Combination of bexarotene and the retinoid CD1530 reduces murine oral-cavity carcinogenesis induced by the carcinogen 4-nitroquinoline 1-oxide

Significance

Oral-cavity squamous-cell carcinoma is one of the most common human cancers in the world. About 60–70% of oral-cavity carcinoma cases are diagnosed only after the tumors have become locally advanced. Therefore, in addition to treatment, prevention of oral cancer is a very important goal. In this study, we found that the combination of the drugs bexarotene and CD1530 was more effective than either drug alone in preventing oral carcinogenesis in our mouse model of human oral and esophageal cancers. We envision that the combination of bexarotene and CD1530 could potentially be applied to humans at a high risk for oral cancer, as a very effective strategy for the prevention and treatment of human oral cancer.

Abstract

We investigated the effects of bexarotene (a retinoid X receptor agonist), CD1530 (a retinoic acid receptor γ selective agonist), and the combination of these two drugs for the prevention of oral carcinogenesis induced by the carcinogen 4-nitroquinoline 1-oxide (4-NQO) in a mouse model of human oral-cavity and esophageal squamous-cell carcinoma previously generated in our laboratory. We observed decreased numbers of neoplastic tongue lesions and reduced lesion severity in the 4-NQO plus CD1530 (4N+C) and 4-NQO plus bexarotene plus CD1530 (4N+B+C) groups compared with the 4-NQO group. RNA-Seq analyses showed increases in transcripts in cell proliferation/cell cycle progression pathways in the 4-NQO vs. the untreated group. In addition, β-catenin and matrix metallopeptidase 9 (MMP9) protein levels and reactive oxygen species (ROS), as assessed by 4-hydroxynonenal (4-HNE) staining, were elevated in tongue tissues 17 wk after the termination of the 4-NQO treatment. The 4N+B, 4N+C, and 4N+B+C groups showed dramatically lower levels of β-catenin, MMP9, and 4-HNE staining compared with the 4-NQO group. The major reduction in 4-HNE staining in the retinoid treatment groups suggests a novel mechanism of action, reduction of ROS, by which bexarotene and CD1530 inhibit carcinogenesis.

Oral-cavity squamous-cell carcinoma (OCSCC) is one of the most common human cancers in the world (1). The two major etiological factors in OCSCC are tobacco and alcohol (2, 3). Oral-cavity squamous-cell carcinoma (SCC) development is a complicated, multistep process that involves genetic, epigenetic, and metabolic changes (4). About 60–70% of oral-cavity carcinoma cases are diagnosed only after the tumors have become locally advanced (3). Therefore, in addition to treatment, prevention of oral cancer is a very important goal.

Retinoids, including vitamin A (retinol) and its metabolites, such as all-trans retinoic acid (RA), regulate cell proliferation and differentiation (5). RA regulates gene expression by binding and activating retinoic acid receptors (RARs α, β, and γ) and retinoid X receptors (RXRs α, β, and γ), transcription factors that heterodimerize and associate with retinoic acid response elements (RAREs) (6–8). Natural and synthetic retinoids have shown efficacy in the prevention and treatment of various human cancers, including leukemia, breast cancer, and lung cancer (8). With respect to OCSCC, treatment with 13-cis RA, which is isomerized to RA, a pan-RAR agonist, resulted in dramatic reductions in the sizes of oral leukoplakias in patients (9, 10).

We previously induced oral-cavity and esophageal SCCs that mimic human oral and esophageal tumors in terms of their morphological, histopathological, and molecular characteristics in mice by adding the carcinogen 4-nitroquinoline 1-oxide (4-NQO) to the drinking water (11–13). We also showed that 4-NQO has profound effects on the stem-cell population in the oral cavity (14). Therefore, this murine 4-NQO model is an excellent one for the evaluation of potential cancer preventive and therapeutic approaches. Bexarotene, a synthetic, pan-retinoid X receptor (RXR) agonist (15), has shown efficacy in the treatment of human T-cell lymphoma and lung cancer and is well tolerated by patients (16–18). RARγ (Nr1b3) has also shown tumor-growth suppression in mouse epidermal keratinocytes (19). We examined the cancer preventive effects of bexarotene, CD1530 (a synthetic, specific RARγ agonist) (20), and the combination of both drugs on oral-cavity carcinogenesis using this mouse model.

