Fenretinide, Tocilizumab, and Reparixin Provide Multifaceted Disruption of Oral Squamous Cell Carcinoma Stem Cell Properties: Implications for Tertiary Chemoprevention Free

Oropharyngeal cancer, which is a worldwide health problem associated with significant morbidity and mortality, will affect over 53,000 Americans in 2019 (https://www.cancer.org/cancer/oral-cavity-and-oropharyngeal-cancer/about/key-statistics.html). Despite treatment advances, e.g., fluorescence-guided surgery and intraoperative radiotherapy, 5-year survival rates have only modestly improved for human papillomavirus–negative oral squamous cell carcinomas (OSCC) and still hover around 50% (1). Following initial therapy, OSCC patients are managed by close clinical follow-up often supplemented with imaging (CT, PET, or MRI) studies. Despite image-enhanced monitoring and knowledge of the risk factors for recurrence (close margins, immunosuppression, high histologic grade, deep tumor extension), over a third of patients develop recurrent, life-threatening, inoperable OSCC tumors (2, 3). In accordance with its relentless and infiltrative nature, the majority of OSCC deaths are attributable to massive locoregional recurrences that arise from malignant transformation of dysplastic surface epithelium approximating the tumor resection site (2, 3).

Treatment choices for recurrent, inoperable OSCCs are limited and include palliative radiotherapy or chemotherapy with or without biological adjuncts (4, 5). As OSCC tumors are primarily comprised of treatment-responsive transient amplifying cells and postmitotic differentiated cells, tumors initially regress but eventually recur (6, 7). Cancer stem cells (CSC), which comprise only a small percentage of total cells in OSCC tumors, play a prominent role in postchemotherapy OSCC recurrences (6, 7). Notably, CSCs' high levels of phase III/drug egress enzymes endow CSCs with chemotherapeutic resistance (7). Furthermore, due to their growth state reciprocity that facilitates transition to transient amplifying cells, CSCs readily repopulate treated OSCC tumors (6, 7). CSCs' plasticity, which enables the epithelial–myoepithelial transition from an adherent, proliferative epithelial cell to a mobile, invasive myoepithelial phenotype, further facilitates tumorigenesis and metastases (8). Clinical evidence shows OSCC progression coincides with a concurrent rise in CSCs, which implicates a contributory role for CSCs in OSCC carcinogenesis (9). Collectively, these data demonstrate that suppression of CSCs' tumorigenic properties is essential for effective, sustained duration of OSCC tertiary chemoprevention (3, 4, 9).

Currently, following initial management of the primary OSCC tumor, additional treatment is suspended until tumor recurrence or a second primary OSCC is detected (10). Provided the deleterious effects of standard systemic chemotherapy, this delayed treatment approach is logical. In contrast, controlled release local delivery, which can provide therapeutic chemopreventive levels at the site without drug-related systemic toxicities, would enable a proactive tertiary chemoprevention approach (11). Provided OSCC tumors' signaling redundancies that enable signal transduction even if a specific pathway is blocked and ability to modulate their microenvironment, e.g., angiogenesis, local immune suppression, identification of multimodal agents with complementary mechanisms of action is a critical first step (12–14). We have identified a group of three agents, i.e., the vitamin A derivative, fenretinide (https://pubchem.ncbi.nlm.nih.gov/compound/5288209), the humanized monoclonal antibody to the IL6 receptor, tocilizumab, and the CXCR1/CXCR2 inhibitor, reparixin (https://pubchem.ncbi.nlm.nih.gov/compound/reparixin), that function in a coordinated fashion to suppress key CSC tumorigenic mechanisms. The rationale for selection of these agents is as follows. Our labs have demonstrated fenretinide binds with exceptionally high affinity to and induces downstream chemopreventive effects on tyrosine kinases integral for sustained proliferation, migration, and invasion (12, 15). These additional chemopreventive mechanisms supplement fenretinide's well-recognized growth regulation via induction of apoptosis and differentiation (12, 16). The proinflammatory, proangiogenic cytokine, IL6, facilitates tumorigenesis within the tumor microenvironment via influx of reactive species generating inflammatory cells and induction of angiogenesis (17). Clinically, high IL6 sera levels correspond to an overall poor prognosis in OSCC patients (18). In addition, OSCC cells generate very high levels of IL6, and via their concurrent release of sIL6R, tumor cells can function as both IL6 effectors and targets (12, 19). In addition, we have shown that a locally delivered fenretinide–tocilizumab combination treatment provided additive in vivo chemopreventive efficacy in an OSCC xenograft model (12).

