The mouse Lewis lung carcinoma cell line: LLC were purchased from National Collection of Authenticated Cell Cultures (China). The Human Umbilical Vein Endothelial Cells (HUVEC) and mouse endothelial cells (SVEC4-10) were purchased from ATCC. LLC and SVEC4–10 were cultured in DMEM high glucose (Procell) supplemented with 10% fetal bovine serum and 1% antibiotic mixture at 37 °C and 5% CO₂. HUVEC was cultured in endothelial cell medium with 10% fetal bovine serum, 1% antibiotic and 1% endothelial cell growth supplement (ECGS) mixture.

The primary antibodies used in IHC analysis included CD8 anti-mouse (ab217334, Abcam), CD31 anti-mouse (ab182981, Abcam), NG2 anti-rabbit (ab275024, Abcam). The secondary antibodies used in immunofluorescence assay were mouse and rabbit mixed secondary antibody (GK500705, Shanghai Gene Tech Company Limited). Mixtures of probes are made; six probes are purified and then mixed. The sequences of Meox1 probe are shown below.

Human and mouse Meox1 was knocked down using siRNA technology. HUVEC and SVEC4–10 were transfected using siRNA oligonucleotides against Meox1 or nontargeting controls at 50 pmol/5 × 10⁵ cells. Cells were processed 48 hours after transfection. All transfections were performed with Lipofectamine RNAiMAX (ThermoFisher scientific) following the manufacturer’s recommendations. The sequences of siRNA are shown below.

Immunohistochemistry and immunofluorescence (IHC and IF) were performed with the following primary antibodies: anti-CD8 (1:250); anti-CD31 (1:500) and anti-NG2 (1:500) 5μm thick Paraffin-embedded sections were sequentially dewaxed in xylene (times×5 min) and rehydrated through an ethanol gradient (100%, 90%, 80%, and 70%, 3 min each). Sections were rinsed in PBS (3 times×3 min) to remove residual ethanol and equilibrate for subsequent procedures. Heat-induced epitope retrieval was performed using Tris-EDTA buffer (pH 6.0; G1203, Wuhan Xavier Biotechnology Co., Ltd.) diluted 1:20 with distilled water. Sections were microwaved at medium power (8 min), allowed to cool for 8 min, and reheated at medium-low power (7 min). After natural cooling to room temperature, slides were rinsed in PBS on an orbital shaker (3 times×3 min). To quench endogenous peroxidase activity, sections were incubated in 3% hydrogen peroxide (H₂O₂) for 15 min at room temperature under light-protected conditions, followed by PBS washes (3 times×3 min). The BSA was added 30 min and sections were incubated overnight at 4°C with the respective primary antibodies. Next sections were washed in PBS and incubated for 50 min at room temperature with the corresponding goat-anti-rabbit and goat-anti-mouse secondary antibodies (diluted 1:500 (v/v), Cell Signaling Technology, Beverly, MA, USA). Sections were washed 3 times for 5 min on a rocking table after which the DAB solution was added on the sections and staining was followed under the microscope. Positive staining was identified depending on the antibody used, as brow yellow coloring of cataplasm or nuclei. Sections were stained with hematoxylin (2 min), washed in running water (10 min), differentiated in 1% acid alcohol (2 sec), and rinsed in distilled water (3 times×5 min). Nuclear bluing was achieved by immersion in pre-warmed (40°C) deionized water for 5 min. Tissues were dehydrated through an ethanol gradient (70%, 80%, 90%, and 100%, 3 min each) and cleared in xylene (2 times × 5 min) (14). Sections were permanently mounted using a 1:1 mixture of xylene and neutral resin. Slides were air-dried in a fume hood before microscopic analysis. Images were taken by EVIDENT VS200 microscope equipped with a digital camera (Olympus, Japan). Quantification of immunohistochemical staining was performed in two different ways. In case of CD8 cells were counted in 5 randomly chosen sections from each tumor. The number of immunoassayed cells was expressed as percentage of the CD8 positive cells counted. For the quantification of CD31 in the tumor we opted for quantifying staining intensity. The CD31- and CD8-positive cells were quantified by the percentage of positive cells per field of vision (FOV) (15). For this we used the program image 8.0 as described previously (16, 17).

