The present study employed a two-sample MR method. The eQTL datasets of lactylation-related genes were used as exposure variables, and the genome-wide association study (GWAS) summary statistics of lung cancer were used as outcome variables. As described previously (20–23), a catalog of 46 lactylation-related genes was compiled.

The eQTL exposure data were acquired from the IEU OpenGWAS Project (https://opengwas.io/), a global public database. Single nucleotide polymorphisms (SNPs) exhibiting a strong association with the expression of the 46 lactylation-related genes (P<5×10⁻⁸) were selected as IVs. To avoid weak instrument bias, only SNPs with an F-statistic >10 were included as IVs, ensuring a strong association between the IVs and gene expression.

The lung cancer outcome data were sourced from the Finnish database (finngen_R12_C3_lung cancer_EXALLC; http://r12.finngen.fi/pheno/C3_lung_cancer_EXALLC), which included 909 lung cancer cases and 3,788,794 healthy controls. The GWAS summary statistics in this database contained ~21,325,071 SNPs.

The MR analyses were performed using R software (version 4.4.2; http://cran.r-project.org/bin/windows/base/old/4.4.2/) packages ‘TwoSampleMR’ (version 0.5.6; http://CRAN.R-project.org/package=TwoSampleMR),‘ieugwasr’ (version 0.1.7; http://CRAN.R-project.org/package=ieugwasr) and ‘MRInstruments’ (version 0.4.0; http://CRAN.R-project.org/package=MRInstruments). IV selection was conducted by retaining SNPs strongly associated with lactylation-related gene expression (P<5×10-8) and with an F-statistic>10 to avoid weak instrument bias; SNPs were then clumped to remove linkage disequilibrium (LD) using a threshold of r<0.001 and a window size of 1Mb to ensure independence. The processed eQTL data were used as the exposure dataset, and the lung cancer GWAS summary statistics were used as the outcome dataset. Prior to analysis, allele harmonization was performed to ensure consistency of allele orientation and effect direction between exposure and outcome datasets. Strand-ambiguous SNPs and those with conflicting effect directions were excluded. The specific methods employed in the present study included inverse variance weighting (IVW), MR Egger, weighted median, simple mode and weighted mode. For each gene, the causal impact size (β), P-value, standard error, 95% confidence interval (95% CI) and odds ratio were recorded. Subsequently, The heterogeneity and pleiotropy of the odds ratio were calculated using Cochran's Q test (for heterogeneity) and the MR Egger intercept test (for horizontal pleiotropy), followed by multivariate sensitivity analyses including leave-one-out validation and MR-PRESSO outlier detection. The reliability of the MR findings was confirmed using scatter diagrams, forest plots and leave-one-out sensitivity analysis. Funnel plots were generated to assess potential publication bias and the presence of outliers. The screening criteria for core causal genes were set as P<0.05 in the IVW method and uniform OR direction trends across all five analytical approaches.

The multiple testing correction protocol was performed as follows: First, the IVW P-values of all lactylation-related genes were extracted from the MR analysis results to construct a raw P-value dataset. Second, the ‘p.adjust’ function in R software (version 4.4.2; r-project.org/) was used to implement false discovery rate (FDR) and Bonferroni corrections, with the correction base set as the number of effectively analyzed candidate genes; the FDR correction threshold was defined as 0.05 to balance false-positive and false-negative risks, whereas the Bonferroni correction threshold was set to 0.05 divided by the number of effectively analyzed genes as a strict verification standard. Finally, genes with corrected P-values below the corresponding thresholds were identified as significantly associated genes, and the screening criteria for core causal genes were updated to: IVW method P<0.05, consistent OR directions across all five MR analytical methods, and statistical significance after at least one multiple testing correction.

The transcriptomic data, including five Gene Expression Omnibus (GEO) datasets, among which GSE6044 included nine lung cancer tissues and 5 normal lung tissues, GSE11969 included 9 lung cancer tissues and 5 normal lung tissues, GSE40275 included 15 lung cancer tissues and 41 normal lung tissues, and GSE43346 included 23 lung cancer tissues and 42 normal lung tissues. (ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE6044; ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE11969; ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE40275; ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE43346). Expression data from three independent studies: George et al (24), Cai et al (25) and Liu et al (26) were also included.

RNA sequencing data from the GEO datasets underwent differential expression profiling using the R software package ‘limma’ (version 3.58.1; http://www.bioconductor.org/packages/release/bioc/html/limma.html). Log₂ fold change (log₂FC) and adjusted P-value (P.adj) for Stathmin1 (STMN1) were calculated in tumor vs. normal tissues derived from healthy individuals. The differential expression profiles were visualized using box plots. Kaplan-Meier survival analysis, paired with the log-rank test, was used to explore the relationship between STMN1 expression and OS, with all computations conducted in the ‘survival’ R package (version 3.5–8; http://CRAN.R-project.org/package=survival) and survival curves visualized for publication using ‘survminer’ (version 0.4.9; http://CRAN.R-project.org/package=survminer). The findings from multivariate Cox proportional hazards regression analysis were illustrated using a forest plot, which was generated with the ‘forestplot’ R package (version 2.0.5; CRAN.R-project.org/package=forestplot).

