Paraffin-embedded human liver cancer specimens were obtained from 79 patients diagnosed between 2013 and 2015, with written informed consent obtained from all participants. This study was in accordance with the Declaration of Helsinki and approved by the Ethics in Research Committee of People’s Hospital of Lianyungang, China. The THP-1 and Raw246.7 cell lines were procured from Procell Life Science & Technology Co., Ltd. (Wuhan, China). Cellular samples were incubated in an RPMI 1640 solution from Invitrogen, based in Carlsbad, California, supplemented with 10% FBS (Cat. A5256701, Invitrogen, Carlsbad, CA, USA), 100 micrograms per milliliter of streptomycin, and penicillin G at a concentration of 100 units per milliliter. Both THP-1 and Raw246.7 cell lines have undergone mycoplasma infection testing and STR validation. Cells were cultured at a constant temperature of 37 degrees Celsius within an environment containing 5% carbon dioxide. Each treatment group of cell lines has 6 replicates to render sufficient statistical power, and every experiment was performed in triplicate.

Lentiviral vectors for Mettl3 over-expression (METTL3 OE) and Cpt1a knockdown (CPT1A shRNA) were obtained from GeneChem (Shanghai, China), and an empty vector was used as the negative control (NC), as specified in Table 1. Each cell line has three treatment groups: Control, METTL3 OE (Group A), and METTL3 OE + CPT1A shRNA (Group B). The Control group was transfected with empty lentiviral vectors, Group A was transfected with METTL3 OE lentiviral vectors, and Group B was transfected with METTL3 OE and CPT1A shRNA lentiviral vectors. Stable transduced cells were selected using puromycin (Cat.No.: ST551-10mg, Beyotime, Shanghai, China). To reduce batch-to-batch variation, each comparison involved pairs of samples from the same batch. All transfections were performed according to the manufacturer’s instructions.

RNA was entirely extracted from the cells using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA, Cat. No. 15596026) following the manufacturer’s instructions. Subsequently, cDNA synthesis was performed with the PrimeScript RT Master Mix Kit (TaKaRa, Dalian, China, Cat. No. RR036A). Real-time PCR analysis was conducted using SYBR Premix Ex Taq (Tli RNaseH Plus) (TaKaRa, Dalian, China, Cat. No. RR420A). Gene expression levels were normalized to GAPDH, which served as the reference gene. The primers used for amplification are detailed in Table 1. After amplifying cDNA, the quantification is typically done using the $\mathrm{\Delta }\mathrm{\Delta }$Ct method. We first determined the threshold cycle (Ct) values for both the target and reference genes, calculated the difference ($\mathrm{\Delta }$Ct) between them, and then compared the $\mathrm{\Delta }$Ct values of experimental samples to a control ($\mathrm{\Delta }\mathrm{\Delta }$Ct). The relative expression is expressed as 2^{-Δ⁢Δ⁢Ct}.

The THP-1 human monocyte cell line was cultivated in RPMI 1640 medium, enriched with 10% heat-inactivated FBS, at a constant environment of 37 °C and 5% CO₂. To induce macrophage activation, the THP-1 cells were exposed to 5 ng/mL of PMA (Sigma, St. Louis, MO, USA, Cat. No. P8139) for 48 hours. For tumor-associated macrophage (TAM) activation, macrophages derived from THP-1 were treated with a conditional medium derived from tissue lysate, adjusted to a protein concentration of 50 µg/mL, for 48 hours. To induce M2 macrophages, differentiate THP-1 cells with PMA (50–100 ng/mL) (Sigma, Cat. No. P8139) for 24–48 hours, then polarize with IL-4 (10–20 ng/mL) (Sigma, Cat. No. I4269) and IL-13 (10–20 ng/mL) (Sigma, Cat. No. SRP3211) for 24–48 hours.

In vitro macrophage stimulation in Raw246.7 cells is simpler than in THP-1 cells because they are already macrophage-like and do not need a differentiation step using PMA before stimulation. Stimulus steps are the same as THP-1 cells with IL-4 and IL-13 proteins.

