observed visible upregulation of p53 expression, that was accompanied by decreased expression of G₁ cell cycle–associated proteins Cyclin A, Cyclin D1, Cyclin E, CDK2, CDK4, and CDK6 in the MI-773–treated tumors when compared with the vehicle-treated tumors (Supplementary Fig. S2C and S2D). We have also confirmed that MI-773 decreases the presence of ALDH^high CD44^high cells, but does not increase the number of apoptotic cells in UM-HMC-3B tumors (Supplementary Fig. S2E and F).

To define the effect of MI-773 on the fraction of CSCs (i.e. ALDH^highCD44^high cells), we performed immunofluorescence studies using the combination ALDH and CD44 as markers of CSCs in mucoepidermoid carcinoma, as we showed (18). We observed that MI-773 caused a significant decrease in the fraction of CSCs in both, UM-HMC-3A and UM-HMC-3B tumors (Fig. 6A). These data were confirmed by flow cytometry for ALDH activity and CD44 expression from cells retrieved from UM-HMC-3A and UM-HMC-3B tumors treated with MI-773 or vehicle control (Fig. 6B). This is consistent with the in vitro data that also showed significant reductions in the CSC fraction upon treatment with MI-773 (Fig. 2). Interestingly, UM-HMC-3B tumors appear to have a higher baseline fraction of ALDH^highCD44^high cells as compared with UM-HMC-3A tumors (Fig. 6A and B). In an attempt to understand a possible mechanism that explains the decrease in CSCs mediated by MI-773, we performed Western blots in both UM-HMC-3A and UM-HMC-3B–treated tumor tissue lysates that revealed that MI-773 mediates a substantial reduction in expression levels of the key regulator of self-renewal Bmi-1 (Fig. 6C and D). And finally, we observed that MI-773 mediated dose-dependent inhibition of some stem cell transcription factors (i.e., Notch-1, Notch-2), but not Nanog or Oct-4, in mucoepidermoid carcinoma cells (Fig. 6E). Collectively, these data demonstrate that therapeutic inhibition of MDM2 mediates a decrease in the fraction of ALDH^highCD44^high cells in vivo that is correlated with a decrease in the expression of Bmi-1, a key regulator of stem cell self-renewal.

Resistance to chemotherapy and radiation treatments poses a major challenge to effectively treating patients with advanced stage mucoepidermoid carcinomas. Previous research indicates that CSCs play a critical role in resistance to therapy and disease recurrence (12, 36–37). Studies from our laboratory identified CSCs in salivary gland mucoepidermoid carcinoma (18). This aggressive and uniquely tumorigenic subpopulation of cells can be identified as ALDH^highCD44^high cells. Importantly, we found that ALDH^highCD44^high cells are able to self-renew and differentiate into non-CSCs. As CSCs are resistant to treatment, novel therapies aiming the ablation of this uniquely tumorigenic population of CSCs are being proposed. Already, much research has been done to target stem cell–associated pathways aiming to selectively eliminate the CSC population in other cancer types. Such pathways include Notch, Wnt, Hedgehog, IL6, Her-2, and PI3K/AKT (38–41). For many of these pathways, lack of specific targeted agents poses a barrier to effective elimination of the CSC population. Here, we present studies that provide the conceptual framework for targeted inhibition of the MDM2–p53 interaction as a strategy to ablate CSCs in mucoepidermoid carcinomas.

The tumor suppressor protein p53 plays a critical role in regulating the cell cycle and senescence as well as inducing apoptosis upon oncogenic stress. Importantly, research also suggests that p53 plays an important role in normal stem cell function by actively initiating or repressing several stem cell–associated proteins such as Nanog and Oct-4, making it an intriguing pathway to study in the context of CSCs (42). Studies suggest that in the absence of p53, both normal and tumor cells acquire dedifferentiated phenotypes (22–27). MDM2, the main regulator of p53, functions as a E3 ligase signaling p53 for degradation and can also bind p53 to block the transactivation domain of the protein. Inhibition of MDM2 binding to p53 prevents its degradation and activates p53 to function in the cell. Many groups have sought to develop a small-molecule inhibitor to block the MDM2/p53–binding interaction. However, many of these inhibitors lack specificity or clinical translatability to be effective therapeutically (43–47).

