Targeting MDA-9/Syntenin-1 (SDCBP) as a Strategy to Eliminate Head and Neck Squamous Cell Carcinoma Stem Cells

Abstract

Melanoma differentiation associated gene-9 (MDA-9), also known as Syntenin-1 or SDCBP, exhibits elevated expression in multiple cancers, promoting invasion, migration, and tumor cell survival. The TCGA database reveals high MDA-9 expression in late-stage HPV (−) HNSCC tissues, which associates with poor patient survival. Moreover, bioinformatics analyses support a potential role of MDA-9 in promoting cancer stemness in HNSCC tissues. We characterized the functional role of MDA-9 in HNSCC tumorigenesis, particularly in relation to cancer stem cell maintenance. Knockdown of MDA-9 suppresses BMI1, a functional cancer stem cell marker in HNSCC, and the self-renewal of cultured HNSCC cells. Mechanistically, MDA-9 controls expression of BMI1 through NF-κB p65. Targeting MDA-9 with a specific small molecule pharmacological inhibitor, IVMT-Rx-4, inhibits HNSCC growth and metastasis in a HNSCC xenograft mouse model. In vivo lineage tracing reveals that IVMT-Rx-4 potently inhibits self-renewal and tumorigenicity of Bmi1⁺ cancer stem cells in a spontaneous model of HNSCC. Additionally, IVMT-Rx-4 significantly improves treatment with cisplatin by eliminating cisplatin-enriched Bmi1⁺ cancer stem cells, thereby overcoming cisplatin resistance. Taken together, our results demonstrate that targeting MDA-9 with IVMT-Rx-4 is an effective strategy for treating HNSCC by eliminating cancer stem cells, and IVMT-Rx-4 is a promising chemical entity that may be amenable for developing an effective therapy for HNSCC.

Introduction

Head and neck squamous cell carcinoma (HNSCC) accounts for 90% of all head and neck tumors, which are highly invasive and metastatic¹. With a 5-year survival rate of approximately 50%, it remains one of the deadliest tumors worldwide. Depending on the HNSCC region involved, surgical intervention is commonly used for oral cavity cancers, while radiation is more commonly employed for laryngeal and pharyngeal cancers². With the absence of effective targeted therapies, radiotherapy remains the best treatment modality for HNSCC. Chemotherapy is an option for patients with recurrent metastatic disease and replaces radical surgical intervention by centering on neoadjuvant or adjuvant settings³. Moreover, concurrent chemo/radiotherapy has increased the survival rates in clinical settings for patients with inoperable tumors^{4, 5}. At present, cisplatin and platinum-based chemotherapy remain the first-line treatment of many solid tumors, especially for HNSCC patients, and represents the gold standard for patients with pathological high-risk HNSCC⁶. Although HNSCC patients initially respond to cisplatin, patients can progress to cisplatin-resistance within 6 months and relapse, inevitably leading to death^{7, 8}. Consequently, there is an urgent need for approaches to both prevent and effectively treat HNSCC and to overcome cisplatin resistance ^9–12.

Cancer stem cells (CSCs) comprise a small subpopulation of cancer cells responsible for tumor initiation and resistance to conventional chemo/radiotherapy^{10, 11, 13, 14}. Moreover, it is established that cisplatin can enrich the CSCs population, ultimately contributing to chemoresistance^15–17. Accordingly, targeting CSCs may provide an efficient approach to improve HNSCC chemotherapy outcomes and overall patient survival^{11, 13}. To achieve this outcome, it is imperative to elucidate the molecular mechanisms that maintain the functional properties of CSCs and devise new therapeutic strategies for effectively eradicating HNSCC.

Melanoma differentiation associated gene-9 (MDA-9), also known as Syntenin-1 or syndecan binding protein (SDCBP), is a pro-metastatic gene upregulated in multiple cancers and negatively correlates with patient survival and response to chemotherapy. This gene was initially identified using subtraction hybridization from human metastatic melanoma^{18, 19} and shown to be a key metastasis progression gene in melanoma^20–23. Subsequent research documented that MDA-9 is also associated with many other cancers, including breast cancer, lung carcinoma, brain cancer, prostate cancer, urothelial cancer, neuroblastoma, and HNSCC^{11, 20–27}. MDA-9 positively regulates tumor metastasis through its involvement in EMT induction, migration and invasion, angiogenesis, and tumor-cell survival^{12, 24, 28, 29}. In HNSCC, MDA-9 is a promoter and prognostic marker of invasion and metastasis³⁰; a regulator of differentiation and angiogenesis³¹; and to positively control cisplatin resistance³². These discoveries suggest that MDA-9 might serve as a novel therapeutic target for HNSCC treatment.

Considering the documented role of MDA-9 in cancer progression and in exosomes^{25, 33–35}, this protein represents a desirable target for developing potential cancer-specific drugs^{25–27, 36}. Small molecules targeting and inhibiting MDA-9 activity have been identified^{25, 28, 29, 37–43}. MDA-9 is an adaptor (scaffold) protein that binds to specific partner proteins, i.e., MDA-9 (homodimers), Slug, ILK, SRC, EGFR, and many other proteins^{22, 26, 27, 29, 44, 45}. MDA-9 lacks intrinsic enzymatic activity; thus, drug design approaches have focused on targeting the protein interaction interfaces. The small molecule inhibitor PDZ1i was discovered through fragment-based drug discovery informed by NMR, and PDZ1i was shown to disrupt the MDA-9/Syntenin-1/SDCBP protein-protein interactions^{25, 28, 29, 38, 39}. PDZ1i is synthesized by combining two independent synthesis products into a single small molecule²⁸. One of the synthesis products is IVMT-Rx-4, which structurally is predicted to bind to the PDZ1 domain of MDA-9, which we now show unexpectedly displays equivalent MDA-9 inhibitory functions as does the complete parent compound, PDZ1i. Tumor spread/invasion/metastasis are inhibited by PDZ1i in glioblastoma (GBM), prostate cancer, breast cancer, and neuroblastoma^{28, 38, 39, 46}. IVMT-Rx-4 differs in specific physical-chemical properties from PDZ1i, while retaining many of the desired biological properties. Specifically, IVMT-Rx-4 is significantly more soluble and shows less efflux from cells than PDZ1i and can be formulated to be delivered intravenously (IV), whereas PDZ1i is poorly soluble, requires DMSO and can be delivered intraperitoneally (IP) but not IV. This solubility issue makes IVMT-Rx-4 a preferable drug candidate than parental unmodified PDZ1i^{28, 47}. Based on these considerations, we focused on IVMT-Rx-4 as the small molecule of choice to target MDA-9 in HNSCC. To permit a direct comparison with PDZ1i, IVMT-Rx-4 was dissolved in DMSO and administered IP to determine biological properties in HNSCC.