Materials and Methods

Tumor Development in the Mouse Oral Cavity and Drug Treatments.

Six-week-old wild-type C57BL/6 female mice (15 mice per group) were treated with vehicle as a negative control or 100 μg/mL 4-nitroquinoline-1-oxide (4-NQO) for 10 wk, as previously described (11, 12). Two weeks after termination of the carcinogen treatment, mice received various drug treatments at doses based on previous studies (20, 21): bexarotene at 300 mg/kg in the diet, CD1530 at 2.5 mg/100 mL in drinking water, and the combination of both drugs. The care and use of animals in this study were approved by the Institutional Animal Care and Use Committee (IACUC) of Weill Cornell Medical College.

Tissue Dissection, Lesion-Grade Measurement, and Pathological Diagnosis.

The tongues of mice were dissected immediately after cervical dislocation. Gross lesions were identified, photographed, counted, and graded (SI Materials and Methods). The histological diagnosis of squamous neoplasia was performed by a board-certified pathologist (T.S.) on paraffin-embedded, hematoxylin/eosin (H&E)-stained tissue samples in a blinded manner.

RNA-Seq Analysis of the mRNA Transcriptome.

One part of each mouse tongue (same position on each mouse tongue) was snap-frozen in liquid N₂ and stored at −70 °C until total RNA extraction. Total tissue RNA was extracted and subjected to Next-Generation Sequencing (RNA-Seq) at the Genomics Resources Core Facility, Weill Cornell Medical College. Bioinfomatics analyses were performed using Tophat and Cufflink software (SI Materials and Methods).

Immunohistochemistry.

Paraffin-embedded tongue sections were stained with various antibodies (SI Materials and Methods).

Statistical Analyses.

We performed statistical analyses by one-way analysis of variance and subsequently the Bonferroni test or the Tukey test for multiple comparisons. Differences with a P < 0.05 (two-tailed test) were considered statistically significant.

Results

Drug Treatments Reduce Tongue-Tumor Development.

All of the mice tolerated the 10-wk 4-NQO treatment, and almost all of the mice survived the 15-wk post–4-NQO treatment period (Fig. 1A). We determined that the CD1530 in the regular drinking water was stable for at least 7 d at room temperature by HPLC analysis (Fig. S1); thus, we used drinking water as a delivery method because this method is effective and less laborious. During the 15-wk post–4-NQO treatment, consumption of regular drinking water and water that contained CD1530 by the 4-NQO–treated mice was comparable, and the consumption of regular diet and diet with bexarotene by 4-NQO–treated mice was also comparable (data not shown).

Fig. 1.

The combination of bexarotene and CD1530 reduced average gross cancerous lesion numbers and lesion grades in mouse tongues. (A) Diagram of the experimental protocol (Materials and Methods). (B) Representative gross morphology of tongues from mice in this study and the gross tongue lesion grading system (8×), severity 4 > 3 > 2 > 1 > 0. (C) Representative pathological stages of tongue lesions: (a) normal (untreated tongue); (b) hyperplasia; (c) dysplasia; (d) papilloma; (e) invasive squamous-cell carcinoma. (D) Cancerous tongue lesions (number of lesions per tongue). (E) Severity of tongue lesions. A one-way analysis of variance test was used to analyze the differences in the tongue-lesion numbers and severity among all treatment groups (UNT, n = 15; 4-NQO, n = 10; 4N+B, n = 12; 4N+C, n = 13; 4N+B+C, n = 10). Differences with P values of <0.05 between the 4-NQO and the 4N+B, 4N+C, and 4N+B+C groups were considered statistically significant (*P < 0.05; **P < 0.01; ***P < 0.001). UNT, untreated; 4-NQO, 4-NQO treatment; 4N+B, 4-NQO plus bexarotene; 4N+C, 4-NQO plus CD1530; 4N+B+C, 4-NQO plus bexarotene plus CD1530.

