MTA3-SOX2 Module Regulates Cancer Stemness and Contributes to Clinical Outcomes of Tongue Carcinoma

Cancer cell plasticity plays critical roles in both tumorigenesis and tumor progression. Metastasis-associated protein 3 (MTA3), a component of the nucleosome remodeling and histone deacetylase (NuRD) complex and multi-effect coregulator, can serve as a tumor suppressor in many cancer types. However, the role of MTA3 in tongue squamous cell cancer (TSCC) remains unclear although it is the most prevalent head and neck cancer and often with poor prognosis. By analyzing both published datasets and clinical specimens, we found that the level of MTA3 was lower in TSCC compared to normal tongue tissues. Data from gene set enrichment analysis (GSEA) also indicated that MTA3 was inversely correlated with cancer stemness. In addition, the levels of MTA3 in both samples from TSCC patients and TSCC cell lines were negatively correlated with SOX2, a key regulator of the plasticity of cancer stem cells (CSCs). We also found that SOX2 played an indispensable role in MTA3-mediated CSC repression. Using the mouse model mimicking human TSCC we demonstrated that the levels of MTA3 and SOX2 decreased and increased, respectively, during the process of tumorigenesis and progression. Finally, we showed that the patients in the MTA3low/SOX2high group had the worst prognosis suggesting that MTA3low/SOX2high can serve as an independent prognostic factor for TSCC patients. Altogether, our data suggest that MTA3 is capable of repressing TSCC CSC properties and tumor growth through downregulating SOX2 and MTA3low/SOX2high might be a potential prognostic factor for TSCC patients.


INTRODUCTION
More than 500,000 new cases of oral and pharyngeal cancers are diagnosed yearly worldwide (World Cancer report 2014, https://www.who.int/cancer/publications/WRC_2014/en/). Oral cancer is malignant neoplasia which arises on the lip or oral cavity. Although progress has been made in cancer treatments, the oral cancer survival rate has not been improved significantly for decades (1,2). Since ∼90% of these cancers are histologically shown to be originated from squamous cells (3), this subtype of cancers is traditionally defined as oral squamous cell carcinoma (OSCC). OSCC may show different levels of differentiation and has a propensity to coincide with lymph node metastasis (4). Among OSCC, tongue squamous cell cancer (TSCC) has the highest incidence and is usually associated with a poor survival rate. Therefore, TSCC is one of the most lethal types of cancers in the head and neck region (5). Thus, a better understanding of the underlying mechanisms in TSCC development will provide not only a more reliable biomarkers for diagnosis and prognosis but also potential therapeutic targets for the treatment of this cancer.
Cellular plasticity plays critical roles in tumor initiation, progression, and metastasis. It is now well-established that stem cell-like cancer cells or cancer stem cells (CSCs) are responsible for both cell plasticity and treatment (6)(7)(8)(9)(10). CSCs are a small subset of cancer cells and multiple lines of evidence indicate that CSCs are responsible for tumor initiation, indefinite, and progression (7,11,12). Accumulating data also indicate that plasticity of CSCs closely correlates with recurrences and metastasis (13)(14)(15)(16) with poor prognosis in a wide variety of cancers, including tongue cancer (17)(18)(19)(20).
Metastasis-associated protein 3 (MTA3) is a multi-effect coregulatory factor and plays indispensable roles in cell proliferation, tumorigenesis, and metastasis (21)(22)(23)(24)(25)(26). Compelling evidence suggests that MTA3 is a tumor suppressor in many cancer types (26)(27)(28) by serving as an integral subunit of the nucleosome remodeling and histone deacetylase (NuRD) complex (21,25,27). As a transcriptional corepressor (29), MTA3 either directly or indirectly regulates the expression and activity of EMT-associated genes such as Snail and Ecadherin (25,27). Dysregulation of MTA3 has been observed in many different human tumors (26)(27)(28). Reduced levels of MTA3 lead to the upregulation of Snail and subsequently enhance the process of epithelial-mesenchymal transition (EMT) (25,27,30). Consistently, the dysfunctional MTA3 reduces cell-cell adhesion and promotes cancer invasion and metastasis (23,26). Moreover, reduced expression of MTA3 in tumor specimens has been associated with poor survival and therefore the expression of MTA3 has been suggested as an independent predictor of patient prognosis in uterine non-endometrioid carcinomas, gastroesophageal junction adenocarcinoma, glioma, and colorectal cancer (27,28,31,32). However, the role of MTA3 and the underlying mechanism of MTA3's function in TSCC remain largely unknown.
In this study, we found that reduced levels of MTA3 in the patient specimens correlated with poorer clinical outcomes with concurrently increased cancer stemness. We also showed that MTA3 was capable of repressing cancer cell proliferation through inhibiting SOX2 expression. Using a chemical-induced mouse model of TSCC, we demonstrated that MTA3 and SOX2 decreased and increased, respectively, during the process of carcinogenesis and progression. Finally, our findings suggested that MTA3 low /SOX2 high could potentially serve as an independent prognostic factor for TSCC patients. A total of 119 patients with TSCC were recruited at the Affiliated  Tumor Hospital of Shantou University Medical College from  2009 to 2011 and their TSCC were clinically diagnosed and  histologically confirmed. The primary TSCC specimens and their matched non-cancerous tissues were paraffin-embedded. Samples from patients who underwent preoperative radiotherapy or chemotherapy for TSCC were excluded. Clinical research protocols of this study were reviewed and approved by the Ethics Committee of Shantou University Medical College.