The ‘MCPcounter’ (Microenvironment Cell Populations counter) method was used to measure the infiltration abundance of 10 immune cell subsets (version 1.0.0; http://github.com/ebecht/MCPcounter). All analyses were conducted using the George et al (24) dataset. Samples were dichotomized into the low and high STMN1 expression group by taking the median STMN1 expression level as the cut-off value. The abundances of major immune cell infiltration were further characterized using the Cell-type Identification By Estimating Relative Subsets of RNA Transcripts (‘CIBERSORT’ R package, version 1.0; cibersort.stanford.edu/) (27), MCPcounter (28), Estimation of Proportions of Immune and Cancer cells (EPIC) (‘EPIC’ R package, version 1.1.0; http://github.com/GfellerLab/EPIC) (29), Quantitative Transcriptomics for Immune cell Quantification (quanTIseq) (‘quanTIseq’ R package, version 2.1.0; http://icbi-lab.github.io/quanTIseq/) (30) and xCell algorithms (‘xCell’ R package, version 1.1; http://xcell.ucsf.edu/) (31). Moreover, the expression patterns of immune checkpoint molecules were compared between the high and low STMN1 expression groups. The ESTIMATE algorithm (‘ESTIMATE’ R package, version 1.0.13; http://CRAN.R-project.org/package=ESTIMATE) and Spearman's rank correlation analysis (‘stats’ R package, version 4.4.0; http://CRAN.R-project.org/package=stats) were applied to calculate the correlation coefficient (R) and significance (P-value) between STMN1 expression and stromal score, immune score, estimate score and tumor purity.

The association between STMN1 expression and lung cancer-related genomic alterations, including copy number variations (CNVs) and somatic single-nucleotide variants/indels, was evaluated using Spearman's rank correlation analysis and OR analysis. CNVs were obtained from cBioPortal (cbioportal.org/) and classified into amplification and homozygous deletion types. Somatic mutation data, including SNPs, and insertions and deletions, were obtained from the UCSC genome browser (https://genome.ucsc.edu/). The R software package ‘maftools’ was used for the data integration and analysis (version 2.14.0; bioconductor.org/packages/release/bioc/html/maftools.html), and the R software packages ‘read.maf’ and ‘oncoplot’ (integrated in maftools version 2.14.0) were used for file import facilitation and generating waterfall plot visualizations. The enrichment of mutated genes in patients with lung cancer and high or low STMN1 expression in oncogenic signaling pathways was also analyzed. All analyses were conducted using the George et al (24) dataset. Samples were dichotomized into the low STMN1 expression group and high STMN1 expression group by taking the median STMN1 expression level (9.477354) as the cut-off value. The Drug-Gene Interaction database (DGIdb) (32) was used to generate a hypothesis on mutant gene druggability and match the gene lists to drug-gene interactions.

The STMN1-associated proteins with differential expression were identified, and their expression patterns were visualized using volcano plots and heatmaps. A protein-protein interaction (PPI) network was built using Cytoscape (version 3.10.4; cytoscape.org/Cytoscape) and the hub proteins within this network were identified using the cytoHubba plugin. All analyses were conducted using the Liu et al (26) dataset. Additionally, the top 20 proteins co-expressed with STMN1 were selected for in-depth investigation. Enrichment analysis was conducted on these proteins to clarify the potential mechanisms underlying the role of STMN1 in lung cancer. Prior to this analysis, Pearson correlation analysis was used to filter STMN1-associated genes based on the criteria of a correlation coefficient >0.3 and P<0.05. Subsequently, over-representation analysis (ORA) was performed by leveraging datasets from Kyoto Encyclopedia of Genes and Genomes (KEGG; genome.jp/kegg/), Reactome (reactome.org/), WikiPathways (https://www.wikipathways.org/) and Gene Ontology (GO; geneontology.org/Gene Ontology), with support from the R packages including ‘clusterProfiler’ (version 4.10.0; http://bioconductor.org/packages/clusterProfiler/), ‘org.Hs.eg.db’ (version 3.18.0; http://bioconductor.org/packages/org.Hs.eg.db/) and ‘ReactomePA’ (version 1.46.0; http://bioconductor.org/packages/ReactomePA/).

The RP1 cell line is a well-characterized immortalized lung cancer cell line established by immortalizing the primary lung cells isolated from Rb^L/L/Trp53^L/L (Rb1 and Trp53 conditional double knockout) mice. Specifically, the cell line was derived from spontaneous lung cancer tumors induced by intranasal inhalation of adenovirus-mediated Cre recombinase in these knockout mice, which triggers specific deletion of Rb1 and Trp53 genes in lung tissues. Primary tumor cells were subsequently isolated from the induced lung cancer lesions and immortalized to generate the RP1 cell line, which has been verified to maintain stable proliferation and malignant phenotypes over continuous passages in vitro. This cell line was generously donated by the research group of Professor Hongbin Ji (Center for Excellence in Molecular Cell Science, Chinese Academy of Sciences, Shanghai, China) (33). The human lung cancer cell line DMS114, human lung sarcomatoid carcinoma cell line H196 and 293T cell line were purchased from The Cell Bank of Type Culture Collection of The Chinese Academy of Sciences. DMS114 and H196 were cultured in RPMI-1640 medium (Gibco; Thermo Fisher Scientific, Inc.) supplemented with 10% fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc.), 100 IU/ml penicillin and 100 µg/ml streptomycin. 293T cells were cultured in Dulbecco's Modified Eagle Medium (Gibco; Thermo Fisher Scientific, Inc.) supplemented with 10% FBS, 100 IU/ml penicillin, and 100 µg/ml streptomycin. All of the cells were maintained in a humidified incubator at 37°C with 5% CO₂ and were passaged every 2–3 days once they reached 80–90% confluence.