No visible lesions (grade 0) developed in the untreated (UNT) mouse tongues (Fig. 1B). However, we observed obvious multifocal, precancerous, and cancerous lesions during the 17-wk post–4-NQO treatment period in all 4-NQO–treated mice (Fig. 1B). Pathological analyses show that after 4-NQO–treatment mice developed cancerous lesions, ranging from hyperplasia to malignant squamous-cell carcinomas (Fig. 1C), consistent with our previous findings (11, 12, 14, 22). The examination of gross tongue-lesion multiplicity revealed that, compared with an average of 5.9 ± 3.2 tongue lesions observed in the 4-NQO (4-NQO) group, the 4-NQO plus bexarotene (4N+B) group developed 5.3 ± 1.4 (P > 0.05) tongue lesions, the 4-NQO plus CD1530 (4N+C) group developed significantly fewer (3.9 ± 1.7, P < 0.05) tongue lesions, and the 4-NQO plus bexarotene plus CD1530 (4N+B+C) group developed only 2.7 ± 1.2 (P < 0.001) tongue lesions, a significantly lower number (Fig. 1D).

In addition, in the 4-NQO group, the severities of all tongue lesions were at, or greater than, grade 2, and the average severity was 2.8 ± 0.8. The 4N+B group showed an average severity of 1.9 ± 1.2 (P < 0.05), with 50% of the lesions at grade 1. The 4N+C group showed an average severity of 1.6 ± 0.8 (P < 0.01). In the 4N+B+C group, no lesion severity grade was greater than 2, 70% of the lesions were at grade 1, and the average lesion severity was 1.3 ± 0.5 (P < 0.001) (Fig. 1E).

Bexarotene and CD1530 Do Not Elevate Serum-Triglyceride Levels.

We also examined the effects of this drug combination on the triglyceride levels in serum samples because elevated triglyceride levels are associated with cardiovascular risks that could be a major side-effect in terms of a long-term cancer prevention approach. Compared with the UNT group, treatment for 15 wk with CD1530 alone, bexarotene alone, and the CD1530 plus bexarotene combination did not change serum triglyceride levels significantly (Fig. S2), suggesting that these treatments do not elevate serum triglyceride levels.

Drug Treatments Attenuate 4-NQO–Induced Gene-Expression Changes.

We next performed RNA-Seq using tongue samples that contained large tumors in the 4-NQO group compared with tongue samples from untreated mice to assess global changes in mRNA expression profiles. We found that the mRNA levels of a total of 3,379 genes were significantly overexpressed or underexpressed in the 4-NQO–induced tongue tumors vs. untreated tongues, including increases in 1,377 and decreases in 2,002 genes (Fig. S3 A, a). Compared with the 4-NQO group, the transcript levels of 1,050 genes were significantly different in the 4N+B group (436 with increased levels and 614 with decreased levels) (Fig. S3 A, b); 1,518 transcripts were significantly differentially expressed in the 4N+C group (987 genes with greater mRNA levels and 531 genes with lower mRNA levels) (Fig. S3 A, c); and 671 transcripts showed higher levels and 763 transcripts showed lower levels in the 4N+B+C group (Fig. S3 A, d). Heatmap analysis of the total of 3,379 transcripts that differed between the UNT group and the 4-NQO group revealed that, to some extent, all drug treatments mitigated the effects of 4-NQO on the transcript levels of the majority of these genes (Fig. S3B).

Gene-ontology analysis.

Gene ontology (GO) analysis (P value cutoff, 0.00001), using the ConsensusPathDB online tool, was conducted on the Gene Ontology level 3 of the “Biological Process” domain that has predefined, functional sets for enrichment analysis with default parameter settings. GO analysis revealed that among the top categories, with obvious overrepresentation in the 4-NQO–induced tongue tumors vs. untreated tongues, were “cell cycle phase” and “cellular component organization at cellular level” (including increases in ECM component degradation enzymes) (Fig. S4A, white bars). Underexpressed GO categories in the 4-NQO–induced tongue tumors vs. untreated tongues included “energy derivation by oxidation of organic compounds,” “generation of precursor metabolites and energy,” and “electron transport chain” (Fig. S4A, black bars), indicating that the transcript levels of the genes in the tricarboxylic acid cycle (TCA) cycle and oxidative phosphorylation (OX-PHOS) pathway were decreased in tongue tumors from the 4-NQO–treated group compared with tongues from untreated mice.

