Transcriptome datasets, comprising 515 HNSCC and 44 normal tissue samples, were retrieved from the TCGA database via the UCSC Xena database (https://xenabrowser.net) and the GDC portal (https://portal.gdc.cancer.gov). Specifically, the Genomic Data Commons (GDC) TCGA Head and Neck Cancer cohort was accessed through the UCSC Xena database, and the Spliced Transcripts Alignment to a Reference-Fragments Per Kilobase of transcript per Million mapped reads data under the gene expression RNA sequencing module were downloaded. Clinical information, including pathological stage, T stage, N stage and M stage, was extracted from the TCGA-HNSC clinical data files. Tumor staging was based on the American Joint Committee on Cancer staging system (25). A total of 23 m6A-associated genes were selected from the dataset based on findings from previous studies (15,16). All bioinformatics analyses were performed using R software (version 4.2.1; http://www.r-project.org/) and RStudio (version 2023.06.1; https://posit.co/; Posit Software, PBC). The following R packages (all obtained from CRAN, http://cran.r-project.org/) were used: i) ‘ggplot2’ (version 3.5.2) and ‘boxplot’ (version 1.3.1) for constructing heatmaps and boxplots to visualize gene expression patterns; ii) ‘survival’ (version 3.7–0) for Cox regression analysis; ‘forestplot’ (version 3.13) and ‘pheatmap’ (version 1.0.12) for hazard ratio visualization; iii) ‘survminer’ (version 0.5.1) for Kaplan-Meier curves; iv) ‘rms’ (version 6.8–2) for nomogram construction; v) ‘gghalves’ (version 0.1.4) and ‘Hmisc’ (version 5.1-3) for clinical correlation analysis; and vi) ‘ggExtra’ (version 0.10.1), ‘cowplot’ (version 1.1.3), ‘ggpubr’ (version 0.6.0), ‘ggstatsplot’ (version 0.14.0) and ‘ggside’ (version 0.3.1) for Pearson's correlation analysis between hnRNPC and BCAT1. To identify genes associated with hnRNPC expression, patients with HNSCC from TCGA database were stratified into hnRNPC-high and hnRNPC-low groups based on the median expression level of hnRNPC. Differentially expressed genes between the two groups were identified using the ‘limma’ package (version 3.50.0) in R software. Genes with |log2 fold-change|>1 and adjusted P-value (false discovery rate) <0.05 were considered significantly differentially expressed.

Study samples were randomly allocated into a training cohort and a validation cohort at a 3:7 ratio. Univariate and multivariate Cox regression analyses were applied to identify independent prognostic factors for patients with HNSCC. The robustness of these factors was further assessed using the validation cohort. Therefore, three genes, insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2), hnRNPC and YTHDC2, were incorporated into the final Cox risk regression model.

The human HNSCC cell lines CAL-27 and SCC-15 were acquired from The Cell Bank of Type Culture Collection of The Chinese Academy of Sciences. These cell lines were maintained in Dulbecco's Modified Eagle's Medium (DMEM) containing 10% fetal bovine serum (FBS; Corning, Inc.). For stem cell (27-CSC) induction, OSCC cells were cultured in DMEM/F12 medium supplemented with basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF). These induced stem cells were grown as tumor spheres in DMEM/F12 medium with 20 ng/ml each of bFGF (cat. no. HY-P7331; MedChemExpress) and EGF (cat. no. AF-100-15-100; Peprotech, Inc.; Thermo Fisher Scientific, Inc.).

For lentivirus-mediated knockdown, two independent short hairpin RNA (shRNA) sequences targeting distinct regions of hnRNPC and methyltransferase-like 3 (METTL3) transcripts were designed and cloned into the pLKO.1-puro lentiviral vector [Hanheng Biotechnology (Shanghai) Co., Ltd.]. Lentiviral particles were produced by co-transfection of 293T cells with the pLKO.1-shRNA transfer plasmid, the psPAX2 packaging plasmid and the pMD2.G envelope plasmid. Transfection was performed when 293T cells reached 70–80% confluence using Lipofectamine 3000 (Thermo Fisher Scientific, Inc.) with 3 µg of total plasmid DNA (pLKO.1-shRNA:psPAX2:pMD2.G=4:3:1) at 37°C for 6 h, after which the culture medium was replaced. Viral supernatants were collected at 48 and 72 h post-transfection, filtered through 0.45 µm filters and used immediately or stored at −80°C. For transduction, CAL-27 and SCC-15 cells were incubated with viral supernatants supplemented with 8 µg/ml polybrene for 24–48 h at 37°C. The negative control (NC) group was transduced with lentiviral particles carrying a non-targeting scrambled shRNA sequence cloned into the same pLKO.1-puro backbone. Stably transduced cells were selected and maintained in medium containing puromycin. The corresponding shRNA sequences are provided in Table SI.