The cell lines were authenticated by short tandem repeat (STR) profiling, and were confirmed to match reference STR profiles in the American Type Culture Collection (atcc.org/), DSMZ (dsmz.de/), Japanese Collection of Research Bioresources (cellbank.nibio.go.jp/) and ExPASY/Cellosaurus databases (cellosaurus.org/), verifying their identity and ruling out cross-contamination.

Briefly, 293T cells were seeded in 10-cm culture dishes at a density of 5×10⁶ cells/dish 24 h prior to transfection, ensuring cells reached 70–80% confluence. All lentiviral vectors (pLKO.1, pCDH/hygro, psPAX2, pMD2.G) were obtained from Addgene, Inc. Lentiviral packaging was performed using a third-generation packaging system.

For knockdown, short hairpin RNA (shRNA) sequences were inserted into the pLKO.1 lentiviral vector, and an empty vector was used as a negative control. Lentiviral packaging and infection were performed identically for both knockdown and overexpression constructs. For transfection, a total of 6 µg plasmid DNA was used at a ratio of lentiviral vector: psPAX2: pMD2.G=2:1:1. Transfection was performed using polyethylenimine (cat. no. 40820ES, Yeasen Biotech Co., Ltd.) at 37°C in a 5% CO2 incubator for 6 h. Lentiviral supernatants were collected at 48 h post-transfection and filtered through a 0.45 µm filter. RP1, DMS114 and H196 cells were infected with lentivirus at a multiplicity of infection of 10 for 24 h. The culture medium was replaced with fresh complete medium. Cells were cultured at 37°C in a 5% CO2 incubator for an additional 48 h before antibiotic selection. Stable knockdown cell lines were established by selection with 1 µg/ml puromycin for 2 weeks, and maintained in medium containing 0.5 µg/ml puromycin thereafter. The shRNA sequences are listed in Table I.

For overexpression, the human STMN1 gene was amplified using PCR and then ligated with the digested pCDH/hygro lentiviral vector to construct a recombinant plasmid (STMN1-OE). For the negative control, an empty pCDH/hygro vector without the STMN1 insert was used to infect cells in parallel. The successful construction of the recombinant plasmid was then verified through transformation, single colony screening and Sanger sequencing. Subsequently, the recombinant plasmid was co-transfected with packaging plasmids into 293T cells for viral packaging, followed by the collection of lentivirus-containing supernatant. Ultimately, DMS114 and H196 cells were infected with the viral supernatant. Stable STMN1 overexpression cell lines were established by selection with 200 µg/ml hygromycin B for 10–14 days, and maintained in complete medium supplemented with 100 µg/ml hygromycin B.

Total proteins were isolated from cultured lung cancer cells (DMS114, H196, RP1) and mouse subcutaneous tumor tissue using radioimmunoprecipitation assay (RIPA) lysis buffer (supplemented with protease and phosphatase inhibitors). Histones were extracted employing the 0.2 mol/l H₂SO₄ method: Cell pellets were resuspended in histone extraction buffer [10 mmol/l HEPES (pH 7.9), 1.5 mmol/l MgCl₂, 10 mmol/l KCl, 0.5 mmol/l DTT] and incubated on ice for 30 min. 5 mol/l H₂SO₄ to achieve a final concentration of 0.2 mol/l, and the mixture was incubated at 4°C overnight. Following centrifugation at 13,700 × g at 4°C for 10 min, 100% trichloroacetic acid was added to the supernatant to achieve a final concentration of 20%, followed by precipitation at 4°C for 2 h. The precipitate was collected and washed three times with cold acetone, lyophilized under vacuum and dissolved in histone lysis buffer [20 mmol/l Tris-HCl (pH 7.5), 7 mol/l urea, 2 mol/l thiourea, 4% CHAPS]. A total of 20 µg of total protein or 1 µg of histone protein were loaded per lane and separated by 12% SDS-PAGE gels. The proteins were then transferred to PVDF membranes, which were blocked with 5% skimmed milk for 60 min at room temperature to minimize non-specific antibody adsorption. For immunodetection, the blocked membranes were incubated with primary antibodies at 4°C for 16 h, followed by incubation with HRP-linked secondary antibodies for 1 h at room temperature. The protein bands were visualized using the Pierce ECL Western blotting Substrate (Thermo Fisher Scientific, Inc.), and the band optical density was analyzed using ImageJ software (version 1.53t; National Institutes of Health). The primary antibodies used in the present study were as follows: Anti-STMN1 (1:1,000; cat. no. TB2777;), anti-Bcl-2 (1:1,000; cat. no. T40056), anti-Vimentin (1:1,000; cat. no. T55134), anti-GAPDH (1:10,000; cat. no. P60037) (all from Abmart Pharmaceutical Technology Co., Ltd.), global histone lysine lactylation (Pan Kla) (1:1,000; cat. no. 1401), histone H3 lysine 18 lactylation (H3K18la) (1:1,000; cat. no. 1427RM) (both from PTM BIO LLC) and Histone H3 (1:1,000; cat. no. 17168-1-AP; Proteintech Group, Inc.). HRP-linked secondary antibodies used were horseradish peroxidase-conjugated goat anti-rabbit IgG (1:5,000; cat. no. SA00001-2; Proteintech Group, Inc.) and horseradish peroxidase-conjugated goat anti-mouse IgG (1:5,000; cat. no. SA00001-1; Proteintech Group, Inc.).