The comparisons between the 4-NQO group and all 4-NQO plus drug treatment groups (4N+B, 4N+C, and 4N+B+C) demonstrated that the GO categories related to cell cycle, DNA replication, and mitosis were statistically the most reduced in the three 4-NQO plus drug treatment groups (Fig. S4 B–D, black bars). The 4N+C group showed overrepresentation of many GO categories, with some of them involved in lipid metabolism (Fig. S4C, white bars).

Pathway analysis.

Furthermore, pathway enrichment analysis using the ConsensusPathDB online tool (P < 0.00001) of the 3,379 transcripts with altered levels in the 4-NQO–induced tongue tumors was carried out. Similar to the GO analysis, the results suggest that changes in gene expression in the 4-NQO–induced tumors compared with the UNT group are broad and reflect the characteristics of these tumor cells, such as enhanced cell proliferation and mobility and abnormal metabolism (Table S1). In addition to the cell cycle-related pathways that were down-regulated in the three 4-NQO plus drug treatment groups compared with the 4-NQO group (Tables S2–S4), some cytochrome P450 enzymes were significantly elevated only in the 4N+C group (Table S3).

Comparison of mRNA levels of individual genes that play important roles in the pathways described above.

The 4-NQO treatment affected some GO categories and pathways. We then compared the mRNA levels of some individual genes that are important in these GO categories and pathways whose mRNA levels were significantly different between the UNT group and the 4-NQO group. First, we analyzed the transcript levels of Cyp26a1, one member of the cytochrome P450 enzyme family and a direct target of RARγ (23). The Cyp26a1 mRNA level was about 20-fold greater in the 4N+C group than in other groups (Fig. S5A), indicating that the CD1530 administered to mice specifically transcriptionally activated RARγ target genes.

One of the characteristics of tumor development is cell-cycle progression, the transition from the G1 phase to the S, G2, and M phases of the cell cycle (24). With respect to the transcript levels of genes involved in cell-cycle progression, only aurora kinase A and B, CDK 1 and 6, and cyclin A2, B1, B2 and E1 (25, 26) levels were significantly greater in the 4-NQO group than in the UNT tongues. Compared with the 4-NQO group, these transcript levels were significantly lower in all 4-NQO plus drug treatment groups except for cyclin A2, which was significantly lower only in the 4N+B+C group (Fig. 2A). These data correlate with our data on tumor multiplicity and severity.

Fig. 2.

Quantitative analysis of the transcripts involved in cell proliferation from RNA-Seq data. (A) Genes involved in cell-cycle regulation. (B) Genes involved in DNA replication. DNA2, DNA replication helicase 2; LIG1, DNA ligase 1; FPKM, fragments per kilobase of exon model per million mapped reads; MCM, minichromosome maintenance complex; ORC1, origin recognition complex subunit 1, POLA1, DNA polymerase, alpha 1; PRIM2, DNA primase large subunit. Differences with P values of <0.05 between the 4-NQO and the 4N+B, 4N+C, and 4N+B+C groups were considered statistically significant (UNT, n = 5; 4-NQO, n = 3; 4N+B, n = 4; 4N+C, n = 4; 4N+B+C, n = 5; *P < 0.05; **P < 0.01; ***P < 0.001). UNT, untreated; 4-NQO, 4-NQO treatment; 4N+B, 4-NQO plus bexarotene; 4N+C, 4-NQO plus CD1530; 4N+B+C, 4-NQO plus bexarotene plus CD1530.

Compared with the UNT samples, transcripts involved in DNA replication were also significantly increased in the 4-NQO group, including minichromosome maintenance (MCM) complex members 2–7 and 10 that unwind the double-stranded DNA at the origins, recruit DNA polymerases, and initiate DNA synthesis (27); DNA replication helicase 2 (DNA2), a key enzyme involved in DNA replication and DNA repair in both the nucleus and mitochondria; DNA ligase 1 (LIG1); origin recognition complex, subunit 1 (ORC1); DNA polymerase, alpha 1 (POLA1), the catalytic subunit of DNA polymerase; and DNA primase large subunit (PRIM2). The 4N+B, 4N+C, and 4N+B+C groups showed significantly lower transcript levels of the majority of these genes compared with the 4-NQO group (Fig. 2B).

