The mRNA expression data and clinical information were downloaded from The Cancer Genome Atlas (TCGA) database (https://www.cancer.gov/ccg/research/genome-sequencing/tcga) and Genotype-Tissue Expression (GTEx) database (15). The TCGA database includes ccRCC tissues from 531 patients and adjacent normal tissues from 72 cases, while the GTEx database includes RNAseq data from 28 normal tissues (16). The extracted data is presented in Transcripts Per Million (TPM) format. The proteomics data and immunohistochemical staining slides are derived from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) database (17) and The Human Protein Atlas (HPA) database (https://www.proteinatlas.org/) (18). The gene amplification and mutation status of MGLL was obtained from the cBioPortal (http://www.cbioportal.org/) for Cancer Genomics (19). The GEPIA database (http://gepia.cancer-pku.cn/) (20), the UALCAN database (https://ualcan.path.uab.edu/index.html) (21) and the TCGA database were utilized for Kaplan-Meier survival curve analysis. The three independent validation datasets, comprising comprehensive gene expression profiles of both ccRCC and matched normal control samples, were systematically retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). The GSE66270 dataset contains mRNA microarray expression analysis data from 28 pairs of matched malignant and non-malignant kidney tissue samples from 14 ccRCC patients without diagnosed metastasis (22). The GSE40435 dataset comprises comprehensive whole-genome expression profiles derived from 101 paired samples of ccRCC tumors and their corresponding adjacent non-tumor kidney tissues (23). Additionally, the GSE213324 dataset contains high-throughput sequencing data obtained from 21 RCC samples and 20 normal kidney tissue specimens (24). Detailed information regarding these datasets is presented in Table 1 .

A threshold value of 50% was selected to categorize the cohort into high and low MGLL expression groups. The ‘ggplot2’ package of R software (version 4.2.1) is utilized to visualize the results of the differential analysis. The enrichment analysis of gene sets is performed using the ‘clusterProfiler’ package, including Gene Ontology (GO) analysis and Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway analysis. We performed Receiver Operating Characteristic (ROC) analysis on the data and visualized the results using the ‘pROC’ and ‘ggplot2’ packages. The ‘survival’ R package is used to conduct proportional hazards assumption tests and perform survival regression curve analysis, with results visualized using the ‘survminer’ and ‘ggplot2’ packages.

Gene correlation analysis employs the Spearman method, and an absolute correlation coefficient greater than 0.3 indicates a correlation between the two. The STRING database (https://cn.string-db.org/) is used to construct a protein-protein interaction (PPI) network, and then visualize the data using Cytoscape software 3.9.1 (https://cytoscape.org/index.html) (25, 26). We further analyzed the PPI network and performed gene screening and mining using the Mcode plugin.

The immune infiltration algorithm is based on the ssGSEA algorithm provided in the R package ‘GSVA’, which uses markers of 24 types of immune cells to calculate the immune infiltration status of the corresponding data. The stacked bar chart of immune infiltration is based on the CIBERSORT core algorithm. It utilizes markers of 22 types of immune cells provided by the CIBERSORTx website (https://cibersortx.stanford.edu/) to calculate the immune infiltration status of the data. The ‘ggplot2’ and ‘ggalluvial’ packages were used for computation and visualization. 29 m7G RNA modification-related regulator genes were extracted from the literature (27). We performed prognostic Least Absolute Shrinkage and Selection Operator (LASSO) coefficient selection by analyzing the cleaned data with the ‘glmnet’ package to obtain variable lambda values and visualize the results.

ACHN, A498, 786-O and HK-2 cells were purchased from the Chinese Academy of Sciences cell bank. All cells were cultured in the presence of penicillin/streptomycin at 37°C in air containing 5% CO₂. Total RNA was isolated utilizing a commercial RNA extraction kit following the manufacturer’s protocol. Reverse transcription was carried out using the Evo M-MLV RT Premix. Quantitative PCR (qPCR) analysis was performed with the SYBR^® Green Premix Pro Taq HS qPCR kit. All RNA extraction, reverse transcription, and real-time fluorescence quantification reagents were procured from Accurate Biotechnology Co., Ltd. (Changsha, Hunan, China). PCR amplification was conducted using a thermal cycler (Bio-Rad Laboratories, USA), while fluorescence quantification was performed on the LightCycler 480II system (F. Hoffmann-La Roche Ltd, Switzerland). β-actin was used as an endogenous reference to normalize RNA expression. The relative expression of genes was calculated using the 2^−ΔΔCt method. The primer sequences can be found in the Supplementary Table S1 .