For IP, 500 µg total protein (200 µl of cell lysate) extracted from human lung cancer DMS114 and H196 was diluted to 500 µl with IP lysis buffer [20 mmol/l Tris-HCl (pH 7.5), 150 mmol/l NaCl, 1% Nonidet P-40 (NP-40), 1 mmol/l EDTA, supplemented with protease and phosphatase inhibitors]. The protein solution was pre-cleared with 20 µl protein A/G agarose beads (Santa Cruz Biotechnology, Inc.) at 4°C for 1 h to reduce non-specific binding. After centrifugation at 118,000 × g at 4°C for 10 min, the supernatant was incubated with 2 µg anti-STMN1 antibody (cat. no. TB2777; Abmart Pharmaceutical Technology Co., Ltd.) or 2 µg normal IgG (cat. no. 30000-0-AP; Proteintech Group, Inc.) as a negative control at 4°C overnight with gentle rotation. Subsequently, 30 µl protein A/G agarose beads were added and incubated for another 4 h at 4°C. The beads were washed four times with IP wash buffer [20 mmol/l Tris-HCl (pH 7.5), 300 mmol/l NaCl, 1% NP-40, 1 mmol/l EDTA]. After a final centrifugation at 78,400 × g at 4°C for 3 min, immune complexes were eluted by boiling with 2X SDS loading buffer for 10 min. The eluted proteins were then subjected to SDS-PAGE and western blotting as aforementioned.

Cell viability was evaluated using the Cell Counting Kit-8 (CCK-8; cat. no. B34302; Selleck Chemicals). Briefly, DMS114 and H196 cells were seeded onto 96-well culture plates at a density of ~3,000 cells/well. After adhering to the plate, the cells were subjected to the required treatment: Continuous culture for an additional 48 h in a standard 37°C, 5% CO₂ incubator under normal growth conditions. The CCK-8 reagent was then added according to the manufacturer's instructions at 37°C with 5% CO2 for 2 h. To assess cell viability, the absorbance value was measured at 450 nm wavelength using a microplate reader.

For the colony formation assay, DMS114 and H196 cells were seeded onto 6-well plates at a density of ~500 cells/well and maintained in culture for 14 days to allow colony formation. After 14 days, the cells were then washed with phosphate-buffered saline (PBS) to eliminate leftover medium. They were subsequently fixed in 4% paraformaldehyde (PFA) at room temperature for 15 min and stained with 0.1% crystal violet solution at room temperature for 20 min to make colonies visible. The cells were then re-rinsed with PBS to eliminate excess stain and finally air-dried at room temperature for subsequent colony counting and analysis. Colonies were defined as cell clusters containing >50 cells, and counting was performed using ImageJ software (version 1.53t; National Institutes of Health) in a blinded manner.

DMS114 and H196 cells with stable STMN1-knockdown (STMN1-sh) or overexpression (STMN1-OE) were seeded onto 6-well plates. Subsequently, 2 ml culture medium (supplemented with 10% FBS) was added to each well, and the cells were cultured for 24 h to allow adherence and initial growth. Upon reaching 95–100% cell confluence, vertical scratches were made on the confluent cell monolayer using a sterile 200-µl pipette tip. The scratched cells and cell debris were gently rinsed off with PBS to ensure a clear scratch area. The plates were then refreshed with 2 ml serum-free media to minimize the interference of cell proliferation, and the cells were incubated for another 24 h. Cell migration was monitored by capturing images of the scratch areas at 0, 12 and 24 h under a light microscope (×4 magnification). For each experimental group, three random fields were imaged to ensure data reproducibility. The migration rate was calculated using the following formula: Migration rate (%)=[(Scratch width at 0 h-Scratch width at each time point)/Scratch width at 0 h] ×100.

DMS114 and H196 cells were collected by centrifugation at 300 × g at 4°C for 5 min, after which, the supernatant was discarded and the pellet was washed with 1 ml PBS. The cells were then centrifuged again at 300 × g at 4°C for 5 min and the supernatant was discarded. The cell pellet was resuspended in 300 µl Stain Buffer (cat. no. 420201; Biolegend, Inc.), and PI-PE (cat. no. HY-D0815; MedChemExpress) and Annexin V-FITC (cat. no. 640949; Biolegend, Inc.) were added, at 4°C in the dark for 15 min. Cell apoptosis was detected using a FACSCalibur flow cytometer (BD Biosciences). Data were analyzed using FlowJo software (version 10.8.1, BD Biosciences). Early apoptosis and late apoptosis were combined, and the total apoptosis rate was calculated using the following formula: Total apoptosis rate (%)=early apoptotic cells (%) + late apoptotic cells (%).

Cell migration and invasion were evaluated using the Transwell assays (0.4-µm pore-size inserts in 12-well plates). For the cell migration assay, ~3×10⁴ DMS114 or H196 cells were suspended in 200 µl serum-free medium. This cell suspension was seeded into the upper chambers of Transwell inserts, whereas 600 µl medium supplemented with 10% FBS was added to the lower chambers as a chemoattractant. After incubating the cells at 37°C for 48 h, the non-migrated cells were removed from the upper surface of the inserts using cotton swabs. The migrated cells adhering to the lower surface were fixed in 4% PFA for 20 min and then stained with 0.1% crystal violet for 20 min at room temperature, followed by rinsing with PBS. Images of the cells were captured under a light microscope at ×10 magnification across three random fields, followed by cell counting.