Leukocytes infiltrating tumors were isolated following previously established methods [8]. In summary, non-essential tissue like fat, connective tissue, and necrotic material was discarded. The remaining tissue was finely chopped into pieces of 1–2 mm in RPMI 1640 (Invitrogen, CA), then transferred to conical tubes of 15 or 50 mL. These samples underwent a digestion process using a medium containing three enzymes: DNase at 30 U/mL, hyaluronidase at 0.1 mg/mL, and collagenase at 1 mg/mL (Invitrogen). This digestion occurred over 2 hours at ambient temperature with mild agitation. The tissues were then suspended in 10 mL of RPMI 1640 and strained through a 70-µm cell strainer (BD Pharmingen, San Diego, CA, USA). Any tissue retained by the strainer was individually placed into wells with 1 mL of T-cell growth medium in a 24-well plate, for further isolation or analysis via flow cytometry.

For flow cytometry and sorting, the isolated cells were first washed and then resuspended in FACS buffer (1$\times$ PBS, 2% inactivated Fetal Calf Serum (iFCS), 2 mM EDTA) for counting in a Neubauer chamber with Trypan blue (C1313S, Beyotime, Shanghai, China). Between 2 and 3 million cells were used for FACS staining. The cell suspensions were treated with a viability dye (1:400) in 1$\times$ PBS for 20 minutes. Following a wash with FACS buffer, the cells were incubated with FcR-block (1:100, C1752S, Beyotime, Shanghai, China) for 10 minutes and then sequentially stained with antibodies listed as below: CD45 (1:300, BD Pharmingen, CA, 553081), F4/80 (1:150, BioLegend, San Diego, CA, USA, 157305), CD206 (1:100, BioLegend, CA, 141719). After another wash, the samples were stained with streptavidin (A0305-0.2ml, Beyotime, Shanghai, China) for 10 minutes, fixed with 4% paraformaldehyde (PFA) (P0099-100ml, Beyotime, Shanghai, China) for 30 minutes, and then washed again with FACS buffer. The stained cells are then analyzed using a flow cytometer, where gates are set based on forward scatter (FSC) and side scatter (SSC) parameters to exclude debris and dead cells. LSRII Flow Cytometer from BD Biosciences (Franklin Lakes, NJ, USA) was utilized to assess the expression of the specific markers. The focus is on populations that are positive for CD45 and F4/80, with an emphasis on CD206 expression to identify M2 macrophages. The data was processed using FlowJo software (v10.0.7) from Tree Star Inc (Ashland, OR, USA).

Ninety 3-week-old C57BL/6 mice that consistently weigh around 13~15 g (Gemphamatec, Nanjing, China) were purchased and first subjected to a high-fat diet for eight weeks per previously successfully built animal model [13]. The Non-Alcoholic Fatty Liver Disease (NAFLD) Activity Score (NAS) for evaluating NAFLD in mouse models includes scoring steatosis (0–3), lobular inflammation (0–2), hepatocellular ballooning (0–2), and fibrosis (0–4) [14]. An orthotopic hepatocellular carcinoma (HCC) model was established using both NAFLD and non-NAFLD mice with the mouse HCC cell line Hep1-6. NAFLD mice were divided into 3 groups: Control, Group A, and Group B. To anesthetize the mice, propofol was injected into the abdomen at a dosage of 50 mg/kg. Eight weeks post-orthotopic transplantation with 3 $\times$ 10⁶ cells. In adoptive transfer therapy using mouse macrophages, bone marrow-derived monocytes were isolated from the femurs and tibias of C57BL/6 mice and differentiated into bone marrow-derived macrophages (BMDMs) over 7 days using complete medium (RPMI 1640 + 10% FBS) supplemented with 20 ng/mL M-CSF. Lentiviral packaging was then performed by co-transfecting HEK293T cells with either the METTL3 overexpression vector (pLVX-METTL3) or CPT1A shRNA vector along with packaging plasmids (psPAX2/pMD2.G) using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA, Cat. No. L3000001). Viral supernatants were collected after 48 hours and concentrated. Differentiated BMDMs were pretreated with polybrene (C0351-1ml, Beyotime, Shanghai, China) (8 µg/mL) and subsequently infected with lentivirus. Finally, 5 $\times$ 10⁶ cells were intravenously injected into tumor-bearing mice via the tail vein for therapeutic evaluation. For euthanasia, pentobarbital was injected into mice at 150 mg/kg dosage. Immediately following euthanasia, a midline laparotomy was performed, and the liver was carefully excised. For tumor volume measurement, the liver was weighed. For multiple immunohistochemistry (mIHC) staining and Hematoxylin and Eosin (H&E) staining, portions of the liver were fixed in 10% neutral buffered formalin. For other analyses (e.g., RNA and protein extraction), additional liver tissue samples were snap-frozen in liquid nitrogen and stored at –80 °C. Each group had 6 mice and every experiment was performed in triplicate.