Male BALB/c nude mice (6 weeks old; 18–22 g) were purchased from Cyagen Biosciences, Inc. A total of 15 mice were randomly assigned to three groups (n=5 per group: sh-NC control, sh-hnRNPC#1 and sh-hnRNPC#2). Mice were housed in a specific pathogen-free animal facility under controlled environmental conditions: Temperature, 20–26°C; relative humidity, 40–70%; a 12/12 h light/dark cycle; and free access to food and water. The present study was approved by the Ethics Committee of Zhenjiang Stomatological Hospital (approval no. PJ2023-08-10; Zhenjiang, China). SCC-15 cells stably expressing sh-NC or sh-hnRNPC were collected, resuspended at a density of 1×10⁷ cells per 100 µl in a 1:1 mixture of PBS and Matrigel (as recommended by ATCC for subcutaneous tumor formation in nude mice; http://www.atcc.org/products/crl-1623), and injected subcutaneously into the right flank of each mouse. Tumor growth was monitored every 3 days starting from 1 week post-injection; tumor volume was calculated using calipers with the formula: Volume=(length × width²)/2. Humane endpoints were predefined as: i) Tumor diameter >20 mm; ii) tumor ulceration; iii) body weight loss >20%; or iv) signs of severe distress. No mice reached these endpoints before day 21 and tumor growth remained within humane limits throughout the experiment. At 21 days post-injection, mice were euthanized by gradual CO₂ inhalation (30–50% in an airtight chamber) followed by cervical dislocation after loss of consciousness. Death was confirmed by cessation of heartbeat and respiration, absence of corneal/pedal reflexes and graying of mucous membranes. Tumors were then dissected, weighed and processed for subsequent analyses. All procedures complied with institutional animal welfare guidelines.

For reagent treatment, the indicated cells were treated with various concentrations (5, 10, 20 and 30 µM) of ERG240 (cat. no. HY-W193545A; MedChemExpress). Dimethyl 2-oxoglutarate were purchased from MedChemExpress (cat. no. HY-44134). The m⁶A RNA methylation assay kit were purchased from Abcam (cat. no. ab185912) and the assays were conducted according to the manufacturer's instructions.

Total RNA was isolated from cultured cells with the Axygen^® RNA Extraction Kit (cat. no. AP-MN-MS-RNA-250; Corning, Inc.). Subsequent to extraction, the RNA samples were converted into complementary DNA via reverse transcription, a process carried out using the Takara PrimeScript^™ RT Master Mix Kit (cat. no. RR036A; Takara Bio, Inc.) according to the manufacturer's protocol (37°C for 15 min, followed by 85°C for 5 sec). For the analysis of gene expression, reactions were run in triplicate using the Takara SYBR^® Premix Ex Taq^™ II kit (cat. no. RR036A; Takara Bio, Inc.) on an Applied Biosystems^™ 7500 instrument (Thermo Fisher Scientific, Inc.). The thermocycling conditions were as follows: 95°C for 30 sec, followed by 40 cycles of 95°C for 5 sec and 60°C for 34 sec, with a final dissociation stage to generate melting curves. The 2^−ΔΔCq algorithm was employed to determine the relative expression levels of each target gene (26). β-actin served as the endogenous reference gene to normalize the data. Detailed information regarding the primers utilized in the qPCR assays is provided in Table SII.

SDS-PAGE and western blotting procedures were conducted according to a previous study (27). For cell lysis, RIPA buffer (cat. no. P0013D; Beyotime Biotechnology) was used, supplemented with both protease inhibitors (cat. no. 43002700; Roche Diagnostics) and phosphatase inhibitors (cat. no. 43002700; Roche Diagnostics). Protein concentrations was determined using the BCA Protein Assay kit (cat. no. K813-2500; BioVision, Inc.; Abcam) according to the manufacturer's protocol. Equal protein aliquots were separated by SDS-PAGE and transferred to 0.45 µm PVDF membranes (cat. no. IPVH00010; MilliporeSigma; Merck KGaA). Following membranes were blocked with skim milk before incubation with primary antibodies, followed by a 1-h incubation with horseradish peroxidase-conjugated secondary antibodies (Beijing Zhongshan Jinqiao Biotechnology Co., Ltd.). Immunoreactive bands were visualized using an ECL detection kit (MilliporeSigma; Merck KGaA), with β-actin serving as the loading control. Primary antibodies included: Rabbit anti-hnRNPC (cat. no. ab133607; 1:1,000; Abcam), rabbit anti-BCAT1 (cat. no. ab197941; 1:1,000; Abcam) and mouse anti-β-actin (cat. no. sc58673; 1:10,000; Santa Cruz Biotechnology, Inc.).