In this report, we focused on the relationship between MDA-9 expression and the maintenance of CSCs in HNSCC. We document a role of MDA-9 expression in regulating invasion, stemness (potentially through BMI1, a functional CSC marker)^42–45, resistance to cisplatin, and tumorigenesis in HNSCC. Targeting MDA-9 using the small molecule inhibitor IVMT-Rx-4 suppresses the functional properties of Bmi1⁺ CSCs in a HNSCC spontaneous mouse model. IVMT-Rx-4 effectively eradicates CSCs, thereby inhibiting HNSCC tumorigenesis and overcoming cisplatin resistance in vivo. Taken together, our study provides new insights into the regulation of CSC maintenance in HNSCC by MDA-9 and identifies this gene/protein as a novel direct drug target and establishes IVMT-Rx-4 as a promising chemical probe that may provide a scaffold for developing more effective therapeutics for HNSCC.

Results

High expression of MDA-9 is associated with poor outcome in patients with HNSCC

We first analyzed the clinical data available from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) to determine the potential role of MDA-9 in the survival outcome of HNSCC patients. Our analysis showed that MDA-9 is highly expressed in HPV-negative HNSCC compared to HPV-positive cases for both TCGA dataset⁴⁸ and GEO datasets GSE65858⁴⁹ and GSE117973^{50, 51} (Fig. 1A). Further analysis of the TCGA dataset revealed that, in comparison with normal tissues, high expression of MDA-9 is mainly observed at later stages of HPV-negative HNSCC, particularly at stage IV (Fig. 1B). Consistent with these transcriptomic findings, analysis of the CPTAC HNSCC proteome dataset (PDC000221) also demonstrated elevated MDA-9 protein expression in HPV-negative tumors relative to HPV-positive tumors and normal tissues (Fig. 1C).⁵² Additionally, Kaplan-Meier survival analysis demonstrated that MDA-9 overexpression was associated with poor survival of HNSCC patients (Fig. 1D). Patients with high MDA-9 expression (≥ median) had a shorter mean overall survival (719.2 vs. 828.4 days) and more deaths (18 vs. 7) than those with low MDA-9 expression (< median). Consistent with these findings, both the log-rank test (p = 0.0244) and Wilcoxon test (p = 0.0270) indicated significantly reduced overall survival among patients with elevated MDA-9 expression. MDA-9 expression was higher in patients denoted as “deceased” in the clinical follow-up compared to patients denoted as “alive” in these datasets (Fig. 1E). These clinical data encouraged us to further investigate the mechanistic and functional roles of MDA-9 in HNSCC progression and the consequence of inhibition.

Figure 1: MDA-9 overexpression correlates with poor clinical outcome of HNSCC patients.

(A) Box plots indicate that average MDA-9 expression is elevated in HPV-negative (HPV-N) primary tumors compared to HPV-positive (HPV-P) ones in TCGA, GSE65858 and GSE117973 datasets (p-values of Student’s t-tests are <0.001, 0.0014 and 0.0108 respectively). In the TCGA cohort, average MDA-9 expression in HPV-P tumors is lower than in normal samples (t-test, p=0.0054). (B) Expression of MDA-9 at different clinical stages of HNSCC in both HPV-positive and HPV-negative subsets of the TCGA dataset. MDA-9 is significantly overexpressed at stage IV HPV-N HNSCC compared to normal tissue (p-value of Student’s t-test is 0.02). (C) Analysis of the Clinical Proteomic Tumor Analysis Consortium (CPTAC) HNSCC proteome dataset (PDC000221) demonstrated higher MDA-9 (SDCBP) protein abundance in HPV-negative (HPV-N) HNSCC compared with HPV-positive (HPV-P) HNSCC. Only primary tumors with available HPV status determined by p16 testing were included in the HPV-stratified analysis. Differences among groups were significant by Kruskal–Wallis test (p = 0.0287). Pairwise Wilcoxon comparisons showed higher MDA-9 protein abundance in HPV-N tumors than in HPV-P tumors (p = 0.0160) and normal tissues (p = 0.0391). (D) Kaplan-Meier plot for HPV-N HNSC patients in TCGA dataset demonstrates reduced overall survival for MDA-9-high patients (red line) vs. MDA-9-low patients (blue line). (E) On average, MDA-9 expression was higher in patients recorded as “Deceased” compared to those recorded as “Alive” at clinical follow-up (t-test, p=0.0009). The chart notes the recorded treatment response of patients, describing their outcome after treatment as per NCI Genomic Data Commons (GDC) “Clinical Data Standardization” page.

MDA-9 promotes invasive growth, colony formation, and self-renewal of HNSCC cells

To investigate the role of MDA-9 in HNSCC tumorigenesis, we performed phenotypic cell-based assays to examine whether knockdown of MDA-9 could affect HNSCC cell invasion or colony formation abilities. As shown in Fig. 2A and Supplementary Fig. 1A, siRNA-mediated depletion of MDA-9 significantly inhibited cell invasion capability of UM-SCC-1, HN6, and UM-SCC-22B HNSCC cell lines. Additionally, clonogenic assays demonstrated that the cloning ability of these cells was also significantly inhibited by knockdown of MDA-9 (Fig. 2B and Supplementary Fig. 1B). Next, to further investigate the effects of MDA-9 in promoting HNSCC tumorigenesis and stemness, we established UM-SCC-1, HN6, and UM-SCC-22B cell lines that stably express MDA-9. Overexpression of MDA-9 increased invasion (Fig. 2C and Supplementary Fig. 1C) and colony formation ability (Fig. 2D and Supplementary Fig. 1D) of these cells. To determine whether MDA-9 promotes the self-renewal ability of HNSCC cells, we performed an in vitro tumor sphere formation assay, but with limiting dilution to ensure the accuracy of the measurements. Notably, a significant increase in the sphere re-initiating cell frequency was also significantly enhanced by overexpression of MDA-9, measured by tumor sphere formation assays (Fig. 2E and Supplementary Fig. 1E). Moreover, the size of tumor sphere was also increased by over expression of MDA-9 (Fig. 2F).These results reveal that MDA-9 plays an essential role in HNSCC tumorigenesis.