Protein extraction was performed using RIPA lysis buffer (Beijing Solarbio Science & Technology Co., Ltd, China), followed by quantification of protein concentration through the Bicinchoninic Acid Assay (BCA) method, employing a gradient of standard samples to establish a calibration curve. Polyacrylamide gel electrophoresis (PAGE) was conducted using a gel preparation kit (Yzyme Biotech Co., Ltd, Shanghai, China), with protein samples (30 μg per lane) resolved on 6-15% sodium dodecyl sulfate-polyacrylamide gels and subsequently transferred to polyvinylidene fluoride (PVDF) membranes (MilliporeSigma, USA). Primary antibodies against MGLL (Catalog No.: ab124796, Abcam Inc., USA) and β-actin (Proteintech Group, Inc., China) were utilized, along with horseradish peroxidase-conjugated goat anti-rabbit/mouse secondary antibodies (Proteintech Group, Inc., China). Electrophoresis and transfer buffers were obtained from Yzyme Biotech Co., Ltd, Shanghai, China. Membranes were blocked with 5% non-fat milk, incubated overnight with primary antibodies, and subsequently probed with horseradish peroxidase-conjugated secondary antibodies. Protein bands were visualized using enhanced chemiluminescence (ECL) detection solution (MilliporeSigma, USA) in a gel imaging system.

Stable MGLL knockdown cell lines were established in ccRCC cell lines A498, 786-O and ACHN. To achieve stable transfection, si-MGLL was subcloned into the lentiviral vector (LV) GV493 (hU6-MCS-CBh-gcGFP-IRES-puromycin), along with the negative control virus (NC, CON313), both purchased from Shanghai Genechem Co., Ltd., China. Puromycin was used to screen for stable cell lines. Three RNA interference (RNAi) sequences were prepared for LV-si-MGLL: PSC91228-1 (CAACTCCGTCTTCCATGAAAT), PSC91229-1 (CCAGGACAAGACTCTCAAGAT), and PSC91230-1 (CCAATCCTGAATCTGCAACAA). The control insert sequence was TTCTCCGAACGTGTCACGT.

The CCK-8 solution was prepared by diluting the CCK-8 reagent (Dojindo Molecular Technologies, Inc., Japan) in incomplete medium at a 10:1 volumetric ratio. Cells were seeded in a 96-well plate with five technical replicates per group and maintained in a standard cell culture incubator. Absorbance at 450 nm was quantified using a microplate reader at 0, 24, 48, and 72-hour time points. The experiment was performed in triplicate, and a standard curve was established for quantification. For the colony formation assay, cells were plated at a density of 600 cells per well in 6-well plates and cultured until microscopic examination revealed the formation of >50 single-cell colonies. Subsequently, cells were washed, fixed, and stained with 0.1% crystal violet (Beijing Solarbio Science & Technology Co., Ltd., China), followed by air-drying and photographic documentation. For the migration assay, cells were seeded in the upper chamber of a Transwell system (Corning Incorporated, NY, USA) containing 200 µL of serum-free medium, while the lower chamber was filled with complete medium supplemented with 20% fetal bovine serum as a chemoattractant. After 48 hours of incubation, cells in the upper chamber were fixed, stained with crystal violet for 30 minutes, and the membrane was mounted on glass slides, sealed, dried, and imaged for analysis.

All statistical analyses were performed using R software (version 4.2.1) alongside GraphPad Prism 9 software. Statistical significance for differences between groups was determined using the t-test. For comparisons among multiple groups, one-way analysis of variance (ANOVA) was used. A p-value of less than 0.05 was considered statistically significant. The following p-values were considered: *p < 0.05, **p < 0.01, and ***p < 0.001.