The final agent, reparixin, was selected for its abilities to interfere with IL8 signaling (20). The inflammatory cytokine IL8 promotes angiogenesis and tumor cell proliferation, while also enabling epithelial–mesenchymal transition (EMT; refs. 20, 21). Furthermore, as OSCC cells respond to and produce IL8, the potential for an intracrine growth loop within the tumor microenvironment is feasible (22). Clinically, high levels of IL8 at the OSCC tumor-invasive front are associated with poor prognosis (23). IL8 enhances OSCC cancer stem cell proliferation and survival via increased expression of its cognate receptors CXCR1 and CXCR2 (20). IL8–CXCR1/2 binding activates tyrosine kinase (PI3K-Akt, PI3K-Src, JAK2, or MAPK) mediated signaling cascades that facilitate expression of a bank of tumor-promoting genes (24). An international phase II clinical trial to assess the effects of combination reparixin + paclitaxel on disease-free survival in patients with metastatic triple-negative breast cancer is ongoing (https://clinicaltrials.gov/ct2/show/NCT02370238; https://www.centerwatch.com/clinical-trials/listings/70211/metastatic-breast-cancer-double-blind-study-paclitaxel-combination/). In this trial, reparixin was specifically selected to disrupt the breast cancer stem cells. Previous studies by Schott et al. (25) confirmed the safety and potential efficacy of this drug combination in patients with HER-2 negative metastatic breast cancer (MBC).

The goals of the study were 2-fold. First, to develop and validate a CSC-enriched (CSCE) OSCC cellular model. Second, to assess the effects of these three bioactive agents, singularly and in combination, on key CSC tumorigenic activities. Our data reveal that although single agents interfere with CSC essential activities, e.g., gratuitous growth factor signaling, the triple agent combination conveys the greatest chemopreventive impact as demonstrated by significant reductions in IL6 and IL8 release, stem cell–associated gene expression, and invasion of a synthetic basement membrane. Consistent with its ability to associate with high affinity to signal transduction binding sites, fenretinide significantly suppressed Wnt3–β-catenin signaling.

OSCC tumor tissues were obtained in accordance with Ohio State University Institutional Review Board approval. JSCC-1, JSCC-2, and JSCC-3 cells, which were isolated from OSCCs of tonsil, tongue, and floor of mouth, respectively, were cultured in Advanced DMEM supplemented with 1X Glutamax and 5% heat-inactivated FBS (GIBCO; Life Technologies; “complete” medium). All OSCC tumors from which the JSCC cell lines were derived represented primary resections that had not been exposed to chemotherapy. Furthermore, none of the OSCC tumors used to generate the OSCC cell lines showed the histologic features consistent with oncogenic human papillomavirus DNA (see Supplementary Fig. S1). The highly tumorigenic CAL27 ATTC CRL-2095 human tongue OSCC cell line (2095sc), which has been well characterized by our lab (12, 15), was also evaluated and used for some in vivo explant studies. An immortalized, nontumorigenic cell line derived from E6/E7-transduced normal human oral keratinocytes [ScienCell; HOK#2610 (EPI)] was also used for selected experiments. Sera were omitted (“base” medium) during experiments to assess endogenous or growth factor–related effects. The most recent cell lines authentication was performed via short tandem repeats profiling analyses at the Genetic Resources Core Facility (Johns Hopkins University, Baltimore, MD) in December 2018. Mycoplasma status was not assessed. SEL cell cultures were used between PDL4 to PDL8, whereas nonselected cell lines were passed approximately 5 times. Clinical parameters of the primary OSCC tumors, such as the tumor–node–metastasis classification, perineural and vascular invasion, and IHC characterizations have been previously reported (15).

Two recognized properties of CSC, i.e., chemotherapeutic resistance and anchorage-independent growth, were employed to obtain CSCE populations. Preliminary dosing studies to optimize paclitaxel treatment [(2, 5, and 10 nmol/L, paclitaxel dissolved in DMSO, final [DMSO] > 0.01%)] on cell proliferation (hemocytometer counting) and viability (trypan blue exclusion) revealed that 5 nmol/L paclitaxel (fresh drug added q 24 hours × 72 hours) dramatically reduced cell numbers yet retained a small cell subset of viable cells (see Supplementary Fig. S2). This protocol, i.e., fresh 5 nmol/L paclitaxel, q24h × 3 days, was then used for the initial selection step. Paclitaxel-selected cells were then grown to confluence in complete drug-free medium and then transferred to Ultra low-attachment culture flasks (Corning, Fisher Scientific) with complete medium. After 2 days in the low-attachment flasks, cells were centrifuged, resuspended, and returned to Ultra low-attachment flasks × 3 to generate tertiary tumorspheres. Increases in recognized stem cell markers (RT-PCR analyses) and functional assays (ALDH activity) were used to confirm a CSCE phenotype. CSCE cultures are designated by SEL following cell line, e.g., JSCC2 SEL, 2095scSEL.