For the cell invasion assay, the insert chambers were pre-coated with 50 µl Matrigel (diluted to 1:8 in ice-cold serum-free culture medium) at 37°C for 60 min to construct a matrix barrier layer. The subsequent protocol was identical to that described for the migration assay.

6-8-week-old male C57BL/6 mice (weight, 18~22 g) were purchased from the Laboratory Animal Center of Huazhong University of Science and Technology (Wuhan, China). A total of 15 mice were included All mice were housed at a temperature of 22±2°C, with a relative humidity of 50–60%, a 12 h light/12 h dark cycle, and free access to standard chow and water. In order to establish subcutaneous tumors, the stably transduced RP1 cells (transduced with NC, STMN1-sh1 and STMN1-sh2) were respectively inoculated into the right axilla (~1×10⁶ cells in 100 µl PBS/mouse, n=5/group). Tumor dimensions were recorded every 4 days for a total of 24 days. Tumor volume was calculated using the following formula: Tumor volume=1/2 × length × width². Humane endpoints were set as follows: i) tumor volume exceeding 2,000 mm³; ii) >20% body weight loss; iii) severe signs of distress or impaired mobility. No mice met these endpoints On day 24, the mice were euthanized by cervical dislocation. Death was confirmed by two indicators: i) Complete cessation of thoracic movement and breathing for ≥1 min; and ii) absence of hindlimb withdrawal reflex when gently pinched with forceps. After confirming death, the tumor tissues were dissected, weighed and processed: Tissues were fixed in 4% PFA) at room temperature for 24 h, followed by paraffin embedding and sectioning at a thickness of 4 µm. Some tissue were lysed with RIPA lysis buffer (supplemented with protease and phosphatase inhibitors) for protein extraction and western blotting.

The paraffin-embedded sections of subcutaneous tumor tissues were prepared. After gradient deparaffinization with xylene and rehydration with ethanol at room temperature, the sections were stained with hematoxylin for 10 sec at room temperature. Subsequently, the tissue sections were differentiated in 1% hydrochloric acid-ethanol at room temperature for 20 sec, blued in 0.5% ammonia water for 2 min and counterstained with eosin for 3 min at room temperature. The tissue sections were then dehydrated with gradient ethanol, cleared with xylene and mounted using neutral balsam at room temperature. Images were captured using a light microscope (Olympus BX53, Olympus Corporation).

The paraffin-embedded subcutaneous tumor tissue sections were deparaffinized with xylene, rehydrated with a graded series of ethanol and subjected to antigen retrieval using citrate buffer (pH 6.0) at 95°C for 15 min. The sections were then permeabilized with 0.2% Triton X-100 at room temperature for 10 min and blocked with 5% bovine serum albumin (BSA, cat. no. A1933, Sigma-Aldrich; Merck KGaA) at room temperature for 1 h. Subsequently, the sections were incubated with a primary antibody against Vimentin (cat. no. T55134; Abmart Pharmaceutical Technology Co., Ltd.) at a dilution of 1:200 at 4°C overnight, followed by incubation with Cy3-conjugated goat anti-rabbit IgG (cat. no. GB21303; Wuhan Servicebio Technology Co., Ltd.) at a dilution of 1:400 for 1 h at room temperature. The cell nuclei were counterstained with DAPI at room temperature for 5 min. Images of the cells were captured under a fluorescence microscope, and the mean fluorescence intensity was analyzed and semi-quantified using ImageJ software (version 1.53t; National Institutes of Health).

The paraffin-embedded subcutaneous tumor tissue sections were deparaffinized with xylene, rehydrated with a graded series of ethanol. Antigen retrieval was performed using citrate buffer (pH 6.0) under high-pressure at 121°C for 15 min to restore antigen epitopes. To eliminate endogenous peroxidase interference, the prepared sections were incubated with 3% hydrogen peroxide (H₂O₂) at room temperature for 10 min, and to block the non-specific antibody binding, they were treated with 5% BSA at room temperature for 60 min. The sections were then incubated with primary antibodies, including anti-Ki67 (cat. no. TW0001; Abmart Pharmaceutical Technology Co., Ltd.), anti-Caspase-3 (cat. no. 19677-1-AP; Proteintech Group, Inc.) and anti-Cleaved-Caspase-3 (C-Caspase-3; cat. no. AF7022; Affinity Biosciences) overnight at 4°C. The next day, the sections were incubated with HRP-conjugated secondary antibodies (cat. no. SA00001-2; Proteintech Group, Inc.) at room temperature for 1 h. DAB chromogen (cat. no. DA1010; Beijing Solarbio Science & Technology Co., Ltd.) was used for visualization of protein expression with subsequent counterstaining with hematoxylin at room temperature for 10 sec. Under an optical microscope, three random high-power fields (×40 magnification) were assessed for each section. Positive cells were semi-quantified using Image-Pro Plus 6.0 software (Media Cybernetics, Inc.). The percentage of positive cells was calculated using the following formula: Positive cells (%)=[(Number of positive cells in each field/Total number of cells in each field)] ×100, to assess protein expression levels. Integrated optical density (IOD) of positive staining was quantified using Image-Pro Plus 6.0 software (Media Cybernetics, Inc.). IOD represents the total optical density of positively stained areas, reflecting the overall expression level of the target protein. For C-Caspase-3, the IOD value was normalized to the IOD of total Caspase-3 to indicate the level of apoptotic activation.