To explore the significance of MGLL in tumor tissues, we first analyzed its expression in pan-cancer tissues. The results indicated statistically significant differences in MGLL expression in 26 of the 33 tumor types compared to normal tissues ( Figure 1A ). MGLL was significantly upregulated in diffuse large B cell lymphoma (DLBC), acute myeloid leukemia (LAML), ovarian serous cystadenocarcinoma (OV), pancreatic adenocarcinoma (PAAD), and stomach adenocarcinoma (STAD), while it was downregulated in breast invasive carcinoma (BRCA), head and neck squamous cell carcinoma (HNSC), lung squamous cell carcinoma (LUSC), bladder urothelial carcinoma (BLCA), and skin cutaneous melanoma (SKCM). RNA-seq data from the TCGA and GTEx databases show that MGLL mRNA levels are significantly elevated in ccRCC compared to normal tissues ( Figures 1B, C ). We used ROC curve analysis to evaluate the effectiveness of MGLL in distinguishing between tumor and non-tumor tissues. The MGLL gene demonstrated an area under the curve (AUC) value of 0.787 in initial analyses ( Figure 1D ). Subsequent validation across three independent GEO datasets revealed good diagnostic performance, with ROC curve analyses yielding AUC values of 0.934 (GSE66270) ( Figure 1E ), 0.922 (GSE40435) ( Figure 1F ), and 0.921 (GSE213324) ( Figure 1G ). These consistently high AUC values (> 0.9) across multiple datasets suggest that MGLL expression exhibits significant potential as a biomarker for distinguishing tumor from non-tumor tissues. Further confirmation through qRT-PCR experiments showed that compared to normal renal tubular epithelial cells HK-2, the mRNA level of MGLL was significantly elevated in ccRCC cell lines ACHN, A498 and 786-O ( Figure 1H ). The CPTAC protein database indicates that the protein level of MGLL is significantly higher in ccRCC tissues compared to normal tissues ( Figure 1I ). Immunohistochemical staining in the HPA database indicates that the protein level of MGLL in ccRCC tissues is higher than that in normal renal tissues ( Figure 1J ). In addition, to explore the mutation level of MGLL in ccRCC, we analyzed its genome and copy number. Analysis of the MGLL gene’s OncoPrint profile in ccRCC patients using the cBioPortal database revealed that truncating, deep deletion, and profiled mutations in MGLL were present in less than 8% of cases ( Figure 1K ).

The baseline data of ccRCC patients from the TCGA database were statistically analyzed ( Table 2 ). The data was divided into two groups: 270 ccRCC patients with low MGLL expression and 271 with high MGLL expression. Significant differences were observed between the two groups in histologic grade (p = 0.007) and gender (p < 0.001). No significant differences were found in TNM staging and age. Logistic regression analysis was conducted to examine the correlation between MGLL expression and the clinical pathological features of ccRCC ( Table 3 ). Similar to the baseline statistics, the logistic regression analysis showed differences between the two groups in terms of the histologic grade (p = 0.043) and gender (p < 0.001).

After confirming the increased expression of MGLL in ccRCC tissues, an exploration was conducted on the relationship between the expression of MGLL and 11 clinical variable characteristics within the ccRCC patient cohort. The variables included age, gender, T stage, N stage, M stage, pathologic stage, histologic grade, OS event, serum calcium, hemoglobin levels, and race ( Figure 2 ). The results indicated that MGLL expression correlated with gender ( Figure 2B ), histologic grade ( Figure 2G ), OS event ( Figure 2H ), and hemoglobin ( Figure 2J ). Further analysis through disease-free survival curves indicated that high levels of MGLL expression were associated with decreased survival rates ( Figure 2L ), suggesting that the elevation of MGLL is related to clinical prognosis.

We categorized ccRCC patients into two groups based on MGLL expression levels: high and low. We then performed a differential analysis. The screening criteria were established as follows: protein-coding genes, |logFC| > 1, and p.adj < 0.05. We screened a total of 751 differentially expressed genes, which included 22 upregulated genes and 729 downregulated genes ( Figure 3A ). Detailed information on all genes can be found in Supplementary Table S2 . Next, we analyzed the co-expressed genes related to MGLL expression in the TCGA ccRCC dataset. The screening criteria were set as follows: |Spearman correlation| >0.4 and p.adj < 0.05. A total of 2008 genes were obtained, including 1 negatively correlated gene and the rest being positively correlated genes. For specific information on genes, please refer to Supplementary Table S3 . The heatmap of the 10 genes with Spearman correlation > 0.6 co-expressed with MGLL is shown in Figure 3B . After conducting GO clustering analysis on the 2008 co-expressed genes of MGLL, the top five major biological processes involved are: proteasomal protein catabolic process, proteasome-mediated ubiquitin-dependent protein catabolic process, macroautophagy, Golgi vesicle transport, and response to endoplasmic reticulum stress (ERs) ( Figure 3C ). The key cellular components include the vacuolar membrane, lytic vacuole membrane, lysosomal membrane, focal adhesion, and intrinsic component of the organelle membrane ( Figure 3D ). The involved molecular functions mainly include: transcription coregulator activity, cadherin binding, ubiquitin-like protein ligase binding, GTPase binding, and transcription coactivator activity ( Figure 3E ). The main KEGG pathways involved include: Endocytosis, Protein processing in ER, Shigellosis, Hepatitis B, and Sphingolipid signaling pathway ( Figure 3F ). The detailed results of the GO analysis and KEGG pathway analysis can be found in Supplementary Table S4 .