Chemopreventive treatment doses [4HPR (Cedarburg Pharmaceuticals, generous donation), tocilizumab (Ohio State University James Cancer Hospital Pharmacy), and reparixin (Sigma Chemical Company)] were derived from our previous studies (12, 15), literature values (26), and our combined viability and proliferation studies.

Twenty-four hours prior to ALDH1 assay [ALDH Activity Colorimetric Assay (Abcam)], cells were treated in the following experimental groups: (1) 5 μmol/L fenretinide, (2) 1 μg/mL tocilizumab, (3) 10 μmol/L reparixin, (4) fenretinide + tocilizumab, (5) fenretinide + reparixin, (6) tocilizumab + reparixin, (7) fenretinide + tocilizumab + reparixin, (8) DMSO control (<0.01%). Combinations employed the same drug levels as single treatment. Functional ALDH activity was determined in cell lysates at 450 nm using FLUOstar Omega (BMG Labtech) and analyzed according to the manufacturer's protocol in mU/mL. An NADH standard curve (r > 0.995) was conducted with every assay, and the samples' activities (reported as nmol/L NAD(P)H/mg protein) were within the linear range of the standard curve.

Total RNA was isolated from cultured OSCC cells using the mirVana miRNA isolation Kit (Thermo Fisher Scientific). RNA concentrations were measured via Nanodrop ND-1000 (Nanodrop). Integrity of total RNA was evaluated using capillary electrophoresis (Bioanalyzer 2100, Agilent Technologies) and quantified using a Nanodrop 1000 (Nanodrop). The total RNA was reverse-transcribed to cDNA using a High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems). An ABI 7500 RealTime PCR System (Applied Biosystems) was used for amplification and detection of the qRT-PCR products. Data were normalized using the endogenous GAPDH control. NCBI database sequence information (https://www.ncbi.nlm.nih.gov/tools/primer-blast/) was used to design the intron-spanning primers.

Male BALB/c mice (n = 10, Ohio State Institutional Animal Care and Use Committee–approved) were obtained from Ohio State's Shared Resources Target Validation Murine facility (https://cancer.osu.edu/research-and-education/shared-resources/target-validation). To control for any interanimal tumor growth discrepancies, every mouse received injections of 1 × 10⁶ log growth 2095sc cells (control, left flank) and 1 × 10⁶ 2095scSEL (right flank) cells suspended in 100 μL Matrigel (Corning). Measurements were obtained daily (length × width), and mice were euthanized 14 days after OSCC cell implantation.

OSCC tumors and CSCE tumors were stained with hematoxylin and eosin and the IHC proliferation marker Ki-67 (Cell Signaling (1:400 dilution, Cat#9027; Clone#D2H10). All histopathologic analyses were conducted by an investigator blinded to the experimental groups. Hematoxylin and eosin histologic sections were used to assess two-dimensional tumor sizes (tumor center, KR887 stage micrometer) and the highest mitotic index (#mitoses/400× field). Images were captured via an Olympus BX50 microscope (Olympus) and Moticam 10/10.0 MP digital camera (Motic). Ki-67–stained slides were scanned by Aperio Digital Pathology Slide Scanner (Leica). Ki-67 IHC staining intensity was quantitatively analyzed using Image-Pro Premier software (Media Cybernetics). The investigator outlined the tumor representative areas to exclude areas of necrosis and peripheral murine stroma.