Paraffin-embedded tumor tissue sections were deparaffinized in xylene and rehydrated through a graded ethanol series. Following antigen retrieval using proteinase K (20 µg/ml) for 20 min at 37°C, the sections were incubated with the TUNEL reaction mixture at 37°C for 60 min in the dark, according to the manufacturer's instructions (cat. no. C1088; Beyotime Biotechnology). Subsequently, the sections were mounted with an anti-fade mounting medium containing DAPI to visualize cell nuclei. Images were captured using a fluorescence microscope equipped with appropriate filter sets for FITC and DAPI detection (Olympus BX53, Olympus Corporation, Tokyo, Japan). To ensure representative quantification, at least 3 randomly selected non-overlapping fields of view per section were observed. The apoptotic index was calculated as the ratio of TUNEL-positive cells (green fluorescence) to the total number of cells (blue fluorescence).

Cellular lactate concentrations were measured using Lactate Colorimetric Assay Kit (cat. no. K607-100, BioVision, Inc., according to the manufacturer's instructions. DMS114 and H196 cells with stable STMN1 knockdown or overexpression were seeded in 6-well plates at a density of 1×106 cells/well and cultured for 48 h. After washing twice with ice-cold PBS, cells were lysed in lactate assay buffer and centrifuged at 12,000 × g for 10 min at 4°C to remove cell debris. The supernatant was collected and incubated with the lactate detection reagent mix at 37°C for 30 min in the dark. The absorbance was measured at 570 nm using a microplate reader, and lactate concentrations were calculated based on a standard curve generated with known lactate concentrations.

Pearson correlation coefficient was used to assess linear associations between STMN1 expression and continuous variables with approximately normal distributions, such as the expression levels of co-expressed genes. Spearman's rank correlation coefficient was applied to evaluate non-linear or non-normally distributed relationships, including those between STMN1 expression and immune cell infiltration scores, stromal/immune/ESTIMATE scores, tumor purity and genomic alteration frequencies. ORA was utilized for functional enrichment assessments. For the in vitro and in vivo experimental data, the results are presented as the mean ± standard deviation. Comparisons between two groups were performed using unpaired Student's t-test, whereas those among multiple groups were performed using one-way analysis of variance (ANOVA). When ANOVA results were significant, Bonferroni's multiple comparisons test was used for post-hoc analysis, comparing each experimental group with the control group. All in vitro experiments were performed with three biological replicates. P<0.05 was considered to indicate a statistically significant difference. For MR analyses, additional multiple testing corrections were performed using the FDR and Bonferroni methods, with FDR-corrected P<0.05 regarded as statistically significant.

From the eQTL data of 46 lactylation-related genes, the IVs meeting strict quality control criteria for MR analysis were identified: 63 SNPs for the platelet-type phosphofructokinase (PFKP) gene (F-statistic=315.6), 8 SNPs for the SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 4 (SMARCA4) gene (F-statistic=92.7) and 7 SNPs for the STMN1 gene (F-statistic=109.9). All of the IVs satisfied the strong instrument threshold (F>10). Following allele harmonization, no strand-ambiguous SNPs or SNPs with conflicting effect directions were detected. The detailed information on the finally included IVs is provided in Table II.

To investigate the causal relationships between these three core lactylation-related genes and lung cancer, MR analyses were conducted using multiple statistical approaches, including IVW, MR Egger, weighted median, simple mode and weighted mode. Among these methods, IVW served as the main model for causal inference ('I), and the detailed quantitative results of all five MR analytical methods for each gene are summarized in Table III. IVW analysis revealed that PFKP (OR=0.909, 95% CI: 0.845–0.979) and SMARCA4 (OR=0.633, 95% CI: 0.409–0.979) were associated with a lower risk of lung cancer, whereas STMN1 (OR=1.741, 95% CI: 1.182–2.564) was associated with an elevated risk.

Scatter diagrams were drawn to illustrate the detailed MR analyses results. The calculation of individual causal impacts demonstrated that increased SNP-mediated effects on STMN1 expression were associated with elevated lung cancer risk (Fig. 1A). Conversely, enhanced SNP effects on PFKP and SMARCA4 expression were linked to more pronounced protective effects against lung cancer (Fig. 1B and C). The effects of SNPs corresponding to each of the three core exposure factors (STMN1, PFKP and SMARCA4) on lung cancer risk were assessed by generating forest plots. Examining the effects of SNPs linked to each of the three key genes revealed consistent patterns across all analytical frameworks (Fig. 1D-F).

A total of three genes with valid MR results underwent multiple testing correction. STMN1 was the only gene significantly associated with lung cancer risk in both correction methods. PFKP and SMARCA4 were only significant after FDR correction, but not after Bonferroni correction (Table IV).

To enhance the reliability of the MR analyses findings, the IVs underwent rigorous screening standards, and all the samples were confined to a European ancestry cohort to reduce the risk of false-negative results and population stratification bias. The heterogeneity across SNPs linked to the three exposure factors (STMN1, PFKP and SMARCA4) was evaluated using Cochran's Q test and the MR Egger approach. The results showed no notable heterogeneity among the SNPs corresponding to the three genes, thus verifying the stability of the selected IVs (Table V).