Figure 2. MDA-9 promotes invasive growth and colony formation of HNSCC cells.

(A) An invasion assay shows that knockdown of MDA-9 results in a significant decrease in invasive abilities of both UM-SCC-1 and HN6 cells. Data represents mean ± SD of triplicate experiments. *P<0.05 and **P<0.01 by Student’s t-test (n=3). (B) A clonogenic assay shows that knockdown of MDA-9 results in a significant decrease in colony formation of both SCC1 and HN6 cells. Data represents mean ± SD of triplicate experiments. (C-E) Overexpression of MDA-9 increased invasive growth (C), colony formation (D), and sphere formation (E) abilities of UM-SCC-1 and HN6 cells. Data represents mean ± SD of triplicate experiments. *P<0.05 and **P<0.01 by Student’s t-test (n=3). Scale Bar: 50μm.

MDA-9 regulates cancer stemness gene BMI1 expression in HNSCC

CSCs are responsible for maintaining cancer progression in many tumors, including HNSCC^{13, 53}. Our phenotypic assays implied that CSCs are involved in MDA-9-mediated HNSCC tumorigenesis (Fig. 2), prompting us to perform further investigations. We focused our attention on BMI1, a stemness gene that has been suggested to maintain functional properties of CSCs in HNSCC ^{17, 54}, and observed that transient depletion of MDA-9 by siRNA dramatically suppressed the expression of this gene in UM-SCC-1, HN6, and UM-SCC-22B cell lines (Fig. 3A, 3B, Supplementary Fig. 2A and 2B). Moreover, MDA-9 overexpression in all three HNSCC cell lines elevated BMI1 protein levels (Fig. 3C and Supplementary Fig. 2C) as well as BMI1 mRNA levels (Fig. 3D and Supplementary Fig. 2D). Taken together, these findings reveal a key association between MDA-9 and BMI1 in regulating HNSCC tumorigenesis via maintenance of CSCs.

Figure 3. MDA-9 regulates expression of BMI1 through NF-κB p65 in HNSCC.

(A) Western blotting results show selective and efficient knockdown of MDA-9 protein levels in UM-SCC-1 and HN6 cells. BMI1 protein levels are decreased with knockdown conditions. Tubulin was used as the loading control. (B) Knockdown of MDA-9 decreased BMI1 transcription in HNSCC cells. (C) Overexpression of MDA-9 increases BMI1 expression in HNSCC cells. (D) RT-qPCR analysis revealed that overexpression of MDA-9 increased BMI1 transcription in HNSCC cells. Data represents mean ± SD of triplicate experiments. *P<0.05 and **P<0.01 by Student’s t-test (n=3). (E) The ChIP-seq signal of p65 was highly enriched at BMI1 in UM-SCC-1 cells. (F) RT-qPCR analysis revealed that knockdown of RELA (p65) suppressed BMI1 transcription in HNSCC cells. Data represents mean ± SD of triplicate experiments. *P<0.05 and **P<0.01 by Student’s t-test (n=3).

MDA-9 regulates expression of BMI1 through NF-κB p65

MDA-9 has been implicated in promoting tumorigenesis across multiple cancer types by modulating NF-κB signaling^{55, 56}. Studies, including our own, have demonstrated that NF-κB p65 (RELA/p65) plays a critical role in sustaining CSCs in HNSCC^57–59. To explore the potential role of p65 in regulating BMI1 expression, we reanalyzed our previously published ChIP-seq data⁵⁷ and confirmed p65 enrichment at the BMI1 promoter in UM-SCC-1 cells (Fig. 3E). Moreover, depletion of p65 with siRNA significantly suppressed BMI1 expression in both UM-SCC-1 and HN6 cells (Fig. 3F), indicating that BMI1 is directly regulated by p65 in HNSCC. In addition, western blot analysis reveals that depletion of MDA-9 leads to a marked reduction in p65 expression (Fig, 3A and Supplementary Fig. 2A), whereas MDA-9 overexpression dramatically increases p65 expression in UM-SCC-1, HN6 cells, and UM-SCC-22B cells (Fig. 3C and Supplementary Fig. 2C). Taken together, these findings strongly suggest that MDA-9 regulates BMI1 expression through p65.

MDA-9 inhibitor IVMT-Rx-4 suppresses malignant phenotypes of HNSCC cells

IVMT-Rx-4 is a specific MDA-9 inhibitor that targets the PDZ1 domain of MDA-9 (Fig. 4A). This molecule is a synthesis intermediate of PDZ1i²⁸ (Supplementary Fig. 3). We characterized the interaction between MDA-9 and IVMT-Rx-4 using a fluorescence-based titration assay, which relies on the quenching of intrinsic Trp fluorescence emission upon ligand binding. As shown in Fig. 4B, the binding affinity of IVMT-Rx-4 for MDA-9 was determined using this assay, yielding a K_D of 0.581 ± 0.086 μM. ADME and PK studies comparing IVMT-Rx-4 and PDZ1i indicate enhanced solubility of IVMT-Rx-4 (160 μg/ml vs. 10 μg/ml, in PBS, pH 7.4), greater lipophilicity of IVMT-Rx-4 (1.49 vs. 0.5 Log D by shake flask at pH 7.4), and reduced efflux of IVMT-Rx-4 (0.92 vs. 300 in Caco-2 bidirectional assay)⁶⁰. Additionally, both compounds have minimal degradation in liver microsomes from human, monkey, dog, rat and mouse with both NADPH and UDPGA added (in all species: IVMT-Rx-4, >3.5 hr. t_1/2, PDZ1i, >4 hr. t_1/2), minimal CYP inhibition (1A2, 2A6, 2B6, 2C8, 2C9, 2D6, 2E1, 3A4 M, 3A4 T), no time dependent inhibition (TDI) of CYP4A, and hERG assay by manual patch-clamp system (>30 mM)⁶⁰.

Figure 4. Putative docking structure of IVMT-Rx-4 with the PDZ1 region of MDA-9.