Matrix metalloproteinases (MMPs) contribute to extracellular matrix (ECM) breakdown and cancer cell migration (28), tenascin C (TNC) is involved in cell migration (29), and Slug plays a role in epithelial to mesenchymal transition and cell migration in human head and neck squamous-cell carcinoma cells (30). The transcript levels of MMPs 3, 9, 10, and 12–14, TNC, and Slug were significantly elevated in the 4-NQO group compared with the UNT group. We also showed that all 4-NQO plus drug treatment groups displayed significantly lower mRNA levels of the majority of these genes and that the 4N+B+C combination reduced the transcript levels of all these genes (Fig. 3A).

Fig. 3.

Quantitative analysis of some transcripts identified from RNA-Seq as involved in extracellular matrix (ECM) breakdown and cell migration, HIF1α signaling, and oral cancer. (A) Genes involved in the ECM breakdown and cell migration. (B) HIF1α signaling pathway members. (C) Human oral cancer markers. FPKM, fragments per kilobase of exon model per million mapped reads; HIF1α, hypoxia-inducible factor 1α; HMMR, hyaluronan-mediated motility receptor; GLUT1, glucose transporter 1; MCT4 (Slc16a3), monocarboxylate transporter 4; MMP, matrix metalloproteinase; Ndufa4l2, NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4-like 2; PTGS2, prostaglandin-endoperoxide synthase 2; TNC, tenascin. Differences with P values of <0.05 between the 4-NQO and the 4N+B, 4N+C, and 4N+B+C groups were considered statistically significant (UNT, n = 5; 4-NQO, n = 3; 4N+B, n = 4; 4N+C, n = 4; 4N+B+C, n = 5; *P < 0.05; **P < 0.01; ***P < 0.001). UNT, untreated; 4-NQO, 4-NQO treatment; 4N+B, 4-NQO plus bexarotene; 4N+C, 4-NQO plus CD1530; 4N+B+C, 4-NQO plus bexarotene plus CD1530.

Hypoxia-inducible factor 1α (HIF1α) is involved in the switch from tricarboxylic acid (TCA) cycle and oxidative phosphorylation (OX-PHOS) pathways to glycolysis (31). Because we observed significantly decreased mRNA levels of many genes of the TCA cycle and oxidative phosphorylation, we ascertained the transcript levels of HIF1α and several HIF1α targets that participate in the modulation of glycolysis and oxidative phosphorylation, such as glucose transporter 1 (GLUT1) (32), monocarboxylate transporter 4 (Slc16a3, MCT4) (32), and NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4-like (Ndufa4l2) that inhibits oxidative phosphorylation (33). Transcripts of HIF1α, Glut1, MCT4, and Ndufa4l2 were significantly increased in the 4-NQO group compared with the UNT group, and the mRNA levels of these genes in the 4N+B+C group were significantly lower than in the 4-NQO group (Fig. 3B). Moreover, the transcript levels of oral-cancer markers, such as keratin 1 (12) and prostaglandin-endoperoxide synthase 2 (PTGS2), also known as cyclooxygenase-2 (COX-2) (34), were significantly lower in the 4N+B+C group than the 4-NQO group (Fig. 3C).

Additionally, we performed quantitative real-time PCR on some of the genes described above to validate the RNA-Seq data (Fig. S5B). Collectively, the RNA-Seq analyses correlate with the tongue tumor multiplicity and severity data (Fig. 1 D and E), suggesting that the drug treatments, especially the combination of bexarotene and CD1530, may suppress/reduce tongue carcinogenesis via reduction of transcript levels of genes involved in cell-cycle progression and cell migration.

Oxidative-Stress Level, as Assessed by 4-Hydroxynoneal, Is Lower in the Tongues from the 4-NQO and Subsequent Drug-Treatment Groups.

