Pan-cancer analysis was performed by UALCAN (https://ualcan.path.uab.edu/) (38). Gene expression, correlation analysis and survival analysis was from GEPIA (http://gepia.cancer-pku.cn/) (39). Survival analysis was from Kaplan-meier-plotter (https://kmplot.com/analysis/) (40). All survival analysis data were obtained from the Kaplan-meier-plotter website. The survival difference between groups was assessed using the log-rank test. Hazard ratios (HR), 95% confidence intervals (CI), and P-values were calculated using the Cox proportional hazards regression model. m⁶A site prediction was performed by SRAMP (http://www.cuilab.cn/sramp) (41).

Authenticated A549, murine lewis lung cancer cells (LLC), H1437, Beas-2b, H446 and H460 cell lines (Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences) were cultured in DMEM medium (Gibco, Gran Island, NY, USA). All these cells were cultured with 10% fetal calf serum (Gibco), 4 mM glutamine (Gibco), 100 U/ml penicillin and streptomycin (Invitrogen Life Technologies, Carlsbad, CA, USA) at 37°C in a 5% CO₂ incubator. Cells were transfected with plasmid DNA and siRNAs using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions.

Cells were lysed with RIPA buffer (Beyotime, Shanghai, China), and protein extracts were separated on SDS-PAGE gels and transferred onto PVDF membranes (Bio-Rad, Richmond, CA) as previously described (42). Nuclear and cytoplasmic proteins were separated using a nuclear protein extraction kit (Solarbio, Beijing, China). Rabbit antibodies against KCTD10 (27279-1-AP, Proteintech, Wuhan, China) (1:1000), β-catenin (ET1601-5, HUABIO, Hangzhou, China) (1:1000), PD-L1 (ab228415, Abcam, Waltham, USA) (1:1000), E-cadherin (A3044, ABclonal, Wuhan, China) (1:1000), N-cadherin (A0433, ABclonal) (1:1000), METTL14 (ab309096, Abcam) (1:1000), YTHDF2 (ab220163, Abcam) (1:1000), Lamin B1 (A1910, ABclonal) (1:1000), Tubulin (AF7010, Affinity Biosciences, Changzhou, China) (1:5000) and Flag-Tag (F2555, Sigma) (1:1000) were used Supplementary Table 1 . Mouse monoclonal anti-ubiquitin (sc-8017, Santa Cruz Biotech, Texas, USA) were used. HRP-conjugated goat anti-rabbit and goat anti-mouse secondary antibodies were from ABclonal. Western blots were independently replicated for at least three times.

Lung and brain tissues, along with corresponding tumor tissues were examined. Polyformalin-fixed paraffin-embedded (FFPE) tissues were processed through an alcohol gradient. Antigen retrieval was performed using citric acid solution (Service Biotechnology, Wuhan, China). Rabbit antibodies against KCTD10 (HPA014273, Sigma) (1:200, GAR), β-catenin (1:200), PD-L1 (1:200), c-Myc (380784, Zenbio, Chengdu, China) (1:200), VEGFR2 (A5609, ABclonal) (1:200), E-cadherin (1:200) and N-cadherin (1:200) were used. Mouse monoclonal antibodies against CD31 (ab9498, Abcam) (1:200), CD8α (70306, Cell Signaling, Massachusetts, USA) (1:200), Vimentin (240140, Zenbio) (1:200) were used. The primary antibodies were incubated overnight after blocking, and the HRP-conjugated goat anti-rabbit and goat anti-mouse secondary antibodies (Service Biotechnology) and DAB detection kit (Sevice Biotechnology) were incubated sequentially with the tissues, and the nuclei were counterstained with hematoxylin dye solution (Sevice Biotechnology). Sections were visualized under an Olympus BX53 microscope (Japan) after neutral resin sealing. The pathology information of human lung cancer tissues was shown in Supplementary Tables 2 and 3 . HE staining was performed following standard protocols. These experiments were approved by the Human Ethics Committee of Hunan Normal University (2021–017).