Wnt binding to the CRT domain of the transmembrane protein Frizzled initiates both canonical (β-catenin) and noncanonical “alternative” (YAP-TAZ) Wnt signaling, (27). The Frizzled protein structure was obtained from the Protein databank 4F0A (The Protein Data Bank: http://www.rcsb.org/pdb/home/home.do; ref. 28]. The missing loop was modeled using Molecular Operating Environment (29). The structure was optimized with Yasara using the default minimization algorithm (30). All ligands were constructed in Spartan10 (31) and minimized using MMFF (32). The optimized protein structure and ligands were docked using AutoDock Vina (33) using an exhaustiveness of 500. All calculations were repeated 3 times to ensure a thorough exploration of the binding site. Previous studies have shown that flexible amino acid side chains provide for optimal results; therefore, the side chains for amino acids, 71, 72, 74, 75, 77, 78, 121, 122, 125, and 127–130, were made flexible to ensure a more realistic binding mode for all calculations. Although the calculations were conducted using the Frizzled 8 protein, the high structural homology among all of the Frizzled proteins implies these structure-functional modeling data would be applicable to other Frizzled receptors (34).

Calculated binding free energies were used to determine a binding affinity (Ka) and dissociation constant (Kd) to compare with experimental data. ΔG = RT ln(Ka) or kA = e^−(DG/TR) and Kd = 1/Ka. During binding, the Wnt protein undergoes modification at Ser-187 by addition of a palmitoleic acid (PAM; ref. 35). The hydrophobic tail of the PAM lipid residue then binds to a specific groove in the Frizzled-CRT domain (35). Frizzled's lipid binding site was tested with PAM, fenretinide, and 4-oxo-fenretinide. Because PAM is very flexible, it did not bind in the same orientation as the crystal structure. The binding analyses were repeated with a rigid PAM molecule in the crystal conformation. Binding energies were also evaluated using a segment of Wnt-PAM employing both rigid and flexible PAM structures.

Fenretinide's effects on Wnt–β-catenin signal transduction were evaluated by a TCF/LEF luciferase reporter assay (Qiagen). Cells were transfected (Lipofectamine, 3000; Invitrogen/Fisher Scientific) with both the LEF/TCF luciferase reporter that contained concatemers of the wild-type TCF/LEF binding sites plus the constitutively expressed internal control CMV-renilla luciferase (Qiagen). Base medium cultured cells were challenged with human recombinant Wnt3 (100 ng/mL, Abcam) 24 hours after transfection. Experimental groups consisted of: (1) 4HPR (5 μmol/L) + Wnt3, 4HPR 5 μmol/L (no Wnt3), DMSO vehicle control (<0.01%) along with positive and negative controls. Fenretinide was added to the cells 1 hour prior to challenge with Wnt3 to enable fenretinide–FZD interactions. Luciferase activity was measured after 24 hours using the dual-luciferase reporter assay (Promega).

Levels of Wnt3 release were determined using a Wnt3a ELISA (R&D) with results reported as pg per 10⁶ cells. Chemopreventive effects on IL6 and IL8 release were evaluated by ELISAs (R&D Systems; pg/10⁶ cells).

SCC2095sc, SCC2095sc SEL, JSCC2, and JSCC2 SEL cells were cultured on Lab-Tek chamber slides (Thermo Fisher Scientific) at a 1 × 10⁵/chamber in Advanced DMEM + 1x GlutaMax + 5% FBS media. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Tween (YAP/TAZ-0.1% Triton X-100), blocked, and then incubated with vimentin (1:250, Abcam; Cat# ab92547; Clone# EPR3776), YAP/TAZ (1:100, Cell Signaling Technology; Cat#8418; Clone# D24E4), and pancytokeratin MNF16 (1:25, Abcam; Cat#ab756; Clone#MNF16) antibodies, followed by incubation with Alexa Fluor 488–conjugated secondary antibodies (1:1,000, ThermoFisher Scientific; Cat#A11029). Nuclei were stained with 4′,6′-Diaminidino-2-phenylindole dihydrochloride (DAPI, VECTOR Laboratories; Cat#H-1200). Fluorescence microscopy images were obtained by using an Olympus BX51 microscope (Olympus), Nikon DS-Fi1 digital camera (Nikon), and NIS-Elements software (Nikon Instruments Inc.).

Initial studies employed hemocytometer cell counts with trypan blue exclusion to assess the effects of the chemopreventives (singularly and in combination, 5 μmol/L 4HPR, 10 μmol/L reparixin, 1 μg/mL tocilizumab) on cell proliferation and viability over time. Chemopreventive doses were determined from our previous data or the literature (12, 15, 26).

The single and combined effects of fenretinide, tocilizumab, and reparixin on YAP, TAZ, and ZEB (all antibodies purchased from Cell Signaling Technology), total and phosphorylated YAP and TAZ, and total ZEB protein levels by image analyzed immunoblotting were conducted by methods in use in our lab (12, 15).

