Six tissue samples (C1/N1, C2/N2, and C3/N3) were collected from three OSCC patients at the First Affiliated Hospital of Shandong Second Medical University (formerly Weifang Medical University). The criteria and procedures for tissue collection were detailed in our previous studies, and the study was approved by the Ethics Committee of Shandong Second Medical University (formerly Weifang Medical University) [19, 20]. Human OSCC cell lines (HN30, CAL-27, HN6, SCC-25 and SCC-9) were obtained from Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine. The cDNA samples of normal primary human oral mucosal epithelial cells were provided by the same institution. All procedures related to these samples were conducted under institutional ethical approval (Approval No. SH9H-2019-TK235-1) with informed consent obtained from donors. The characterization of these cells has been previously reported [21, 22]. All cell culture conditions and protocols were consistent with those used in our earlier publication [20]. All procedures in this study were performed in accordance with the Declaration of Helsinki. All cell lines were recently verified prior to use by short tandem repeat (STR) analysis performed by Shanghai Biowing Applied Biotechnology Co., Ltd. All cell lines tested negative for mycoplasma.

The sequencing process was described in a previous study [19]. In brief, the MeRIP-seq technique involves the use of an m7G-specific antibody to enrich m7G-modified lncRNA fragments, followed by high-throughput sequencing to locate and quantify m7G modifications on the RNA fragments.

A quality score threshold of Q30 was applied for quality control following sequencing on the Illumina NovaSeq 6000 platform. Cutadapt software (v2.10; Marcel Martin, Max Planck Institute for Molecular Genetics, Berlin, Germany) was utilized to generate high-quality, clean reads [23]. Clean reads from all samples were mapped to the reference genome using HISAT2 (v2.0.4, Johns Hopkins University, Baltimore, MD, USA) [24]. Methylated sites were called using MACS3 (v3.0.3, Harvard University, Cambridge, MA, USA) with the default parameters, and the p values of the peaks were calculated based on the Poisson distribution [25]. Differentially methylated sites were called using DiffReps software (v1.55.6, The University of Texas MD Anderson Cancer Center, Houston, TX, USA), with p values computed via negative binomial regression [26]. Peaks located in the exons of the lncRNAs were extracted for annotations, and differentially methylated lncRNAs were identified using the thresholds of fold change (FC) $>$2 and adjusted p 0.05.

For transcriptomic analysis, high-quality reads were mapped to the human reference genome (UCSC HG19) using HISAT2 (v2.0.4) [24]. Transcript-level raw counts were generated using HTSeq (v0.9.1, European Molecular Biology Laboratory, Heidelberg, Germany), and lncRNA expression profiles were analyzed [27]. Differential expression analysis between OSCC and normal tissues was performed using edgeR (v3.16.5, Walter and Eliza Hall Institute of Medical Research, Melbourne, VIC, Australia) [28], with thresholds set at FC $>$2 and adjusted p 0.05.

p-values for both MeRIP-seq and RNA-seq analyses were adjusted for multiple testing using the Benjamini-Hochberg method, and an adjusted p-value 0.05 was considered statistically significant.

GO and KEGG pathway enrichment analyses on genes associated with differentially m7G-modified and differentially expressed lncRNAs were analyzed using the online tool, Database for Annotation, Visualization, and Integrated Discovery (DAVID, https://davidbioinformatics.nih.gov/) [29]. The analyses were conducted based on the Homo sapiens genome reference (HG19). A statistically meaningful difference was defined as p 0.05. The top GO and KEGG terms were ranked by enrichment score [–log₁₀ (p)] and presented accordingly.

To visualize the chromosomal distributions of m7G-modified lncRNAs in OSCC versus normal tissues, the RCircos package (v1.2.2, National Cancer Institute, Bethesda, MD, USA) in R was utilized [30]. In addition, sequence motif characterization of m7G modifications in lncRNAs was performed with DREME (v5.5.9, University of Washington, Seattle, WA, USA) software [31].

Total RNA was isolated from cultured cells using TRIzol reagent (M5101, NCM Biotech, Co., Ltd., Suzhou, Jiangsu, China). cDNA was generated by reverse transcription with the Evo M-MLV RT Kit (AG11705, Accurate Biology Co., Ltd., Changsha, Hunan, China). The target RNAs were amplified with the SYBR Green Pro Taq HS qPCR Kit (AG11701, Accurate Biology Co., Ltd., Changsha, Hunan, China). Gene expression levels were normalized to the $\beta$-actin internal control using the 2^{-Δ⁢Δ⁢ct} method. The sequences of all primers used are provided in Supplementary Table 1.

RNA from nuclear and cytoplasmic fractions was extracted using the PARIS™ Kit (AM1921, Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer’s instructions. $\beta$-actin was used as the internal control for cytoplasmic RNA, while U6 served as the reference for nuclear RNA. Detailed primer sequences are provided in Supplementary Table 1.