(A) Structure of MDA-9/Syntenin (PDB ID: 1W9E) with PDZ1 (blue) and PDZ2 (raspberry) domains. IVMT-Rx-4 (pose 1, green sticks) is docked within the PDZ1 pocket. Receptor-derived pharmacophore features generated in Schrödinger are shown as spheres colored by type: aromatic (orange ring), anionic (red), H-bond donor (blue), and H-bond acceptor (pink). Arrows indicate feature directionality, and key interacting residues are shown as labeled sticks. (B) Measurement of the binding affinity of IVMT-Rx-4 for MDA-9. Change in fluorescence as a percentage of the initial fluorescence (% ΔF) was plotted as a function of the concentration of the titrant. Error bars represent SD. The solid line shows a nonlinear fit to the data to obtain the KD value from triplicated experiments.

To further characterize the molecular basis of IVMT-Rx-4 binding to PDZ1, pharmacophore modeling and molecular docking were performed (Fig. 4A). This analysis revealed several key interactions within the PDZ1 binding pocket that are consistent with the experimental binding profile of IVMT-Rx-4. The amphipathic nature of K124 and K179 creates a local aliphatic region that favors an aromatic feature, corresponding to the 2-amino-1,4-dimethylbenzene moiety observed in the lowest-energy docked conformations, with an MM/GBSA predicted free energy of binding of −40.2 kcal/mol. L178 further contributes to the formation of this hydrocarbon-rich microenvironment. A unidirectional hydrogen-bond donor feature was oriented toward the positively charged nitrogen of K119, aligning with the amine group on the aromatic ring and supporting a favorable electrostatic interaction. Two anionic features were positioned toward the solvent-exposed surface near K124 and K179, suggesting opportunities to enhance binding affinity through the introduction of additional negatively charged substituents in this region. A bidirectional donor feature was directed toward the backbone atoms of L127 and aligned with the conjugated nitrogen–oxygen motif of IVMT-Rx-4, indicating stabilization via backbone-mediated hydrogen bonding. In contrast, the bidirectional acceptor feature extended toward the solvent-exposed region near I125, where no corresponding interaction was observed in the docked poses, identifying this region as a potential site for further ligand optimization. Molecular docking of PDZ1i also predicted that PDZ1i and IVMT-Rx-4 occupy the same general binding region and preserve a common core interaction fingerprint including K1214, G126, Q142, H75, and K179 (Supplementary Fig. 4A & 4B). Both ligands maintain strong contacts across the hydrophobic groove of the PDZ1 pocket. PDZ1i does have a higher free energy of binding on average (−48.4 kcal/mol) compared to IVMT-Rx-4 (30.8 Kcal/mol), however, the average ligand efficiency (−1.2 kcal/mol and 1.2 kcal/mol, respectively) is closer, but actually slightly higher for IVMT-Rx-4, indicating that the difference in free energy is due to ligand size. Additionally, PDZ1i extends deeper into the pocket towards PDZ2, forming consistent interactions near L140, V141, and G248 (Supplementary Fig. 4B & 4C). Ultimately, these two compounds bind similarly in the pocket anchored by the hydrophobic PDZ1 pocket motif, however, they do differ in periphery interactions and size, both of which may explain the improved free energy of binding of PDZ1i.

To test the therapeutic potential of IVMT-Rx-4 in suppressing HNSCC tumorigenesis, we first investigated whether IVMT-Rx-4 could inhibit the invasive growth and self-renewal abilities of cultured HNSCC cells. Treatment of UM-SCC-1, HN6, and UM-SCC-22B cells IVMT-Rx-4 significantly inhibited cell invasion and sphere formation in a dose-dependent manner (Fig. 5A–5C, Supplementary Fig. 5A–5C). Western blotting analysis indicated that both p65 and its target BMI1 expression were dramatically inhibited by IVMT-Rx-4 (Fig. 5D and Supplementary Fig. 5D). Of note, MDA-9 protein expression remained unaffected following IVMT-Rx-4 treatment, even at 100 μM IVMT-Rx-4 (Fig. 5D). Finally, ChIP-qPCR analysis revealed that the enrichment of p65 at the BMI1 promoter was significantly suppressed by IVMT-Rx-4 in both UM-SCC-1 and HN6 cells (Fig. 5E). These results support the conclusion that IVMT-Rx-4 can inhibit the transformed phenotype of HNSCC cells in vitro through inhibiting MDA-9/p65 axis.

Figure 5. MDA-9 inhibitor (IVMT-Rx-4) suppresses invasion and self-renewal abilities of HNSCC cells.

(A) IVMT-Rx-4 inhibits the invasive growth of UM-SCC-1 and HN6 cells. Data represents mean ± SD of triplicate experiments. *P<0.05 and **P<0.01 by Student’s t-test (n=3). (B) CSC frequency of control or IVMT-Rx-4 treated HNSCC cells was measured by limiting dilution assay in vitro. Frequency and probability estimates were computed with ELDA software. (C) The size of tumor spheres was significantly inhibited by IVMT-Rx-4. (D) IVMT-Rx-4 inhibits BMI1 and p65 expression in HNSCC cells. (E) ChIP-quantitative real-time PCR showed that IVMT-Rx-4 significantly reduced the recruitment of p65 to the BMI1 promoter in UM-SCC-1 and HN6 cells. Data represents mean ± SD of triplicate experiments.

Resistance to cisplatin remains a major obstacle to effective HNSCC therapy. Given the established role of p65 in promoting tumor progression and chemotherapy resistance across multiple cancer types, including HNSCC, we next investigated whether IVMT-Rx-4 could enhance the therapeutic response to cisplatin. Importantly, IVMT-Rx-4 significantly potentiated cisplatin-induced apoptosis in UM-SCC-1 cells (Supplementary Fig. 5E), supporting its potential as a novel therapeutic agent to overcome chemoresistance in HNSCC.

IVMT-Rx-4 inhibits HNSCC tumorigenesis and metastasis in vivo

To extend our findings and provide insights into potential clinical utility, we examined the effect of IVMT-Rx-4 on HNSCC tumorigenesis using a HNSCC orthotopic xenograft mouse model^{17, 57}. UM-SCC-1 cells were inoculated into the tongues of nude mice (n=5/group). After establishing tumors, the tumor-bearing mice were injected intraperitoneally daily with either a control vehicle or IVMT-Rx-4 daily (30mg/kg) for 2 weeks (Fig. 6A). As shown in Fig. 6B & 6C, IVMT-Rx-4 treatment significantly inhibited UM-SCC-1 growth in vivo. Additionally, H&E staining of the tongues also confirmed the reduction in tumor tissue growth (Fig. 6D). Moreover, histological analysis revealed that the lymph node (LN) metastasis was also significantly suppressed by IVMT-Rx-4 treatment (Fig. 6E and 6F). Similar phenomena were observed in the HN6 xenograft model (Supplementary Fig. 6). These results suggest that targeting MDA-9 with IVMT-Rx-4 may represent an effective strategy to suppress tumorigenesis and metastasis of HNSCC.