Cell lines previously determined to undergo rapid directed migration (2095sc, JSCC1, and EPI) were transfected with siRNA (50 nmol/L) of YAP, ZEB, and the combination of YAP/ZEB in complete medium for 12 hours for the scratch-wound assays (siRNA sequences: YAP1:GGUCAGAGAUACUUCUUAAAUCACA; ZEB1: GUUFFAFAAUAAUCAAGCCAAUCTT).

Confluent cells were wounded with a pipette tip, washed with PBS, and provided complete medium. Three images of every well (left, middle, right) were obtained daily (Apotome Fluorescence Microscope), and AxioVision software (Carl Zeiss) until the wound was completely closed in the siRNA-free cultures. ImagePro software (Media Cybernetics, Inc.) was used to quantitate wound closure.

Fifty thousand 24-hour sera-deprived cells/well were plated on type IV collagen-coated microporous polyester membrane (InnoCyte cell invasion kit, Calbiochem) with the following treatments: (1) 5 μmol/L 4HPR (2095sc log and SEL cells), 10 μmol/L 4HPR (JSCC2 log and SEL cells), (2) 1 μg/mL tocilizumab, (3) 10 μmol/L reparixin, (4) fenretinide + tocilizumab, (5) fenretinide + reparixin, (6) tocilizumab + reparixin, (7) fenretinide + tocilizumab + reparixin, (8) DMSO control. Preliminary studies confirmed all cell line viabilities remained ≥96%, and cell numbers remained comparable during all treatments. JSCC3 conditioned medium was used as the chemoattractant (15). After 16 hours of invasion (37°C, 5% CO₂), cells were formalin fixed and stained with 0.1% v/v crystal violet solution. Images were captured by Nikon DS-Ri1 using NIS Elements (Nikon), followed by target pixelation analyses by image segmentation [ImagePro software (Media Cybernetics, Inc.)].

All statistical analyses used GraphPad Prism 6 software (GraphPad). Data normality (Shapiro–Wilk normality test) determined whether to employ parametric or nonparametric analyses. The Kruskal–Wallis ANOVA followed by the Dunns Multiple Comparison post hoc test was used to analyze the following studies: ALDH functional activity, TCF/LEF β-catenin signaling, siRNA wound migration, Wnt3a, IL6 and IL8 ELISAs, RT-PCR analyses, and invasion. The OSCC xenograft data mitotic activity and Ki67 data were analyzed using an Unpaired t test, whereas the Ki-67 data were evaluated using a Mann–Whitney U two-tailed test. “N numbers” for experiments were determined by conduction of sufficient repetitions to allow all cell lines to be evaluated for at least three separate experiments. The number of mice for the tumor implant analyses was based upon our previous in vivo studies (12). As the researchers were directly involved in conduction of the experiments, they were not blinded to the experimental groups.

Cancer stem cell enrichment altered gene expression toward stem cell proteins (2095sc SEL) and protumorigenic proteins (2095sc SEL, JSCC2 SEL; Fig. 1A). Tumorsphere and CSCE cultures also demonstrated significantly higher activity of, and greater capacity to upregulate the NADPH generating enzyme, ALDH (Fig. 1B). The CSCE phenotype was retained for at least eight population doubling levels following CSC selection (see Supplementary Fig. S3).

CSCE xenografts showed statistically significant greater mitotic activity (P = 0.014, n = 10 both groups) and higher proliferation as indicated by Ki-67 quantitative analyses (P = 0.049; Fig. 1C, panels 5 and 6) relative to non-selected OSCC xenografts. The tumors also showed qualitative intergroup differences. Although the OSCC parental/control cells possessed a more differentiated phenotype that exhibited increased cytoplasm and keratin production, the 2095sc CSCE tumor nests revealed more nuclear pleomorphism and higher nuclear to cytoplasmic ratios (Fig. 1C, panels 3–6). Sizes of the tumor xenografts are presented in Supplementary Table SI.