Full-length and truncated fragments of KCTD10 were cloned into pCMV-HA and pCMV-Myc vectors (Invitrogen), respectively. Full-length and deletion constructs of β-catenin were generated as described (43). The pcDNA3.1-(HA-Ub) was generated by inserting human UBC into pCDNA3.1-HA vector (Addgene, Massachusetts, USA). The HA-K0 and KCTD10 5′UTR was synthesized by Sangon Biotech. The K27R, K33R and K48R mutations were amplified from the pcDNA3.1-(HA-Ub) plasmid and the K27O, K33O and K48O mutations were amplified from the pcDNA3.1-(HA-K0) plasmid using splicing overlapping extension polymerase chain reaction (SOE-PCR) site-directed mutagenesis. DNA fragments containing K11R, K63R, K11O and K63O were digested with SalI and HpaII, and ligated to the pcDNA3.1-(HA-Ub) and pcDNA3.1-(HA-K0) digested with the same enzymes, respectively. The KCTD10 CDS, 3′UTR and 5′UTR were inserted into to the pGL3 reporter plasmid (Promega, MA, USA). The primers used are listed in Supplementary Table 4 . All constructs were verified using the Sanger method (Sangon Biotech).

KCTD10 lentivirus expression vectors and packaging plasmids (pHelper 1.0 and pHelper 2.0) (Genechem, Shanghai, China) were cotransfected into 293T cells, and viral supernatants were harvested, filtered, concentrated and titrated (44). The lentiviral expression vector GV365 included GFP protein and puromycin resistance gene for the observation of infection efficiency and screening of stable cell lines. Cells were placed in 6-well plates and infected at MOI=10 and observed for fluorescence four days post-infection. stable cell lines were selected with complete medium with 2 μg/ml puromycin (Solarbio), and maintained in medium with reduced puromycin concentration.

For colony formation, 1,000 cells were placed per well in 6-well plates. Cells were cultured for two weeks. Cells were fixed with methanol, stained with 0.5% crystal violet, and counted. For MTT assays, 10, 000 cells were placed per well in 96-well plates, incubated with MTT reagent for 4 h and dissolved in DMSO. Absorbance at 492 nm was measured using a spectrophotometer (UV-2102C, Unico, Changsha, China).

For wound healing assays, cells were cultured in 24-well plates until reaching 90% confluence. A 10-μl pipette tip was used to generate wounds. After wound generation, cells were changed to medium containing 2% serum. Three wound areas in each well were marked on the bottom of the plates and imaged at 0, 24 and 48 h after wound formation, which was photographed with the microscope at each time point. For cell migration, the chamber (Corning, New York, USA) was inserted into a 24-well plate, 2×10⁴ infected cells were distributed in the upper chamber with 10% or 15% FBS in the lower chamber. Cells were fixed, stained with 0.5% crystal violet (Sangon Biotech, Shanghai, China) and imaged under the microscope. Invasion assays were performed using the chamber covered with Matrigel glue (Corning), which were carried out as previously described (45).

4-week-old nude mice were purchased from the Hunan SJA Laboratory Animal Corporation (Changsha, China). 6-week-old C57BL/6J mice were purchased from Jackson Laboratories (BarBarbor, ME). 5-week-old CDH5 ^CreERT2/+ and KCTD10 ^flox/flox mice were obtained from GemPharmatech (Nanjing, China) (28) ( Supplementary Table 5 ). Mice were maintained in a 12 h light/dark cycle and regularly fed with chow and water in an SPF room. All procedures were approved by Hunan Normal University (2021–017).

1x10⁷ cells were injected subcutaneously into 4-week-old nude mice, with tumor growth measured twice weekly until tumor volume reached 1,000 mm³. For lung colonization assays, 5×10⁵ LLC cells were injected via the tail vein of 6-week-old C57BL/6J mice or 4-week-old nude mice, and lungs were removed after 4 weeks. For brain metastases, 5×10⁵ LLC cells were injected intracranially into 6-week-old C57BL/6J mice and brains were harvested 4 weeks after injection as previously described (46). Anti-PD-1 therapy (100 µg/mouse, RPM1-14, BioXCell, New Hampshire, USA) was injected intraperitoneally on days 11/14/17/20 in LLC models. Tumor weight, volume, survival and tissue analysis were recorded and analyzed. All mouse experiments were repeated at least twice.