Proteins were extracted from cultured cells using a Protein Extraction Kit (Kaiji, Nanjing, China). These proteins were then separated using SDS-PAGE gels and subsequently transferred onto PVDF membranes (Millipore, Burlington, MA, USA) with a pore size of 0.45 µm. For the detection process, anti-rabbit (A0208) or anti-mouse (A0216, Beyotime, Shanghai, China) horseradish peroxidase (HRP)-conjugated secondary antibodies were applied before electrochemiluminescence (ECL) detection. Multi-spectral images of the stained slides were captured using the Vectra 3 System (PerkinElmer, MA), ensuring consistent exposure times for each scan.

The process of mIHC staining on tissue sections was conducted using the PANO 5-plex immunohistochemistry kit (Catalog #10080100100, Panovue, Beijing, China), following the manufacturer’s instructions. The procedure began with the treatment of paraffin slices in a microwave, followed by incubation with one primary antibody. These slices were then incubated with horseradish peroxidase (HRP) conjugated secondary antibodies and subjected to aminide signal amplification. The antibodies utilized included anti-METTL3 (1:50, 15073-1-AP, Proteintech, Rosemont, IL, USA), anti-CD68 (1:200, 25747-1-AP, Proteintech), anti-iNOS (1:100, 22226-1-AP, Proteintech), anti-CPT1A (1:100, 15184-1-AP, Proteintech), and anti-$\alpha$-Tubulin (1:500, 80762-1-RR, Proteintech) and Biotin-conjugated Goat Anti-Rabbit IgG (H + L) (1:200, SA00004-2, Proteintech). Following the antibody staining and washes in the mIHC procedure, tissue sections were counterstained with 4^′,6-diamidino-2-phenylindole (DAPI) to visualize cell nuclei. Specifically, slides were incubated with a DAPI solution at a concentration of 1 µg/mL in PBS for 5 minutes at room temperature. After incubation, slides were washed twice with PBS to remove excess DAPI. Finally, coverslips were mounted using a fluorescence mounting medium with antifade properties to preserve signal and image quality during microscopy. DAPI fluorescence was excited using a UV light source and detected in the blue emission channel. Based on full film scanning, three high-magnification fields were randomly selected from each tissue point for image capture, maintaining the same exposure time for consistency. Cell counts and fluorescence intensity were assessed using InForm Tissue Finder (version 2.0) from PerkinElmer, Waltham, MA, USA. This software also facilitated automated tissue segmentation to distinguish between tissue and blank regions.

Liver tissues from the normal diet and high-fat diet mice were fixed, dehydrated by gradient ethanol and xylene, and then immersed in wax. The processed tissues were then sliced at 5 µm, followed by dewaxing and hydration by xylene and gradient ethanol. The slides were then stained with hematoxylin and eosin (Cat. A3429, Sigma-Aldrich, St. Louis, MO, USA).