Molecular modeling data reveal that fenretinide binds with appreciably higher affinity (∼158-fold) relative to the endogenous Wnt-PAM ligand at the Wnt-Frizzled binding site (Fig. 2A). Fenretinide's oxidized metabolite, 4-oxo-fenretinide, also binds with greater affinity than Wnt-PAM, albeit at reduced levels (∼15-fold; Fig. 2A). Signaling assays to assess functional effects revealed fenretinide significantly suppressed β-catenin activation in the 2095scSEL (P < 0.05) and JSCC2SEL (P < 0.01) CSCE lines (Fig. 2B). All cell lines released Wnt3a (Fig. 2C). Further, all cell lines demonstrated constitutive YAP-TAZ nuclear translocation (Fig. 2D), which persisted following chemopreventive treatment. YAP and TAZ protein levels were refractory to addition of Wnt3a and chemopreventives (Fig. 2E) that were not affected by chemopreventives. Further, all lines released Wnt3a (Fig. 2E).

Stem cell selection affected intermediate filament expression (Fig. 3A). Although the 2095sc SEL cultures demonstrated enhanced cytokeratin, the JSCC2 SEL cells showed cytokeratin reduction with a concurrent increase in vimentin (Fig. 3A).

Proliferation studies revealed appreciable differences among the cell lines, with the growth rate of the 2095sc cells appreciably higher than the JSCC2 cells (Fig. 3B). Furthermore, although stem cell enrichment increased proliferation rate in the 2095sc SEL cells, the JSCC2 SEL cells demonstrated a reduced proliferation rate relative to their parent cell line. Fenretinide was the single most effective proliferation-reducing agent across all cell lines (Fig. 3B). Although single-agent tocilizumab showed no effect or a slight growth enhancement, fenretinide+tocilizumab and fenretinide+tocilizumab+reparixin combinations were highly effective in growth reduction. During all treatments, cell viabilities remained comparable with control cultures ≥96% (see Supplementary Table SII).

Although all cell lines released IL6 and IL8, parent cell line and CSCE differences were apparent (Fig. 4A). Triple chemopreventive agent treatment significantly suppressed IL6 and IL8 release in all lines (Fig. 4A) and also significantly reduced expression of stem cell genes in the CSCE cultures (Fig. 4B).

Results of the siRNA ZEB and YAP studies revealed that ZEB and YAP are both necessary for directed migration in OSCC cell lines (see Fig. 5A, P < 0.01 for both ZEB and YAP siRNA, n = 5; Fig. 4A). The nontumorigenic immortalized oral keratinocytes transduced with HPV16 E6 and E7 proteins (Epi) only showed significant wound-healing inhibition following siRNA treatment for YAP (P < 0.01, n = 5). Combination ZEB and YAP siRNA treatment significantly inhibited all cell lines' wound healing (Fig. 5A). Although siRNA treatment reduced cell numbers relative to control cultures, viability (trypan blue exclusion) remained >95% in all groups.

Chemopreventive treatments reduced levels of ZEB and YAP proteins in the 2095sc SEL cells. Selective treatments also increased levels of the inactive phosphorylated-YAP in the 2095sc SEL cells (Fig. 5B).

Intercell line and CSCE differences were observed with regard to cell invasiveness (Fig. 6). Although the JSCC2 SEL cells demonstrated enhanced invasion relative to the JSCC2 parental cells, the converse, i.e., enhanced invasion by the parental cell line, occurred in the 2095sc cells. Only the SEL cell lines showed significant invasion inhibition by single or dual treatments (fenretinide, fenretinide + tocilizumab, fenretinide+ reparixin). Triple treatment with fenretinide, tocilizumab, and reparixin significantly inhibited invasion across all cell lines including the CSCE cultures (Fig. 6).

Although targeted immune response upregulation studies, e.g., T-cell immune checkpoint inhibitors are ongoing (36), their efficacy in OSCC management is not yet delineated (37). Despite these new treatment options, prevention of OSCC recurrence via tertiary chemoprevention is the better option. In order to assess putative tertiary chemopreventives, OSCC CSC enrichment is essential. Our postselection results showed appreciably higher levels of baseline and inducible ALDH activity, enhanced proliferation in vitro and in vivo (2095sc SEL cells), greater in vitro invasion (JSCC2 SEL cells), increased expression of Oct-4 and CD24, increased expression of the proinflammatory, proproliferative cytokines IL6 and COX-2, and appreciably higher release of IL6 or IL8 protein were consistent with CSC enrichment. Our data's standard errors indicated population heterogeneity within the SEL cultures. These findings mirror the in vivo situation where CSCs comprise a small but significant cell population of the OSCC tumor (6, 7). Importantly, there may be a synergistic, tumor-promoting relationship between non-CSC and CSC tumor cells (6, 7). Studies in human breast cancer cells showed IL6 release by non-CSC cells activated an IL6-JAK1-STAT3-Oct-4 cascade that facilitated conversion of non-CSCs to CSCs (38).