Cells were seeded in 6-well plates at a density of 1,000 cells per well and maintained in a 5% CO₂ atmosphere at 37°C for 14 days in adherent culture (without soft agar). Following PBS washes, cells were fixed with 4% paraformaldehyde at room temperature for 15 min and stained with 0.5% crystal violet at room temperature for 15 min. Colony images were captured using a Nikon bright-field microscope. Colony number and area were quantified using ImageJ software (version 1.54f; National Institutes of Health). A distinct cluster was scored as a positive colony only if it contained ≥50 cells. To ensure the accuracy of the analysis in distinguishing adjacent or overlapping colonies, all software-based quantification was subjected to manual verification. Results are expressed as mean ± standard deviation (SD) from three independent biological replicates.

Relative cell proliferation was evaluated through viability measurements using the Cell Counting Kit-8 (CCK-8; cat. no. CK04; Dojindo Laboratories, Inc.) and 5-ethynyl-20-deoxyuridine (EdU) incorporation assays with the EdU Cell Proliferation Assay Kit (Guangzhou RiboBio, Co., Ltd.), following previously described methods (28).

Isolated stem cells were plated in 96-well plates at varying densities: 5, 10, 20, 50, 100 or 200 cells per well. After an 8-day incubation period, tumor sphere formation in each well was examined. Stem cell frequency was calculated using the extreme limiting dilution analysis web tool (http://bioinf.wehi.edu.au/software/elda).

Oxygen consumption rate (OCR) was assessed using the Seahorse XFe24 extracellular flux analyzer (Agilent Technologies, Inc.) following the manufacturer's protocol. Cells were seeded onto collagen-coated 24-well Seahorse plates at a density of 1×10⁵ cells per well and allowed to adhere overnight at 37°C. OCR measurements were performed in basal assay medium containing 1 mM pyruvate, 10 mM glucose and 2 mM glutamine. This was followed by sequential injections of 1 µM oligomycin, 2 µM FCCP and 0.5 µM rotenone/antimycin A into each well.

BCAA and glutamate production were measured using the BCAA Assay Kit (cat. no. ab83374; Abcam) and Glutamate Assay Kit (cat. no. MAK004-1KT; MilliporeSigma; Merck KGaA), respectively, following the manufacturer's instructions. BCAA was determined by subtracting the measured BCAA concentration in the medium from the initial concentration. All results were normalized to cell number.

ATP levels were quantified using the ATP Assay Kit (colorimetric/fluorometric; cat. no. ab83355). The ATP Assay kit employs a standard protocol for fluorometric and colorimetric ATP determination involving three key reactions: i) Glycerol phosphorylation by glycerol kinase (ATP-dependent) to form glycerol 3-phosphate; ii) conversion of glycerol 3-phosphate to glycerone phosphate and hydrogen peroxide by glycerol phosphate oxidase; and iii) peroxidase-mediated reaction of hydrogen peroxide with a probe to generate a red color (λ=570 nm) and fluorescence (excitation/emission=535/587 nm). Succinate levels were measured using the Succinate Assay Kit (colorimetric; cat. no. ab204718, Abcam) following the manufacturer's instructions. This assay detects succinate through its conversion (along with ATP and CoA) to succinyl-CoA, ADP and inorganic phosphate by succinyl-CoA synthase, with colorimetric detection at 450 nm. Fumarate concentrations were determined using the Fumarate Assay Kit (cat. no. MAK492; Merck KGaA), which utilizes a colorimetric method (λ=450 nm) for quantification. GSH levels were measured with the GSH Assay Kit (colorimetric; cat. no. ab239709; Abcam), where GSH concentration is determined by absorbance readings at 412 nm.