Macrophages were harvested, washed twice, and fixed with 4% paraformaldehyde for 30 minutes. They were then stained with Oil Red O (Cat. Ab223796, Abcam, Cambridge, UK). After a 10-minute incubation, the cells were washed five times with double-distilled water and immediately examined under a microscope. The Oil Red O dye was extracted from the macrophages using isopropanol (563935, Sigma-Aldrich, St. Louis, MO, USA).

Total RNA was isolated from infiltrated macrophages from the normal diet and high-fat diet using TRIzol Reagent. RNA sequencing (RNA-seq) analysis was performed at LC Biotech Inc. (Guangzhou, China).

Metabolite analyses were performed using Gas Chromatography-Mass Spectrometry (GC-MS). For the examination of intracellular metabolites in THP-1 and THP-1-METTL3 cells, samples containing 10 million cells each were initially quenched. The metabolites were then derivatized using methoxyamine (15 mg/mL in pyridine) for 90 minutes at a temperature of 37 °C. Following this, the derivatives underwent further treatment with Bis(trimethylsilyl) trifluoroacetamide, including 1% chlorotrimethylsilane, and 20 µL of n-hexane, for a duration of 60 minutes at 70 °C.

The metabolomics instrumental analysis was carried out using an Agilent 7890A gas chromatography system linked to an Agilent 5975 C inert Mass Selective Detector (MSD) system (Agilent Technologies, Inc., Santa Clara, CA, USA). For the separation of these derivatives, an HP-5ms fused-silica capillary column measuring 30 m in length, 0.25 mm in diameter, and with a film thickness of 0.25 µm, from Agilent J&W Scientific was utilized. Mass spectra were collected within an m/z range of 50–600, under the selected reaction monitoring mode.

To interpret the metabolic variations among the different experimental groups, Partial Least Squares-Discriminant Analysis (PLS-DA) was applied to the data of intracellular metabolites. The Variable Importance in the Projection (VIP) scoring method was employed to determine the overall impact of each variable on the PLS-DA model. Metabolites that showed significant differences were identified based on two criteria: a p-value less than 0.05 obtained from t-tests and a VIP score exceeding 1 in the PLS-DA analysis.

The RNA Immunoprecipitation (RIP) assay was executed in line with the protocol of the RIP Kit (Geneseed Guangzhou, China). Briefly, first, cells are harvested, washed, and resuspended in a lysis buffer containing RNase inhibitors, followed by incubation on ice to facilitate lysis. The lysate is then pre-cleared with protein A beads to minimize non-specific binding before anti-mettl3 antibodies (1:50, Cat. ab195352, Abcam) are added for overnight incubation at 4 °C. After this, pre-washed beads are added to capture the antibody-antigen complexes, which are subsequently washed multiple times to remove unbound materials. After the washing phase, the complexes were eluted from the beads, followed by a purification process. The final step involved analyzing these purified complexes through RT-qPCR (TaKaRa, Dalian, China, Cat. No. RR036A). To determine the level of RNA enrichment, the quantities of precipitated RNAs were normalized against the input controls used in the experiment.

Cells were initially plated in 6-well culture dishes (Cat.140685, Thermo Fisher, Waltham, MA, USA). Following seeding, they were treated with actinomycin D (Sigma, Cat. No. A9415) at a concentration of 5 µg/mL for three distinct durations: 0 hours, 3 hours, and 6 hours. This treatment was conducted as a preparatory step before RNA extraction. After extraction, the focus was on determining the stability of CPT1A mRNA by measuring its half-life. For accurate quantification, CPT1A mRNA levels were normalized to $\beta$-Actin, which served as the reference standard in this assessment.

Statistical analyses in the study were carried out using two different software tools: R language (version 3.5.2, https://www.r-project.org/) and GraphPad Prism (version 9.0.0, GraphPad Software, Inc., San Diego, CA, USA). The embedded Kolmogorov-Smirnov (KS) test was used to test the normality of all data before comparison. For the purpose of comparing two distinct groups, a normality test was performed and followed by two-tailed Student’s t-tests. In contrast, the chi-square test was applied to evaluate categorical data.