Fluorescence-labeled probes targeting DPY19L1P1, U6, and 18S rRNA were commercially designed and produced by RiboBio (Guangzhou, China). FISH was performed using the Ribo™ Fluorescent In Situ Hybridization Kit (C10910, RiboBio Co., Ltd., Guangzhou, Guangdong, China). Briefly, HN30 cells were cultured on sterile glass coverslips, fixed with 4% paraformaldehyde for 10 min, and permeabilized with 0.5% Triton X-100. After pre-hybridization, cells were incubated with fluorescent probes at 37 °C overnight. The next day, excess probes were removed by washing, and nuclei were counterstained with DAPI. Coverslips were mounted using an antifade reagent, and fluorescence images were captured using a TCS SP8 laser scanning confocal microscope (version 3.7.5, Leica Microsystems, Germany). Image processed with LAS X software (Leica Microsystems, Wetzlar, Germany) to determine the subcellular localization of DPY19L1P1.

MeRIP was conducted using the GenSeq® m7G MeRIP Kit (GS- ET-004A, CloudSeq Inc., Shanghai, China) to assess the m7G modification levels of the specific transcripts. Initially, total RNA was treated with mRNA Decapping Enzyme (M0608S, New England Biolabs, Ipswich, MA, USA) to remove the 5^′ cap structure. The RNA was then fragmented, and 3 µg of the fragmented RNA was reserved as the input control. The remaining RNA was mixed with either anti-m7G antibody or control IgG, both pre-bound to Protein A/G magnetic beads were added, and the mixture was incubated at 4 °C for 1 hour. After immunoprecipitation, the beads were washed with 1$\times$ IP buffer, and the bound RNA was purified using RLT buffer. Both input and m7G-enriched RNA fractions were reverse transcribed, followed by qPCR to quantify m7G enrichment.

siRNAs targeting METTL1 and WDR4 (si-METTL1 and si-WDR4), as well as Smart Silencer RNAs targeting DPY19L1P1 (SS-DPY19L1P1), were synthesized by RiboBio (Guangzhou, China). The target sequences are provided in Supplementary Table 2. Expression plasmids for METTL1, WDR4 and DPY19L1P1 were constructed by GeneChem Co., Ltd. (Shanghai, China). Cells were transfected using the Lipo8000TM reagent (Beyotime, Shanghai, China) following the protocol. qPCR was used to evaluate transfection efficiency.

Cells (1.5 $\times$ 10³ cells/well) transfected with the SS-DPY19L1P1 or DPY19L1P1 plasmid for 24 h were plated in triplicate into 96-well plates. Following incubation, 10 µL of CCK-8 solution (CK04, Dojindo Laboratories, Kumamoto, Japan) was added to each well, and the plates were incubated at 37 °C with 5% CO₂ for 2 hours. Cell viability was assessed daily for five consecutive days by measuring the optical density at 450 nm with a microplate reader (Thermo Fisher Scientific, Waltham, MA, USA).

The cells (2.5 $\times$ 10⁴ cells/well) transfected with the SS RNA/expression plasmid for 24 h were seeded into the upper chamber in 150 µL of serum-free DMEM (YRPHE, Shanghai, China). Subsequently, the lower chamber was filled with DMEM containing 10% fetal bovine serum to support cell growth, and the cells were incubated at 37 °C and 5% CO₂ for 24–48 h. After fixation with 4% paraformaldehyde solution (Beyotime), the cells were stained with 0.5% Crystal Violet Staining Solution (C0121, Beyotime Biotech Inc., Shanghai, China). Five fields of view were randomly selected for imaging using a Leica S8 APO Stereomicroscope (Leica Microsystems CMS GmbH, Wetzlar, Germany).

Twelve 4-week-old BALB/c nude mice (Jinan Pengyue Experimental Animal Breeding Center, SCXK [Lu] 2019-0003) were randomly assigned into two groups (n = 6 per group). All animal studies received approval from the Animal Care and Use Committee of Weifang Medical University (Approval No. 2023SDL332). Due to institutional restructuring, the committee has been renamed the Animal Care and Use Committee of Shandong Second Medical University, and the original approval remains valid under the same protocol.

To establish the subcutaneous xenograft tumor model, CAL-27 cells (2 $\times$ 10⁷ cells/mL) were injected subcutaneously (100 µL) into the dorsal area near the forelimbs on both sides of each mouse. One week after inoculation, mice received intratumoral injections of 5 nmol/50 µL cholesterol- and methylation-modified DPY19L1P1-targeting antisense oligonucleotide (ASO-DPY19L1P1) or ASO-NC every 4 days, for a total of five injections. The ASO-DPY19L1P1 sequence was derived from the most effective candidate identified in the SS-DPY19L1P1 screening and was synthesized by RiboBio (Guangzhou, China) with cholesterol and methylation modifications to enhance in vivo stability and transfection efficiency. This method of in vivo transfection in animal models has been documented in the previous studies [32].

Following 28 days, euthanasia was performed on mice under anesthesia via an intraperitoneal injection of sodium pentobarbital (150 mg/kg; prepared as a 50 mg/mL solution, Sigma-Aldrich, St. Louis, MO, USA), with death confirmed by the absence of heartbeat and respiration. Cervical dislocation was applied as a secondary measure if necessary. All procedures adhered to the AVMA Guidelines for the Euthanasia of Animals (2020).