Cells in 10 cm dishes were grown to 80% confluence and transfected with 5 μg KCTD10 and β-catenin (full length or truncated). After 30 h, cells were lysed and whole cell extracts were immunoprecipitated using rabbit polyclonal antibodies against Myc-tag (C3956, Sigma) or HA-tag (05-904, Sigma) and protein A/G plus beads (K1305, APEXBIO, Texas, USA). Immunoprecipitates were resolved by 10% SDS-polyacrylamide gels and detected by HA-tag or Myc-tag. Rabbit preimmune IgG (ab37355, Abcam) served as a negative control. For mass spectrometry, proteins from KCTD10-overexpressing A549 cells were separated in 10% SDS-PAGE gels after immunoprecipitation using either IgG or Flag-tag (F7425, Sigma), and stained using a Protein Fast Silver Stain Kit (Leagene Biotechnology, China) and analyzed by mass spectrometry (Novogene, Beijing, China) (47, 48).

To assess protein stability, cells and KCTD10-overexpressing cells were treated with 50 μg/mL CHX (Selleck Chemicals, Texas, USA) for 0, 2, 8 h, respectively. The proteins were extracted and detected through Western blotting. To identify protein degradation pathway, transfected cells were treated with 20 µM MG132 for 10 h before harvesting. Cells were lysed and analyzed by Western blotting with antibodies against Flag-Tag, β-catenin and Tubulin. For ubiquitin experiments, cells were transfected either with the expression plasmids pCMV-Myc-β-catenin alone or with pCMV-HA-KCTD10 and pcDNA3.1-(HA-Ub) or its mutants. 24 h after transfection. Myc-β-catenin was immunoprecipitated with rabbit polyclonal anti-β-catenin antibodies and these immunoprecipitates were subjected to Western blotting with ubiquitin to detect β-catenin-ubiquitin conjugation.

Tissue sections were dehydrated and processed for antigen retrieval. For cell immunofluorescence double labeling, the cells were fixed with methanol. The tissue sections and cells were blocked and incubated overnight with the first primary antibodies, followed by incubation with the specific secondary antibodies. After antigen repair, the second primary antibodies were added for overnight incubation, and another species-specific secondary antibodies were added for incubation. Primary antibodies are KCTD10 (1:100), β-catenin (1:100), CD31 (1:200), α-SMA (250104, Zenbio) (1:100). Alexa Fluor 488 phalloidin goat anti-rabbit (A-11008, Thermo Fisher Scientific, Massachusetts, USA) (1:2000) and Alexa 594 goat anti-mouse antibodies (A11005, Thermo Fisher Scientific) (1:2000) were used as secondary antibodies. The nucleus was stained with Hoechst 33258 (Beyotime). The fluorescence signals were analyzed with a fluorescence microscope (Zeiss Axioskop-2).

Kctd10^flox/flox and CDH5^CreERT2/+ mice were intercrossed and screened to generate homozygous Kctd10^flox/floxCDH5^CreERT2/+ mice. To obtain the CDH5^CreERT2/+Kctd10^-/- mice, the mice were administered intraperitoneally with tamoxifen (75 mg/kg body weight) for one week. For the LLC mouse models, 1×10⁶ tumor cells were injected via the tail vein into tamoxifen-injected 6-week-old female Kctd10^flox/flox mice or CDH5^CreERT2/+ Kctd10^-/-mice.

Total RNA was extracted from the cell lines using TRIzol reagent (Thermo Fisher Scientific) and reverse transcribed into cDNA using MMLV RTase and random primers (Sangon Biotech). SYBR green (Invitrogen)-based real-time PCR was performed using an ABI 7900 thermocycler (Thermo Fisher Scientific). Reactions were incubated in a 96-well plate at 95°C for 5 min, followed by 35 cycles of 95°C for 20 sec and 60°C for 30 sec. The PCR primers are listed in Supplementary Table 4 . The relative expression levels of genes were calculated by the 2^−ΔΔCt method compared to β-actin.

m⁶A modification sites were predicted on the SRAMP website. For MeRIP assays, RNA was isolated and incubated with 2 μg m⁶A (A19841, ABclonal) or IgG antibodies. Precipitated RNA was reverse transcribed, amplified and then subjected to agarose gel electrophoresis as described (49). m⁶A-related siRNAs were purchased from genepharma (Shanghai, China) ( Supplementary Table 6 ).