CAL-27 and SCC-15 cells (4×10⁵) were seeded in 6-well plates at 37°C overnight and treated with 5 µg/ml actinomycin D (cat. no. HY-17559; MedChemExpress) at 37°C for various durations (0, 2, 4 and 6 h). At each designated time point, cells were harvested for RNA isolation and subsequent RT-qPCR analysis as aforementioned.

RIP assays were performed using the PureBinding^® RNA Immunoprecipitation Kit (cat. no. P0101/P0102; Guangzhou Geneseed Biotech. Co., Ltd.) according to the manufacturer's protocol. Briefly, 5 µg of hnRNPC antibody (cat. no. ab133607; 1:1,000; Abcam) was bound to 200 µl protein A/G magnetic beads (provided in the kit) at room temperature for 30 min, followed by three washes with wash buffer (prepared from the kit components). The antibody-conjugated beads were then incubated with pre-cleared nuclear extracts in RIP buffer at 4°C overnight with rotation (10 rpm). For each IP reaction, 400 µl of pre-cleared nuclear lysate was used. After immunoprecipitation, the beads were washed three times with wash buffer (prepared from the kit components) with vortexing for 2 min each wash, followed by magnetic separation using a magnetic stand. Protein complexes were eluted by boiling in 1X loading buffer for 5 min and isolated via magnetic separation. RNA was purified using the DR Columns and RC Columns provided in the kit, with centrifugation steps performed according to the manufacturer's protocol (including 10,000 × g for 1 min for DNA removal, 10,000 × g for 30–60 sec for column washing, and 12,000 × g for 3 min for RNA elution). Total RNA served as the input control and extracted RNA was analyzed by RT-qPCR, which included U1 small nuclear RNA as a negative control. All RIP-qPCR data were normalized to U1 levels to account for non-specific background binding. All RIP experiments were conducted in triplicate using three biological replicates. Primer sequences are provided in Table SIII.

The m⁶A modification status of a specific gene was analyzed using the MeRIP Kit (MilliporeSigma; Merck KGaA) following previously described procedures (2). Briefly, 5 µg of anti-m⁶A antibody (cat. no. CS220007; MilliporeSigma; Merck KGaA) or normal mouse IgG (cat. no. CS200621; MilliporeSigma; Merck KGaA) was prewashed and then incubated with Magna chromatin IP protein A/G magnetic beads (cat. no. CS203152; MilliporeSigma; Merck KGaA) for 30 min at room temperature. This antibody-bead complex was then incubated with purified poly-(A) RNA. The degree of enrichment of m⁶A-containing mRNA was subsequently quantified using RT-qPCR as aforementioned, with the relevant primer sequences listed in Table SIII.

To evaluate the correlation between hnRNPC and BCAT1 expression across different cancer types and normal tissues, expression data were analyzed using the GEPIA web server (http://gepia.cancer-pku.cn/), which integrates RNA-sequencing expression data from the TCGA and GTEx projects. Pearson's correlation analysis between hnRNPC and BCAT1 expression was performed using the correlation analysis module of GEPIA. Specifically, the pan-cancer correlation was assessed using TCGA tumor samples across all available cancer types, and the pan-tissue correlation was assessed using GTEx normal tissue samples. Scatter plots were generated with log2(TPM+1)-transformed expression values, and the Pearson's correlation coefficient (R) and corresponding P-value were calculated by GEPIA to quantify the strength and significance of the association.

All experiments were performed with at least three independent biological replicates (n ≥3), unless otherwise indicated in the figure legends. Data are presented as mean ± SD. Normality of data distribution was assessed using the Shapiro-Wilk test. Homogeneity of variances was assessed using Levene's test. For comparisons between two groups, a two-tailed Student's t-test was applied when both normality and variance homogeneity assumptions were met; otherwise, the Mann-Whitney U test was used. For comparisons among three or more groups, one-way ANOVA followed by Tukey's post hoc test was used for normally distributed data with equal variances; the Kruskal-Wallis test followed by Dunn's post hoc test was used for non-normal distributions or unequal variances. All statistical analyses were performed using GraphPad Prism (version 9.0; Dotmatics) or R software (version 4.1.0; R Development Core Team). P<0.05 was considered to indicate a statistically significant difference.