Following euthanasia, tumors were excised, photographed, and measured to compare tumor volume and weight between the two groups. The tumor tissues underwent fixation in 4% paraformaldehyde for 24 hours, followed by dehydration, paraffin embedding, and sectioning. Histological analysis was conducted using Hematoxylin and Eosin (H&E) staining, while immunohistochemistry with a Ki-67 antibody was employed to assess cellular proliferation. The stained tissue sections were visualized and captured using an upright light microscope.

To evaluate the splicing efficiency of DPY19L1P1, qPCR was performed using two sets of transcript-specific primers, one amplifying unspliced precursor RNA (targeting intronic regions or exon-intron junctions) and the other amplifying spliced mature RNA (targeting exon-exon junctions). Splicing efficiency was quantified as the ratio of mature RNA to precursor RNA, and changes in this ratio were used to indicate alterations in splicing efficiency.

ATP levels in CAL-27 and HN30 cells were measured using an ATP Content Assay Kit (Cat# BC0300, Solarbio, Beijing, China) according to the manufacturer’s instructions. In brief, the cells were harvested and resuspended in the lysis buffer (1 mL per 10⁷ cells). After homogenization and full lysis on ice, the samples were centrifuged at 10,000 $\times$g for 10 minutes at 4 °C, after which the supernatants were collected. For quantification, 100 µL of 0.625 µmol/mL ATP standard solution or cell lysate was added to a 1 mL quartz cuvette containing 640 µL of working reagent and 260 µL of reagent I. The reaction mixture underwent incubation at 37 °C for 10 min. Absorbance was then measured at 340 nm using a UV spectrophotometer. ATP concentrations were derived from a standard curve and are expressed as µmol per 10⁶ cells.

Glucose uptake was quantified with a commercial Glucose Assay Kit (A154-1-1, Nanjing Jiancheng Bioengineering Institute, Nanjing, China) following the manufacturer’s protocol. Cells were seeded in 6-well plates and subjected to the indicated treatments. After 24–48 h, culture supernatants were collected and centrifuged at 12,000 $\times$g for 10 min. Glucose concentrations were measured using the glucose oxidase-peroxidase (GOD-POD) method, and absorbance was detected at 505 nm. Glucose consumption was determined as the difference between the concentration in blank medium and the residual concentration in the experimental groups, and was normalized to cell number or total protein content.

Lactate levels in culture media were quantified using the L-Lactic Acid (LA) Colorimetric Assay Kit (E-BC-K044-M, Elabscience Biotechnology Inc., Wuhan, Hubei, China) following the manufacturer’s protocol. After treatment or transfection, culture supernatants were collected and centrifuged at 12,000 $\times$g for 10 minutes to remove debris. Dilute an appropriate amount of the supernatant, then add lactate assay buffer and color development reagent. Incubate at 37 °C for 10 minutes. Measure absorbance at 530 nm using a microplate reader. Calculate lactate concentration based on the standard curve and normalize to cell number or total protein concentration.

Western blotting was carried out as previously described [22]. Antibodies employed in this study are listed in Supplementary Table 3. HRP-conjugated goat anti-mouse IgG (H+L) (Cat# P8001) and goat anti-rabbit IgG (H+L) (Cat# P8002) were purchased from New Cell and Molecular (NCM) Biotech (Suzhou, Jiangsu, China) and used as secondary antibodies at a 1:10,000 dilution. Immunoreactivity was visualized with enhanced chemiluminescence (ECL) Ultra reagent (NCM Biotech, Suzhou, Jiangsu, China). Band intensities were measured using ImageJ software, and the relative protein expression levels were analyzed accordingly.

All statistical analyses were performed using SPSS version 16.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism version 10.0 (GraphPad Software, San Diego, CA, USA). For comparisons between two groups, an unpaired Student’s t-test was used. Differences among more than two groups were assessed using one-way analysis of variance (ANOVA) followed by post hoc tests as appropriate. Pearson’s correlation analysis was conducted to evaluate the relationships between gene expression levels. All data are expressed as mean $\pm$ standard deviation (SD), and p-values 0.05 were considered statistically significant unless otherwise noted.

To explore the m7G methylation landscape in OSCC, m7G MeRIP-seq analysis was performed on three OSCC tissue samples (C1, C2, C3) and three matched adjacent normal tissue (N1, N2, N3). For high-confidence identification, only m7G peaks and lncRNAs detected in at least two biological replicates per group were included in subsequent analyses.

In OSCC tissues, a total of 1672 m7G peaks and 5651 modified lncRNAs were commonly detected across three samples (Fig. 1A,B). In comparison, 1220 shared m7G peaks and 5903 m7G-modified lncRNAs were identified in normal tissues (Fig. 1C,D).

Fig. 1.