Cells were co-transfected with recombinant pGL3 vectors bearing CDS, 3′UTR and 5′UTR of the KCTD10 gene or specific siRNAs. After the transfection, β-galactosidase and luciferase activities were measured using the Luciferase Assay System (Promega, Madison, WI) in a TD-20/20 luminometer (Turner Design, Sunnyvale, CA) as previously reported (36).

Data are expressed as mean ± SD of at least three independent experiments. Statistical analysis was performed using GraphPad Prism 7 (San Diego, California, USA) and SPSS 22.0 (SPSS Inc., Chicago, Illinois, USA). The significance of the differences between groups was determined using Student’s t-test and multi-group comparisons was determined using one-way ANOVA with post hoc tests. Values of P<0.05 were considered statistically significant. *p < 0.05, **p < 0.01, ***p < 0.001.

To assess the expression of KCTD10 in lung cancer, we analyzed the TCGA Pan-Cancer and GEPIA database and found lower KCTD10 expression in lung cancer tissues than in normal tissues ( Figure 1A , Supplementary Figure 1A ). Western blots further confirmed reduced KCTD10 expression in lung cancer cell lines compared with bronchial epithelial Bears-2b cells ( Figure 1B ). IHC analysis revealed that KCTD10 expression was markedly lower in high-stage lung cancer patients ( Figures 1C, D ), with a particularly pronounced reduction in LUAD ( Figures 1E, F ). Kaplan-Meier Plotter survival analysis revealed that patients with high expression of KCTD10 exhibited significantly prolonger overall survival (OS) and post-progression survival (PPS) ( Figures 1G, H ), suggesting a potential tumor-suppressive role of KCTD10. Interestingly, the hazard ratio (HR) for KCTD10 was lower in LUAD, implying a notable role for KCTD10 in this lung cancer type ( Figure 1I ). As demonstrated in the GEPIA database analysis, elevated KCTD10 expression correlates with prolonged disease-free survival (DFS) in patients, suggesting its potential role in inhibiting tumor metastasis ( Figure 1J ). Thus, KCTD10 expression is inversely correlated with lung cancer stage and is associated with a favorable prognosis for lung cancer patients.

To further investigate the functional role of KCTD10, we constructed a stable A549 lung cancer cell line overexpressing KCTD10 by lentiviral transduction and demonstrated successful overexpression of KCTD10 by fluorescence imaging and Western blots ( Figures 2A, B , Supplementary Figure 1B ). MTT assays revealed that overexpression of KCTD10 reduced cell viability while enhancing cisplatin-induced cytotoxicity ( Supplementary Figure 1C ). And colony formation assays further supported the suppressive effect of KCTD10 on the growth of A549 cells ( Figure 2C ). Subsequently, subcutaneous tumorigenesis assays showed that KCTD10-overexpression A549 cells formed significantly smaller tumors compared to controls ( Figures 2D, E , Supplementary Figure 1D ). HE staining revealed decreased tumor cell density in Kctd10-overexpressing tumors ( Figure 2F , Supplementary Figure 1E ). These results indicate that KCTD10 inhibits lung cancer growth both in vitro and in vivo.