Immunohistochemistry (IHC)
Tissue sections (4 µm) from the formalin-fixed paraffinembedded clinical specimens or 4NQO-induced tongue tumor tissues were processed and immune-stained with antibodies against MTA3 (Catalog No. A300-160A, Bethyl, 1: 600), SOX2 (Catalog No. 23064, Cell Signaling, 1: 200), each with at least two cores of the primary tumor as well as two cores of normal tongue tissue. Sections immune-labeled with rabbit IgG or mouse IgG as the primary antibody were used as negative controls, known MTA3 and SOX2 positive slides were used as a positive control.
The tissue sections were scored by two pathologists blind to the clinical outcomes. Receiver operating characteristic (ROC) curves were employed to define an optical cut-off score, which was closest to the point with maximum sensitivity and specificity. The cases with scores lower than or equal to the cut-off value were designated as low expression group and those with higher scores were categorized as high expression group.

Histological Analysis
For histological analysis, tissues were fixed in 4% neutral buffer formalin, embedded in paraffin, sectioned (4 µm) and stained with hematoxylin and eosin.

Immunofluorescence Staining
FFPE tissue sections were deparaffinized and dehydrated in xylene and graded ethanol solutions in preparation for MTA3 and SOX2 double immunofluorescence (IF) staining. All slides were subjected to heat-induced epitope retrieval in Citrate Buffer [pH = 6.0]. Endogenous tissue peroxidases were blocked by incubating the slides in 3% hydrogen peroxide solution and blocking buffer before incubation with MTA3 (Catalog A300-160A, Bethyl, 1: 600), SOX2 (Catalog No. #23064, Cell Signaling, 1: 200) as primary antibodies. And HRP-conjugated streptavidin as the secondary antibody. The signal in IF labeled slides were visualized with AlexaFluor 488 and AlexaFluor 594 Tyramide Super Boost kits (Invitrogen, Carlsbad, CA), and nuclei were visualized with Prolong Diamond Antifade Reagent with 4' ,6-diamidino-2-phenylindole (DAPI; Invitrogen, Carlsbad, CA). Primary and secondary antibodies were stripped using Citrate Buffer [pH = 6.0] in the microwave. Known MTA3 and SOX2 positive slides were used as a positive control. Immunofluorescence staining was analyzed using the PerkinElmer Vectra analysis platform to estimate the cell numbers. The percentage of positive cells was estimated by two pathologists.

Cell Culture
Human TSCC cell lines (SCC-25 and SCC-4) were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The cells were cultured in Dulbecco's Modified Eagle Media (DMEM, Gibco/Invitrogen) supplemented with 10% FBS (Gibco/Invitrogen) at 37 • C in a humidified atmosphere containing 5% CO 2 .