Moreover, the MR Egger intercept test was conducted to detect the occurrence of horizontal pleiotropy. The respective intercept values were 0.012, 0.007 and 0.016 for PFKP, SMARCA4 and STMN1 genes, respectively, with no statistical significance. This result indicated that MR estimates were not biased by directional pleiotropy. The MR-PRESSO test did not identify any outliers (Table V).

A leave-one-out sensitivity assessment was additionally performed to evaluate the impact of individual SNPs on the overall IVW estimate. Excluding a single SNP had a minimal effect on the overall causal estimate (Fig. 1G-I). In addition, funnel plots were symmetric (Fig. 1J-L), indicating no significant publication bias or outlier interference. This observation confirmed that no individual SNP exerted a disproportionate influence on the outcomes, thus enhancing the credibility and stability of the MR results.

Among the three core lactylation-related genes identified by MR analysis, STMN1 was the only gene that exhibited a significant causal association with increased lung cancer risk after both FDR and Bonferroni corrections, and its upregulation in lung cancer tissues was correlated with poor patient prognosis.

Expression analyses of the GEO datasets (GSE6044, GSE43346, GSE11969 and GSE40275) revealed that STMN1 was significantly upregulated in lung cancer tissues compared with in normal tissues (Fig. 2A). Furthermore, the patients exhibiting high STMN1 expression had a shorter OS time (Fig. 2B), and high STMN1 expression was associated with a higher risk of poor prognosis (Fig. 2C).

Functional enrichment analyses were performed to investigate the biological roles of STMN1. KEGG pathway enrichment analysis (Fig. 2D) demonstrated significant enrichment in pathways, including ‘spliceosome’ and ‘ATP-dependent chromatin remodeling’. Reactome enrichment analysis (Fig. 2E) highlighted its enrichment in processes, including ‘processing of capped intron-containing pre-mRNA’, ‘mRNA splicing major pathway’ and ‘mRNA splicing’. WikiPathways enrichment analysis (Fig. 2F) revealed its role in the ‘mRNA processing’ pathway.

Immune cell infiltration profiling revealed that the infiltration scores of B lineage cells, CD4+ T cells, fibroblasts, monocytic lineage cells and neutrophils were significantly lower in the high STMN1 expression group as compared with those in the low STMN1 expression group (Fig. 3A). The correlation analysis of STMN1 expression and immune checkpoint molecules further revealed an inverse association between STMN1 and most immune checkpoint molecules (Fig. 3B), suggesting a potential role of STMN1 in modulating checkpoint-mediated immune suppression. In order to validate the association between STMN1 and key antitumor immune cells, the infiltration abundances of CD4+ and CD8+ T cells and M1 macrophages were quantified using five independent algorithms (CIBERSORT, MCPcounter, EPIC, quanTIseq and xCell). All of the algorithms demonstrated a weak inverse correlation between STMN1 expression and the infiltration of these three cell types (Fig. 3C-E), highlighting the association between STMN1 and reduced antitumor immune cell infiltration. ESTIMATE algorithm-derived immune scores combined with Spearman's rank correlation analysis confirmed that high STMN1 expression was weakly negatively correlated with the overall immune score (r=−0.276; Fig. 3F). These results indicated that STMN1 high expression may be associated with the characteristics of an immunosuppressive TME in lung cancer, suggesting a potential role of STMN1 in regulating TME immune status.

Waterfall plots visualized the genetic mutation patterns of lung cancer samples across STMN1 expression subgroups. In the high STMN1 expression group (Fig. 4A), all 35 lung cancer samples (100%) harbored genetic mutations. Moreover, 34 of 36 samples (94.44%) in the low STMN1 expression group (Fig. 4B) exhibited genetic alterations, with mutation types comparable to those in the high STMN1 expression subgroup. OR analysis was performed to quantify the association between STMN1 expression and gene mutation status. Numerous genes, including TRIM58, GCN1L1, KIAA1239, LRP1, SLITRK3, APOB, GPRC6A and TP53, showed an OR of >1, indicating a higher mutation probability in the high STMN1 expression group. On the other hand, genes such as MYPN, COBL and LRFN5, had an OR of <1, suggesting a lower mutation probability with high STMN1 expression (Fig. 4C).

As compared with in the low STMN1 expression group (Fig. 4E), the high STMN1 expression group (Fig. 4D) was relatively highly enriched in the TP53 signaling pathway (94.12% vs. 83.78%), Hippo signaling pathway (79.41% vs. 72.97%), cell cycle (79.41% vs. 70.27%) and WNT signaling pathway (64.71% vs. 56.76%). By contrast, the enrichment levels of the RTK-RAS (76.47% vs. 78.38%) and NOTCH (76.47% vs. 75.68%) signaling pathways demonstrated no statistically significant difference between the two groups. The druggable categories in the lung cancer subgroups (high and low STMN1 expression groups) were analyzed using the DGIdb. In the high STMN1 subgroup, druggable categories included ‘serine threonine kinase’ (such as TTN), ‘transporter’ (such as RYR2) and TP53-related histone modification pathways, with ‘druggable genome’ involving COL11A1 and RB1 (Fig. 4F). The druggable categories in the low STMN1 expression subgroup included ‘serine threonine kinase’ (such as TTN), TP53-related RNA-directed DNA polymerases, ‘fibrinogen’ (TNR) and ‘drug resistance’ (TP53), with ‘druggable genome’ involving ABCA13 and COL11A1 (Fig. 4G). Despite the overlapping core elements, gene composition in druggable categories was different, suggesting that STMN1 expression may shape druggable target landscapes in lung cancer and guide subgroup-specific therapies.