To investigate the impact of m⁶A regulatory genes on HNSCC, the present study analyzed transcriptome data from 515 patients with HNSCC and 44 normal tissues from TCGA database. Expression profiles of 23 m⁶A-associated genes were extracted, among which 18 demonstrated aberrant expression in HNSCC (Fig. 1A), suggesting a close association between m⁶A modification and HNSCC development. Feature selection of m⁶A regulators was performed using univariate Cox regression analysis (Fig. 1B), identifying five genes (hnRNPC, ALKBH5, IGF2BP1, IGF2BP2 and YTHDC2) significantly associated with HNSCC survival. Subsequent multivariate Cox regression analysis of these candidates revealed that three genes (hnRNPC, IGF2BP2 and YTHDC2) were significantly associated with survival (Fig. 1C). Univariate analysis (Fig. 1D) was also conducted on clinical variables from HNSCC samples and significant clinical factors (including sex, age and tumor stage) along with the three candidate genes were incorporated into a multifactorial Cox model (Fig. 1E). The results indicated that hnRNPC, IGF2BP2 and YTHDC2 could serve as independent prognostic factors for HNSCC. Specifically, YTHDC2 was identified as a protective factor, while IGF2BP2 and hnRNPC were risk factors. Patients were stratified into high- and low-risk groups based on a risk score derived from the model, with the optimal cut-off value (4.64) determined by maximally selected rank statistics based on the log-rank test (Fig. S1). Survival analysis demonstrated that patients in the high-risk group had significantly shorter overall survival (OS; Fig. 1F), a finding consistent in the test set (Fig. 1G and H). A nomogram and calibration curves based on the final multivariate Cox model further supported these results (Fig. 1I and J). Although the roles of YTHDC2 and IGF2BP2 in HNSCC have been previously reported (29,30), the present study focused on hnRNPC. Survival analysis based on hnRNPC expression alone indicated that high expression was associated with worse prognosis (Fig. 1K). Furthermore, hnRNPC expression was positively associated with stage III and IV and T stage (Fig. 1L and M).

Prior to functional studies, the present study examined the endogenous mRNA expression level of hnRNPC in a panel of six commonly used OSCC cell lines (CAL-27, SCC-15, SCC-9, SCC-25, HSC-3 and FaDu) by qPCR. Among these, CAL-27 and SCC-15 exhibited the markedly highest basal levels of hnRNPC mRNA (Fig. S2A), the present study thus knocked down hnRNPC expression in both cell lines using two distinct shRNA sequences to verify hnRNPC function (Fig. 2A and B). Results from CCK-8 and colony formation assays demonstrated that hnRNPC knockdown significantly inhibited the proliferative capacity of OSCC cells compared with the NC group, confirming the oncogenic function of hnRNPC (Fig. 2C-F). Furthermore, EdU incorporation assay indicated that hnRNPC knockdown significantly suppressed the proliferation rate of OSCC cells (Fig. 2G and H). Consistently, nude mouse subcutaneous xenograft experiments demonstrated that hnRNPC silencing led to a significant reduction in both volume and weight of subcutaneously formed OSCC tumors (Fig. 2I-K). Furthermore, hnRNPC knockdown significantly decreased the frequency of tumor sphere formation by CAL-27-derived stem cells, as determined by an in vitro limiting dilution assay (Fig. 2L). The present study analyzed the expression levels of key stemness genes, including Oct4, Sox2 and Nanog, which are well-known in promoting tumor cells with self-renewal capabilities, using qPCR. hnRNPC knockdown resulted in a significant downregulation of these genes (Fig. 2M). Therefore, hnRNPC emerges as a key regulator of OSCC cell proliferation and the maintenance of stem-like properties.

To investigate the mechanism underlying hnRNPC-mediated regulation of OSCC progression, the present study analyzed transcriptomic data of HNSC from TCGA database. The results revealed that the BCAA metabolic pathway was active in HNSC tissues with high hnRNPC expression (Fig. 3A). Since the BCAA metabolic pathway is involved in energy production and redox homeostasis, the present study also observed changes in pathways associated with oxidative phosphorylation and stress (Fig. 3B and C). Consistently, BCAA uptake capacity was significantly impaired in OSCC cell lines with hnRNPC knockdown (Fig. 3D). Since BCAAs generate large amounts of glutamate and BCKAs through catabolic pathways, where glutamate contributes to the synthesis of reduced GSH for antioxidant defense, and BCKAs promote the TCA cycle and oxidative phosphorylation via acetyl-CoA production, the present study examined these indicators in OSCC cells (31,32). The present study findings demonstrated that hnRNPC knockdown led to significantly decreased glutamate levels, reduced OCR and lower GSH levels in OSCC cells compared with the negative control group (Fig. 3E-J). In accordance, gene set enrichment analysis (GSEA) based on TCGA data also revealed a significant enrichment of GSH metabolism pathway in hnRNPC highly expressed OSCC tissues (Fig. 3K and L). These results indicated that hnRNPC can enhance tumor energy production and antioxidant capacity by promoting the BCAA metabolic pathway, thereby facilitating oxidative phosphorylation and increasing GSH levels, which likely represents a key mechanism underlying its tumor-promoting effects.