To further explore the genes that are strongly co-expressed with MGLL and their functional roles, we constructed a PPI network using the STRING database for 65 genes with a Spearman correlation greater than 0.55 ( Supplementary Table S3 ), and visualized it using Cytoscape software ( Figure 3G ). We conducted module analysis on the network using the MCODE plugin to identify highly interconnected clusters of nodes. These clusters were scored based on node connectivity density and the number of nodes, facilitating the discovery of functionally related modules. The results show a module composed of 11 genes: ACLY, EIF4G1, PDIA5, ACAD9, SUN2, PKM, SEC61A1, UMPS, CALM3, RAB1B, and MGLL ( Figure 3G ). Then, the expression of these genes in ccRCC and normal tissues was analyzed through the TCGA and GTEx databases, revealing that, except for RAB1B, the other 10 genes exhibited significant differential expression (p < 0.05) ( Figure 3H ). This indicates that genes strongly associated with MGLL play important roles in ccRCC.

To explore the correlation between MGLL expression levels and tumor immune response, TCGA database was used to investigate immune infiltration in ccRCC with different MGLL expression levels. Analysis of 24 immune cell types revealed that MGLL expression in ccRCC patients is positively correlated with several immune cells, including Neutrophils, Th17 cells, Eosinophils, Mast cells, NK cells, dendritic cells (DCs), and NK CD56dim cells. Conversely, MGLL expression is negatively correlated with Cytotoxic cells, NK CD56bright cells, and CD8 T cells (p < 0.05) ( Figure 4A ). Further studies showed that MGLL expression was significantly positively correlated with Neutrophils (p < 0.001), Th17 cells (p < 0.001) and Eosinophils (p < 0.001) infiltration levels ( Figures 4B-D ). Based on the median expression level of MGLL, ccRCC patients were divided into high and low expression groups, and then the distribution of 22 immune cell subtypes (including 7 types of T cells, naive and memory B cells, plasma cells, NK cells, and bone marrow subpopulations) in the two groups was analyzed ( Figure 4E ). We also observed significant differences in the infiltration levels of Neutrophils, Th17 cells, Eosinophils, DC, NK CD56dim cells, NK cells, and Mast cells between the high and low MGLL expression groups (p < 0.05) ( Figures 4F-M ).

The functional enrichment of genes co-expressed with MGLL includes biological processes such as mRNA processing, RNA catabolism process, RNA localization, RNA splicing, and RNA transport (p.adj < 0.001), suggesting that MGLL expression may be linked to RNA methylation, which plays a vital role in splicing, export, processing, translation, and degradation. m7G methylation, one of the most prevalent RNA modifications, has recently garnered significant attention as an emerging area of research (28–30). However, its specific molecular role in ccRCC is still lacking in-depth research. We aim to explore the differential expression and functional role of m7G methylated genes associated with MGLL in ccRCC. We first evaluated the co-expression heatmap of MGLL and 29 m7G modification genes, finding statistically significant correlations for 25 of these genes (Spearman p < 0.001) ( Figure 5A ) ( Supplementary Table S5 ). The expression of MGLL was significantly positively correlated with NCBP2 (r = 0.498, p < 0.001) ( Figure 5B ), EIF4E2 (r = 0.486, p < 0.001) ( Figure 5C ), CYFIP1 (r = 0.476, p < 0.001) ( Figure 5D ), SNUPN (r = 0.472, p < 0.001) ( Figure 5E ), LARP1 (r = 0.452, p < 0.001) ( Figure 5F ), NUDT16 (r = 0.424, p < 0.001) ( Figure 5G ), and GEMIN5 (r = 0.42, p < 0.001) ( Figure 5H ). The scatter plot of genes with a correlation coefficient r > 0.3 related to MGLL expression can be found in Supplementary Figure S1 . Subsequently, the expression of 18 genes with a correlation of r > 0.3 was analyzed in ccRCC and normal tissues using the TCGA and GTEx databases, revealing that 13 genes exhibited significant differential expression (p < 0.01) ( Figure 5I ). We also observed the prognostic associations of NUDT16, NCBP1, NUDT4, and NSUN2 in ccRCC ( Figures 5J-M ), suggesting a potential regulatory relationship with MGLL, which may affect patient prognosis.