After paraffin sections dewaxed to water, tissue sections were circumscribed with a hydrophobic barrier pen. Sections were then treated with Proteinase K (20 μg/mL) at 40°C for antigen retrieval, followed by sequential rinsing with deionized water and three 5 min PBS washes. Prehybridization buffer was subsequently applied and incubated at 37°C for 1 h in a humidified chamber. The hybridization procedure was initiated by replacing prehybridization buffer with probe-containing hybridization solution (60 μL/section), followed by overnight incubation at 40°C in a thermostatically controlled hybridization oven. Post-hybridization washes were performed under stringent conditions: 40°C 2 × SSC (10 min), two 5-minute washes with 1 × SSC at 40°C, and a final 10 min 0.5 × SSC wash at ambient temperature. For specimens exhibiting nonspecific hybridization, formamide (25% v/v) was incorporated into wash buffers. Following desiccation by gentle agitation, sections received 60 μL preheated branching probe hybridization solution and were incubated horizontally in a humidity chamber (40°C, 45 min) containing 50 mL 2 × SSC moisture reservoir. Subsequent post-hybridization processing included sequential 5-minute rinses with temperature-equilibrated (40°C) 2 × SSC, 1 × SSC, 0.5 × SSC, and 0.1 × SSC buffers. Signal amplification was achieved through application of fluorophore-conjugated detection probes (1:200 dilution in hybridization buffer) with 3 h incubation at 40°C. Stringency washes were subsequently performed using the following regimen: 40°C 2 × SSC (10 min); Two times 5 min 1 × SSC washes at 40°C; Three times 5 min 0.5 × SSC washes at 40°C. Following nuclear counterstaining with DAPI (5 μg/mL, 8 min at 40°C), sections were mounted using anti-fade fluorescence mounting medium and stored at 4°C protected from light.

Expression matrix was processed from raw data following the recommended protocol of the dnbc4 software. For normalization, clustering, and cell type annotation, the Seurat R package was utilized with support from ACT (http://xteam.xbio.top/ACT/). The Pyscenic Python package was then employed to quantify transcription factor (TF) regulon activities using default parameters. The inferred activity matrix was imported into the Seurat object and stored as an assay for subsequent analysis. Data visualization was primarily performed using the scplotter and scCustomize R packages.

The animal studies were approved by the Institutional Animal Use and Care Protocol of Guangzhou Laboratory. (GZLAB-AUCP-2024-07-A11). Six-week-old male C57BL/6J (18–20 g) mice were purchased from the GemPharmatech Co., Ltd. For the first animal study evaluating the effect of siRNA. Mice were injected subcutaneously with mouse colon cancer LLC cells (5 × 10⁵) mixture with siMeox1 (50 pmol). The mice were sacrificed after the two weeks; In the Meox1 KO mouse model, LLC cells were inoculated at a dose of (5 × 10⁵), and tumor size was recorded starting on day 4 until necropsy on day 22. Tumor tissues were removed from the body for IHC staining. For the second animal study evaluating the effect Meox1 inhibition on BMS-1 blockade therapy. Mice were injected subcutaneously with mouse colon cancer LLC cells (5 × 10⁵) mixture with siMeox1. Tumor-bearing mice were randomly assigned into four groups: control group, siMeox1-induced group, BMS-1 group, and siMeox1 + BMS-1 group. BMS-1 treatment was given 10 µg/mL (100 μL) once every 3 days for a total of three times by intraperitoneal injection. Vehicle is PBS. The tumor volume (V) was measured using a slide caliper and calculated using the following formula: V (mm³) = 0.5 × ab², where a and b represent the long diameter and perpendicular short diameter (mm) of the tumor, respectively.

Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software). Data were presented as mean ± standard error of the mean (SEM). For quantification of immune cell density, five fields of tumor sections with IHC staining were randomly selected and the positive cells were counted. Student’s t-test was used to compare the differences between the two groups. One-way ANOVA was used to compare the differences between the four groups. Statistical significance for all tests was defined as p < 0.05, with non-significant results indicated as ‘ns’; specific data are detailed in the text or figure legends.

To systematically characterize transcription factor dynamics in both malignant and immune cell populations, we performed scRNA-seq on tumor specimens and subsequently conducted TFs regulon analysis through the SCENIC pipeline (PYSCENIC) to identify cell-type-specific regulatory networks ( Figure 1A ). UMAP dimensionality reduction visualization of integrated single-cell transcriptomes (processed with SCTransform normalization and PCA feature selection) demonstrated the following cellular subpopulations within tumor tissue. To resolve hierarchical cellular organization within the TME, we implemented graph-based clustering on integrated single-cell data, revealing 8 transcriptionally distinct TME subsets, including Cancerous_G2M; Cancerous_S; Cancerous_G1; Fibroblast (FB); Endothelial cell (Endo); Macrophage (MP); DC cell (DC); T cell (T) ( Figures 1B-D ).

We performed tissue distribution analysis (18) and observed endothelial cell specific high expression of TFs (Nfib, Sox17, Meox1) in the analyzed ( Figure 1E , Figure 2A ). A comprehensive review of the extant literature reveals that TFs, Nfib (19, 20) and Sox17 (21, 22) have been extensively investigated for their functional involvement in mediating angiogenesis. Current evidence demonstrates their critical regulatory roles in modulating pathological vascularization processes during oncogenesis. Concomitantly, we conducted systematic profiling of the transcriptional regulatory network governed by Nfib, Sox17 and Meox1 transcription factors, integrating multi-omics datasets to delineate their downstream target genes with potential angiogenic regulatory functions. ( Figures 2B–D ). Strikingly, Gene Ontology (GO) term enrichment analysis of the Meox1 regulon-associated genes revealed its pivotal role in biological processes (BP) such as endothelial cell development, vasculogenic and so on. Meanwhile, analysis of molecular functions (MF) identified extracellular matrix binding as the most significantly altered term ( Figure 2E ). Emerging evidence suggests that the mechanistic contribution of Meox1 to tumor vascularization processes remains poorly characterized, thus underscoring the critical need for systematic investigation into its functional interplay with oncogenic angiogenesis pathways in subsequent experimental analyses.