A key modulator of cancer stem cells is the Wnt–β-catenin signaling pathway, which regulates CSC proliferation, survival, and differentiation (39). Wnt–Frizzled (FZD) interactions occur extracellularly at cysteine-rich domains (CRD) dominated by hydrophobic interactions (39), which represent an optimal environment for the very lipophilic fenretinide. Our modeling data show fenretinide binds with approximately 158-fold higher affinity relative to Wnt at the FZD8 CRD-binding site. Furthermore, as the Wnt-binding CRDs are highly conserved (39), fenretinide likely interferes with Wnt interactions with other FZD family members. Our data also confirm a downstream effect as fenretinide significantly inhibited the canonical β-catenin signaling in CSCE cell lines. Wnt also signals through an alternative YAP-TAZ–mediated pathway (40). All OSCC lines demonstrated nuclear YAP-TAZ in sera-free medium, consistent with constitutive activation of the FZD/ROR–Gα12/13–Rho GTPases–Lats1/2 alternate Wnt signaling pathway (40). As all OSCC cell lines released Wnt3a, an intracrine loop may be at least partially attributable for the constitutive intranuclear YAP-TAZ. Although the selected chemopreventives did not suppress nuclear translocation or total YAP-TAZ protein levels, they did mitigate a downstream event of the alternate Wnt pathway, i.e., invasion (41). Fenretinide's capacity to inhibit Wnt-Frizzled binding and its subsequent downstream signaling combined with its ability to bind with higher affinity than the corresponding natural ligands at tyrosine kinases associated with stemness and is often deregulated during carcinogenesis, e.g., STAT3, FAK, c-Myc, c-Abl (12, 15, 42) substantiates an integral tertiary chemopreventive role for fenretinide.

Cancer cell stemness, EMT entrance, and tumorigenecity are interrelated (6–8). Subsequently, OSCC CSCs possess the inherent plasticity to transition between preferentially migratory or proliferative phenotypes that fulfill cooperative roles (7). Although “mesenchymal/migratory” cells establish new foci, the “epithelial/proliferative” cells populate the tumor (43). The two CSCE cultures isolated in this study were at divergent ends of the EMT spectrum. The 2095sc SEL population, which expressed more cytokeratin, high proliferation indices, elevated IL6 release with low vimentin, and modest mobility, was more epithelial-like. In contrast, the JSCC2 SEL cells possessed increased vimentin, reduced cytokeratin, showed excellent invasion, produced high levels of IL8, had lower IL6 release and proliferation indices, and were more mesenchymal-like. Notably, the three selected agents collectively suppressed protumorigenic properties in both of these diverse CSCE populations. Furthermore, due to their membrane-associated receptor blocking mechanism of action, tocilizumab and reparixin will likely demonstrate enhanced efficacy in vivo.

Despite the cell line differences in proliferation, the chemopreventives showed similar growth-suppressive responses. Fenretinide routinely suppressed cell proliferation across all cell lines. These data correspond to fenretinide's inhibition of growth-promoting tyrosine kinases FAK, STAT3, and c-Src, and its ability to induce apoptosis and differentiation (12, 15, 16). Single-agent tocilizumab did not affect cell growth. These data are not particularly worrisome as tocilizumab demonstrated excellent in vivo chemopreventive efficacy (12), and our strategy is to employ multiple agents in patients. Importantly, agent combinations did not antagonize fenretinide's growth-suppressive effects.

Elevated IL6 and IL8 levels within the tumor microenvironment support proliferation, inflammation, and angiogenesis (20, 21, 38, 44), and at the systemic level correlate with a less favorable prognosis in persons with OSCC (18, 23). IL6 facilitates OSCC tumorigenesis by EMT induction and promotes metastases via activation via the JAK–STAT3–SNAIL signaling cascade (45). Our data confirm OSCC cells release appreciable levels of both IL6 and IL8. Blockade of the cognate receptors (IL6R, sIL6R via tocilizumab, CXCR1/CXCR2 via reparixin, respectively) in vivo could therefore suppress both intracrine and paracrine loops in tumor cells and the tumorigenic stroma. Combined treatment with fenretinide, reparixin, and tocilizumab significantly suppressed IL6 and IL8 release in all cell lines, findings that suggest a comparable reduction of angiogenesis, inflammation, and proliferation in vivo.