For survival analysis, the Kaplan-Meier method was employed to generate survival curves. To further assess survival data, both univariate and multivariate Cox regression analyses were conducted. Additionally, comprehensive gene analysis was performed using R language (version 3.5.2), which included Gene Ontology (GO) analysis (https://www.geneontology.org/), Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis (https://www.genome.jp/kegg/), and gene set enrichment analysis (GSEA) (https://www.gsea-msigdb.org/gsea/index.jsp). In these analyses, a p-value of less than 0.05 was considered statistically significant, guiding the determination of the significance of the results obtained.

A high-fat diet has been implicated in the facilitation of lipid metabolism alterations and the potential advancement of hepatocellular carcinoma through metabolic reprogramming within macrophages associated with liver tumors. This hypothesis was explored using an orthotopic liver cancer model and a non-alcoholic fatty liver disease (NAFLD) model induced by a high-fat diet. Observational data indicated that a high-fat diet markedly augmented the proliferation of hepatic carcinoma. The tumor size of the normal diet mice was significantly smaller than the high-fat diet (HFD) mice, and so was the tumor weight (p 0.05) (Fig. 1A,B). Concurrently, the NAFLD score was employed to assess the substantial impact of a high-fat diet on hepatic pathology. Hematoxylin and eosin (H&E) staining further substantiated the diet’s facilitative role in tumor development. Morphological analysis revealed that a high-fat diet-induced NAFLD-related changes in hepatic tissue were notably characterized by an increase in vacuole-like hepatocytes. This hepatocellular vacuolation was predominantly pronounced in peritumoral regions (Fig. 1C), which is consistent with the the significant higher NAS score of HFD mice (p 0.001, Fig. 1D). Subsequent extraction and flow cytometric assessment of macrophages from hepatic carcinoma tissues showed that the F4/80-CD206 double positive rate of HFD mice was significantly higher than the normal diet mice (p 0.001), indicated a higher prevalence of M2 macrophages within the tumor environment of NAFLD-afflicted mice (Fig. 1E).

Fig. 1.

Effects of high-fat diet on liver cancer growth and M2-type macrophage prevalence. (A) Construction of liver cancer models in mice through in-situ implantation, comparing normal diet and high-fat diet groups (n = 6). (B) Quantitative comparison of liver cancer tissue weights in both diet-based liver cancer models (n = 6). (C) Morphological analysis of in-situ liver cancer in mice (the area outlined by the black coil), comparing those on high-fat versus normal diets, as visualized by hematoxylin and eosin (H&E) staining ($\times$50), Scale bar = 200 µm. (D) Assessment of Non-Alcoholic Fatty Liver Disease (NAFLD) index in the two dietary groups (n = 6). (E) Comparative analysis of M2-type macrophages in liver cancer tissues from normal and high-fat diet mice (n = 6). All experiments were performed in triplicate, and statistical significance was considered at p 0.05. *p 0.05; ***p 0.001. HFD, high-fat diet.