Lung cancer metastasis remains a major clinical challenge (50). To determine whether KCTD10 influences metastatic ability of lung cancer, we performed wound healing, cell migration and invasion assays and found that KCTD10 overexpression significantly inhibited the migration and invasion ability of A549 cells ( Figures 2G, H , Supplementary Figure 1F ). In an in vivo metastasis model, tail vein injection of KCTD10-overexpressing A549 cells into nude mice resulted in a significant reduction in lung nodule formation ( Figure 2I ). HE staining further confirmed a lower lung tumor burden in the KCTD10-overexpressing group compared to the NC group ( Figure 2J ). Since tumor metastasis is closely linked to an EMT process (19), we examined EMT markers in A549 cells and subcutaneous tumors. Western blots and IHC analyses showed that overexpression of KCTD10 increased the epithelial marker E-cadherin while decreased the stromal markers N-cadherin and β-catenin ( Figures 2K, L , Supplementary Figures 2A, B ). Moreover, KCTD10-overexpressing cells exhibited a morphology resembling epithelial cells ( Supplementary Figure 2C ). Additionally, CD31 expression, an angiogenesis marker, was downregulated in KCTD10-overexpressing subcutaneous tumors, indicating reduced tumor angiogenesis ( Figure 2L ). Similar results were observed in LLC and H1437 cells ( Supplementary Figures 3A–F ). These findings suggest that KCTD10 inhibits lung cancer metastasis by suppressing EMT and tumor angiogenesis.

To determine the molecular mechanism of KCTD10 in lung cancer, we performed IP followed by silver staining and MS. The identified differential bands revealed potential interacting KCTD10-interacting proteins ( Figure 3A ) and their predicted subcellular localization ( Supplementary Figure 4A ). KEGG enrichment analysis indicated strong associations between KCTD10-interacting proteins and metabolism, cancers and immune system ( Supplementary Figure 4B ). GO enrichment confirmed that KCTD10 highly correlated with protein binding, as previous reported ( Supplementary Figure 5A ) (51). KOGs analysis highlighted its involvement in post-translational modifications, protein turnover, chaperones and signal transduction ( Supplementary Figure 5B ). Based on protein-peptide scores, we identified β-catenin as a top interactor of KCTD10 ( Figure 3B ). β-catenin is well-established oncogene implicated in EMT and tumor progression (22). GEPIA database analysis showed that β-catenin is highly expressed in LUAD ( Figure 3C ). IHC analysis further showed a positive correlation between β-catenin expression and lung cancer stage ( Figures 3D, E ), particularly in LUAD ( Figure 3F ). The IHC scores revealed a negative correlation between KCTD10 and β-catenin expression ( Figure 3G ). Moreover, Kaplan-Meier survival analysis indicated that high β-catenin expression is associated with poor prognosis ( Figures 3H–K ). Importantly, overexpression of KCTD10 reduced β-catenin and its downstream effector, PD-L1, in lung cancer cells ( Figure 3L , Supplementary Figure 6 ).

To verify the interaction between KCTD10 and β-catenin, co-IP assays were performed. Endogenous β-catenin was detected in immune complexes of HA-KCTD10, whereas control IgG failed to precipitate any band ( Figure 4A ). To map the binding regions, we constructed truncated plasmids of KCTD10 and β-catenin, we found that the BTB region of KCTD10 binds to the Armadillo repeat region (1–9) of β-catenin ( Figures 4B–D , Supplementary Figures 7A–C ). We next examined whether KCTD10 promoted the degradation of β-catenin using cycloheximide (CHX), which revealed accelerated β-catenin degradation in KCTD10-overexpressing cells ( Figure 4E ), and this was rescued by the proteasome inhibitor MG132 ( Figure 4F ), suggesting KCTD10 mediates β-catenin degradation via the ubiquitin-proteasomal pathway. Immunofluorescence and nucleocytoplasmic fractionation experiments confirmed that KCTD10 primarily degrades β-catenin in the cytoplasm ( Figures 4G, H ). Ubiquitination assays demonstrated that co-transfection of KCTD10 and ubiquitin plasmids promoted the ubiquitination of β-catenin ( Figure 4I ). To identify the specific polyubiquitin linkage, we constructed a series of lysine-linked ubiquitin active and mutant site plasmids and found that KCTD10 facilitates polyubiquitination of β-catenin via K48-linked ubiquitin chains ( Figures 4J, K ), supporting its role in the ubiquitin-proteasomal degradation of β-catenin. Similar results were observed in LLC and H1437 cells ( Supplementary Figures 8A-D ). Taken together, our results demonstrated that KCTD10 directly binds with β-catenin and promotes its K48-linked polyubiquitination and proteasomal degradation.