Comprehensive landscape of m7G-modified lncRNAs in OSCC and normal tissues. (A,B) Venn diagrams showing the number of total m7G peaks (A) and m7G-modified lncRNAs (B) identified in OSCC tissues, based on the intersection of three OSCC samples (C1–C3), as analyzed by methylated RNA immunoprecipitation sequencing (MeRIP-seq). (C,D) Venn diagrams showing the number of total m7G peaks (C) and m7G-modified lncRNAs (D) identified in normal tissues (N1–N3), based on the overlap of three normal samples (N1–N3), also determined by MeRIP-seq. (E,F) Venn diagrams showing the overlap and specificity of m7G peaks (E) and m7G-modified lncRNAs (F) between OSCC and normal tissues. (G) Circos plot showing the chromosomal distribution of m7G-modified lncRNAs in OSCC and normal tissues. (H) Bar graph showing the number of m7G peaks across each chromosome in OSCC and normal tissues. (I) Distribution of the number of m7G peaks per lncRNA transcript in cancer and normal tissues. (J) Box plot comparing the length of m7G peaks (log₁₀-transformed) between OSCC and normal tissues. (K) Top six significantly enriched sequence motifs within m7G peaks identified in normal and OSCC tissues, along with associated p-values and E-values. m7G, N7-methylguanosine; lncRNA, long non-coding RNA; OSCC, oral squamous cell carcinoma. ****p 0.0001.

Upon integrating the OSCC and normal datasets, we found that 5486 m7G peaks and 5135 lncRNAs were specific to OSCC, while 2312 m7G peaks and 8646 m7G-modified lncRNAs were shared between the two groups (Fig. 1E,F). This finding highlights that m7G methylation in OSCC encompasses both conserved and tumor-specific modification patterns, suggesting its potential involvement in OSCC-specific transcriptional regulation.

A Circos plot, generated using the Circos package in R, revealed the chromosomal localization of m7G peaks in both OSCC and normal tissues (Fig. 1G). These peaks were located predominantly on chromosomes 1, 17, and 19 (Fig. 1H). To further characterize the m7G modification landscape, we analyzed the distribution of m7G peaks per lncRNA. In normal tissues, the majority of m7G-modified lncRNAs (approximately 75%) harbored only a single peak. In contrast, although 50.13% of lncRNAs in OSCC tissues also carried a single m7G peak, a substantially higher proportion exhibited multiple peaks—nearly 50% had two or more. This shift suggests an increase in methylation complexity in lncRNAs during OSCC tumorigenesis (Fig. 1I). Additionally, the peak length distribution differed significantly between OSCC and normal tissues (p 0.0001), with OSCC tissues exhibiting longer peak regions (Fig. 1J).

Motif analysis was conducted using DREME software to identify conserved sequence motifs within the m7G peaks. A total of six significantly enriched motifs were detected in each group (Fig. 1K). Among these, a conserved RGAAR motif (R = G/A) was predominantly enriched in OSCC tissues (p = 2.4 $\times$ 10^-23, E = 1.9 $\times$ 10^-18), whereas the CUGKR motif (K = G/U, R = G/A) was more frequently observed in normal tissues (p = 4.9 $\times$ 10^-40, E = 4.2 $\times$ 10^-35). Although some motifs were shared between cancer and normal groups, several were uniquely enriched in each, suggesting both common and distinct m7G recognition patterns. These findings are consistent with previously reported m7G motifs, further supporting their biological relevance in epitranscriptomic regulation.

Based on their genomic locations, m7G-modified lncRNAs can be classified into six categories: exon sense-overlapping, intronic antisense, intergenic, natural antisense, intron sense-overlapping, and bidirectional. Among these, the majority of m7G-modified lncRNAs in both OSCC and normal tissues originated from exon sense-overlapping and intergenic regions (Fig. 2A). Comparative analysis of the distribution patterns revealed that m7G-modified lncRNAs in OSCC tissues (C1–C3) were significantly less enriched in exon sense-overlapping regions and more frequently derived from intergenic regions compared to those in normal tissues (N1–N3) (Fig. 2B).

Fig. 2.

Integrated characterization of differentially m7G-methylated lncRNAs in OSCC. (A,B) Genomic feature distribution of m7G-modified lncRNAs in normal (top) and OSCC (bottom) tissues. Pie charts (A) and bar graphs (B) showing the proportion of lncRNAs derived from 6 genomic regions. (C) Proportional analysis showing that 65.98% of differentially modified lncRNAs were hyper-m7G, while 34.02% were hypomethylated (hypo-m7G) in OSCC compared to normal tissues. (D) Genomic origin categorization of hyper- and hypo-m7G lncRNAs in OSCC tissues, revealing distinct distributions between the two methylation states. (E) Chromosomal distribution of hyper- and hypo-m7G lncRNAs across all chromosomes, with a broad but uneven genomic distribution. (F) Length distribution of lncRNAs with aberrant m7G modification. (G,H) GO and KEGG enrichment analyses of genes associated with hyper-m7G-modified lncRNAs in OSCC tissues. GO, Gene Ontology; BP, biological process; CC, cellular component; MF, molecular function; KEGG, Kyoto Encyclopedia of Genes and Genomes. **p 0.01, ***p 0.001, ns, not significant.

To further investigate group-specific methylation differences, differentially m7G-methylated lncRNAs were identified using DiffReps software, with thresholds set at FC $>$2 and p 0.05. A total of 15,085 hyper-m7G lncRNAs (65.98%) and 7779 hypo-m7G lncRNAs (34.02%) were identified in OSCC tissues compared to normal controls, indicating that m7G hypermethylation is the predominant epitranscriptomic alteration in OSCC (Fig. 2C). Further analysis of the genomic distribution showed that hyper-m7G lncRNAs were less frequently located in exon sense-overlapping regions and more enriched in intergenic regions compared to hypo-m7G lncRNAs (Fig. 2D), consistent with the overall group-wise differences shown in (Fig. 2B).