Figure 6. IVMT-Rx-4 inhibits HNSCC growth in vivo.

(A) Experimental plan used to study the effect of IVMT-Rx-4 on HNSCC growth in vivo. (B) Representative images of mice tongues from control group and IVMT-Rx-4 treatment group. Scale Bar: 1 cm. (C) Tumor volumes were significantly smaller in IVMT-Rx-4 treatment group. Data represents mean ± SD of triplicate experiments. *P<0.05 and **P<0.01 by Student’s t-test (n=5). (D) Representative images of H&E staining of xenograft tumors from control group and IVMT-Rx-4 treatment group. Scale Bar: 200 μm. (E) Immunostaining of metastatic cells in cervical lymph nodes (LN) using anti-pan-cytokeratin (Pan-CK). Scale Bar: 200 μm. (F) IVMT-Rx-4 suppressed HNSCC LN metastasis. P=0.1, by Fisher’s exact test.

IVMT-Rx-4 eliminates Bmi1⁺ CSCs and suppresses tumorigenesis of HNSCC

To investigate whether targeting MDA-9 (with IVMT-Rx-4 could suppress HNSCC growth through inhibition of the functional properties of HNSCC CSCs, we adopted the 4NQO-induced HNSCC Bmi1^CreER; Rosa^tdTomato spontaneous mouse model (Fig. 7A)^{17, 57}. This model allows lineage tracing of Bmi1⁺ CSCs in vivo in an unperturbed environment using tamoxifen-induced Cre-mediated recombination. In brief, the Bmi1^CreER; Rosa^tdTomato mice were treated with 4NQO in their drinking water for 16 weeks before returning to normal drinking water for another 6–10 weeks, as described previously^{17, 57}. At 22 weeks from the initial 4NQO treatment, the tumor-bearing mice were randomly divided into 2 groups (2 more groups will be described below). The mice in each group were given the control vehicle or IVMT-Rx-4 every day, respectively, for 4 weeks. Tamoxifen was administered on day 1 of treatment to label the Bmi1⁺ CSCs (Fig. 7A). To evaluate the effect of IVMT-Rx-4 treatment on HNSCC progression, tumor-bearing mice were grown for 26 weeks from the initial 4NQO treatment to reach the study endpoint (Fig. 7A). As shown in Fig. 7B and 7C, the tumor growth was significantly inhibited by administration of IVMT-Rx-4. The lineage tracing showed the Tomato⁺ cells (descendants of Bmi1⁺ CSCs) were profoundly eliminated in the regressed HNSCC, indicating that CSC-derived tumor growth was also suppressed by IVMT-Rx-4 treatment (Fig. 7D and 7E). Of note, the tumor tissues were defined by immunofluorescent (IF) staining of pan-Keratin (marker of HNSCC) and DAPI.

Figure 7. IVMT-Rx-4 suppresses CSC-derived tumor growth and overcomes cisplatin resistance in HNSCC mouse model.

(A) Experimental plan used to trace Bmi1⁺ cells in 4NQO-induced mouse HNSCC upon IVMT-Rx-4 and/or cisplatin treatment. (B) and (C) Representative images of dissected mice tongues from different treatment conditions and statistical analysis of the corresponding tumor lesion areas. Tongue tumors were significantly smaller in IVMT-Rx-4 and cisplatin+IVMT-Rx-4 treatment groups. Scale Bar: 1 cm. (D) and (E) Representative images of immunofluorescence staining of tongue tumors from different treatment conditions alongside the average Tomato fluorescence intensities of each treatment condition that are indicative of the propagation of CSCs. Bmi1⁺ cells (Red). Data represents mean±SD of four replicate experiments. *P<0.05, **P<0.01 and ***P<0.001 by Student’s t-test (n=4). Scale Bar: 50μm.

To determine the direct effect of IVMT-Rx-4 on the elimination of the Bmi1⁺ CSCs, we developed a different treatment strategy like the plan above, except that Tamoxifen was administered on day 5 of treatment to label the Bmi1⁺ CSCs. The mice were sacrificed 24 hours after the administration of tamoxifen (Fig. 8A). As shown in Fig. 8B & 8C, IVMT-Rx-4 significantly inhibited the Bmi1⁺CSCs population, suggesting that the self-renewal ability of CSCs was suppressed by targeting MDA-9. Taken together, these results support the hypothesis that targeting MDA-9 with IVMT-Rx-4 can potently inhibit HNSCC tumorigenesis by effectively eliminating CSCs.

Figure 8. IVMT-Rx-4 effectively eliminates cisplatin-enriched CSCs in HNSCC.

A) Experimental plan used to trace Bmi1⁺ cells in 4NQO-induced mouse HNSCC upon IVMT-Rx-4 and/or cisplatin treatment. (B) Percentage of PCK+Tomato+ cells were significantly diminished with IVMT-Rx-4 treatment. (C) Representative images of immunofluorescence staining of tongue tumors from different treatment conditions. Bmi1+ cells (RedData represents mean±SD of triplicate experiments. *P<0.05 and **P<0.01 by Student’s t-test (n=3). Scale Bar: 50μm.

IVMT-Rx-4 effectively eliminates cisplatin-enriched Bmi1⁺ CSCs of HNSCC

Resistance to cisplatin remains a major obstacle to effective HNSCC therapy⁶¹. Previously, we showed that cisplatin can induce Bmi1⁺ CSC populations, which are responsible for developing cisplatin resistance, and targeting Bmi1⁺ CSCs is an effective strategy to overcome cisplatin resistance in HNSCC¹⁷. Therefore, we also investigated whether targeting MDA-9 with IVMT-Rx-4 could overcome cisplatin resistance by eliminating cisplatin-enriched Bmi1⁺ CSCs and CSC-derived tumor growth. Therefore, in the aforementioned two treatment plans, the tumor-bearing Bmi1^CreER; Rosa^tdTomato mice were also treated with cisplatin alone, or cisplatin with IVMT-Rx-4 in parallel (Fig. 7A and 8A). IVMT-Rx4 treatment also significantly inhibited cisplatin-induced Bmi1⁺ CSC populations (Fig. 8B) and CSC-derived tumor growth (Fig. 7B–7D). Taken together, our data indicates that targeting MDA-9 (Syntenin-1, SDCBP) with IVMT-Rx-4 may overcome cisplatin resistance of HNSCC by effectively eliminating CSCs, providing further support for potential clinical applications of IVMT-Rx-4.