Virus Production and Transduction
The full-length cDNA of MTA3 was PCR amplified from SCC-25 cells and cloned into the pCDNA3.1-flag plasmid. The shRNA targeting human MTA3 (target sequence: GAGGATACCTTCTTCTACTCA) was cloned into pBabe/U6 plasmid. The SOX2 overexpression plasmid and SOX2 short hairpin RNA (shSOX2) plasmid (shSOX2 target sequence: GGTTGACACCGTTGGTAATTT) were obtained from GeneCopoeia. Transfection of plasmid was performed using Lipofectamine 3000 (Thermo Fisher Scientific, catalog no. L3000015) according to the manufacturer's instructions. Stable cells were selected by culturing the cells in the medial with puromycin for 2 weeks.

ALDEFLUOR Assay
ALDEFLUOR assay kit (Stem Cell Technologies TM , Vancouver, BC, Canada) was used to determine ALDH1 activity according to the manufacturer's protocol. Cells were suspended in ALDEFLUOR assay buffer containing 1 µM per 1 × 10 6 cells of the ALDH substrate, boron-dipyrromethene-aminoacetaldehyde (BAAA), and incubated for 50 min at 37 • C. Each sample was treated with 50 mM of an ALDH-specific inhibitor, and diethylaminobenzaldehyde (DEAB) as a negative control. Stained cells were analyzed by BD FACSAria TM II (BD Biosciences, San Jose, CA, USA). To evaluate cell viability the cells were stained with 1 mg/ml of propidium iodide prior to analysis.

Proliferation and Survival Assays
Real-time cell analysis (RTCA) was performed to estimate cell proliferation using the xCELLigence DP device (ACEA Biosciences) as described in the supplier's instructions. In brief, 3,000 cells were seeded in E-plates, and the plates were locked into the RTCA DP device supplied with humidified air with 5% CO 2 at 37 • C. The proliferative ability was monitored by the xCELLigence RTCA Analyzer (Roche Applied Science, Mannheim, Germany) (34,35).

Animals and Carcinogen Treatment
Male and female wild-type C57BL/six mice were supplied by Beijing Vital Laboratory Animal Technology (Beijing, China). To induce tumorigenesis in the tongue, 4NQO (Sigma-Aldrich, St. Louise, MO) was added to the drinking water (100 mg/mL) for the 6-week-old young adult mice for 16 weeks. The mice were sacrificed when the bodyweight loss >1/3, otherwise they were sacrificed at the indicated time. After sacrifice, the tongue surface was photographed. The tissues were resected for histopathological examination and immunohistochemistry (IHC) analyses. Animals were housed in pathogen-free conditions at the Animal Center of Shantou University Medical College in compliance with Institutional Animal Care and Use Committee (IACUC) regulations (SUMC2014-148). All animal experiments were performed according to protocols approved by the Animal Care and Use Committee of the Medical College of Shantou University.

Gaussia Luciferase Assay
The pEZX-PG04 plasmid carrying double-expression cassette for Gaussia luciferase under the control of the SOX2 promoter (+270 to −1038), and secreted Alkaline Phosphatase (SeAP) under the control of the CMV promoter was obtained from GeneCopoeia (Catalog No. HPRM15202). Cells were seeded in 24-well plates, and transiently transfected with the above plasmid using Lipofectamine 3000 (Thermo Fisher Scientific, catalog no. L3000015) according to the manufacturer's instructions. After 72 h of transfection, the culture medium was collected for analysis of Gaussia luciferase and secreted Alkaline Phosphatase (SeAP) activities using a Secrete-PairTM Dual Luminescence Assay Kit (GeneCopoeia, SPDA-D010) according to the manufacturer's instructions. Gaussia luciferase activity was normalized on the basis of seAP activity.

Statistical Analyses
All statistical analyses except for microarray data were carried out using the statistical software package SPSS 17.0 (SPSS, Inc., Chicago, IL, USA). The comparisons between two groups were performed with Student's t-test. The correlation between MTA3 expression and clinicopathological data of patients was analyzed with Pearson χ2 test. Survival curves were plotted with the Kaplan-Meier method and compared by log-rank test. Survival data were evaluated by univariate and multivariate Cox regression analyses. The correlations of the histoscore between MTA3 and SOX2 was determined by Spearman's rank test. Two-way ANOVA followed by a Tukey-Kramer post hoc test was performed to compare the difference of proliferation affected by MTA3 and SOX2 among four groups. All data were presented as the mean ± SEM. The P < 0.05 was considered statistically significant.