Together, our findings demonstrate that LLC tumor tissues exhibit cell type-specific transcriptional regulatory all cellular compartments, with particular biological significance emerging from the endothelial compartment where the angiogenesis-modulating capacity of Meox1 positions it as a novel therapeutic target warranting mechanistic dissection of vascular niche reprogramming.

While Meox1 is recognized as a pivotal regulator in tumor-associated endothelial cells, its molecular underpinnings remain undefined. To delineate its functional repertoire, we employed RNA interference-mediated Meox1 silencing through siRNA transfection. Phenotypic validation revealed that two specificity-controlled siRNA duplexes (si_mMeox1#2/si_mMeox1#3) robust antiproliferative activity across LLC cellular models ( Figures 3A–C ). si_mMeox1#1 shows no inhibitory effect on tumors (data not shown). To verify the knockdown effect of siRNA in tumor tissues, molecular validation of siRNA penetration efficacy was conducted via RNAscope™ multiplex in situ hybridization. Quantitative analysis revealed pronounced Meox1 transcript depletion (p<0.0001) in siRNA-perfused tumor regions ( Figures 3D, E ). We also used a Meox1-KO mouse model and subcutaneously inoculated LLC. Compared with the ctrl group, the subcutaneous tumors in the KO group were significantly reduced (P = 0.0383) ( Figures 3F, G ). It is also noteworthy that a notable reduction in CD31 density was similarly unobserved in the NC group, consistent with the morphological presentation of twisted and irregular CD31+ luminal structures ( Figures 3H, I ). CD31⁺NG2⁺ double-positive cells represent tumor vascular normalization (23). The results indicate that Meox1 knockdown promotes tumor vascular normalization, facilitating the entry of immune cells into tumors to kill tumor cells ( Supplementary Figure 1 ). scTenifoldKnk computational knockout of Meox1 identified enrichment of GO: BP “Nuclear Chromosome Segregation” and Reactome “Cell Cycle Checkpoints”. Literature suggests dysregulation of these pathways induces cytosolic DNA accumulation, activating the cGAS-STING pathway and recruiting CD8⁺ T cells into the tumor microenvironment to elicit antitumor immunity ( Figure 3J ). The IHC results further showed that CD8⁺ cells were significantly increased in the group of si_mMeox1 compared to the control mice ( Figures 3K, L ).Previous studies have suggested that tumor-associated vascular neovascularization leads to increased tissue interstitial pressure, which in turn reduces the entry of antitumor agents while rejecting immune cells (24).

In brief, these data demonstrate that RNA interference-mediated Meox1 ablation significantly attenuates neoplastic cell proliferation through dual mechanisms involving angiogenic suppression and immunological remodeling through enhanced CD8 cytotoxic T lymphocyte tumor infiltration.

To investigate the functional involvement of Meox1 in angiogenic processes, we conducted in vitro endothelial tube formation assays employing primary human umbilical vein endothelial cells (HUVECs) (25) and SV40-immortalized endothelial cells (SVEC4-10) (26) as dual-species model systems ( Figure 4A ). Our functional screening revealed distinct angio-regulatory effects among Meox1-targeting constructs: murine-specific siRNA (si_mMeox1#1) showed negligible impact on endothelial tubulogenic, whereas both si_mMeox1#2 and si_mMeox1#3 demonstrated potent anti-angiogenic efficacy. Parallel human-specific siRNA (si_hMeox1#1) significantly attenuated HUVEC network formation in standardized Matrigel assays ( Figures 4B, E ). Furthermore, quantitative morphometric analysis demonstrated that validated siRNA constructs (si_hMeox1, si_mMeox1#2/3) significantly attenuated vascular arborization complexity, evidenced by reduction in total branches length and decrease in number master segments compared to non-targeting controls ( Figures 4C, D, G ). Taken together, these data suggest Meox1 critically regulates vascular endothelial cell sprouting and tubulogenic.

BMS-1 is an inhibitor of PD-1/PD-L1 protein/protein interactions (27, 28). A murine LLC cancer model was employed to investigate the effect of Meox1 consumption on PD-1/PD-L1 blockade therapy. The knockdown of Meox1 by specific siRNA was found to potentiate the sensitivity of LLC tumors to BMS-1, leading to a dramatically reduction in tumor burden ( Figures 5A–G ). Interestingly, mice treated with Meox1 siRNA alone exhibited significantly retarded tumor progression compared with the control group. The IHC results further showed that CD8⁺ T lymphocytes was significantly increased in the group of combined Meox1 knockdown and BMS-1 treatment ( Figures 5H, I ). Collectively, our findings indicated that Meox1 reduction enhanced sensitivity of LLC tumors to PD-1/PD-L1 checkpoint blockade.