Additionally, analysis of the chromosomal localization of differentially m7G-modified lncRNAs revealed that hyper-m7G-modified lncRNAs were consistently more abundant than hypo-m7G-modified lncRNAs across all chromosomes (Fig. 2E). This distribution pattern was broadly consistent with the previously observed distribution of total m7G peaks, indicating potential chromosomal preferences of m7G modifications and possible links to transcriptional activity or regional genomic function. Moreover, analysis of transcript lengths revealed that lncRNAs shorter than 2000 nucleotides exhibited a higher frequency of both hyper- and hypo-m7G methylation events (Fig. 2F), suggesting that shorter lncRNAs may be more dynamically regulated by m7G methylation in OSCC (Fig. 2F).

To elucidate the potential biological functions of differentially m7G-methylated lncRNAs in OSCC, GO and KEGG enrichment analyses were performed based on their predicted target genes. The analysis focused on genes associated with hyper-m7G lncRNAs. All enrichment terms presented met the significance threshold of p 0.05 and were ranked according to both gene count and enrichment score. GO analysis was performed across three categories: Biological Process (BP), Cellular Component (CC), and Molecular Function (MF), with the top 10 enriched terms in each category presented in Fig. 2G. KEGG pathway analysis (Fig. 2H) revealed the top 20 significantly enriched pathways. Genes associated with hyper-m7G lncRNAs were significantly showed strong enrichment related to cytoskeletal remodeling, cell adhesion, signal transduction, transmembrane transport, membrane localization, and metabolic regulation. These findings highlight their potential involvement in cellular migration, signal transduction, and metabolic alterations in tumors. Similarly, GO and KEGG pathway analyses of genes associated with hypo-m7G lncRNAs revealed significant enrichment in synaptic signaling, membrane transport, cytoskeletal structure, and calcium-related signal transduction pathways (Supplementary Fig. 1A,B).

As shown in Table 1, the top ten lncRNAs exhibiting the most significant changes in m7G methylation, both hypermethylated and hypomethylated, are listed.

Differential expression analysis of lncRNAs was conducted using HTSeq software, identifying 266 upregulated and 303 downregulated lncRNAs based on the thresholds of p 0.05 and $|$log₂ FC$|$ $>$2 (Fig. 3A,B). By intersecting the transcriptome and epitranscriptome datasets, we identified 80 lncRNAs that exhibited concurrent hyper-m7G modification and upregulation, while 103 lncRNAs were both hypomethylated and downregulated (Fig. 3C).

Fig. 3.

Integrated analysis of differentially expressed and hyper-m7G-modified lncRNAs in OSCC. (A) Volcano plot displaying the differentially expressed lncRNAs in OSCC tissues compared to adjacent normal tissues. Red and blue dots represent significantly upregulated and downregulated lncRNAs, respectively ($|$log₂ FC$|$ $>$2, p 0.05). (B) Proportion of significantly upregulated (266; 46.75%) and downregulated (303; 53.25%) lncRNAs between tumor and normal tissues. (C) Venn diagrams illustrating the overlap between hyper-m7G-modified and upregulated lncRNAs (top), and between hypo-m7G-modified and downregulated lncRNAs (bottom), suggesting potential regulatory associations between m7G methylation and lncRNA expression. (D,E) GO and KEGG enrichment analyses of genes associated with significantly upregulated lncRNAs, indicating their involvement in metabolic processes, immune regulation, signaling pathways, and tumor-related pathways. (F,G) GO and KEGG enrichment analyses of genes associated with both hyper-m7G-modified and upregulated lncRNAs, highlighting enriched pathways related to metabolism, protein processing, stress response, and cancer-related signaling.

GO and KEGG pathway enrichment analyses were conducted for genes associated with dysregulated lncRNAs. The results showed that genes associated with upregulated lncRNAs were significantly enriched in diverse metabolic processes, pathways related to cell adhesion and migration, along with a variety of signaling pathways (Fig. 3D,E). Meanwhile, GO and KEGG analyses of genes associated with downregulated lncRNAs, showing significant enrichment in muscle system processes, cytoskeletal organization, and cardiomyopathy-related pathways (Supplementary Fig. 2A,B).

Furthermore, GO and KEGG enrichment analyses of genes associated with hyper-m7G-modified and upregulated lncRNAs revealed significant enrichment in various biological pathways, including retinoic acid metabolism, xenobiotic and drug metabolism, the proteasome pathway, and cell adhesion-related processes such as focal adhesion and adherens junction. These findings suggest that such lncRNAs may be involved in regulating cellular metabolic activity, protein degradation mechanisms, and tumor cell adhesion and migration, potentially contributing to OSCC progression (Fig. 3F,G). In parallel, functional annotation (GO and KEGG) of the target genes linked to the concurrent hypo-m7G and downregulated lncRNAs highlighted a prominent involvement in pathways related to neuronal development, intracellular signaling and transport, calcium signaling, and chromatin remodeling (Supplementary Fig. 2C,D).

Taken together, the data suggest that lncRNAs exhibiting both hypermethylation and upregulation are closely associated with oncogenic pathways, implicating them as potential contributors to OSCC progression.

