Integrated multi-omics reveals glycolytic gene signatures of lung adenocarcinoma brain metastasis and the impact of Rac2 lactylation on immunosuppressive microenvironment

Abstract

Purpose

Tumor Metabolic Behavior modulates the immunosuppressive microenvironment through multiple pathways, thereby compromising anti-tumor immune responses. To date, there have been limited studies assessing the role of metabolic plasticity or immunometabolism in the tumor microenvironment (TME) during metastasis. Notably, emerging evidence suggests the presence of an immunosuppressive niche in brain metastases. This research aims to delineate distinct metabolic signatures in brain metastatic, investigate the impact of tumor-associated glycolysis on the development of brain metastases in lung adenocarcinoma, and characterize the lactylation regulation in this immunosuppressive microenvironment.

Methods

The GSE131907 and GSE198291 datasets were retrieved for bioinformatic analysis. Combined with the results of proteomic and transcriptomic sequencing conducted on the lung adenocarcinoma brain metastasis model, differentially expressed signaling pathways were systematically identified through KEGG and GO functional annotations. A multimodal approach encompassing immunohistochemical (IHC) staining, immunofluorescence (IF) imaging, enzyme-linked immunosorbent assay (ELISA) quantification, and co-immunoprecipitation (Co-IP) assays was employed to experimentally validate the characteristics of the immunosuppressive microenvironment and the levels of tumor lactate/lactylation. Rescue experiments were performed by adding a lactylation-specific inhibitor (LDHi) or an H3K18la site-specific inhibitor. Finally, immunohistochemical staining was used to verify the expression level of H3K18la in clinical samples.

Results

A total of 86,215 cells were extracted from the GSE131907 dataset, and the metabolic profiles of different cell types were analyzed. The results showed that glycolysis plays a dominant role in tumor cell metabolism. Further analysis revealed that early-stage primary lesions exhibit an inflammatory response signature, while advanced-stage primary lesions and brain metastatic lesions display an immunosuppressive signature. Elevated glycolytic flux showed a significant positive correlation with both the progression of brain metastasis and the immune evasion capacity of brain metastatic lesions. Pathological evaluation of tumor tissues from the LLC-BM (Lewis Lung Cancer Brain Metastasis) model confirmed its immunosuppressive characteristics. Additionally, obvious hypoxia was observed in the tumor tissues, accompanied by intratumoral vascular malformation and dysfunction. Significant lactate accumulation was present in the tumor microenvironment of LLC-BM tumors, and prominent lactylation modifications were detected in the tumor regions. In this model, Rac2 was identified as a potential core mediator of lactylation modification in macrophages, promoting the M2 polarization of macrophages. Meanwhile, CD40, TNFSF13 and CCL22 were identified as key immunoregulatory factors regulated by lactylation signaling. Notably, H3K18la was significantly highly expressed in lung cancer brain metastatic lesion samples.

Conclusions

The glycolytic pathway plays a critical role in the metabolic reprogramming of tumor cells during lung adenocarcinoma brain metastasis. Tumor glycolysis is closely associated with lung cancer progression, brain metastasis, and immune evasion. The Rac2 could be affected by lactylation, and then facilitate the formation of an immunosuppressive tumor microenvironment by induce the M2 polarization of macrophages.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07207-6.

Introduction

Lung adenocarcinoma, the most common subtype of non-small cell lung cancer (NSCLC), has a high propensity for central nervous system (CNS) metastasis. Approximately 20% of patients present with brain metastases (BM) at the time of diagnosis, and about 25%-50% develop BM during disease progression, among which mutations in EGFR (epidermal growth factor receptor)/ALK (anaplastic lymphoma kinase) are key risk factors [1]. Therapeutic approaches include local treatments (surgery/stereotactic radiosurgery (SRS) for oligometastases, whole-brain radiotherapy (WBRT) for multiple metastases) and systemic therapies (targeted therapies such as osimertinib/lorlatinib for EGFR/ALK mutations, PD-1 inhibitors for patients without driver gene mutations, etc.). The natural survival time is only 3–6 months, and the overall median survival time after treatment is 8–16 months [2]. Although the tumor immune microenvironment has been confirmed as a key factor influencing the therapeutic efficacy of lung cancer brain metastasis—and it is precisely this factor that leads to the limited effectiveness of current treatments [3]—it is not in a static state, but is tightly regulated by metabolic processes within the brain metastatic niche.

Omics sequencing technologies have provided preliminary insights into the immune microenvironment of lung cancer brain metastases [4]. Cellular metabolism directly modulates immune microenvironment and therapeutic efficacy [5, 6]. Recent years have witnessed growing research focus on metabolic dynamics during disease progression. During cytoskeletal remodeling in epithelial-mesenchymal transformation (EMT), glycolytic enzymes may be released to enhance glycolytic flux [6]. Circulating tumor cells (CTCs) have been shown to upregulate the production of nicotinamide adenine dinucleotide phosphate (NADPH) and lactate uptake, which redirects glucose-derived carbons into oxidative pathways and enhances their antioxidant capacity. Bloodborne cytokines and metabolites may influence immune surveillance of CTCs and promote tumor immune evasion [7]. Primary and metastatic lung adenocarcinoma (LUAD) tissues demonstrate patient-specific and intratumoral heterogeneity in pathways regulating tumor progression, chemoresistance, and metabolism. Brain metastasis subpopulations display differential expression of genes associated with type II alveolar cell differentiation, chemoresistance, glycolytic-oxidative phosphorylation switching (metabolic reprogramming), and EMT activation [8]. Furthermore, cerebrospinal fluid analysis reveals distinct abundance profiles of 43 metabolites between primary and metastatic central nervous system (CNS) tumors, indicating alterations in glycine/choline/methionine degradation, biogenic amine biosynthesis, and glycolytic pathways [9]. Mechanistic studies evaluating EMT-associated metabolic plasticity and immune metabolism in metastatic processes remain scarce. To date, the metabolic landscape of lung cancer brain metastasis remains undefined. This study aims to investigate the metabolic characteristics in lung cancer brain metastatic, and explore the regulatory role of the glycolytic pathway in the immune microenvironment of lung cancer brain metastasis.

Methods and materials

Bioinformatics analysis

Dataset integration & joint analysis

The single-cell RNA sequencing (scRNA-seq) dataset (GSE131907) was acquired from the Gene Expression Omnibus (GEO) database [8], consisting of 11 primaries human LUAD tissue samples and 10 brain metastasis specimens. Raw sequencing data were converted into expression matrices and processed through the Seurat pipeline (v2.3.4), involving data filtering, normalization via the LogNormalize method in the NormalizeData function, dataset integration, followed by cluster analysis and gene expression profiling. Highly variable genes were identified across all datasets using the FindVariableFeatures function. Nonlinear dimensionality reduction and visual cluster analysis were performed through Uniform Manifold Approximation and Projection (UMAP), with resultant cellular clusters visualized in UMAP plots.