MTA3 Is Reduced in Human TSCC
To estimate the expression MTA3, we first assessed the mRNA levels of MTA3 in OSCC from GEO database (https://www. ncbi.nlm.nih.gov/geo/) GSE30784 (36) and GSE25099 (37). We found that the MTA3 mRNA levels were significantly lower in OSCC when compared with the normal controls (P < 0.001 and 0.01, respectively; Figure 1A and Supplementary Figure 1A). Since TSCC is the highest incidence of all oral squamous cell cancers (5), we focused on the role of MTA3 in TSCC. Data from both datasets GSE78060 (38) and GSE34105 (39) revealed higher MTA3 mRNA levels in normal tongue tissues than in TSCC tissues (P = 0.014 and 0.003, respectively; Figure 1B and Supplementary Figure 1B). Next, we examined the MTA3 expression at protein levels in TSCC of 119 patient specimens using immunohistochemistry (IHC). Representative Frontiers in Oncology | www.frontiersin.org photomicrographs for MTA3 IHC scores of level 0, 4, 6, 9, and 12 are shown in Figure 1C (left panel). TSCC showed significantly (P < 0.001, n = 119) lower levels of MTA3 protein in the primary tumors compared to the corresponding normal tissue (Figure 1C, Right panel). These findings demonstrate that MTA3 is downregulated in TSCC tissues compared to normal controls.

Downregulation of MTA3 Correlates With Clinical Outcomes in TSCC Patients
We next assessed the prognostic impact of MTA3 on the outcome of TSCC patients. An optimal cutoff value was identified using Receiver operator characteristic (ROC) analysis which categorized 51.3% (61/119) of the patient cohort into a high MTA3 group and the remainder into a low MTA3 group (Figure 2A). Then Kaplan-Meier survival analyses were performed and showed that patients with low MTA3 were associated with shorter overall survival than those with high MTA3 (P = 0.002, Figure 2B).
Univariate analyses found that MTA3 expression, pTNM stage, pN status, and tumor depth were significantly related to TSCC patient outcome ( Table 1). However, after multivariate Cox regression analysis only MTA3 expression (HR 0.420; 95% CI 0.218-0.810; P = 0.010) and pTNM stage (HR 3.029; 95% CI 1.075-8.538; P = 0.036) were independently significant with overall survival ( Table 1). These results reveal that reduced expression of MTA3 may be an independent prognostic factor for the overall survival of patients with TSCC.

MTA3 Inhibits Key TSCC Plasticity Regulator SOX2
The plasticity of CSC plays an important role in oncogenesis and progression (7,11,12,40). Therefore, we next explored the relationship between MTA3 level and cancer cell stemness using gene set enrichment analysis (GSEA) of published human TSCC expression profiles (GSE78060) and found that a cancer stemness related gene signature (BOQUEST_STEM_CELL_UP) was significantly enriched in TSCC with low MTA3 expression (P < 0.001, Figure 3A). Interestingly, SOX2, a key regulator in the plasticity of cancer stemness, was also closely associated with poor prognosis in TSCC patients (20). To study the relation between MTA3 and SOX2, we measured the (co-)expression   of MTA3 and SOX2 in adjacent non-tumor tissues (ANT) and TSCC tissues from the 119 patients. Indeed, a highly significant inverse correlation between the MTA3 and SOX2 levels (r 2 = 0.534, P < 0.001 and r 2 = 0.624, P < 0.001, respectively; Figures 3B,C) was observed. Moreover, a low level of SOX2 was accompanied by high expression of MTA3, and vice versa, as indicated by double immunofluorescent staining ( Figure 3D). In addition, two TSCC cell lines, SSC-25 and SSC-4, were transfected with shMTA3 or the plasmid overexpressing MTA3, respectively. The mRNA and protein levels of MTA3 were then assessed by qRT-PCR and western blot analysis, respectively. When compared to the controls, the expression level of MTA3 was obviously enhanced in cells transfected with pcMTA3 (P < 0.001), while it was significantly decreased in cells transfected with shMTA3 (P < 0.001) (Figures 3E,F). Knockdown MTA3 dramatically increased, whereas overexpression of MTA3 decreased, the expression levels of SOX2 (Figures 3E,F), suggesting that the MTA3 inhibits SOX2 expression in TSCC cells.