Since KCTD10 downregulates β-catenin and PD-L1, and PD-L1 expression is known to promote tumor immune evasion in lung cancer (52), we investigated the impact of KCTD10 on lung cancer immunotherapy. Kaplan-Meier Plotter survival analysis revealed that high KCTD10 expression significantly correlated with improved prognosis in lung cancer patients with CD8⁺ T cell infiltration ( Figure 5A , Supplementary Figure 9A ), indicating a potential role in immune system activation. To assess the therapeutic potential of KCTD10 in combination with immune checkpoint blockade, KCTD10-overexpressing LLC cells were injected into 6 week-old C57BL/6J mice, followed by anti-PD-1 antibody treatment ( Figure 5B ). Fluorescence analysis and Western blots confirmed successful Kctd10 overexpression ( Figures 5C, D ). Survival analysis demonstrated that both Kctd10 overexpression and anti-PD-1 treatment individually prolonged the survival of C57BL/6 mice compared to the NC group. Notably, the combination of KCTD10 overexpression and anti-PD-1 therapy exhibited an additive effect, with 50% of the combination group surviving beyond 50 days, compared to fewer than 30 days in the NC group ( Figure 5E ). HE staining further showed that Kctd10 overexpression and PD-1 blockade reduced lung tumor burden, independently, with the combination treatment exerting the most pronounced effect ( Figure 5F ). IHC and IF analysis indicated that Kctd10 overexpression suppressed β-catenin and Pd-l1 expression while improved CD8a⁺ T cell infiltration. The combined treatment led to the most strongly improvement in CD8a⁺ T cell infiltration ( Figures 5G, H , Supplementary Figures 9B, C ). Therefore, overexpression of Kctd10, in conjunction with PD-1 blockade, effectively inhibits lung tumor metastasis and augments anti-tumor immunity.

Brain metastases are among the most common distant metastases in lung cancer (53). To investigate the effect of Kctd10 on brain colonization, we intracranially injected LLC cells into 6 week-old C57BL/6J mice and assessed the impact of Kctd10 overexpression combined with anti-PD-1 blockade ( Figure 6A ). Both Kctd10 overexpression and anti-PD-1 treatment independently prolonged the mouse survival compared to the NC group. Remarkedly, the combination therapy showed the greatest survival benefic, with 80% of the mice in this group survived beyond 60 days, whereas those in the NC group die at approximately 23 days ( Figure 6B ). HE staining revealed that both Kctd10 overexpression and anti-PD-1 treatment independently reduced brain tumor burden, with the combined treatment exhibiting the most profound effect ( Figure 6C ). IHC analysis of brain sections revealed that overexpression of Kctd10 downregulated β-catenin and Pd-l1 but increased CD8a⁺ T cell abundance, indicating the similar effects in the lungs. Notably, the combined therapy led to the most substantial increase in CD8a⁺ T cell infiltration, further enhancing the anti-tumor immune response ( Figures 6D, E , Supplementary Figures 9D, E ). These results highlight the therapeutic potential of KCTD10 overexpression in improving the efficacy of immune checkpoint blockade for lung cancer metastases.