Cell type annotation

The SingleR package was employed to annotate individual cells within identified clusters. Cellular expression profiles were compared against reference datasets (Human Primary Cell Atlas, Blueprint, ENCODE) through Spearman’s correlation analysis, where label-specific scores were calculated as fixed quantiles of correlation distributions. For each unannotated cell, reference cell types with the highest label-specific scores were assigned as definitive annotations. Log-normalized count matrices generated by Seurat were imported into SingleR for this annotation process. Epithelial cells, including tumor cells annotated across non-tumor, primary tumor, and brain metastasis samples, were subsequently re-integrated and subclustered using Seurat. This refined epithelial population was then subjected to copy number variation (CNV) analysis via the inferCNV package to discriminate malignant from benign epithelial subsets based on chromosomal arm-level aberrations.

Differential expressed genes (DEG) analysis & functional enrichment

Differential gene expression analysis between cellular communities (MoDMs vs. CRMs) was conducted using Varimax factor clustering, prioritizing genes with top-ranked average log2-fold change (avg_logFC) values indicative of subpopulation-specific expression enrichment. The clusterProfiler package performed KEGG pathway enrichment analysis on differentially expressed genes (DEGs), with pathways meeting the significance threshold (adjusted p-value < 0.05) designated as biologically relevant. For comparative analysis of primary and metastatic lung cancer lesions, transcriptomic data were normalized to transcripts per million (TPM) formats. Glycolysis-related genes demonstrating inter-group differential expression (DEG criteria: nominal p-value < 0.05, |log2FC|≥1) were extracted, followed by pathway enrichment evaluation where pathways with false discovery rate (FDR)-adjusted p-values < 0.05 were considered statistically significant.

Cell culture

The mouse-derived lung adenocarcinoma cell line LLC cells and RAW246.1 mouse macrophages were purchased from Pro-cell (Wuhan, China) and cultured in high-glucose DMEM medium supplemented with 10% fetal bovine serum and 1% antibiotic solution. Cultures were maintained in a humidified incubator at 5% CO2. LLC cells were stably transfected with lentiviral vector GV260 (Genomeditech, Shanghai, China) carrying a luciferase reporter gene (-luc). RAW246.1 cultured with L-Lactate (15mM, psaitong, TianJin, China) for 24 h [10]. P300 inhibitor C646 (20 μm, Selleck, Texas, America) for 6 h, 12 h, 24 h, respectively [11]. LDHi (oxamate, 20mM, Selleck).

LLC-BM animal model

The animals used in this study were C57BL/6J male mice, aged 6–8 weeks and weighing 19–21 g, purchased from Ji Cui Medicine Co., Ltd. (Nanjing, China), and housed at the Experimental Animal Center of Nanchang University under a 12-hour light/dark cycle at a temperature of 25 °C with ample food and water. All experimental procedures were approved by the Institutional Animal Ethics Committee of Nanchang University (Approval No. NCULAE-20221031140).

LLC-luc tumor cells were prepared for injection after trypsin digestion and washing with PBS. Cell concentration in PBS for each injection was adjusted to 5 × 10^5 cells using Trypan Blue staining for live cell counting. Mice were anesthetized with 4% isoflurane inhalation. Under sterile conditions, the skin of the mice was disinfected with ethanol, and a skin incision (1 cm) was made to expose the skull for positioning at the bregma. LLC labeled with luciferase were injected using a stereotactic apparatus (coordinates: 1 mm anterior and 3 mm right lateral to the bregma, depth 3 mm), with a total volume of 5 µl of Luciferase-LLC (5 × 10^5 cells/ml) [12], to establish a NSCLC brain metastasis model (LLC-BM), evaluated for tumor growth via live imaging and histological staining.

Cell hypoxia model

Tumor cell culture dishes are placed in a specialized hypoxic incubator, into which a gas mixture (1% oxygen, 5% carbon dioxide, balance nitrogen) [13]. The cells are then cultured at a constant temperature of 37 °C for 24 h.

Transcriptome sequencing

Total RNA was extracted from control and BM mouse tissues (3v3) using TRIzol, followed by quality control (RIN > 7.0). Ribosomal RNA was depleted (for total RNA) or poly(A) enrichment performed (for mRNA). Libraries were prepared via fragmentation, cDNA synthesis, and adapter ligation (Illumina TruSeq kit), then sequenced on NovaSeq 6000 (150 bp paired-end). Raw reads were quality-trimmed (Fastp), aligned to the reference genome (HISAT2/STAR), and quantified (featureCounts). Differential expression analysis was performed (DESeq2/edgeR) with |log2FC| >1 and FDR < 0.05 as cutoffs. Functional enrichment (GO/KEGG) used clusterProfiler.

Proteomics sequencing

Proteins were extracted from control and BM mouse tissues (3v3) using RIPA buffer (with protease inhibitors), quantified (BCA assay), and digested with trypsin (overnight, 37 °C). Peptides were desalted (C18 columns), labeled (TMT/iTRAQ for quantitative analysis), and separated by nanoLC (Dionex Ultimate 3000). Mass spectrometry (Q-Exactive HF-X) analyzed peptides in data-dependent acquisition (DDA) mode. Raw data were processed (MaxQuant/Proteome Discoverer) against UniProt databases, with FDR < 1%. Differentially expressed proteins (|fold change| >1.5, p < 0.05) were subjected to pathway analysis (KEGG/STRING).

ELISA

The lactate, citrate and succinate level of tissue samples or cell culture supernatants were analyzed using a quantitative sandwich ELISA (abcam). Briefly, 96-well plates were coated with capture antibody (overnight, 4 °C), blocked with 5% BSA (1 h, RT), and incubated with samples/standards (2 h, RT). Detection was performed using biotinylated detection antibody (1 h, RT), streptavidin-HRP (30 min, RT), and TMB substrate (15 min), with reaction termination by 2 N H₂SO₄. Absorbance was measured at 450 nm (reference 570 nm) using a microplate reader, with analyte concentrations calculated against standard curves. All samples were run in triplicate with blank and negative controls.

Immunofluorescence (IF) staining

Fresh-frozen or formalin-fixed tissue Sect. (5–7 μm) were fixed in 4%PFA (15 min), permeabilized with 0.1% Tween-20/PBS (5 min), and blocked with 5% BSA/10% normal serum (1 h). Primary antibodies (e.g., ki67, 1:200; anti-F4/80, 1:100, Servicebio, Wuhan, China) were applied overnight at 4 °C, followed by species-matched fluorophore-conjugated secondary antibodies (1:200, 1 h, dark, Servicebio, Wuhan, China). Nuclei were counterstained with DAPI (1 ug/mL, 5 min), and slides were mounted with antifade medium. Imaging was performed using a confocal microscope, with fluorescence intensity quantified via ImageJ. Controls included isotype-matched IgG and unstained samples.

Immunohistochemistry (IHC) staining

Formalin-fixed, paraffin-embedded tissue Sect. (4 μm) were deparaffinized, rehydrated, and subjected to antigen retrieval in citrate buffer (pH 6.0) using a pressure cooker (20 min). Endogenous peroxidase activity was blocked with 3% H₂O₂ (10 min), followed by protein blocking with 5% BSA (30 min). Primary antibodies (e.g., anti-CD3, 1:200; anti-CD8, 1:100; anti-CD31, 1:200; anti-IBA1, 1:200; Servicebio, Wuhan, China; H3K18la,1:200, Jingjie PTM BIO, China) were incubated overnight at 4 °C, then detected using HRP-conjugated secondary antibodies (1:500, 1 h; Servicebio, Wuhan, China) and DAB chromogen (5 min). Slides were counterstained with hematoxylin, dehydrated, and mounted. Staining was evaluated by two pathologists using light microscopy, with scoring based on intensity and distribution.