We conducted a comparative transcriptomic analysis via RNA sequencing to delineate the transcriptional modifications in macrophages of 79 patients who adopted two dietary regimens. The analysis revealed that dietary composition precipitated marked transcriptional disparities in macrophages. Applying an adjusted p-value threshold of 0.05 to define statistical significance, we identified that a high-fat diet instigated the upregulation of 546 genes and the downregulation of 561 genes within hepatic carcinoma-associated macrophages (Fig. 2A). Genes exhibiting significant upregulation encompassed Mettl3, Cpt1a, Slc6a12, Eomes, among others, whereas those demonstrating notable downregulation included Igfbp5, Fcn1, Stc1, among others (Fig. 2B). In contrast to the control cohort. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis underscored substantial modulations in fatty acid oxidation (FAO) pathways (Fig. 2C). Concurrently, Gene Ontology (GO) enrichment analysis indicated a pronounced accumulation of gene sets associated with RNA modification as a consequence of the high-fat diet (Fig. 2D). Specifically, we observed pronounced alterations in RNA modification pathways subsequent to the administration of a high-fat diet. Through differential gene expression analysis, quantified by log-fold change (LogFC), we delineated the genes implicated in RNA methylation. We noted a marked increase in the transcriptional activity of METTL3, an enzyme responsible for N6-methyladenosine (m6A) RNA modification, suggesting its upregulation in response to lipid-rich nutritional intake (Fig. 2E). Concurrently, through comprehensive gene sequencing of FAO-related genes, it was ascertained that CPT1A, an enzyme critical to the FAO pathway, emerged as the most substantially upregulated (Fig. 2E). Additionally, the upregulation of these genes in tissue samples from the high-fat diet group was validated using Real-time PCR analysis (Fig. 2F). This observed correlation suggests a hypothesis where a high-fat diet could potentially increase the stability of Cpt1a mRNA (p 0.01) through the upregulation of Mettl3 (p 0.001), thereby augmenting FAO metabolism within tumor-associated macrophages.

Fig. 2.

Impact of high-fat diet on gene expression and metabolic pathways in liver cancer tumor-associated macrophages. (A) Heat maps illustrating the transcriptional effects on hepatoma tumor-associated macrophages under high-fat versus normal diet conditions. (B) Volcano plots highlighting the most significantly altered genes in hepatocellular carcinoma-associated macrophages, comparing high-fat and normal diets. (C) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of differentially expressed genes. (D) Gene Ontology (GO) enrichment analysis of differentially expressed genes. (E) Comparative presentation of differentially expressed genes in the “RNA Modification” and “Fatty Acid Metabolism” pathways, across the two diet groups. (F) Validation of Mettl3 (p 0.001) and Cpt1a (p 0.01) gene expression in macrophages from liver cancer models on different diets, using real-time PCR. All experiments were performed in triplicate, and statistical significance was considered at p 0.05. **p 0.01; ****p 0.0001. ND, no difference.

Subsequent to RNA sequencing (RNA-seq) analysis, we verified the co-expression pattern of Mettl3 and Cpt1a in tumor-associated macrophages across diverse experimental conditions. This validation ascertained a direct proportionality in the expression levels of Mettl3 and Cpt1a. Employing both human macrophage cell line THP-1 and murine macrophage cell line Raw 264.7, we engineered cellular models to either overexpress Mettl3 or to possess knockdown constructs of Mettl3 with concurrent Cpt1a suppression (Supplementary Fig. 1). The mRNA and protein expression of METTL3 was upregulated in Mettl3 overexpressing cells, which were diminished by Cpt1a knockdown (Fig. 3A,B, p 0.05). First, metabolomic profiling of THP-1 macrophages, with and without Mettl3 overexpression, was conducted using metabolite set enrichment analysis (MSEA), Principal Component Analysis (PCA), and metabolic composition analysis. This revealed distinct metabolic reprogramming in tumor-associated macrophages, characterized by significant enrichment in fatty acid metabolism (Fig. 3C–E). To further explore the potential role of METTL3 in enhancing macrophage free fatty acid metabolism via regulation of CPT1A, we employed the aforementioned THP-1 and RAW246.7 cell models. The lipid metabolism levels in various cell populations were assessed using Oil Red O staining. Our results indicated a significant promotion of free fatty acid metabolism in macrophages mediated by METTL3. Notably, this augmentative effect was entirely abrogated following the knockdown of the Cpt1a gene and the percentage of oil O red cells of METTL3 OE-CPT1A shRNA is even higher than wildtype (p 0.001, Fig. 3F).

Fig. 3.