To identify the proteins associated with varying STMN1 expression levels, the protein expression patterns were visualized using volcano plots and heatmaps. These analyses revealed 452 differentially expressed proteins, including 72 upregulated and 380 downregulated proteins (Fig. 5A and B). Next, a PPI network was drawn, and Cytoscape software was used to select hub proteins from the top 20 upregulated and top 20 downregulated proteins (Fig. 5C-F).

Subsequent GO enrichment analysis showed that STMN1-associated proteins were involved in the ‘regulation of G₀ to G₁ cell cycle transition’ and ‘negative regulation of messenger RNA catabolic process’ (Fig. 5G). These results indicated that STMN1 could regulate cell cycle progression and maintain mRNA stability in lung cancer.

The effects of STMN1 expression were further assessed using cells transduced with NC, STMN1-sh1, STMN1-sh2, vector and STMN1-OE. Western blotting was performed to verify the efficiency of STMN1 knockdown and overexpression in DMS114 and H196 cells (Fig. 6A-D). In both cell lines, STMN1-sh1 and STMN1-sh2 significantly reduced STMN1 protein expression compared with the negative control group (Fig. 6A and C). Conversely, OE-STMN1 markedly increased STMN1 protein levels relative to the vector control (Fig. 6B and D). These results confirmed the successful construction of stable STMN1-knockdown and -overexpression cell lines for subsequent functional assays. Subsequently, CCK-8 assays demonstrated that STMN1 knockdown significantly suppressed cell proliferation in both DMS114 and H196 cells, whereas STMN1-OE transduction significantly accelerated cell proliferation (Fig. 6E-H). Moreover, wound healing assays showed that STMN1 knockdown markedly impaired cell migration in both DMS114 and H196 cells, whereas its overexpression significantly accelerated cell migration (Fig. 6I-L).

To characterize the functional role of STMN1 in lung cancer, its effects on apoptosis, colony formation and invasion/migration were assessed in DMS114 and H196 cells. Flow cytometric analysis showed that STMN1 knockdown significantly enhanced apoptosis in both cell lines. Total apoptosis rate was significantly higher in the STMN1-sh groups compared with the control group. By contrast, STMN1 overexpression modestly decreased apoptosis, indicating that the pro-apoptotic effect of STMN1 seems to be mainly achieved through its knockdown rather than overexpression (Fig. 7A-D). Colony formation assays revealed that STMN1 knockdown markedly diminished colony-forming ability, which was evidenced by fewer and smaller colonies compared with that in the NC group (Fig. 7E and F). By contrast, STMN1-OE transduction significantly enhanced this capacity. Additionally, Transwell assays demonstrated that STMN1 knockdown impaired invasive and migratory capacities (with fewer membrane-penetrating cells), whereas its overexpression significantly elevated these abilities (Fig. 7G-J).

A murine shRNA model targeting STMN1 was established to assess the in vivo effects of STMN1. As compared with in the NC group, both STMN1-sh1 and -sh2 groups exhibited reduced tumor volumes (Fig. 8A), decreased tumor weights (Fig. 8B) and slower tumor growth kinetics (Fig. 8C). The maximum tumor volume was 1,654 mm³, corresponding to a maximum diameter of 15 mm. H&E staining and immunohistochemical analysis revealed that STMN1 knockdown was associated with reduced expression of the proliferation marker Ki67, and elevated levels of the apoptosis marker C-Caspase-3/Caspase-3 in tumor tissues (Fig. 8D-F). Immunofluorescence staining demonstrated that Vimentin expression was reduced in the STMN1-sh1 and -sh2 groups, indicating impaired epithelial-mesenchymal transition (EMT) progression (Fig. 8G and H). The TUNEL assay further confirmed a significant increase in apoptotic cells in STMN1-knockdown tumors (Fig. 8I and J). Western blotting validated these findings, showing downregulated Vimentin and Bcl-2 (anti-apoptotic protein) expression levels in the STMN1-sh1 and -sh2 groups (Fig. 8K). These results were consistent with the enhanced apoptosis and suppressed EMT. STMN1 knockdown efficiency was also confirmed in tumor tissues (Fig. 8K). Collectively, these results indicated that STMN1 knockdown could induce tumor regression by suppressing cell proliferation, enhancing apoptosis and impairing EMT in vivo.

To elucidate the effect of STMN1 on histone lactylation, histones were extracted from DMS114 and H196 cells following STMN1 knockdown or overexpression. The levels of Pan Kla and H3K18la were decreased upon STMN1 knockdown and increased upon STMN1 overexpression, demonstrating that STMN1 promotes histone lactylation (Fig. 9A-D).

To explore the underlying mechanism, cellular lactate concentrations were measured and it was revealed that STMN1 knockdown significantly reduced lactate levels, whereas STMN1 overexpression increased lactate production (Fig. 9E-H), indicating that STMN1 may regulate histone lactylation by modulating cell lactate generation.

To determine whether STMN1 itself undergoes lactylation, co-IP and western blotting was performed. The results confirmed that STMN1 was lactylated in both DMS114 and H196 cells, and its lactylation level was positively associated with its expression: STMN1 knockdown decreased its lactylation, whereas STMN1 overexpression increased it (Fig. 9I-L). Collectively, these findings indicated that STMN1 may not only promote cellular lactate production and histone lactylation, but also undergoes lactylation itself.