Co-immunoprecipitation (Co-IP)

Cells were lysed in RIPA buffer (with protease/phosphatase inhibitors) and pre-cleared with Protein A/G beads (1 h, 4 °C). Lysates (500 ug protein) were incubated with primary antibody (anti-lactylation, 2 ug) or IgG control overnight at 4 °C, followed by Protein A/G bead capture (2 h). Beads were washed 4× with lysis buffer, resuspended in 2× Laemmli buffer, and boiled (10 min). Eluted proteins were analyzed by Western blot for target interactors (Rac2). Input lysates (10%) served as loading controls.

Statistical analyses

Experimental data are expressed as mean ± SD derived from three or more independent replicates. For comparisons between two groups, statistical significance was assessed using paired Student’s t-test. Multi-group analyses were conducted via one-way analysis of variance (ANOVA) followed by post-hoc tests. All statistical procedures were implemented using GraphPad Prism 9.0 (GraphPad Software, Inc.) or IBM SPSS Statistics 23 (IBM Corp.), with significance thresholds defined as: *P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0. 0001.Three independent experimental repeats were performed to ensure reproducibility.

Results

Metabolic features of the tumor microenvironment in early-stage primary, advanced-stage primary and brain metastatic of lung cancer

The mechanisms by which lung cancer progresses to brain metastasis remain poorly understood. To investigate metabolic differences in the TME between early-stage、advanced-stage primary lesions and brain metastases, we extracted 86,215 cells from the GSE131907 dataset. UMAP dimensionality reduction and clustering analysis categorized cells into B lymphocytes, endothelial cells, epithelial cells (annotated as tumor cells), fibroblasts, mast cells, myeloid cells, NK cells, oligodendrocytes and T lymphocytes (Fig. 1A). Comparative analysis of cellular proportions revealed that early-stage primary lung lesions contained the highest abundance of T lymphocytes and the lowest proportion of epithelial cells. In contrast, advanced primary lesions and brain metastases exhibited similar cellular distributions, dominated by epithelial cells, followed by myeloid cells and T lymphocytes (Fig. 1B). Subsequent metabolic profiling across cell types revealed that epithelial cells exhibited marked upregulation of multiple anabolic pathways compared to other populations, including phenylalanine/tyrosine/tryptophan biosynthesis, steroid hormone synthesis, amino acid metabolism, fatty acid synthesis and glycolysis. Notably, B lymphocytes, NK cells and T lymphocytes showed minimal activation of metabolic pathways, suggesting a “metabolically quiescent” state in these immune cells (Fig. 1C). To delineate the role of glycolysis in tumor metabolism, we further analyzed the expression levels of glycolytic genes across cell populations, including LDHB, LDHA, PKM, ENO1, PGK1, TPI1 and ALDOA. Epithelial cells showed the strongest enrichment of glycolytic genes, followed by myeloid cells, endothelial cells and fibroblasts. Notably, all cell types displayed conserved expression patterns of these glycolytic genes (Fig. 1D). These findings collectively indicate that during the progression of lung cancer primary lesions and metastasis, immune cells (e.g., T and B lymphocytes) gradually decrease in abundance, whereas tumor cells expand numerically and exhibit heightened metabolic activity. Notably, the glycolytic pathway is an important metabolic pathway that maintains the dominant position of tumor cells in the TME.

Fig. 1.

Metabolic features of early-stage primary, advanced-stage primary and brain metastatic in lung cancer from the GSE131907 dataset. (A) UMAP clustering analysis reveals cell populations in early-stage primary, advanced-stage primary and brain metastatic. (B) Stacked bar plot illustrates the proportional distribution of cell populations. (C) KEGG enrichment analysis demonstrates the metabolic profiles across cell types. (D) Bubble plot highlights the enrichment patterns of glycolytic genes in diverse cell populations

Tumor glycolysis demonstrates significant associations with lung cancer progression, brain metastasis and immune evasion

Our analysis revealed that tumor cells exhibited the highest metabolic activity. We therefore isolated tumor cells from the dataset for further investigation, obtaining 7,270 cells from early-stage primary lesions, 6,582 from advanced-stage primary lesions, and 15,463 from brain metastases. UMAP dimensionality reduction and clustering analysis revealed multiple tumor cell subtypes both within individual lesions and across primary versus metastatic lesions, demonstrating significant intratumoral heterogeneity (Fig. 2A). Subsequent tumor associations phenotype analysis using the Molecular Signatures Database (MsigDB) revealed striking divergence in signaling pathway activation across early-stage primary lesions, advanced-stage primary lesions and brain metastases. In early-stage primary lung lesions, tumor cells exhibited marked upregulation of multiple signaling pathways including NOTCH, WNT, IL2/STAT5, KRAS, TGF-β, complement activation, p53 signaling, inflammatory response, autophagy and IL6/JAK/STAT3, suggesting a pro-inflammatory tumor microenvironment characteristic of early tumorigenesis. Advanced-stage primary lung lesions were dominated by cell cycle-related pathways, including G2/M checkpoint, E2F targets, MYC activation. Brain metastases, in contrast to early-stage primary lesions, displayed downregulated inflammatory signaling pathways, with predominant enrichment of estrogen response, peroxisome activity, bile acid metabolism, hypoxia, angiogenesis, oxidative phosphorylation, glycolysis, fatty acid metabolism, PI3K/AKT/mTOR signaling, DNA repair, and MTORC1 activation. These findings suggest a hypoxic, hypermetabolic, and immunosuppressive niche in brain metastases. (Fig. 2B). Subsequent KEGG enrichment analysis delineated the metabolic landscape across early-stage primary, advanced-stage primary and brain metastatic lesions. Early-stage primary lesions were primarily characterized by linoleic acid metabolism, arachidonic acid metabolism, sulfur metabolism and histidine degradation. Advanced-stage primary lesions transitioned to phenylalanine metabolism, glycerolipid metabolism and the pentose phosphate pathway. In contrast, brain metastases exhibited hypermetabolic activity, marked by tyrosine metabolism, pyruvate metabolism, fatty acid degradation, oxidative phosphorylation and glycolysis (Fig. 2C). We further analyzed the expression levels of glycolytic genes in tumor cells from early-stage primary lesions, advanced-stage primary lesions and brain metastases. Compared to early-stage primary tumors lesions, advanced-stage primary lesions and brain metastases exhibited significant upregulation of core glycolytic genes, including LDHA, PKM, ENO1, PGK1, TPI1 and ALDOA, with the most pronounced upregulation observed in brain metastases (Fig. 2D). These findings collectively demonstrate substantial heterogeneity among tumor cells from Early-stage primary lesions, advanced-stage primary lesions and brain metastases. Early-stage primary lesions were characterized by an inflammatory response signature, while advanced-stage primary lesions and brain metastases displayed immunosuppressive reprogramming. Notably, brain metastases exhibited hypoxia, angiogenic activation and hypermetabolic activity, with marked overexpression of glycolytic genes, including LDHA, PKM, ENO1, PGK1, TPI1, ALDOA, strongly suggesting that glycolytic is critically associated with lung cancer brain metastasis progression. Compared with bone metastases and adrenal metastases, the expression of glycolytic genes appears to be more significant in brain metastasis (see Supplementary files).