Role of methyltransferase-like 3 (METTL3) in regulating fatty acid metabolism in macrophages. (A) Real-time PCR analysis for the transcription of relevant genes in various macrophage cell lines subjected to treatments as depicted (n = 6, p 0.05). (B) Western blot assays to assess the protein expression levels in macrophage cell lines following the described treatments (p 0.05). (C) Differential metabolite analysis using metabolite set enrichment analysis (MSEA). (D) A PCA plot illustrating the metabolomic profile of macrophages with high Mettl3 expression. (E) Comparative analysis of METTL3’s effects on macrophage metabolism, represented in a composition diagram (n = 6). (F) Oil Red O staining to visualize lipid accumulation in macrophages under different treatment conditions (n = 6). Scale bar = 100 µm. METTL3 over-expression (METTL3 OE) + CPT1A shRNA cells have drastically more lipid accumulation than wild-type control (p 0.001). All experiments were performed in triplicate, and statistical significance was considered at p 0.05. ns, no statistical significance; *p 0.05; ***p 0.001. CPT1A, carnitine palmitoyltransferase 1A.

Upon differentiation into M2 macrophages under canonical M2-polarizing conditions (interleukin-4 (IL-4) and interleukin-13 (IL-13) stimulation), we observed a pronounced augmentation in M2 phenotype specification in cells with heightened Mettl3 expression. Conversely, cells with Cpt1a ablation did not exhibit a significantly different percentage of M2 macrophages than the wild type, indicating a substantial reduction in M2 differentiation propensity (p $>$ 0.05, Fig. 4A). These findings implicate METTL3 in the facilitation of M2 macrophage differentiation, potentially through the stabilization or upregulation of Cpt1a mRNA. Beyond in vitro standard induction experiments, we extended our investigation to in vivo models to determine whether METTL3 facilitates tumor growth through the Cpt1a gene. Flow cytometry was used to isolate macrophages, with a successful gating strategy that identified M2 macrophages with CD206 among the F4/80 macrophage population isolated from leukocytes by applying gating based on CD45 expression (Supplementary Fig. 2). Macrophages were isolated from bone marrow and subjected to lentiviral-mediated gene transfection to overexpress the Mettl3 gene in wild-type macrophages, concurrently with the knockout of the Cpt1a gene. Following adoptive cell therapy via tail vein injection, it was observed that macrophages overexpressing Mettl3significantly accelerated tumor growth (p 0.001, Fig. 4B). However, this tumorigenic effect was nullified when the Cpt1a gene was knocked down in conjunction with Mettl3 overexpression since the tumor weight of Mettl3 OE-Cpt1a shRNA mice is not significantly different than wild type (p $>$ 0.05, Fig. 4B). A similar trend was noted in the analysis of type II tumor-infiltrating macrophages, where Mettl3 overexpression favored M2 differentiation, but this effect ceased with the Cpt1a knockdown, evidenced by the observation that the increase in M2 percentage of METTL3 OE-CPT1A-shRNA was not significant than control (p $>$ 0.05, Fig. 4C). Consequently, our study indicates that METTL3 enhances tumor growth and type II macrophage differentiation by amplifying the effects of CPT1A in tumor-associated macrophages.

Fig. 4.

Mechanism of METTL3 in enhancing macrophage fatty acid metabolism through N6-methyladenosine (m6A) modification of Cpt1a gene. (A) Flow cytometry analysis demonstrating the impact of various treatments on the proportion of M2-type macrophages (n = 6). (B) Investigation of the influence of different macrophage treatments on liver cancer growth, utilizing adoptive cell transfer and in-situ implantation models (n = 6). (C) Flow cytometry assessment of M2-type macrophage ratios in the liver cancer models described in part B (n = 6). (D) Real-time PCR analysis showing expression levels of Mettl3 and m6A RIP-Seq enrichment of CPT1A/Cpt1a gene mRNAs (n = 6, p 0.001). (E) Assessment of Igf2bp3 knockdown effects on Cpt1a mRNA stability in THP-1 and RAW264.7 cells. Cells were treated with 5 µg/mL actinomycin D for 0, 3, or 6 hours before RNA extraction (n = 6). All experiments were performed in triplicate, and statistical significance was considered at p 0.05. ns, no statistical significance; ***p 0.001. RIP-Seq, RNA immunoprecipitation sequencing.