To explore the specific mechanism by which hnRNPC promotes tumor progression via regulation of the BCAA metabolism, the present study analyzed differentially expressed genes between hnRNPC-high and -low expressing samples from the HNSC transcriptome database. The present study reported that BCAT1 was significantly upregulated in OSCC tissues with high hnRNPC expression, suggesting that hnRNPC may regulate BCAA metabolism by promoting BCAT1 expression (Fig. 4A and B). Consistent with this observation, western blotting and qPCR results demonstrated that BCAT1 expression was significantly suppressed in OSCC cells following hnRNPC knockdown (Fig. 4C and D). Notably, treatment with the BCAT1 inhibitor ERG240 further enhanced the tumor-suppressive effects of hnRNPC knockdown (Fig. 4E and F), indicating that combined targeting of hnRNPC and BCAT1 to interfere with BCAA metabolism could potentially represent a promising therapeutic strategy for OSCC in the future.

Since hnRNPC functions as an m⁶A modification reader protein to regulate mRNA stability of downstream genes, the present study further investigated whether hnRNPC-mediated regulation of BCAT1 is dependent on m⁶A modification. mRNA stability assays demonstrated that BCAT1 mRNA stability was significantly reduced after hnRNPC knockdown (Fig. 4G and H). Using the ‘SRAMP’ algorithm, the present study predicted the top five m⁶A modification sites in BCAT1 mRNA (Fig. 4I). Further MeRIP-qPCR experiments identified the most notable m⁶A peak at position +8,438 within the BCAT1 3′-UTR (Fig. 4J and K). RIP assays confirmed the direct binding of hnRNPC to BCAT1 mRNA (Fig. 4L), while MeRIP analysis revealed that METTL3 knockdown reduced m⁶A methylation on BCAT1 transcripts (Figs. 4M and S2B). Of note, METTL3 silencing also diminished the association between hnRNPC and BCAT1 mRNA, demonstrating that m⁶A marks are required for hnRNPC recognition (Fig. 4N). Furthermore, transcriptomic analysis of TCGA database revealed a significant positive correlation between hnRNPC and BCAT1 expression in HNSC tissues (Fig. 4O). Consistent with this finding, pan-cancer transcriptomic data from TCGA database and pan-tissue transcriptomic data from the GTEx database also revealed the significant positive correlation between hnRNPC and BCAT1 expression (Fig. 4P and Q). Therefore, hnRNPC may promote BCAT1 expression in an m⁶A-dependent manner during OSCC progression, thereby driving the malignant evolution of OSCC dependent on BCAA metabolism.

Since α-KG serves as a key donor for BCAT1-catalyzed glutamate generation in the BCAA metabolic pathway and functions as a key intermediate in mitochondrial TCA cycle, the present study investigated whether exogenous supplementation of α-KG could reverse the hnRNPC-induced suppression of BCAA metabolism and subsequent energy deficiency. Results from the CCK-8 and colony formation assays in the present study demonstrated that α-KG supplementation significantly restored cell viability in hnRNPC-knockdown OSCC cells (Fig. 5A-D). Stem cell viability experiments revealed that α-KG supplementation notably increased the frequency of tumor sphere formation by CAL-27-derived stem cells with hnRNPC knockdown (Fig. 5E), accompanied by significant recovery of Oct4, Sox2 and Nanog expression levels which were downregulated by hnRNPC silencing (Fig. 5F). Furthermore, Seahorse assay results indicated that α-KG supplementation led to significant restoration of OCR and increased ATP levels in OSCC cells (Fig. 5G-K). Therefore, hnRNPC binds to m⁶A-modified BCAT1 mRNA in a METTL3-dependent manner, stabilizing BCAT1 transcripts and enhancing BCAT1 expression. This upregulates BCAA catabolism, producing BCKAs for TCA cycle-driven energy supply and glutamate for GSH-mediated antioxidant defense, ultimately promoting OSCC cell proliferation, stemness and tumor growth (Fig. 6). These findings confirmed that BCAT1-mediated BCAA metabolic pathway for energy production represents one of the key mechanisms underlying hnRNPC-promoted tumor progression.