Fig. 2.

Metabolic landscape of tumor cells in early-stage primary lesions, advanced-stage primary lesions, and brain metastases from the GSE131907 dataset. (A) UMAP clustering reveals tumor cell subpopulations in early/advanced-stage primary lesions and brain metastases. (B) MsigDB enrichment analysis identifies differentially activated signaling pathways in tumor cells from early/advanced-stage primary lesions and brain metastases. (C) KEGG pathway enrichment analysis delineated differential metabolic signaling pathways in tumor cells among early/advanced-stage primary lesions and brain metastases. (D) Bubble plot visualizes the enrichment of glycolytic genes in tumor cell subpopulations across early/advanced-stage primary lesions and brain metastases

We conducted validation analysis on the PT1 cohort from the GSE198291 dataset [14], which comprises five groups: primary lesions tissues, adjacent normal tissues from primary lesions, CTCs in blood, brain metastatic tumor tissues and adjacent normal tissues from metastatic sites. The critical advantage of this dataset lies in its inclusion of CTCs, which capture dynamic alterations during metastatic dissemination. UMAP dimensionality reduction and clustering revealed significant tumor heterogeneity among primary tumors, CTCs and brain metastases (Fig. 3A). Glycolytic activity scores were markedly elevated in CTCs and brain metastases compared to primary tumors (P < 0.05) (Fig. 3B). Further analysis of glycolytic gene enrichment across groups revealed distinct expression patterns: primary lesions exhibited low baseline glycolytic gene expression, while CTCs showed predominant upregulation of TPI1, HK1, ENO1 and ALDOA. Brain metastases were characterized by marked overexpression of GAPDH, ENO1 and LDHA (Fig. 3C). These findings also suggest that different metabolic genes are involved in the process of tumor colonization from primary lesions to blood dissemination to brain metastasis. These results further validate the critical role of the glycolytic pathway in metastasis, with pronounced upregulation of GAPDH, ENO1 and LDHA specifically observed in brain metastases.

Fig. 3.

Glycolysis enrichment analysis of tumor cells in PT1 group from GSE198291. (A) UMAP dimensionality reduction and clustering analysis showing tumor cell distribution in primary lesions (PTT/PTP, tumor/paratumor), blood CTCs (circulating tumor cells) and brain metastasis (MTT/MTP, tumor and paratumor). (B) Violin plot displaying glycolysis-related geneset score in PTT (primary lesion), CTCs and MTT (metastatic lesion). (C) Bubble plot illustrating enrichment levels of glycolysis-related genes in PTT (primary lesion), CTCs and MTT (metastatic lesion)

Furthermore, a total of 517 lung adenocarcinoma tumor cells and 502 lung squamous carcinoma tumor cells were obtained from the TCGA database. The cells were then grouped according to clinicopathological grading and subjected to a comparative analysis of changes in glycolysis-related genes expression levels across various stages. The results indicated a correlation between elevated pathological grades (S1 to S4) and augmented glycolysis-related genes expression levels in lung adenocarcinoma, while this phenomenon was not significant in lung squamous carcinoma, indicating the critical role of glycolysis-related genes in lung adenocarcinoma progression (Fig. 4A-B). Subsequently, an analysis was conducted to examine the association between glycolysis-related gene expression levels and lung cancer prognosis. Using the GSCA online tool (bioinfo.life.hust.edu.cn), glycolysis genes were input to analyze DFI (disease-free interval), DSS (disease-specific survival), OS (overall survival), and PFS (progression-free survival) in lung adenocarcinoma and squamous carcinoma patients, with FDR < 0.05. It was found that elevated levels of glycolysis-related gene expression correlated with high-risk of DFI, DSS, OS and PFS in lung adenocarcinoma, particularly for the PKFP and LDHA genes (Fig. 4C). These findings indicate that elevated tumor glycolysis (especially PKFP and LDHA) is associated with disease progression and poor prognosis in lung adenocarcinoma, further highlighting the significant role of tumor glycolysis in lung cancer development and brain metastasis.

Fig. 4.

GSCA-based analysis revealed the link between glycolysis and lung cancer progression and prognosis. (A) KEGG analysis of glycolysis-related gene enrichment across different pathological grades in lung adenocarcinoma. (B) KEGG analysis of glycolysis-related gene enrichment across different pathological grades in lung squamous carcinoma. (C) Bubble plot visualizing GSCA-based prognostic correlations between glycolysis-related genes and lung cancer survival endpoints (DFI, DSS, OS, PFS). FDR < 0.05. HR denotes a higher risk of death. The redder color of the bubbles represents a higher risk, while bubble size represents the magnitude of statistical significance. DFI, Disease Free Interval. DSS, Disease Specific Survival. OS, Overall Survival. PFS, Progression free survival

Through the aforementioned analysis, we found that tumor cells in brain metastasis exhibit highly active glycolytic metabolism. We further investigated the specific role of glycolysis in these cells. First, UMAP dimensionality reduction and clustering analysis revealed significant heterogeneity within brain metastatic tumor cells (Fig. 5A). These tumor cells were divided into high-glycolysis and low-glycolysis groups based on the median glycolysis score. A comparison was made of differential gene expression between the two groups, and volcano plots were generated, using thresholds of |log2FC| ≥ 0.5 and FDR < 0.05, identified 16 downregulated genes and 30 upregulated genes. Glycolysis-related genes such as LDHA and PGK1 were significantly upregulated in high-glycolysis tumor cells (Fig. 5B). KEGG enrichment analysis of the differentially expressed genes revealed that high-glycolysis tumor cells exhibited positive correlations with central nervous system diseases, biosynthesis, proteasome, and pyruvate metabolism, suggesting a potential association between elevated glycolysis and disease progression. Conversely, negative correlations were observed with transcriptional dysregulation, TNF signaling pathway, chemokine signaling pathway and Th17 cell differentiation, implying that high glycolysis might be linked to immune evasion (Fig. 5C).

Fig. 5.