Excessive reactive oxygen species (ROS) accumulation, caused by carcinogen exposure, may play a role in human oral carcinogenesis because ROS causes oxidative modifications of cellular macromolecules such as DNA, proteins, and lipids (35). Therefore, we next examined the levels of 4-hydroxynonenal (4-HNE), an α,β-unsaturated hydroxyalkenal produced by lipid peroxidation in cells during oxidative stress; 4-HNE is a marker of oxidative stress caused by ROS (36). The 4-NQO samples showed a large increase in 4-HNE levels in the precancerous tongue epithelium (Fig. 4, Upper) and cancerous lesions (Fig. 4, Lower) compared with the UNT samples (Fig. 4, arrows). It is important to note that these 4-NQO samples were assessed for 4-HNE levels 17 wk after the cessation of the 4-NQO administration, indicating that a higher ROS level is a feature of carcinogen treatment. Moreover, the tongue samples from the 4N+B, 4N+C, and 4N+B+C groups exhibited lower 4-HNE levels than those from the 4-NQO group (Fig. 4), indicating that the treatments with bexarotene, CD1530, and bexarotene plus CD1530 resulted in lower oxidative stress. The lower oxidative stress, as assessed by lower 4-HNE levels (Fig. 4), may be one of the mechanisms through which the retinoid drug treatments reduced oral carcinogenesis although further testing is required.

Fig. 4.

Shown is 4-hydroxynonenal (4-HNE), an indicator of oxidative stress, in tongue epithelium. The tongues were fixed, embedded in paraffin, and sectioned. Then, the tissue sections were stained with an antibody against 4-HNE. (Magnification: 200×.) Four to five representative areas of each tongue section from two mice per group were photographed and analyzed. Two samples per group are shown (Upper and Lower in each column). UNT, untreated; 4-NQO, 4-NQO treatment; 4N+B, 4-NQO+bexarotene; 4N+C, 4-NQO+CD1530; 4N+B+C, 4-NQO+bexarotene+CD1530. T, tumor. Darker brown indicates more 4-HNE adducts.

β-Catenin Levels Are Lower in the Tongues from the 4-NQO and Subsequent Drug-Treatment Groups Compared with the 4-NQO Group.

Oxidative stress can activate the β-catenin/Wnt signaling pathway (37), and the increase in β-catenin protein level can lead to increased cell proliferation in human head and neck cancer cells (38). Increased β-catenin levels have been observed during human oral squamous-cell carcinoma development (39). We detected β-catenin protein primarily in the basal layer of tongue epithelium in the UNT samples (Fig. 5A, arrows), and 4-NQO treatment resulted in both an increase in the β-catenin level and an expansion of β-catenin staining to the suprabasal layers of the tongue epithelium (Fig. 5A, Upper) and tumors (Fig. 5A, Lower), consistent with our previous findings (40). Compared with the 4-NQO group, all 4-NQO plus drug treatment groups (4N+B, 4N+C, and 4N+B+C) showed lower β-catenin protein levels in the tongue epithelium, primarily limited to the basal layer, even in the regions of tumor (Fig. 5A, Lower).

Fig. 5.

β-Catenin and MMP9 proteins in tongue epithelium. The tongues were fixed, embedded in paraffin, and sectioned. Then, the tissue sections were stained with various antibodies. (Magnification: 200×.) Four to five representative areas of each tongue section from two to four mice per group were photographed and analyzed. Two samples per group are shown (A and B, Upper and Lower). (A) β-catenin. (B) MMP9. UNT, untreated; 4-NQO, 4-NQO treatment; 4N+B, 4-NQO plus bexarotene; 4N+C, 4-NQO plus CD1530; 4N+B+C, 4-NQO plus bexarotene plus CD1530. T, tumor.

All Drug Treatments Result in Lower Matrix Metallopeptidase 9 Protein Levels.

High matrix metalloproteinase 9 (MMP9) protein levels have been observed in human oral cancers, and MMP9 is a marker of malignant human oral cancer (41, 42). We discovered that the MMP9 protein levels were low in untreated tongue epithelium, and that 4-NQO treatment resulted in an increase in MMP9 protein levels in tongue epithelium and tongue tumors (Fig. 5B, Lower and Fig. 5B, arrows). All 4-NQO plus drug treatment groups showed lower MMP9 protein levels in tongue epithelia compared with the 4-NQO group (Fig. 5B). Combined with the RNA-Seq data on the pathways of ECM breakdown and cell migration (Fig. 3A), our data indicate that the retinoid drug treatments limited the breakdown of ECM and potentially could reduce cell migration and metastasis.

Discussion

Our data show that, for the prevention of murine oral carcinogenesis, the combination of bexarotene and CD1530 is more efficacious than bexarotene alone or CD1530 alone (Fig. 1 D and E). The combination of bexarotene and CD1530 was also more effective than either drug alone in preventing the 4-NQO–induced changes in the transcript levels of genes important for tumor development (Figs. 2 and 3).