MTA3 Suppresses TSCC Cell Stemness and Proliferation Via SOX2
To determine whether SOX2 plays a key role in MTA3mediated plasticity of CSC and cell growth, we first established TSCC cells with knockdown of MTA3 or SOX2 alone or in combination using specific shRNA (Figure 4A) or stably overexpressing MTA3 or SOX2 alone or in combination ( Figure 4B). As expected, knockdown MTA3 significantly promoted the percentage of cells expressing ALDH1, a major CSC marker (Figure 4C). MTA3-depletion-mediated increase of the number of ALDH1-positive cells were significantly reduced when SOX2 is removed suggesting SOX2 plays a crucial role in MTA3-repressed CSC properties ( Figure 4C). Next, the effects of overexpressed MTA3 individually or in combination with SOX2 were examined. Figure 4D showed that overexpression of MTA3 and SOX2 reduced and increased the number of ALDH1-positive cells, respectively. However, overexpressed SOX2 counteracted MTA3-repressed the number of ALDH1-positive cells (Figure 4D). In addition, real-time proliferation assays were conducted to assess changes in cell dynamics. Silencing SOX2 inhibited MTA3-depletion-induced cell proliferation ( Figure 4E) and forced SOX2 expression was capable of counteracting MTA3-repressed cell proliferation ( Figure 4F). These data altogether demonstrate that in TSCC cells MTA3 represses CSC property and proliferation by targeting SOX2.

Dysregulated MTA3/SOX2 Axis Is Associated With Tumor Progression
The results above indicate a role of MTA3/SOX2 in TSCC progression. Therefore, we next investigated the role of MTA3/SOX2 in a mouse model of tongue tumorigenesis induced by 4 nitroquinoline 1-oxide (4NQO). Exposure to 4NQO caused a temporal progression from hyperplasia to invasive carcinoma in the murine tongue, resembling human tongue carcinogenesis and development (41,42). Mice were exposed to 4NQO in daily drinking water for 16 weeks followed by 4NQO-free drinking water for 12 additional weeks (Figure 5A), which resulted in tongue carcinogenesis and progression (Figure 5B) similar to what has been published previously (41,42). Next, we studied the expression of MTA3/SOX2 in normal tongue tissue samples, hyperplasia, carcinoma in situ, early invasive carcinoma and invasive carcinoma by immunohistochemistry. As shown in Figure 5C, the expression of MTA3 gradually decreased during the process of carcinogenesis and progression, in contrast to SOX2. This is indicative of an inverse relation between MTA3 and SOX2 associated with cancer occurrence and development, which was confirmed in mouse TSCC by double immunofluorescent staining ( Figure 5D). Thus, the levels of MTA3 and SOX2 may change dynamically during tongue carcinogenesis and progression.

Dysregulated MTA3/SOX2 Axis Associated Strongly With Poor Prognosis
To validate our findings in a clinical setting, we assessed the levels of MTA3 and SOX2 in 119 TSCC samples and related this to the overall survival rate. Receiver operator characteristic (ROC) analysis identified an optimal cutoff value and categorized 39.5% (47/119) of the patients into a high SOX2 group and the remainder into a low SOX2 group ( Figure 6A). Indeed, patients with low levels of MTA3 but high levels of SOX2 (MTA3 low /SOX2 high ) had significantly shorter overall survival rate (P < 0.001; Figure 6B) than those with high levels of MTA3. Moreover, low levels of SOX2 (MTA3 high /SOX2 low ), were associated with an even shorter overall survival compared to TSCC patients with any other expression pattern (MTA3 high /SOX2 high , MTA3 low /SOX2 high , MTA3 low /SOX2 low ) (P < 0.001; Figure 6C). These data altogether suggest that combined MTA3 low -SOX2 high expression had stronger correlation with worst patient prognosis than that of the individual components. Indeed, univariate and multivariate Cox regression analysis (P < 0.001 and P < 0.001, HR = 4.044, 95 %CI = 1.925-8.495, respectively) indicated that combined MTA3 low /SOX2 high expression is an independent prognostic factor of TSCC as it was significantly associated with prognosis ( Table 2). Overall, these data indicate that dysregulated MTA3/SOX2 axis may contribute to patient outcomes and could be of value as a predictive biomarker for TSCC prognosis.