Given the well-established association between m7G hypermethylation and OSCC progression, we further analyzed the set of 80 lncRNAs that exhibited both increased m7G methylation and upregulated expression in OSCC tissues. Among these lncRNAs, 20 showed a fold change in m7G methylation greater than 100 (Table 2). To ensure the specificity of non-coding RNA candidates, we examined all transcripts generated by each gene and retained only eight lncRNAs whose transcripts were classified as non-coding, whereas the remaining genes contained one or more protein-coding transcripts.

These eight lncRNAs were further analyzed using multiple public databases, including GEPIA3 (https://gepia3.bioinfoliu.com/) [33], ENCORI (https://rnasysu.com/encori/panCancer.php) [34], cProSite (https://cprosite.ccr.cancer.gov) [35], TIMER3 (https://compbio.cn/timer3/) [36], to evaluate their expression patterns in HNSCC. The results revealed that DPY19L1P1, BLACAT1, ANKRD36BP2, and MIR17HG were highly expressed in the TCGA-HNSC cohort (Fig. 4A–D; Supplementary Fig. 3A,B). To further explore regulatory mechanisms, the correlation between these candidate lncRNAs and METTL1/WDR4 was assessed. Among them, DPY19L1P1 and MIR17HG exhibited a significant positive correlation with METTL1 expression (Fig. 4E; Supplementary Fig. 3C). Notably, DPY19L1P1 showed the most prominent m7G hypermethylation and was thus prioritized as the top candidate for further investigation. In addition, analysis using the cProSite dataset revealed that DPY19L1P1 expression was also positively correlated with both METTL1 and WDR4 (Fig. 4F,G), suggesting a potential regulatory relationship.

Fig. 4.

Clinical characterization and expression patterns of DPY19L1P1 in OSCC. (A–D) DPY19L1P1 expression was significantly upregulated in head and neck squamous cell carcinoma (HNSCC) tissues compared to normal tissues, as shown by GEPIA3 (A), ENCORI (B), cProSite (C), and TIMER3 (D) databases. (E) Pearson correlation analysis based on the ENCORI database revealed a strong positive association between METTL1 and DPY19L1P1 expression levels in HNSCC samples. (F,G) Correlation analysis using cProSite showed a positive relationship between DPY19L1P1 expression and METTL1 (F) or WDR4 (G) expression in tumor tissues. (H) Analysis of the GDC (TCGA) database further confirmed the elevated expression of DPY19L1P1 in HNSCC tissues (n = 520) versus normal controls (n = 44). (I) Receiver operating characteristic (ROC) curve analysis revealed that DPY19L1P1 has good diagnostic performance for HNSCC, with an AUC of 0.7848 (95% CI: 0.7242–0.8454, p 0.0001). (J) DPY19L1P1 expression was significantly higher in stage IV tumors compared to stage I–III tumors and in poorly differentiated tumors compared to well- and moderately differentiated tumors, based on UCSC Xena data. AUC, area under the ROC curve. *p 0.05, ***p 0.001, ****p 0.0001.

Expression data from the GDC portal (https://portal.gdc.cancer.gov) [37] and the UCSC Xena Browser (http://xena.ucsc.edu/) [38] further confirmed that DPY19L1P1 is significantly upregulated in HNSCC tissues (Fig. 4H; Supplementary Fig. 3D,E). Moreover, receiver operating characteristic (ROC) curve analysis yielded an AUC of 0.7848, indicating that diagnostic potential of DPY19L1P1 in HNSCC (Fig. 4I). Elevated DPY19L1P1 expression correlated significantly with higher clinical stage (Stage IV) and poor tumor grade (Fig. 4J).

Collectively, these results suggest that DPY19L1P1 is a hyper-m7G-modified lncRNA with potential clinical utility as a diagnostic and prognostic biomarker in HNSCC.

The expression level of DPY19L1P1 was evaluated across commonly used OSCC cell lines. The results showed that DPY19L1P1 was markedly upregulated in OSCC cells, with CAL-27 and HN30 exhibiting the highest expression levels (Fig. 5A). Consequently, these two lines were adopted as model systems to elucidate how the m7G methyltransferases METTL1 and WDR4 influence both the expression and m7G modification of DPY19L1P1. In OSCC cell lines, knockdown or overexpression of METTL1 or WDR4 resulted in consistent changes in DPY19L1P1 expression, indicating a regulatory relationship (Fig. 5B,C).

Fig. 5.