Discussion

Discovering new therapeutic targets and strategies for HNSCC treatment is of utmost importance and urgency. Although new cancer immunotherapies, such as anti-PD-1/PD-L1 reagents, have achieved clinical success in recent years in the management of various cancers, they have not been as successful for HNSCC, which display only limited response rates and median response duration. Consequently, cisplatin chemotherapy remains the first-line treatment option for HNSCC, particularly for late-stage HNSCC cases. This is particularly true for HPV-negative HNSCC, since the majority (>55%) of these cases are diagnosed at later stages (stage III and IV). However, acquired resistance to cisplatin is a well-documented phenomenon that leads to poor survival outcomes for those patients. It is established that cisplatin enriches certain CSC populations, which highlights the need to develop therapies targeting CSCs to improve HNSCC patient survival^{13, 61}.

Following its discovery by subtraction hybridization in metastatic melanoma cells induced to terminally differentiate and lose tumorigenic properties^{18, 19}, MDA-9/Syntenin-1/SDCBP has been studied extensively and shown to be an important scaffold-protein regulating many essential physiological properties^{21, 27}. A large body of evidence emphasizes MDA-9’s role as a pro-metastatic protein in multiple cancers^{23, 26, 27, 36, 45, 62}. MDA-9 promotes cancer progression through modulation of EMT^{63, 64}, angiogenesis^{31, 65}, invasion^{30, 36 25, 28, 66}, migration^{46, 67, 68}, tumor cell survival in the circulation through anoikis resistance^{11, 69}, metastasis^{26, 30, 36, 38, 39, 46}, and extracellular vesicles (exosomes) ^33–35. Additionally, this multifunctional MDA-9 protein also plays a pivotal role in other disease-specific molecular mechanisms and processes^69–72. A key property of MDA-9 involves its role in stemness and chemoresistance^{12, 32, 72}. MDA-9 has been shown to be a regulator of stemness in glioma⁶⁹, prostate cancer¹², and HNSCC³². This protein is also a major contributor to chemoresistance in these three tumor types^{12, 32, 69}. Here we provide evidence demonstrating that MDA-9 is highly expressed in late-stage HPV (−) HNSCC patients, inversely correlates with survival, and we describe a novel inhibitor of MDA-9 protein-protein interactions. We confirm that MDA-9 plays a decisive role in promoting HNSCC cell invasion and proliferation. Furthermore, our investigations suggest that MDA-9 positively controls the expression of BMI1, a well-known HNSCC cancer stem cell marker, supporting a prominent role in HNSCC cancer stemness and cisplatin resistance.

Key components of the cancer paradigm that contribute to tumor development and metastasis are stem cells^{10, 11}. Accordingly, it is critical to define ways of targeting stem cells therapeutically ^{10, 11}. In the context of glioma stem cells (GSCs), MDA-9 is a key enabler and determining factor of glioma stemness and survival.²⁴ Multiple stemness genes that include Nanog, Oct4 and Sox2 are orchestrated by MDA-9 through activation of STAT3. MDA-9 influences the survival and growth of GSCs by activation of the NOTCH1 pathway, which affects Src phosphorylation and DLL1. When activated, NOTCH1 is cleaved and controls the expression of C-Myc via RBP JK, thereby augmenting GSC growth. When the expression of MDA-9 is genetically inhibited, NOCH1/C-Myc and p-STAT3/Nanog pathways are blunted, culminating in a loss of stemness and decreased survival of GSCs. In another study by Talukdar et al.¹², MDA-9 was shown to be an essential gene in defining chemoresistance, survival, and stemness in prostate cancer stem cells (PCSCs). As observed in GSCs, MDA-9 synchronized the expression of multiple stem-regulatory genes in PCSCs, and genetic suppression of MDA-9 rendered PCSCs sensitive to the toxic effects of a spectrum of chemotherapeutic agents operating through diverse mechanisms. These latter observations suggest that MDA-9 may impact multiple drug resistance pathways, which is a possibility since MDA-9 activates STAT3, which modifies chemoresistance in PCSCs by influencing protective autophagy and altering MDR1 expression on the surface of PCSCs²⁴. In HNSSC, MDA-9 plays a central role in cell growth, migration, invasion, cellular differentiation, and angiogenesis programs and has both clinical and prognostic significance in patient biopsies^30–32. This protein is also a key regulator of chemotherapy (cisplatin) resistance and stemness in HNSCC³². In total, these studies support the importance of MDA-9 in stem cell biology, glioma, prostate cancer, and HNSCC, and highlight the potential therapeutic importance of this gene/protein as a viable target to therapeutically treat CSCs.

Kegelman et al.²⁸ used fragment-based drug-discovery combined with NMR to identify the first-in-class small molecule inhibitor of the PDZ1 domain of MDA-9, PDZ1i (IVMT-Rx-1), which inhibits functions of this protein by altering protein–protein interactions. PDZ1i has been shown to be a fundamental contributor to the major pathways in metastasis, including invasion from a primary tumor (intravasation), migration, survival in the circulation (anoikis resistance), angiogenesis, attachment at a secondary site (regulation of integrin expression), invasion (extravasation), and immune suppression^{26, 27, 29, 36, 38, 39, 69}. PDZ1i suppresses glioblastoma, neuroblastoma, prostate cancer, and breast cancer progression by inhibiting interactions of specific partner proteins with the PDZ1 domain of MDA9^{28, 38, 39, 46}.

In the present study, we focused on a synthesis product of PDZ1i, denoted IVMT-Rx-4 (Fig. 4B), which has identical biological properties but differs in subtle ways pharmacologically. Both inhibitors are non-toxic to normal or cancer cells in vitro, have similar efficacy in inhibiting metastasis in vivo in multiple animal models, show good pharmacological/PK properties, and similarly inhibit binding to the PDZ1 domain of MDA-9^{25, 28, 47}. One major difference between these two inhibitors involves solubility, which prohibits PDZ1i from IV injection, whereas IVMT-Rx-4 can be injected IV. This improved pharmacokinetic property is extremely important as one moves pharmacological agents from the lab to the clinic to treat humans.