Given the role of METTL3 as an N6-methyladenosine (m6A) methyltransferase, it was hypothesized that METTL3 may mediate m6A modifications on Cpt1a mRNA. To investigate this, m6A and Mettl3 RNA immunoprecipitation sequencing (RIP-Seq) were employed to analyze macrophage transcripts. The results indicated effective binding and subsequent m6A modification of Cpt1a mRNA by Mettl3. Complementary RIP assays confirmed the specificity of the CPT1A-METTL3 interaction (Fig. 4D, p 0.001). Additionally, actinomycin D assays were conducted, which corroborated that METTL3 modulates the stability of Cpt1a mRNA across two distinct cell lines (Fig. 4E).

At last, we utilized a tissue array comprising 79 pairs of cancerous and adjacent non-cancerous cells to investigate m6A modifications in liver cancer-associated macrophages and their impact on macrophage fatty acids metabolism and differentiation. This was assessed using multispectral staining for CD68, iNOS, METTL3, CPT1A, and CD163 (Fig. 5A). Initially, we compared the expression levels of METTL3 and CPT1A in macrophages between tumor and adjacent tissues, finding no significant differences (Fig. 5B,C). Subsequent examination of the correlation between METTL3 expression and M1/M2 macrophage polarization revealed that macrophages expressing high levels of METTL3 exhibited a reduced M1 phenotype compared to those with low METTL3 expression. Conversely, the proportion of M2 macrophages was markedly elevated in the high METTL3 expression group relative to the low expression group, a trend consistent in both cancerous and adjacent non-cancerous tissues (Fig. 5D,E). Further studies indicated a significant positive correlation between the expression of Mettl3 and Cpt1a in macrophages, both in tumor tissues and adjacent non-tumor tissues (p 2.2 $\times$ 10^-6, Fig. 5F). Concurrently, we observed that the death rate of patients exhibiting high macrophage expression of Mettl3 was markedly lower compared to those with low Mettl3 expression (p 0.01, Fig. 5G). Additionally, we employed univariate Cox regression analysis after selecting significant variables by multivariate regression on the liver tissues from 79 patients. Cox regression results showed that high expression of Mettl3 in macrophages tend to have larger tumor size and later tumor node metastasis (TNM) staging (Fig. 5H).

Fig. 5.

Correlation between Mettl3 expression in macrophages and M1/M2 ratio with prognostic implications in human hepatocellular carcinoma. (A) Multicolor immunostaining displaying the protein expression levels of CD68, iNOS, METTL3, CPT1A, and CD163 in human liver cancer (METTL3 high and low) and adjacent tissues (Scale bar = 100 µm, upper panel, Scale bar = 200 µm). (B) Comparative analysis of METTL3 and CPT1A expression in hepatoma tumor-associated macrophages within cancerous versus adjacent non-cancerous tissues. (C) Differential expression of METTL3 and CPT1A in NAFLD-associated hepatocellular carcinoma tumor-associated macrophages, contrasting cancerous with adjacent non-cancerous tissues. (D,E) Evaluation of the proportion of M1 and M2 macrophages in relation to varying levels of METTL3 expression in macrophages, compared between cancerous and adjacent non-cancerous tissues. (F) Linear correlation assessment between METTL3 and CPT1A expressions in tumor-associated macrophages in both cancerous and adjacent non-cancerous tissues. (G) Survival analysis of patients stratified by different levels of macrophage METTL3 expression. (H) Correlation analysis between macrophage METTL3 expression and clinical parameters, utilizing Cox regression. All experiments were performed in triplicate, and statistical significance was considered at p 0.05. *p 0.05, **p 0.01. TNM, tumor node metastasis; HBV, hepatitis B virus; AFP, alpha-fetoprotein; CEA, carcinoembryonic antigen; ALT, alanine aminotransferase.