Differential enrichment analysis of high-glycolysis and low-glycolysis tumor cells in brain metastasis from the GSE131907 dataset. (A) UMAP clustering analysis of tumor cell subpopulations in brain metastasis from the GSE131907 dataset; tumor cells stratified into high-glycolysis and low-glycolysis groups based on glycolysis score. (B) Volcano plot displaying differential gene expression between high-glycolysis and low-glycolysis tumor cells in brain metastasis. (C) KEGG enrichment analysis of differentially enriched signaling pathways between high-glycolysis and low-glycolysis tumor cells in brain metastasis

Lactate accumulation in the LLC-BM model with an immunosuppressive tumor microenvironment

Previously, we analyzed the critical role of the glycolytic pathway in the development, progression, and brain metastasis of lung adenocarcinoma from public databases. Lactate dehydrogenase (LDH), a key enzyme in aerobic glycolysis, catalyzes the final step of glycolysis by converting pyruvate to lactate. Most cancer cells are characterized by accelerated tumor glycolysis rates to meet the energy demands of rapidly proliferating cells, thereby increasing lactate production. Excessive lactate accumulation leads to extracellular acidosis, which promotes invasion, angiogenesis, metastasis and influences immune responses. To further explore the immune microenvironment in driver gene-negative lung adenocarcinoma brain metastasis and the role of lactate/lactylation in BM, we first established the LLC-BM model, brain tissues of murine model were collected after 3 weeks. Given that lactate, citrate, and succinate are the three major tumor-associated metabolites [15], we measured their expression levels in BM tumor tissues through ELISA, only lactate showed significant elevation in BM tumor tissues (P < 0.001) (Fig. 6A), consistent with our previous findings of upregulated tumor glycolysis in lung cancer BM. Since lactate accumulation promotes lactylation, we subsequently detected lactylation levels in murine BM tissues via immunofluorescence, revealing extensive lactylation in tumor regions and M2-type macrophages (Fig. 6C). These results indicate predominant lactate accumulation and marked lactylation within LLC-BM tumor tissues.

Fig. 6.

Expression of lactate/lactylation and immune microenvironment in driver gene-negative LLC-BM. (A) ELISA quantification of lactate, citrate, and succinate levels in LLC-BM tumor and paratumor tissues. (B) IHC staining of CD31, bar: 100 μm. (C) IF staining of LLC-BM brain tissues: blue (DAPI), yellow (pan-lactylation), red (INOS, M1-type macrophages), green (Arg-1, M2-type macrophages). Scale bar: 100 μm. (D) IHC staining of CD3 (T cells), CD8 (T cells) and IBA-1 (macrophages) in LLC-BM tissues. Scale bars: 50 μm; IF staining: blue (DAPI), green (Hypoxia marker Hif-1α). Scale bar: 100 μm. (E) Transcriptomic sequencing KEGG enrichment analysis of tumor and paratumor tissues: differentially enriched signal pathways (vascular smooth muscle contraction) and gene expression levels of vascular maturation-associated gene S1PR1, pro-angiogenic genes VEGFA, Angpt2 and EGF. (P < 0.05, ns: not significant)

Then we performed IHC staining to assess the distribution of vascular endothelial cells (CD31) in LLC-BM model, revealing intra-tumoral vascular disorganization (Fig. 6B). Furthermore, immunohistochemical analysis of CD3 (T cells), CD8 (T cells), and IBA-1 (macrophages) and Hypoxia (Hif-1α) revealed minimal T cell and macrophage infiltration in normal murine brain tissues, whereas brain metastatic lesions regions exhibited fewer T cells but significant macrophage (IBA-1⁺) infiltration and markedly hypoxic. Immunofluorescence staining of M1/M2 macrophage polarization (iNOS⁺/Arg-1⁺) indicated predominant M2 macrophage infiltration (Fig. 6C-D). These findings collectively suggest that LLC-BM tumor regions are characterized by scarce T cell infiltration, dominant M2 macrophage presence, and marked hypoxia.

Transcriptomic sequencing of brain tumor tissues and brain paratumor tissues from the LLC-BM model identified differential expression of vascular-related signaling pathways and molecules. KEGG enrichment analysis demonstrated significant downregulation of vascular smooth muscle contraction-related signaling pathways in the BM group (P = 0.042). Vascular maturation-associated gene S1PR1 was downregulated, while pro-angiogenic gene VEGFA was upregulated in BM tissues. Although the upregulation of pro-angiogenic genes Angpt2 and EGF in BM tissues did not reach statistical significance, an increasing trend was observed (Fig. 6E). These findings align with our previous analysis of public datasets showing enrichment of hypoxia and angiogenesis-related pathways in brain metastatic lesions. Collectively, these results suggest enhanced angiogenesis but impaired vascular contractile function in brain metastatic lesions.

In summary, we found that in brain metastases, there is lactate accumulation and significant lactylation. CD8 + T cells are reduced, M2-type macrophages are increased, the tumor region is markedly hypoxic, blood vessels are abnormal, and the tumor microenvironment exhibits an immunosuppressive state.

Elevated lactylation levels of Rac2 protein in macrophages

These findings suggest a correlation between tumor glycolytic activity and the immunosuppressive tumor microenvironment. To further investigate the relationship between lactate accumulation and immunosuppressive tumor microenvironment, we performed proteomic sequencing combined with transcriptomic analysis on LLC-BM tumor and paratumor tissues (3 vs. 3). Proteomic sequencing detected 8,099 proteins, of which 202 were differentially expressed (146 downregulated, 56 upregulated), primarily enriched in signal transduction and post-translational modification (Fig. 7A). PPI (protein-protein interaction) network analysis of these proteins revealed that Rac2 protein interacts with multiple proteins, including chemokines such as CCL12, CCL2, CXCR6, and CCL5 (Fig. 7C). We therefore hypothesize that Rac2 protein is a key regulator of chemokine expression.

Fig. 7.

Lactylation of Rac2 occurs in macrophages. (A) GO enrichment analysis of LLC-BM tumor and paratumor tissues. (B) ELISA quantification of lactate levels in LLC cells under hypoxia versus control groups. (C)PPI network analysis of differentially expressed proteins in LLC-BM tumor and paratumor tissues. (D) WB of pan-lactylation: [1] Control: macrophage control group; [2] + tumor: macrophages cultured with tumor supernatant group; [3] + tumor-hypoxia: macrophages cultured with hypoxic tumor supernatant group; [4] + lactate: macrophages supplemented with exogenous lactate group. (E) WB of Rac2 protein in macrophages (grouping as in Panel D). (F) Co-IP of Rac2 lactylation and WB quantification. (G) Statistical analysis of WB results in E. (H) the IHC result of IBA-1 and the immunofluorescence staining of Rac2 (green) and Lactylation (red) (*P < 0.05, **P < 0.01, ****P < 0.0001)

Previous studies have reported that the Rac2 gene encodes a member of the Ras superfamily of small guanosine triphosphate (GTP)-metabolizing proteins. The encoded protein localizes to the plasma membrane, where it regulates diverse cellular processes, including secretion, phagocytosis, cell polarization and immunodeficiency [16]. Its inactivation directly impairs motility and function in most cell types [17]. In our sequencing results, Rac2 protein expression showed no significant difference between LLC-BM tumor and paratumor tissues (Table 1). Combined with our earlier immunofluorescence findings of extensive lactylation in tumor regions and predominant macrophage infiltration in the immunosuppressive microenvironment, and previous studies indicate that Rac2 directly regulates macrophage polarization and secretory functions, then we hypothesized that Rac2 protein in macrophages might undergo lactylation [18]. To validate this hypothesis, we first established a tumor hypoxia model to induce lactate accumulation (Fig. 7B). Tumor hypoxic supernatant was mixed 1:1 with RAW264.7 macrophage culture medium and incubated with macrophages for 24 h. Western blotting (WB) of pan-lactylation and Rac2 protein expression levels revealed marked lactylation in macrophages cultured with hypoxic tumor supernatant, exogenous lactate supplementation further enhanced lactylation (Fig. 7D). Rac2 protein levels showed no significant increase in macrophages treated with hypoxic tumor supernatant but were elevated upon exogenous lactate addition (Fig. 7E, G). Additionally, Co-IP assays confirmed significant Rac2 lactylation in macrophages exposed to hypoxic tumor supernatant (Fig. 7F). Furthermore, we performed immunofluorescence staining on mouse brain tissue samples from the LLC-BM model. Combined with the previous IHC staining results for IBA-1, we found that Rac2 is mainly expressed in macrophages surrounding the tumor and undergoes significant lactylation (Fig. 7H). These findings suggest that Rac2 may serve as a critical regulator of macrophage lactylation.