Tumor angiogenesis is critical for tumor growth and metastasis, while β-catenin can promote tumor angiogenesis via VEGF signaling (54). Immunofluorescence analysis showed that Kctd10 is co-localized with endothelial markers Cd31 and Cd34 in multiple subcutaneous tumors ( Supplementary Figures 10A–F ), suggesting that Kctd10 is associated with tumor angiogenesis. Given that endothelial cells in the tumor microenvironment are dispensable for tumor angiogenesis (55), we generated Kctd10 ^flox/flox CDH5 ^CreERT2/+ mice that specifically delete Kctd10 in vascular endothelial cells ( Figures 7A, B , Supplementary Figure 10G ). Following injection of 1x10⁶ LLC cells ( Figure 7B ), we observed significantly increased lung tumor burden in Kctd10-knockout mice ( Figure 7C ). HE staining confirmed a large lung tumor area in these mice compared to controls ( Figure 7D ). Immunofluorescence analysis revealed a decrease in normal blood vessels, as indicated by Cd31 labeling, whereas the number of tumor-associated blood vessels was markedly increased in Kctd10-knockout mice ( Figure 7E ). Normal vessels have a higher pericyte density than tumor vessels (56). Additionally, pericyte coverage, assessed by α-sma staining, was reduced in tumor regions, indicating a shift toward an abnormal tumor vasculature phenotype. And the expression of β-catenin was upregulated in Kctd10 knockout mice ( Figure 7E , Supplementary Figures 10H , 11A ). The IHC results showed that the expression of Vegfr2 correlated with Cd31 levels, suggesting that Kctd10 deficiency promotes pathological angiogenesis while impairing normal angiogenesis. Furthermore, the upregulation of EMT-associated proteins indicated an enhanced metastatic phenotype and tumor angiogenesis ( Figure 7F , Supplementary Figure 11B ). These results suggest that endothelial-specific loss of Kctd10 promotes tumor angiogenesis and metastasis while inhibits normal vascular development.

To investigate the upstream regulatory mechanism of KCTD10 expression in lung cancer, we explored the m⁶A modification, the most common form of mRNA modification for the regulation of mRNA stability (57). Previous studies have shown that m⁶A modification plays a tumor-suppressive role in lung cancer (58, 59). Bioinformatic analysis predicted potential m⁶A binding sites within the CDS region of KCTD10 ( Supplementary Figure 12 ), which was validated by MeRIP assays ( Figure 8A , Supplementary Figure 13A ). Knockdown of m⁶A-related genes in A549 and LLC cells demonstrated that METTL14 knockdown significantly reduced KCTD10 expression ( Figure 8B , Supplementary Figure 13B ). The GEPIA database revealed lower METTL14 expression in lung cancer tissues ( Supplementary Figure 13C ), and the Kaplan-Meier plotter analysis indicated that high METTL14 expression was associated with improved prognosis ( Figures 8C, D ). A positive correlation between KCTD10 and METTL14 expression was also observed ( Figure 8E ). Luciferase reporter assays confirmed that knockdown of METTL14 specifically decreased KCTD10 CDS reporter activity, but not other regions ( Figure 8F , Supplementary Figure 13D ). Next, RIP experiments showed that METTL14 can enrich RNAs corresponding to the KCTD10 CDS region ( Figure 8G , Supplementary Figure 13E ). Transcription inhibitor actinomycin D treatment revealed that METTL14 knockdown reduced KCTD10 mRNA stability ( Figure 8H , Supplementary Figure 13F ). Furthermore, Western blots showed that knockdown of METTL14 decreased KCTD10 and E-cadherin levels while increased β-catenin expression ( Figure 8I , Supplementary Figures 13G, H ), suggesting that METTL14 regulates the KCTD10/β-catenin axis.

We then knocked down m⁶A-associated readers and found that YTHDF2 interference reduced KCTD10 expression ( Figure 8J , Supplementary Figure 14A ). Although YTHDF2 expression was not significantly different between lung tumors and non-tumors ( Supplementary Figure 14B ), high YTHDF2 levels were associated with improved survival ( Figures 8K, L ) and positively correlated with KCTD10 expression ( Figure 8M ). Luciferase reporter assays showed knockdown of YTHDF2 reduced the KCTD10 reporter activity ( Figure 8N , Supplementary Figure 14C ). RIP experiments demonstrated that YTHDF2 enriches the KCTD10 CDS regions ( Figure 8O ), and actinomycin D treatment indicated that silencing YTHDF2 decreased KCTD10 mRNA stability ( Figure 8P ). Western blots showed that knockdown of YTHDF2 reduced KCTD10 and E-cadherin levels while increased β-catenin expression ( Figure 8Q , Supplementary Figures 14D, E ). These results suggest that the METTL14-YTHDF2 axis stabilizes KCTD10 mRNA via m⁶A modification, contributing to KCTD10 downregulation in lung cancer.