Based on the construction of the PPI network and the identification of m7G methylation genes related to MGLL, as well as the analysis using the TCGA and GTEx databases, a total of 23 genes including MGLL were found to have differential expressions in ccRCC and normal tissues (p < 0.05) ( Figures 3H and 5I ). We further conducted LASSO coefficient screening for the 23 genes in combination with clinical prognosis using the ‘glmnet’ R package. Through the LASSO regression analysis, a risk signature model based on minimum criteria was constructed with 15 optimal and most powerful prognostic markers, which are: ACAD9, ACLY, AGO2, CALM3, CYFIP1, EIF4E3, GEMIN5, LSM1, MGLL, NSUN2, NUDT16, NUDT4, PKM, SEC61A1, and UMPS ( Figures 6A, B ). We further observed the performance of these genes in the TCGA database through diagnostic ROC curves to assess the diagnostic value of the genes in ccRCC. The results indicate that the AUC values for ACLY, CALM3, NSUN2, NUDT16, NUDT4, and PKM exceed 0.7, demonstrating high accuracy and sensitivity ( Figures 6C, D ). To verify the validity of their expression in ccRCC tissues, we switched databases and conducted a repeated validation using the dataset GSE66270 from the GEO database. After data cleaning, organization, and analysis, the GSE66270 dataset contained 9,465 upregulated genes and 6,931 downregulated genes ( Figure 6E ). The heatmap showed that compared to normal tissues, ACLY, CALM3, NSUN2, PKM, and MGLL were significantly upregulated, while NUDT16 and NUDT4 were significantly downregulated in ccRCC tissues ( Figure 6F ), confirming the results of the previous analysis.

We conducted qRT-PCR experiments on the genes selected based on the prognostic LASSO coefficients. The results showed that, compared to normal renal tubular epithelial cells HK-2, the expression levels of ACLY, CYFIP1, EIF4E3, and UMPS were significantly upregulated in ccRCC cell lines ACHN, A498, and 786-O. GEMIN5, NSUN2, PKM, and SEC61A1 were significantly upregulated in A498 and 786-O, and CALM3 was also upregulated in 786-O ( Figure 6G ), which is consistent with our previous analysis. In contrast, the expression of NUDT4 and NUDT16 in ccRCC cell lines was found to be increased, which contradicts our consideration of them as downregulated genes. Since both genes are involved in m7G methylation and belong to the nudix hydrolase family, the mRNA of RNA methyltransferases may be tagged by themselves or other enzymes, which could promote degradation or inhibit translation (31). Additionally, several alternative splicing transcriptional variants of NUDT4 have been described (32), and its selective splicing or translation variants may also lead to reduced protein levels. In summary, the qRT-PCR experiments validated the results of our previous analysis.

The LASSO-derived prognostic risk score stratified patients into distinct high-risk and low-risk cohorts, with Kaplan-Meier survival analysis demonstrating significantly poorer outcomes in the high-risk group ( Figure 7A ). Subsequent integration of the risk score with key clinical parameters, including Pathologic Stage and TM classification, through univariate and multivariate Cox regression analyses revealed the risk score as an independent prognostic indicator (p < 0.001) ( Table 4 ), underscoring its clinical utility. Based on the multivariate regression analysis, the C-index at different time points showed that the model has relatively stable predictive performance ( Figure 7B ). Time-dependent ROC curve analysis yielded AUC values exceeding 0.7 at 1-, 3-, and 5-year intervals, indicating superior sensitivity and specificity ( Figure 7C ). The prognostic nomogram, incorporating both clinical variables and risk score, demonstrated accurate prediction of patient survival probabilities at 1, 3, and 5 years ( Figure 7D ), with calibration curves confirming excellent model fit ( Figure 7E ). The Sankey diagram illustrates the flow and proportional relationships of the multi-level association data among relevant clinical variables, survival, and risk scores ( Figure 7F ).

To elucidate the functional role of MGLL, we performed lentivirus-mediated knockdown experiments and assessed its biological implications. qPCR analysis of three LV-RNAi constructs in ccRCC cell lines (A498, 786-O, and ACHN) identified the PSC91228–1 sequence as the most effective in suppressing MGLL expression, achieving an inhibition rate of 80%-90% ( Figures 8A-C ). This construct was subsequently utilized for further functional assays. CCK-8 assays demonstrated that MGLL knockdown significantly reduced the proliferative capacity of A498, 786-O, and ACHN cells ( Figures 8D-F ). Consistent with these findings, plate colony formation assays revealed that MGLL silencing markedly impaired the colony-forming ability of A498 and 786-O cells ( Figures 8G, H ). Moreover, Transwell migration assays indicated that MGLL knockdown effectively suppressed the migratory potential of A498, 786-O, and ACHN cells ( Figures 8I-L ). To validate these observations, we conducted Western blot analysis in ACHN and A498 cell lines, which exhibited the highest MGLL expression levels based on qPCR results ( Figure 1H ). The Western blot data confirmed that MGLL protein expression was significantly elevated in ACHN and A498 cells compared to normal renal tubular epithelial cells (HK-2) ( Figure 8M ), corroborating our previous findings.