We demonstrate that IVMT-Rx-4 strongly suppresses in vitro transformation (malignant) and the stem cell phenotypes of HNSCC and inhibits the expression of BMI1 and represses in vivo tumor growth. We focused on CSCs as these are critical mediators of HNSCC tumorigenesis and cisplatin-resistance^{6, 17, 61}. By using our unique Bmi1⁺ CSC spontaneous mouse model^{17, 57}, we now show that IVMT-Rx-4 can effectively eliminate cisplatin-enriched Bmi1⁺ CSCs. Mechanistically, we discovered that MDA-9 promotes HNSCC CSC marker BMI1 expression through NF-κB/p65. NF-κB/p65 is a key driver of tumorigenesis across multiple cancer types, with well-recognized roles in CSC maintenance and drug resistance^{57–59, 73–75}. Notably, our recent studies revealed that p65 cooperates with the Mediator complex and Bromodomain-containing Protein 4 (BRD4) to establish super-enhancers at key stemness- and resistance-associated genes, including FOSL1, SNAI2, BIRC3, CD44, MET, and TP63, thereby sustaining high transcriptional output to maintain the functional properties of CSCs in HNSCC⁵⁷. However, despite its central role, clinically effective p65 inhibitors are lacking, and directly targeting p65 poses significant challenges, as NF-κB signaling is also essential for inflammation, immune regulation, and cell survival. These concerns have hindered the translational and clinical development of p65-directed therapies⁷⁶. Importantly, our findings highlight an alternative strategy to safety and selectively suppress p65-driven transcriptional programs to eliminate CSCs, as IVMT-Rx-4 demonstrates potent activity against CSCs while remaining non-toxic to normal cells.

Collectively, our investigations imply that targeting MDA-9 may represent a practical strategy to eliminate CSCs to overcome cisplatin resistance of HNSCC. This finding may have profound clinical implications since resistance to cisplatin currently remains a major obstacle to effective therapy of HNSCC^{61, 77}. Overall, our study provides convincing evidence to demonstrate that MDA-9 is an attractive therapeutic target, and the MDA-9 inhibitor, IVMT-Rx-4, represents a promising chemical entity that may lead to more effective therapies for HNSCC.

Materials and Methods

IVMT-Rx4 synthesis and pharmacological properties

The predicted interaction between IVMT-Rx-4 and the PDZ1 domain of MDA-9 is shown in Fig. 4. The synthesis scheme for IVMT-Rx-4 is shown in Supplementary Fig. 1. IVMT-Rx-4 as well as IVMT-Rx-1 (PDZ1i) were synthesized by WuXi AppTec (Shanghai, China). ADME and Pharmacokinetics studies for IVMT-Rx-4 and IVMT-Rx-1 (PDZ1i) were performed by Pharmaron (Waltham, MA). IVMT-Rx-4 at >96% purity was used for the studies described in this paper.

Cell Culture

The human HNSCC cell line HN6 was obtained from ATCC (Rockville, MD, USA). UM-SCC-1 and UM-SCC-22B cells were obtained from the University of Michigan. HN6 and UM-SCC-1 cells were cultivated in DMEM (Dulbecco’s modified Eagle’s medium) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin-streptomycin. PlasmoTest Mycoplasma contamination detection kit (InvivoGen) was utilized to monitor potential mycoplasmal contamination of the cell lines. All the cells were incubated at constant temperature (37°C) in a humidified atmosphere containing 5% CO₂. For siRNA transfection, HNSCC cells were plated at 40–50% confluence and incubated overnight. The cells were then transfected with various amounts of siRNA using Lipofectamine RNAiMAX (Life Technologies) following the manufacturer’s instructions for 48 hours. The siRNA oligonucleotide sequences were the following: MDA-9 (GAACAUAACAUCUGUGAAA and CAAAAUAGUUGUAGCCUUA).

RT-qPCR

Total RNA was isolated from cells using NucleoSpin^® RNA II RNA purification kit (Machery-Nagel; #740955.20). RNA concentrations were measured using Nanodrop One (Fisher Scientific). For RT-qPCR, cDNA was synthesized from 1000 ng of total RNA using random primers and M-MuLV Reverse Transcriptase (New England Biolabs; #M0253L). Quantification of mRNA was determined using a SYBRGreen supermix (Bio-Rad; #1708880) on a CFX Connect Real-Time PCR Detection System (Bio-Rad). GAPDH levels were used as a loading control for real-time PCR. The primer sequences used for RT-qPCR are listed in Supplementary Table 1.

Western Blotting

Total protein was extracted after lysis of cells with CelLytic buffer (Sigma-Aldrich, #C3228). The protein extracts were then loaded into SDS-PAGE with SDS sample loading buffer. The proteins were transferred into a PVDF membrane after separation. The membranes were immersed into 5% milk for blocking of nonspecific binding for 1 hr. and then incubated with primary and secondary antibodies. The overnight incubation of primary antibody was followed by 2 hr. incubation of secondary antibody with TBS-T washing in-between and after transfer. The membrane was then developed for chemiluminescence, and the images were recorded accordingly. The antibodies used for Western Blotting were the following: MDA-9 (SDCBP1) (Abnova, #H00006386-M01), BMI1 (Cell Signaling, #5855), NF-κB p65 (Millipore Sigma, #06–418), and Tubulin (Sigma Aldrich, #T9026).

Invasion, Clonogenic, and Sphere Formation Assays

The in vitro cell invasion assays were measured using the Corning^™ BioCoat^™ Matrigel^™ Invasion Chamber (Corning; #354480) following the manufacturer’s instructions. In brief, cells were seeded in the upper chambers, and culture medium with 1% FBS was added to the lower chambers and treated according to the experimental design. After 24 hr. incubation, cells that migrated to the reverse side of inserts were stained with Hema 3 Staining Kit (Fisher, #123–869) and quantified with Image-J analysis. For the clonogenic assay, 5000 cells were seeded in a 12-well plate 24 hr. after siRNA transfection. The cells were allowed to grow for 10 days with media being replenished every other day. On the last day, cells were stained using 0.5% crystal violet and imaged using a scanner. For the in vitro limiting-dilution sphere formation assay, HNSCC cells were dissociated into single-cell suspensions and plated in 96-well plates in CSC media at the specified cell densities (50, 500, and 1000). A minimum of 24 wells was seeded for each cell density. Culture medium was refreshed every 4 days. After 2 weeks of incubation, the number of spheres formed was determined, and the frequency of sphere-forming units was estimated using ELDA software, as previously described.⁷⁸.