METTL1/WDR4 regulates the expression and m7G modification of DPY19L1P1 in OSCC. (A) qPCR analysis of DPY19L1P1 expression in normal oral epithelial cells and OSCC cell lines (CAL-27, HN30, HN6, SCC9, and SCC25). (B) qPCR showing that knockdown of METTL1 or WDR4 significantly reduced DPY19L1P1 expression in CAL-27 cells. (C) qPCR showing that overexpression of METTL1 or WDR4 significantly increased DPY19L1P1 expression in HN30 cells. (D) Integrative Genomics Viewer (IGV) visualization of MeRIP-seq reads mapping to the DPY19L1P1 locus in OSCC tissues (C1–C3) and normal tissues (N1–N3), revealing enriched m7G peaks specifically in OSCC samples. (E) Schematic illustration of the predicted m7G modification sites in the DPY19L1P1 transcript based on m7G Hub data and MeRIP-seq peaks, highlighting the overlapping region and its sequence. (F) MeRIP-qPCR showing significant enrichment of m7G-modified DPY19L1P1 relative to IgG in CAL-27 and HN30 cells. (G) MeRIP-qPCR showing reduced m7G enrichment of DPY19L1P1 following METTL1 or WDR4 knockdown in CAL-27 cells. (H) MeRIP-qPCR showing increased m7G enrichment of DPY19L1P1 following METTL1 or WDR4 overexpression in HN30 cells. (I) qPCR analysis of precursor DPY19L1P1 transcript level after METTL1 or WDR4 knockdown in CAL-27 cells. (J) qPCR analysis of mature DPY19L1P1 transcript levels following METTL1 or WDR4 knockdown. (K) Splicing efficiency represented by the ratio of mature DPY19L1P1 to precursor DPY19L1P1 transcripts. siRNA, small interfering RNA; NC, Negative Control; IP, immunoprecipitation; pre-DPY19L1P1, precursor DPY19L1P1. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001.

To further characterize the m7G methylation status of DPY19L1P1, m7G peaks within its transcripts were visualized in OSCC and normal tissues using Integrative Genomics Viewer (IGV) (Fig. 5D). On the basis of predictions from the m7G Hub V2.0 database combined with MeRIP-seq data, exon 14 of DPY19L1P1 was identified as the major m7G-modified region of DPY19L1P1, spanning 126 base pairs (Fig. 5E). Primers targeting the peak region were designed accordingly (Supplementary Table 1), and MeRIP-qPCR analysis confirmed that DPY19L1P1 is hyper-m7G-modified in OSCC cells (Fig. 5F). Importantly, m7G enrichment on DPY19L1P1 decreased significantly following METTL1 or WDR4 knockdown, whereas overexpression of either enzyme enhanced m7G methylation (Fig. 5G,H). Together, these findings indicate that the METTL1/WDR4 methyltransferase complex modulates both the expression and m7G methylation level of DPY19L1P1 in OSCC.

To further investigate the regulatory mechanism by which METTL1/WDR4 modulates DPY19L1P1 expression, we examined the levels of precursor and mature transcripts. Knockdown of METTL1 or WDR4 significantly increased the expression of DPY19L1P1 precursor RNA (Fig. 5I), while markedly reducing the levels of mature RNA (Fig. 5J), resulting in a notable elevation of the precursor-to-mature RNA ratio, indicative of impaired splicing efficiency (Fig. 5K). These findings suggest that METTL1/WDR4 regulates DPY19L1P1 at the post-transcriptional level by modulating its splicing efficiency.

Nuclear/cytoplasmic fractionation analysis in CAL-27 cells revealed that DPY19L1P1 was predominantly localized in the nucleus, with a nuclear/cytoplasmic distribution ratio of 75.06 $\pm$ 2.52% (nuclear) versus 24.94 $\pm$ 2.52% (cytoplasmic) (Fig. 6A). Consistently, FISH analysis further confirmed the nuclear enrichment of DPY19L1P1 in HN30 cells (Fig. 6B). To elucidate the functional role of DPY19L1P1 in OSCC, we performed both loss- and gain-of-function assays. Transfection with SS-DPY19L1P1 effectively silenced DPY19L1P1 expression in CAL-27 cells (Supplementary Fig. 4A), whereas transfection with an overexpression plasmid significantly increased its expression in HN30 cells (Supplementary Fig. 4B). CCK-8 assay revealed that knockdown of DPY19L1P1 inhibited CAL-27 cell proliferation (Fig. 6C), whereas its overexpression enhanced proliferation in HN30 cells (Fig. 6D). Similarly, Transwell migration assay revealed that DPY19L1P1 silencing significantly inhibited migration of CAL-27 cells (Fig. 6E), while its overexpression promoted migration in HN30 cells (Fig. 6F). Moreover, among the three ASOs included in the Smart Silencer, ASO-DPY19L1P1-1 was identified as having the most potent knockdown efficiency (Supplementary Fig. 4C). This ASO (hereafter referred to as ASO-DPY19L1P1) was selected for intratumoral injection in an in vivo xenograft model. The results demonstrated that silencing DPY19L1P1 in vivo significantly suppressed tumor growth (Fig. 6G,H), further supporting its oncogenic role in OSCC.

Fig. 6.