Combination treatment with fenretinide, tocilizumab, and reparixin also significantly reduced expression of key stem cell enabling genes, i.e., ALDH, ABCG2, SOX2, and CD24. These genes are associated with xenobiotic/drug detoxification (ALDH, ABCG2, and SOX2), proliferation and survival (ALDH and SOX2), and invasiveness (SOX2 and CD24) that are all essential to sustain the cancer stem cell phenotype. These results likely reflect reduced IL6 (NF-IL6, STAT3-tocilizumab) and IL8 (NF-KB–reparixin) mediated transcription factor activation in combination with fenretinide's interference via active site binding with signaling kinases, e.g., STAT3 and c-Src (12, 15, 21, 24, 45). Similarly, chemopreventive studies by Castro and colleagues demonstrated that sulforaphane suppressed expression of stem cell–specific genes in triple-negative breast cancer cells (46). The authors speculated sulforaphane inhibited CR1 complex formation with ALK4 and GRP78 (46).

Provided the key roles of YAP and ZEB in promoting stemness, suppression of E-cadherin, and enhanced mobility (47), the migration-inhibitory effects of siRNA on these proteins were not surprising. The chemopreventives, however, only modestly reduced ZEB and YAP, and slightly increased inactive phosphorylated YAP in the 2095sc SEL cells. Although the observed migration-inhibitory effects that we have previously reported (15) do not reflect reduction in ZEB/YAP protein, other mechanisms including fenretinide cytoskeletal disruptions (15) or inhibition of downstream ZEB/YAP interactions are likely attributable for our previously observed fenretinide-mediated migration inhibitory effects (15).

Basement membrane invasion is the ultimate step in malignant transformation of dysplastic surface epithelium (48). Because invasion requires coordination of intricate processes that include cell mobility (myofilaments, actin cytoskeleton), cell–ECM interactions (12, 15), and invadopodia formation (actin core, proteases, regulatory components), there are multiple potential sites for chemopreventive intervention (48). Our lab and numerous others have previously reported inhibition of cancer cell invasion by chemopreventives (15, 49–51). Although triple treatment provided the most extensive invasion suppression across all lines, our data show that fenretinide provided a large anti-invasion impact. Probable underlying mechanisms for these invasion-suppression data include: (1) fenretinide's destabilization of actin filaments (16), (2) fenretinide's perturbation of CSCE–ECM interactions via active site obstruction of STAT3, FAK, and Src (12, 15), (3) tocilizumab's suppression of IL6-mediated STAT3 activation (44), and (4) reparixin's inhibition of IL8-initiated invasion (26).

These studies used micromolar chemopreventive levels, which would not be achievable at the target site by systemic drug administration, e.g., high fenretinide systemic pill dosing was ineffective in OSCC secondary chemoprevention (52). The pharmacologic advantage conveyed by local drug delivery, however, can provide therapeutically relevant micromolar target-tissue drug levels (53) without drug-related adverse systemic effects (54). Another essential treatment consideration is the tumor-promoting effects, e.g., growth and angiogenic factor release, and immune suppression of the microenvironment (55). Importantly, locally delivered fenretinide, tocilizumab, and reparixin would also diminish stromal effects. Collectively, these data combined with the prospect of controlled release local delivery formulations introduce a plausible OSCC tertiary chemoprevention strategy.

Although Dr. Schwendeman is a co-inventor of the fenretinide patch, in accordance with the University of Michigan standards, he did not declare a conflict of interest. S.R. Mallery has an ownership interest (including patents) in a patent in review for the fenretinide patch. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.R. Mallery, D. Wang, B. Santiago, P. Pei, S.P. Schwendeman

Development of methodology: S.R. Mallery, D. Wang, B. Santiago, P. Pei, S.P. Schwendeman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Wang, P. Pei, C. Bissonnette, J.A. Jayawardena, R. Spinney

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.R. Mallery, D. Wang, P. Pei, C. Bissonnette, J.A. Jayawardena, R. Spinney

Writing, review, and/or revision of the manuscript: S.R. Mallery, C. Bissonnette, R. Spinney

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.R. Mallery, D. Wang, P. Pei, C. Bissonnette, J. Lang

Study supervision: S.R. Mallery

These studies were funded in part by NIH grants R01CA227273 (to S.R. Mallery) and R01CA211611 (to S.R. Mallery).

The authors would like to acknowledge the assistance provided by Mary Marin and Merritt Bernath for their histotechnologic expertise in sectioning of the fixed tissues.

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.