Table 1.

Macrophage-associated chemokines identified from transcriptomic sequencing (Total = 29)

Gene	baseMean	log2Fold Change	lfcSE	stat	p-value	padj
Il12rb1	326.9358	2.780447	0.624138	4.454859	8.39E-06	0.001447
Ccl12	298.8691	3.71034	0.874483	4.242897	2.21E-05	0.002833
Ccl2	937.1823	6.492821	1.593041	4.07574	4.59E-05	0.004769
Ccl5	253.344	3.683134	0.918738	4.008906	6.10E-05	0.005759
Cxcl1	90.47619	3.376371	0.905653	3.728105	0.000192925	0.012607
Cxcl10	835.4924	4.573757	1.450394	3.153459	0.00161348	0.048172
Tnf	27.99073	3.164975	1.014947	3.118364	0.001818583	0.051378
Ccl22	103.8784	2.623965	0.858656	3.055896	0.002243889	0.058476
Tnfsf4	10.68458	6.778968	2.253759	3.00785	0.002631031	0.064473
Tnfsf18	8.518331	6.45111	2.299555	2.805373	0.005025834	0.096204
Tnfsf8	46.75213	2.411878	0.885987	2.72225	0.006483905	0.1126
Tnfsf14	24.67376	3.476407	1.323636	2.626407	0.008629161	0.135007
Il23a	31.47844	1.74865	0.724445	2.413778	0.015788074	0.189948
Ccl3	76.71529	1.822979	0.762442	2.390973	0.016803781	0.196449
Cd40	107.9063	1.889223	0.821709	2.29914	0.021497003	0.225784
Il1b	128.642	1.489588	0.66615	2.236114	0.025344316	0.246447
Tnfsf15	12.49362	3.972904	1.974422	2.012186	0.044200332	0.331225
Il6	12.72022	1.663373	0.8497	1.957601	0.050276793	0.353161
Tnfsf10	623.4507	0.947777	0.495152	1.914112	0.055605818	0.372223
Tnfsf9	160.118	1.050975	0.571339	1.839495	0.065842484	0.403501
Il27	13.76448	2.480115	1.536114	1.614538	0.106410817	0.499434
Tnfsf13	67.04946	0.826274	0.526761	1.568595	0.116742318	0.52242
Il12b	20.03435	4.525195	3.320409	1.362843	0.172932106	0.60575
Cxcl13	29.00217	1.187173	0.908593	1.306605	0.191346851	0.630078
Il1a	101.7178	0.589685	0.477882	1.233957	0.217219072	0.66001
Il12a	174.9413	0.401257	0.387974	1.034238	0.301024843	0.739518
Tnfsf11	9.569597	1.385627	1.427812	0.970455	0.331819953	0.764963
Il15	142.3159	0.261334	0.399994	0.653344	0.51353477	0.867834
Il18	1787.593	-0.10259	0.270486	-0.3793	0.704468471	0.929748
	CON	CON	CON	BM	BM	BM
Rac2	19.31347	19.2976	19.25819	19.28194	19.03024809	19.3651

Macrophage lactylation induces M2 polarization via inhibiting CD40 and TNFSF13 secretion while promoting CCL22 secretion

Furthermore, RNA sequencing data revealed 750 differentially expressed genes (629 upregulated and 121 downregulated) through GO and KEGG enrichment analyses, and these genes were predominantly enriched in cell receptor interaction, chemokine signaling pathway, and complement pathway (Fig. 8A), suggesting critical roles of cytokines and chemokines in brain metastatic tumors. We extracted all macrophage-associated chemokines (29 in total) and found that the majority (19/29) exhibited differential expression (Table 1). To investigate which chemokine secretions might be influenced by lactylation, we conducted further cellular validation. First, we established a tumor hypoxia model to induce lactate accumulation, tumor supernatant was mixed 1:1 with RAW246.7 macrophage culture medium and incubated with macrophages for 24 h, then qPCR analysis of all macrophage-associated chemokines revealed significant downregulation of four pro-inflammatory factors (IL23, CD40, TNFSF10 and TNFSF13) and marked upregulation of CCL22 in macrophages, and these expression changes were more pronounced in the tumor-hypoxia group (Fig. 8B).

Fig. 8.

Lactate induces M2 polarization of macrophages. (A) KEGG enrichment analysis of transcriptomic sequencing data from LLC-BM tumor and paratumor tissues. (B) qPCR results of cytokines/chemokines (IL23, CD40, TNFSF10, TNFSF13, CCL22): Control: Macrophage control group; Tumor: Macrophages cultured with tumor supernatant; Tumor-hypoxia: Macrophages cultured with hypoxic tumor supernatant. (C) qPCR results of cytokines/chemokines (IL23, CD40, TNFSF10, TNFSF13, CCL22). Control: Macrophage control group; Tumor-hypoxia: Macrophages cultured with hypoxic tumor supernatant; Tumor-Hypoxia-LDHi: Macrophages cultured with hypoxic tumor supernatant supplemented with the LDHi. (D) Immunofluorescence staining was performed on RAW264.7 macrophages treated with tumor hypoxic supernatant, lactate, or lactate plus LDHi, respectively. DAPI (blue) was used for nuclear staining; Arg-1 (red) served as the marker for M2-type macrophages, and iNOS (green) served as the marker for M1-type macrophages. (E) the statistical results of the immunofluorescence images in D. (F) the western blotting results of RAW264.7 macrophages treated with tumor hypoxic supernatant or lactate, respectively. (G) presents the statistical results of F. (H) The western blotting results of RAW264.7 macrophages treated with lactate or lactate plus LDHi, respectively. (I) shows the statistical results of figure H. (*P < 0.05, **P < 0.01, ***P < 0.001, ***P < 0.0001)

LDHA (lactate dehydrogenase A), a key enzyme mediating lactate production and promoting protein lactylation. LDHi reduces lactate levels and protein lactylation [19]. To further confirm the functional significance of lactylation, tumor hypoxia models were treated with LDHi (20 mM) for 24 h, and tumor supernatant was then mixed 1:1 with macrophage culture medium to assess the expression of five cytokines/chemokines again (IL23, CD40, TNFSF10, TNFSF13, and CCL22). Results demonstrated that lactylation inhibition markedly upregulated TNFSF13 (P < 0.0001) and CD40 (P < 0.001), while significantly downregulating CCL22 (*P < 0.0001) (Fig. 8C). These findings suggest that CD40, TNFSF13 and CCL22 are likely key lactylation-regulated factors in LLC-BM. Subsequently, we treated macrophages with tumor hypoxic supernatant, lactate, or LDHi (a lactylation inhibitor), and detected the markers of Arg-1 (M2 phenotype) and iNOS (M1 phenotype) using immunofluorescence. The results showed that macrophages treated with tumor hypoxic supernatant exhibited a significant increase in Arg-1, indicating the occurrence of M2 polarization; M2 polarization became more pronounced after lactate treatment, while it was inhibited following the addition of LDHi (Figs. 8D-E). Consistent results were also obtained through Western Blotting verification (Figs. 8F-I). In conclusion, lactylation in macrophages inhibits the secretion of CD40 and TNFSF13, promotes the secretion of CCL22, and induces M2 polarization of macrophages.