Retinoid X receptors (RXRs) play a key role within the nuclear receptor (NR) superfamily and can form heterodimers with many other nuclear receptors, including RARs, PPARs, liver X receptors (LXRs), farnesoid X receptors (FXRs) (43), and vitamin D receptors (VDRs) (43, 44). Because RXRs participate in many nuclear receptor signaling pathways, they have been a target for drug discovery (43). Our study shows that bexarotene, an RXR agonist, did not affect mouse serum triglyceride levels (Fig. S2), consistent with a previous study (45). CD1530, at the dose and the duration used in our study, also did not elevate triglyceride levels (Fig. S2). Therefore, our data suggest that the combination of bexarotene and CD1530 would not cause cardiovascular risks if used at similar doses in cancer-prevention treatments for those at risk for oral cancer.

Bexarotene is approved by the FDA and has been used in the treatment of human cancer (16–18). Bexarotene inhibits cell proliferation and induces cellular senescence and apoptosis in a mouse breast cancer model (46) and modulates expression of genes related to the cell cycle, differentiation/apoptosis, and cell adhesion/migration in a mouse breast cancer model (47) and in human normal mammary epithelial cells (48). RARγ also mediates RA-induced growth arrest and apoptosis of neoplastic mouse papilloma cell lines (49). Our findings that the mRNA levels of several CDKs, cyclins, and proteins involved in DNA replication were lower in the 4N+B group, 4N+C group, and the 4N+B+C group than in the 4-NQO group (Fig. 2) are consistent with these previous studies. Similar to the mechanisms of human oral carcinogenesis (35), one of the mechanisms of 4-NQO–induced tumorigenesis is the generation of reactive oxygen species (ROS) that leads to the formation of DNA adducts (50, 51). One of the products generated from excessive ROS, 4-hydroxynonenal (4-HNE), modifies cell proteins by forming protein adducts and thereby changes cellular signaling cascades and gene expression (36). Here, for the first time, to our knowledge, our data (Fig. 4) show that both bexarotene, an RXR agonist, and CD1530, an RARγ agonist, suppress excessive ROS production in tongue epithelial cells. We suggest that this inhibition of ROS by bexarotene and CD1530 contributes to the inhibition of tongue carcinogenesis.

RXR agonists, but not RAR agonists, induce β-catenin protein degradation in human cells (52). However, an RARα dominant negative mutant in mouse liver elevates liver β-catenin protein levels (53). The increase in β-catenin protein leads to increased cell proliferation in human head and neck cancer cells (38). Therefore, compared with the 4-NQO group, the lower β-catenin protein levels in tongue epithelia observed in all 4-NQO plus drug treatment groups (Fig. 5A) may contribute to lower tumor numbers and severity in these groups (Fig. 1 D and E). The mechanisms by which bexarotene and CD1530 modulate β-catenin levels need further investigation.

Treatment with O-hydroxyphenantrene, a synthetic pan RXR agonist, results in decreases in both the expression and the activity of MMP2 and MMP9 in human glioblastoma and melanoma cells (54). We show that the RXR and RARγ agonists in combination reduce the mRNA levels of several MMPs and the protein level of MMP9 (Figs. 3A and 5B). Therefore, these data suggest that targeting RXR and RARγ is a useful strategy for the prevention of cancer-cell migration and metastasis in other human cancers.

HIF1α regulates genes involved in tumor initiation, progression, and metastasis (55), and HIF1α overexpression correlates with a poor prognosis in human head and neck cancer (56). We show that RARγ and RXR agonists inhibit the HIF1α signaling pathway in the context of tongue carcinogenesis (Fig. 3B). Although the gene ontology and pathway analyses of our RNA-Seq data suggest that the TCA cycle and OX-PHOS pathways are significantly less active during tongue carcinogenesis, we did not detect significant increases in transcript levels of the glycolysis pathway genes (Table S1). Therefore, glycolysis does not appear to be increased in this 4-NQO model of oral carcinogenesis.

Because bexarotene has been used in human patients (16–18), we envision that the combination of bexarotene and CD1530 could potentially be applied to humans at a high risk for oral cancer (e.g., patients with leukoplakia). This drug combination could potentially be a very effective strategy for the prevention and treatment of human oral cancer.