One promising inhibitor of the MDM2–p53 interaction, MI-773, is significantly more specific and shows improved antitumor efficacy when compared with other small-molecule inhibitors (30). Upon treatment with MI-773, researchers observed a significant induction of p53 signaling in a variety of different cancer types (30). Activation of p53 induced apoptosis, which in turn caused significant tumor shrinkage in in vivo models (30). Importantly, in our studies MI-773 showed significant therapeutic efficacy against both CSCs and differentiated mucoepidermoid carcinoma cells. Upon treatment with MI-773, we observed a significant reduction of ALDH^highCD44^high CSCs in our mucoepidermoid carcinoma models in vitro and in vivo. Interestingly, when we treated both ALDH^highCD44^high and non-CSCs with increasing concentrations of MI-773 in vitro, we saw no difference (between CSCs and non-CSCs) in the total cell number. This suggests that the reduction in the fraction of ALDH^highCD44^high cells is due to induction of their differentiation and/or inhibition of their self-renewal, rather than enhanced susceptibility to apoptosis (in comparison with non-CSCs). To further support this observation, we observed a significant reduction in expression levels of the self-renewal regulator Bmi-1, and an increase in the expression of the differentiation-associated protein p21, in tumors treated with MI-773 in vivo.

Using unsorted cells, we observed that MI-773 induces cell-cycle arrest and apoptosis in UM-HMC-1 and UM-HMC-3A cells (but not in UM-HMC-3B) in vivo and in vitro. A possible explanation for this unexpected finding is that the MDM2 expression is lower at baseline in UM-HMC-3B when compared with UM-HMC-1 and UM-HMC-3A cells. As such, these cells might be intrinsically less dependent upon the prosurvival function of MDM2. On the other hand, we observed an increase in TUNEL-positive cells and a decrease in the fraction of CSCs and in the expression of cell cycle–associated proteins in UM-HMC-3B treated with MI-773 in vivo. This suggests the tumor microenvironment generated in vivo modifies the sensitivity of the UM-HMC-3B cells to MDM2 inhibition. Mechanistic studies are being conducted to test this hypothesis and understand better the reasons for the different behavior of UM-HMC-3B cells in vitro and in vivo.

Although little is known about the role of p53 and MDM2 in salivary gland mucoepidermoid carcinomas, sequencing and IGHC studies suggest that TP53 mutations and loss of heterozygosity are rare events in these tumors (48). Interestingly, we observed the TP53 gene was fully functional in our UM-HMC-1, UM-HMC-3A, and UM-HMC-3B cell lines. This, together with the results presented here, suggests that small-molecule inhibitors of the MDM2–p53 interaction might have a place in the therapeutic “armamentarium” considered for treatment of salivary mucoepidermoid carcinomas.

Collectively, this work demonstrates that therapeutic inhibition of the MDM2–p53 binding targets and ablates ALDH^highCD44^high CSCs in mucoepidermoid carcinomas through activation of p53 and inhibition of stem cell self-renewal in vitro and in vivo. Activation of p53 by inhibition of MDM2 binding also induces cell-cycle arrest and apoptosis in mucoepidermoid carcinoma cells leading to cell death. Notably, a recent phase I clinical trial (NCT01636479) concluded that MI-773 (i.e., SAR405838) was well tolerated in patients with advanced solid tumors (49). These studies suggest that patients with mucoepidermoid carcinoma might benefit from ablation of CSCs using clinically relevant small-molecule inhibitors of the MDM2–p53 interaction.

S. Wang reports receiving commercial research grants and other commercial research support from, holds ownership interest (including patents) in, and is a consultant/advisory board member for Ascentage Pharma Group. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Andrews, S. Wang, J.E. Nör

Development of methodology: A. Andrews, F. Nör, Z. Zhang, J.E. Nör

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Andrews, K. Warner, C. Rodriguez-Ramirez, S. Kerk, J.I. Helman, J.C. Brenner, J.E. Nör

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Andrews, C. Rodriguez-Ramirez, A. Kulkarni, J.C. Brenner, M.S. Wicha, J.E. Nör

Writing, review, and/or revision of the manuscript: A. Andrews, C. Rodriguez-Ramirez, A.T. Pearson, F. Nör, M.S. Wicha, J.E. Nör

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Andrews, Z. Zhang, J.C. Brenner, S. Wang, J.E. Nör

Study supervision: S. Wang, J.E. Nör

We thank the patients who kindly provided the tumor specimens used to generate the mucoepidermoid carcinoma cell lines and xenograft models that enabled this research project. We also thank the surgeons, nurses, and support staff that enabled the process of tumor specimen collection and processing for use in research as well as the University of Michigan Sequencing and Flow Cytometry Cores for their support. This work was funded by the University of Michigan Head Neck SPORE P50-CA-97248 from the NIH/NCI and grants R01-DE21139 and R01-DE23220 from the NIH/NIDCR (J.E. Nör) and K08-DE026500 from the NIH/NIDCR (A.T. Pearson).

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.