H3K18la may be the key lactylation site in macrophages

Based on the results of our previously validated pan-lactylation modified proteins (Fig. 7D), it can be observed that the lactylation level at 15 kDa is the most prominent. Combined with previous studies, this molecular weight region corresponds to that of histone H3. Currently, the most common lactylation modification sites on histone H3 mainly include H3K9la, H3K14la, and H3K18la [20]. Therefore, we detected the protein levels of H3K9la, H3K14la and H3K18la in Rac2 protein enriched from macrophages cultured in lactate. The results indicated that the protein level of H3K18la was significantly increased (Fig. 9A-B). Subsequently, we added inhibitors targeting H3K18la (P300 inhibitor-C646) respectively to macrophages cultured in lactate. At the point of 6, 12, 24 h, proteins were collected to detect the expression level of Arg-1, and the results showed that C646 exhibited the significant inhibitory effect on Arg-1(Fig. 9C-D). This suggests that H3K18la may be the most prominently lactylated site in macrophages. By collecting paraffin blocks of primary lung cancer lesions and brain metastatic lesions from 2 patients with lung cancer brain metastasis, as well as primary lung cancer lesions from 2 patients without brain metastasis in our hospital, we performed H&E staining and immunohistochemical staining for H3K18la. The results showed that the lactylation (H3K18la) level in lung cancer with brain metastasis was significantly higher than that in the group without brain metastasis (Fig. 9E-F). In conclusion, H3K18la is a critical modification happened in macrophages, which is significantly highly expressed in lung cancer brain metastatic lesions.

Fig. 9.

H3K18la may be the key lactylation site. (A) The H3K9la, H3K14la and H3K18la protein levels of RAW264.7 macrophages treated with Lactate. (B) the statistical results of A. (C) the Arg-1 protein levels of RAW264.7 macrophages treated with Lactate plus P300 inhibitor C646. (D) the statistical results of C. (E) the H&E staining and IHC staining of NSCLC (no BM), NSCLC(BM) and Brain metastasis (BM). (F) the statistical results of E. (*P < 0.05, **P < 0.01, ***P < 0.001, ***P < 0.0001)

Discussion

This study represents the first metabolic-focused analysis of glycolysis-related gene signatures in lung adenocarcinoma brain metastasis, integrated analysis of proteomics and transcriptomics revealed that glycolysis promotes lung adenocarcinoma brain metastasis progression and immune evasion, through in vivo and in vitro models, we validated that the lactylation of Rac2 induces macrophage M2 polarization toward an immunosuppressive phenotype by modulating chemokines such as CD40 and TNFSF13, thereby promoting the formation of an immunosuppressive tumor microenvironment that facilitates lung adenocarcinoma brain metastasis. Clinically we found that the lactylation (H3K18la) level in lung cancer with brain metastasis was significantly higher than that in the group without brain metastasis.

Single-cell RNA sequencing has identified an early metastatic subpopulation in breast cancer that co-exists in both primary tumor lesions and lymph node metastases. This subpopulation exhibits marked upregulation of oxidative phosphorylation (OXPHOS) activity, with OXPHOS pathway activity showing an initial increase followed by a decline during metastatic progression, while glycolytic pathway activity demonstrates the opposite trend. These findings reveal the spatiotemporal evolution of breast cancer cells and highlight the association between OXPHOS pathway activation and lymph node metastasis, along with its potential prognostic value [21]. DLBCL-derived exosomal ENO2 accelerated glycolysis via GSK3β/β-catenin/c-Myc signaling pathway to ultimately promote macrophages to an M2-like phenotype, which can promote the proliferation, migration and invasion of cancer [22]. This evidence underscores those metabolic characteristics during tumor metastasis is not static, and targeting tumor metabolism may represent a potential therapeutic strategy. In this study, analysis of the GSE131907 dataset encompassing 10 matched cases of lung primary tumors and brain metastases revealed that early-stage primary lesions were characterized by the highest infiltration of T lymphocytes and the lowest proportion of epithelial cells, while advanced-stage primary lesions and brain metastases exhibited similar cellular distributions, dominated by epithelial cells followed by myeloid cells and T lymphocytes. Metabolic pathways were significantly upregulated in epithelial cells compared to other cell types. Tumor cells in early-stage primary lesions displayed an inflammatory response state, whereas advanced-stage primary lesions and brain metastases adopted an immunosuppressive phenotype, with brain metastases further exhibiting hypoxia, angiogenesis, and markedly enhanced glycolytic metabolism. Additionally, we observed an inverse correlation between high-glycolysis tumor cells and the “non-small cell tumor” signaling pathway, suggesting that the phenotypic features of primary lesions tumor cells gradually diminish during metastatic progression. These finding highlights that the heterogeneity between metastatic and primary lesions may be linked to heightened glycolytic activity.