DPY19L1P1 enhances OSCC progression and functionally cooperates with METTL1/WDR4. (A) The subcellular localization of DPY19L1P1 was determined by nuclear/cytoplasmic fractionation in CAL-27 cells. U6 and $\beta$-actin served as nuclear and cytoplasmic controls, respectively. (B) The subcellular localization of DPY19L1P1 was determined by nuclear/cytoplasmic fractionation in fluorescence in situ hybridization (FISH) in HN30 cells. U6 and 18S rRNA served as nuclear and cytoplasmic controls, respectively. Scale bar: 10 µm. (C,D) The effect of DPY19L1P1 on OSCC cell proliferation was determined via CCK-8 assays in DPY19L1P1-knockdown CAL-27 cells and DPY19L1P1-overexpressing HN30 cells. (E,F) The effect of DPY19L1P1 on OSCC cell migration was assessed using Transwell assays in DPY19L1P1-knockdown CAL-27 cells and DPY19L1P1-overexpressing HN30 cells. Scale bar: 200 µm. (G) Subcutaneous xenograft model in BALB/c nude mice using CAL-27 cells transfected with ASO-NC or ASO-DPY19L1P1. Representative tumor images and tumor weights after 28 days are shown. (H) Histological analysis of xenograft tumors by HE and Ki-67 staining. Images were acquired at 100$\times$ and 400$\times$ magnification. Scale bar: 50 µm. (I,J) CCK-8 assays assessing the effect of DPY19L1P1 overexpression on cell proliferation in CAL-27 cells with METTL1 (I) or WDR4 (J) knockdown. (K,L) Transwell migration assays evaluating the effect of DPY19L1P1 overexpression on cell migration in CAL-27 cells with METTL1 (K) or WDR4 (L) knockdown. Scale bar = 200 µm. SS, Smart Silencer; ASO, antisense oligonucleotide; H&E, hematoxylin and eosin. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001.

It has been well documented that METTL1/WDR4 functions as a key driver in facilitating the growth and motility of OSCC cells [7]. To determine whether DPY19L1P1 functions downstream of METTL1/WDR4, we conducted rescue experiments. Overexpression of DPY19L1P1 partially reversed the reduced proliferative (Fig. 6I,J) and migratory (Fig. 6K,L) abilities caused by METTL1 or WDR4 knockdown. These results support the notion that DPY19L1P1 acts as a downstream effector of the METTL1/WDR4 complex, mediating its oncogenic effects in OSCC progression.

Given that previous GO and KEGG analyses revealed enrichment of hyper-m7G-modified and upregulated genes in functions and pathways related to metabolism and cell adhesion, we further investigated whether DPY19L1P1 is involved in metabolic processes and EMT.

We first assessed changes in intracellular ATP levels following DPY19L1P1 knockdown or overexpression. ATP quantification assays showed that silencing DPY19L1P1 significantly reduced ATP levels in CAL-27 cells, whereas its overexpression markedly increased ATP production in HN30 cells, indicating that DPY19L1P1 is involved in cellular energy metabolism (Fig. 7A). Furthermore, glucose uptake and lactate production assays confirmed that DPY19L1P1 knockdown decreased glucose consumption and lactate generation, while its overexpression enhanced both parameters (Fig. 7B,C). These results suggest that DPY19L1P1 promotes a shift toward glycolytic metabolism. In line with this, qPCR analysis revealed that DPY19L1P1 overexpression upregulated the mRNA level of key glycolytic genes, including HIF1A, GLUT1, HK2, PFKFB3, PKM, LDHA, PDK1, and PGK1 (Fig. 7D), implying a role in promoting aerobic glycolysis (the Warburg effect) and metabolic reprogramming in OSCC cells.

Fig. 7.

DPY19L1P1 promotes glycolytic metabolism and EMT in OSCC cells. (A) Intracellular ATP levels were measured in CAL-27 cells with DPY19L1P1 knockdown and HN30 cells with DPY19L1P1 overexpression. (B) Glucose uptake was assessed in CAL-27 cells with DPY19L1P1 knockdown and HN30 cells with DPY19L1P1 overexpression. (C) Lactate production was quantified in CAL-27 cells with DPY19L1P1 knockdown and HN30 cells with DPY19L1P1 overexpression. (D) qPCR analysis of key glycolytic pathway genes (HIF1A, GLUT1, HK2, PKM, LDHA, etc.) in CAL-27 cells with DPY19L1P1 knockdown and HN30 cells with DPY19L1P1 overexpression. (E) Western blot analysis of glycolysis- and lipid metabolism–related proteins (HIF-1$\alpha$, GLUT1, PDK1, PKM2, LDHA, p-ACC, and ACLY) in CAL-27 cells with DPY19L1P1 knockdown and HN30 cells with DPY19L1P1 overexpression. (F) qPCR analysis of EMT marker genes (E-cadherin, N-cadherin, and vimentin) in CAL-27 cells with DPY19L1P1 knockdown and HN30 cells with DPY19L1P1 overexpression. (G) Western blot analysis of EMT-related proteins (E-cadherin, N-cadherin, and vimentin) in CAL-27 cells with DPY19L1P1 knockdown and HN30 cells with DPY19L1P1 overexpression. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001.

Consistently, Western blot analysis confirmed that DPY19L1P1 positively regulates glycolytic markers (HIF-1$\alpha$, GLUT1, PDK1, PKM2, LDHA) and lipogenesis-related proteins (p-ACC, ACLY), further supporting its functional involvement in tumor metabolic reprogramming (Fig. 7E).

Additionally, we investigated the role of DPY19L1P1 in EMT. qPCR and Western blot analysis revealed that DPY19L1P1 knockdown caused increased E-cadherin and decreased N-cadherin and vimentin expression, while its overexpression produced the opposite effect (Fig. 7F,G). These findings indicate that DPY19L1P1 facilitates EMT progression in OSCC cells, potentially contributing to enhanced migratory and invasive capabilities.