DISCUSSION
Here we demonstrated that MTA3 was a potential independent prognosis factor for TSCC patients. Moreover, we established the negative regulation of SOX2, a key regulator in the plasticity of  cancer stemness, by MTA3 repressed the number of ALDH1positive cells and cell proliferation in TSCC cells. We identified the expression of MTA3 and SOX2 as important markers in a 4NQO-induced TSCC mouse model, human TSCC progression and clinical outcomes.
The biological and clinical significance of the MTA family members in malignancies has been well-established. The levels of these factors have even been proposed as potential diagnostic parameters and targeting each one of the family members could be potential treatments for different cancers (26,(43)(44)(45). However, unlike MTA1 and MTA2 which were mainly involved in cancer progression and metastasis, MTA3 possesses both tumor-suppressing and tumor-promoting properties depending on specific cancer types (21,43). We found that MTA3 was negatively associated with overall survival and could act as an independent prognosis factor in TSCC. To the best of our knowledge, this is the first study to show the expression of MTA3 in TSCC.
The plasticity of CSCs is key mechanisms in oncogenesis and progression of cancers (7,11,12,16) and a predictive factor of poor prognosis in a wide variety of cancers, including tongue cancer (17)(18)(19)(20). Accumulating evidence suggests that transcription factor SOX2 is a key regulator in the plasticity of cancer stemness (46)(47)(48). Moreover, overexpression of SOX2 is often associated with increased cancer aggressiveness, resistance to chemoradiation therapy and decreased survival rate, which has been reported in various cancer types (49,50), including TSCC (51). Previous studies revealed that SOX2 is vital in the regulation of TSCC motility, invasion, tumorigenicity, and upregulated SOX2 is significantly associated with the progression of TSCC (51). SOX2 is detectable in oral pre-invasive lesions, suggesting that SOX2 upregulation may be an early event in TSCC carcinogenesis. We found that MTA3 can inhibit CSC properties and cell proliferation via downregulating SOX2 in TSCC cells. In addition, luciferase reporter assays showed that knockdown or overexpression MTA3 had no significant effect on the luciferase activity (Supplementary Figure 2), suggesting that MTA3 does not regulate the expression of SOX2 by interacting with its proximal promoter. However, these results do not necessarily rule out the possibility that MTA3 is directly involved in the regulation of SOX2 expression by interacting with its enhancer region or the sequence other than the proximal promoter. Further studies will better the understanding about the mechanisms in MTA3-regulated SOX2 expression in tongue cancer.
Given the fact that alterations of MTA3 and SOX2 highly correlate with clinical outcomes and the predictability of prognosis, the levels of MTA3 and SOX2, especially MTA3 low /SOX2 high , could be used as diagnostic parameters. In addition, targeting MTA3 and/or SOX2 could be potential therapeutic strategies in TSCC treatment. Finally, 4NQO-induce TSCC mice model has been widely used in the study of TSCC (41,42). And we used this model to further elaborate the dynamic changes of MTA3 and SOX2 in the occurrence and developmental progression of TSCC.
In summary, consistent with the fact that MTA3 is often silenced in the process of carcinogenesis and development, we found that MTA3 was capable of inhibiting CSC properties and cell proliferation by negatively regulating SOX2. Additionally, we found that TSCC patients in the MTA3 low /SOX2 high group had a poor prognosis. Future research may involve translational research to clinically evaluate the efficacy of this therapeutic strategy using IHC evaluation of low MTA3 and high SOX2 as a companion diagnostic for patient selection.

DATA AVAILABILITY
The datasets generated for this study are available on request to the corresponding author.