Excessive lactate facilitates the establishment of an immunosuppressive microenvironment conducive to cancer cell growth and plays a pivotal role in shaping immune cell functionality. T cells sense extracellular lactate levels, triggering intracellular signaling, functional modulation, and homeostatic regulation. Excessive lactate suppresses T cell-mediated immune responses [23]. Mechanistically, lactate restricts T cell proliferation by altering the NAD(H) redox state, where lactate-rich conditions reduce NAD to NADH, thereby modifying NAD-dependent enzymatic reactions and diminishing glycolytic intermediates essential for proliferation [24]. Tumor-derived lactate directly reprograms cancer-associated macrophages (CAMs) into M2-like polarized cells, promoting tumor growth within the TME. This process is driven by ERK/STAT3 activation, upregulated expression of vascular endothelial growth factor (VEGF) and arginase Arg-1, and stabilization of HIF-1α, which collectively mediate lactate-induced M2 macrophage polarization and their pro-tumorigenic roles in breast cancer [25, 26]. Furthermore, elevated extracellular lactate levels impair lactate excretion in monocyte-derived macrophage precursors, establishing a negative feedback loop that enhances glycolysis and TNF-α release. In TLR-activated monocytes, tumor-secreted lactate amplifies IL23/IL17 transcription, polarizing immune responses toward the pro-tumor Th17 subset while suppressing anti-tumor Th1 responses [27]. Thus, lactate critically undermines T cell effector functions, disrupts macrophage polarization, and dysregulates monocyte precursor activity, ultimately reinforcing immunosuppression in the TME. Beyond its role as an immunosuppressive metabolite, lactate also functions as an epigenetic regulator by modulating gene expression through histone lactylation, a novel epigenetic modification [28, 29]. The metabolic imbalance between glycolysis and the tricarboxylic acid (TCA) cycle, driven by cellular metabolic reprogramming, is critical for increased histone lactylation. Lactate accumulation promotes histone lysine lactylation in cancer cells and immune cells such as macrophages and T cells, playing an essential role in tumor immune evasion [29]. Elevated lactate levels enhance ribosomal translation elongation rates, promoting tumor cell proliferation, acetyltransferase KAT8 significantly upregulates lactylation of eEF1A2, which plays a pivotal role in tumorigenesis [30]. In our study, LLC-BM model exhibited limited T cell infiltration, predominant M2 macrophage polarization, intratumoral vascular abnormalities, hypoxia, and dysfunctional vasculature. ELISA and immunofluorescence staining confirmed significant lactate accumulation and widespread lactylation in tumor regions of LLC-BM model. Multi-omics sequencing of LLC-BM tumor and paratumor tissues revealed 8,099 proteins in the proteomic dataset, with 202 differentially expressed proteins (146 downregulated, 56 upregulated), primarily enriched in signal transduction and post-translational modification. PPI network analysis demonstrated that Rac2 interacts with multiple proteins, including chemokines such as CCL12, CCL2, CXCR6 and CCL5. Interestingly, analysis of the GSE198291 dataset revealed that Rac1 expression is significantly elevated in LUAD compared to normal tissues. High Rac1 expression correlates with poor prognosis and increased metastatic risk. Genes in Rac1-high cells are enriched in adhesion, extracellular matrix (ECM), and VEGF signaling pathways. Rac1 promotes brain metastasis in the H1975 cell xenograft nude mouse model [14]. As members of the small GTPase family, Rac1 and Rac2 regulate monocyte chemotaxis and NADPH oxidase activity. Knockout of Rac1 and Rac2 severely impairs the development of immature CD4+/CD8 + T cells and mature CD4+/CD8 + populations in the thymus and spleen. Developmental defects in Rac1/Rac2 knockout T cells are linked to proliferation impairment and increased apoptosis [69]. Rac2 knockout mice exhibit enhanced B cell secretion [70]. Rac1 and Rac2, but not Rho, govern mature dendritic cell (DC) formation, their short-range migration toward T cells, and T cell priming [31]. Additionally, analysis of the GSE198291 dataset revealed a gradual increase in Rac2 expression across primary lung tumors, CTCs, and brain metastases, suggesting a potential role for Rac2 in metastatic progression despite lacking statistical significance. Collectively, Rac1 and Rac2 are indispensable for the development and proliferation of T cells, B cells, and DCs. RNA sequencing data analysis via GO and KEGG enrichment identified 750 differentially expressed genes (629 upregulated, 121 downregulated), predominantly enriched in cytokine receptor interaction, chemokine signaling pathway, and complement pathway, indicating a pivotal role of cytokines and chemokines in brain metastasis. We then performed cellular experiments to validate the finding that macrophages cultured in tumor hypoxic supernatants underwent significant lactic acidification, which was further amplified by exogenous lactate supplementation, concurrently, Rac2 protein levels increased in these macrophages. Co-IP assays confirmed significant Rac2 lactylation in macrophages exposed to hypoxic tumor supernatant. qPCR analysis of all differentially expressed macrophage-associated chemokines revealed significant downregulation of pro-inflammatory factors (IL23, CD40, TNFSF10, TNFSF13) and marked upregulation of the immunosuppressive chemokine CCL22, with these trends exacerbated under hypoxic conditions. Treatment with the LDHi markedly reversed these effects, upregulating TNFSF13 and CD40 while suppressing CCL22, implicate CD40, TNFSF13, and CCL22 as key lactylation-regulated factors in LLC-BM. Collectively, macrophage lactylation suppresses pro-inflammatory cytokine secretion (CD40, TNFSF13) and promotes TAMs-related CCL22 secretion, thereby driving M2 polarization and establishing an immunosuppressive microenvironment. Then we found that H3K18la is significantly highly expressed in lung cancer brain metastatic lesions and may serve as a potential target. This construction methods of LLC-BM model remains the most widely used modeling approach currently. It offers relatively simple operation and a high tumor formation rate. Since the striatum is a common site for lung cancer brain metastases, the model can specifically simulate the characteristics of metastases in this region, facilitating studies on the impact of the local microenvironment on tumors. The relatively fixed location of tumors allows for convenient subsequent imaging monitoring and pathological analysis. Additionally, it can exclude interference from peripheral metastases, focusing on the interaction between the intracerebral microenvironment and metastatic tumors. Still have limitations: This method disrupts the blood-brain barrier, neglects the processes of metastasis and colonization, and thus fails to recapitulate the complete pathophysiological course of “spontaneous brain metastasis”, making it difficult to evaluate therapeutic strategies targeting the early steps of metastasis [32]. Moreover, the LLC cell line is derived from mice and cannot fully represent the molecular heterogeneity of human LUAD. In the future, we need to integrate more models with high physiological relevance to human lung cancer brain metastasis, including intravenous injection models, patient-derived xenografts (PDX), and organoid models. This will allow for better simulation of clinical scenarios and provide support for mechanistic research and therapeutic exploration.

In this study, we can only suggest that Rac2 is affected by lactylation and induces M2 polarization of macrophages, and H3K18la may play an important role in this process. However, the specific mechanism by which H3K18la acts on Rac2 still needs further exploration. In the future, it is necessary to verified the function and activity of Rac2, and perform lactylome analysis for site identification and verify by establishing models with lactylation deficiency both in vitro and in vivo to futher comfirm the direct role of Rac2 lactylation modification, explore the therapeutic effect on lung cancer brain metastasis by combining with key inhibitors.

Conclusion

The glycolytic pathway plays a critical role in the metabolic reprogramming of tumor cells during lung adenocarcinoma brain metastasis. Tumor glycolysis is closely associated with lung cancer progression, brain metastasis, and immune evasion. The lactylation of Rac2 in macrophages may facilitate the formation of an immunosuppressive tumor microenvironment by induce the M2 polarization of macrophages.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Anwen Liu, Long Huang and Zhimin Zeng contributed to the conception and design of the study. Yali Yi drafting the article or revising it critically for important intellectual content. Daya Luo and Jing Cai and Le Xiong participated in the implementation and discussion of experiments. Yali Yi, Wenjie Xu, Houjian Yu and Yuxi Luo performed experiments. Yali Yi and Fujuan Zeng performed bioinformatics analysis. All authors approved the final version of the manuscript.

Funding

This study was supported by the National Natural Science Foundation of China [grant numbers 82260628, 82060577, to AW L], the National Science Foundation of Jiangxi Province [grant numbers 20232ACB206042, to AW L; Wu Jieping Medical Foundation [grant numbers 320.6750.2023-16-8, to AW L].

Data availability

The data used to support the finding of this study are available from the corresponding author upon request.

Declarations

Ethical approval

The experiment was approved by the Ethics Committee of Second Affiliated Hospital of Nanchang University Medical Research and the Nanchang University Approval for Research Involving Animals (20221031140).

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.