ChIP-seq and ChIP-qPCR

ChIP-seq was reanalyzed using our previously published ChIP-seq data using Integrative Genomics Viewer (IGV_2.3.75)⁵⁷. For each ChIP reaction, approximately 10⁶ cells were used. HNSCC cells were treated with 10 mM dimethyl 3,30-dithiobispropionimidate-HCl (DTBP) (Thermo Fisher, #20665) in PBS at room temperature for 10 min, followed by fixation with 1% formaldehyde at 37°C for an additional 10 min. Total cell lysates were subsequently sonicated to generate 200–400 bp DNA fragments. Chromatin complexes were immunoprecipitated with anti-p65 antibodies (Millipore Sigma, #SAB5600083). The precipitated DNA samples were measured by qPCR. Data are expressed as the percentage of input DNA. The primer sequences used for ChIP-qPCR analysis of the BMI1 promoter were 5′-GGGCCTGACTACACCGACACTAA-3′ and 5′-AACTGACACCGGCTCCAAAATGG-3′.

Molecular Modeling

The MDA-9/Syntenin structure (PDB ID: 1W9E)⁷⁹ was prepared using the Protein Preparation Wizard^{80, 81} by adding hydrogens, assigning bond orders, optimizing protonation states at pH 7 with Epik, and performing a restrained minimization with the OPLS4 force field. Redocking and subsequent docking with the TNEFYF peptide and IVMT-Rx-4 respectively were docked into MDA-9/Syntenin (PDB ID: 1W9E) using Schrödinger Glide v2024.4 with a box center of 15, 35, 50. In the receptor grid generation, the inner box was 20 Å×20 Å ×20 Å and the outer box allowed for 10 Å outside of the inner box, making the total box size 30 Å ×30 Å ×30 Å. Schrödinger suite was used to calculate MM/GBSA free energy of binding and pose 1 with the lowest free energy (−40.2 kcals/mol) was used for figures and analysis of subsequent structure-based pharmacophore modeling. Phase in Schrödinger v2025–1 was used to generate a receptor-based pharmacophore (e-pharmacophore) hypothesis^81–83. The e-pharmacophore was then constructed around the defined binding pocket.

Animal Studies

Bmi1^CreER; Rosa^tdTomato mouse strain were established as previously described.¹⁷ Nude mice were purchased from Taconic (Taconic, #NCR-NU) and housed in a pathogen-free condition. Animals from both genders were used in the studies. All animal experiments were performed following the protocol approved by the Institutional Animal Care and Use Committee (IACUC) at Virginia Commonwealth University. For the HNSCC orthotopic xenograft model, UM-SCC-1 cells were mixed with Matrigel and subcutaneously injected into the tongues of the nude mice. IVMT-Rx4 dissolved in DMSO was administrated IP (30 mg/kg body weight) 14 days post tumor cell inoculations, daily for 14 days. For induction of HNSCC, four-week-old Bmi1^CreER; Rosa^tdTomato mice were treated with drinking water containing 35 μg/ml 4NQO (Sigma-Aldrich Cat#N8141) for 16 weeks and then given normal drinking water for another 6 weeks. Cisplatin (Sigma-Aldrich Cat#479306) was dissolved in saline. At 22 weeks from the initial 4NQO treatment, mice were randomly divided into 4 groups and given the 1) vehicle control, 2) cisplatin (1 mg/kg body weight), 3) IVMT-Rx4 (30 mg/kg body weight, 3 times per week), or 4) cisplatin plus IVMT-Rx4 for 5 consecutive days via IP injection before being sacrificed. For lineage tracing and Bmi1⁺ cell ablation studies, Cre was activated by the IP injection of tamoxifen (Sigma-Aldrich Cat#T5648) at a dose of 9 mg per 40 g body weight, as described in Fig. 7A and 8A. Additionally, a similar experiment was conducted with 30 days of treatment, as indicated in Fig. 7A. At the endpoint of experiments, the tumor tissues were dissected. The tumor volumes were then calculated by the following formula: TV=(TL/TW2)/2 (TV, tumor volume; TL, tumor length; TW, Tumor width). The weight of the tumors was recorded before histological examination.

Immunofluorescence Staining and H&E Staining

Mouse HNSCC tumors were fixed with 4% paraformaldehyde in PBS overnight. Samples were then rinsed with cold PBS, equilibrated in 30% sucrose in PBS overnight and embedded in OCT (Tissue Tek, #25608–930). Cytosections at a thickness of 5 μm were cut by using a Leica cryostat and collected with Superfrost Slides (Fisher Scientific, #12–550-15). The slides were fixed with ice-cold acetone for 10 minutes and then washed twice with 0.4% Triton X-100 in PBS. The sections were incubated with 5% goat serum for an hr. and then incubated with the indicated primary antibody (anti-pan-cytokeratin, Cell Signaling Technology, #4545) overnight. The antigens were visualized using secondary antibodies conjugated with Alexa Fluor^® 488. Sections were counterstained and mounted with VECTASHIELD Vibrance^® Antifade Mounting Medium with DAPI (Vector Laboratories, #H-1800). For H&E staining, the sections were fixed with ice-cold acetone and then stained with Mayer’s hematoxylin for 10 minutes. Subsequently the sections were stained with 0.5% alcoholic eosin solution and washed with deionized water. The slides were then consecutively equilibrated in 50%, 70%, 95% and 100% alcohol and finally equilibrated with xylene before applying the cover slips.

Statistical Analysis

The results are presented as the mean ± SD from replicate experiments unless otherwise indicated. Two-tailed Student’s t-test was performed between two groups, and a difference was considered statistically significant with P<0.05 using Prism 6 (GraphPad Software). Unpaired 2-tailed Student’s t test was used to compare the means of 2 groups. The Fisher’s exact test was used to compare the difference between two category variables using Prism 6.

Statement

Generative AI and AI-assisted technologies were not used in the preparation of this work.

Highlights.

MDA-9 controls expression of the cancer stem cell marker BMI1 through p65 in HNSCC.

MDA-9 inhibitor IVMT-Rx-4 suppresses BMI1 expression and cancer stemness in HNSCC.

IVMT-Rx-4 eliminates cancer stem cells and overcomes cisplatin resistance in HNSCC.

Availability of data and materials

The authors declare that the data supporting the findings of this study are available within the article. All relevant data are available upon request.