Abstract Background Obesity has emerged as a major global public health challenge and poses a significant threat to human health. Despite extensive research, the mechanisms underlying its pathological progression remain elusive. Aim To systematically identify pivotal targets and underlying mechanisms affecting the pathological progression of obesity through integrated strategies. Methods Transcriptomic and single-cell RNA sequencing (scRNA-seq) datasets were downloaded from the Gene Expression Omnibus (GEO) database. Differential expression analysis and weighted gene co-expression network analysis (WGCNA) were performed to identify obesity-associated genes. Moreover, functional enrichment analysis was conducted to elucidate potential mechanisms, and immune cell infiltration was assessed using the CIBERSORT algorithm. Then, macrophage-related genes were screened and subjected to degree centrality assessment, Least Absolute Shrinkage and Selection Operator (LASSO) regression, and random forest to identify hub genes. Furthermore, scRNA-seq was employed to systematically characterize key cell types and their gene expression profiles in the context of obesity. Finally, immunofluorescence (IF) and ELISA techniques were used to validate the expression of specific genes in the adipose tissue of obese mice. Results A total of 535 differentially expressed genes (DEGs) were identified, highlighting their significant role in the modulation of immune responses and inflammation. WGCNA was conducted to identify gene modules strongly correlated with obesity, and integration with differential expression analysis yielded 425 co-expressed DEGs. Pathway enrichment and immune cell infiltration analyses revealed that these genes were closely associated with the expression of macrophages. A total of 81 macrophage-related genes were further screened, and through protein-protein interaction (PPI) analysis combined with two machine learning algorithms, two hub genes (TREM2 and CXCR4) were ultimately identified. The Human Protein Atlas database and single-cell transcriptome analyses validated that TREM2 is specifically expressed in macrophages. Lastly, animal experiments verified the expression pattern of TREM2 in the adipose tissue of obese mouse models. Conclusion This study identified TREM2 as a key effector in the regulation of obesity-related pathophysiological processes, with specific expression in macrophages. These findings collectively position TREM2 as a potential diagnostic biomarker for obesity. Graphical abstract [48]graphic file with name 12967_2025_7096_Figa_HTML.jpg Supplementary Information The online version contains supplementary material available at 10.1186/s12967-025-07096-9. Keywords: Obesity, Bioinformatics analysis, Single cells, TREM2, Macrophages Highlights * Obesity is a chronic inflammatory condition characterized by metabolic dysregulation. * Obesity induces overactivation of adipose tissue macrophages and exacerbates immune-metabolic disturbances. * TREM2 is a critical regulator of obesity-related immune-metabolic dysfunction. Supplementary Information The online version contains supplementary material available at 10.1186/s12967-025-07096-9. Introduction As is well documented, obesity is a chronic progressive disease characterized by the pathological expansion of adipose tissue and accompanied by an imbalance in immune-metabolic homeostasis, representing one of the most pressing challenges to the global public health system [[49]1, [50]2]. In 2025, the evidence-based medical consensus issued by The Lancet Diabetes & Endocrinology Commission formally defined obesity as “a chronic disease that directly causes organ dysfunction due to abnormal accumulation of fat” for the first time, indicating the strategic transformation of obesity diagnosis and treatment from traditional weight management to organ function preservation [[51]3]. Pathophysiological studies have concluded that the key pathogenesis underlying obesity involves the biological dysfunction of adipose tissue, which constitutes an independent risk factor for complications, including metabolic diseases, cardiovascular events, and malignancies [[52]4–[53]6]. The dysregulation of immune-metabolic homeostasis in adipose tissue macrophages (ATMs) is a central driver of obesity-associated pathological processes. Under physiological conditions, ATMs primarily exhibit an M2 anti-inflammatory phenotype, accounting for approximately 5%-10% of the matrix vascular fraction. In obesity, ATMs undergo significant quantitative expansion (> 50%) and phenotypic polarization, triggering a cycle of chronic low-grade inflammation [[54]7–[55]9]. This pathological transition leads to a cascade of adipose tissue dysfunction, including impaired lipid synthesis and storage capacity, adipocyte necroptosis, impaired insulin signaling, and collagen deposition-mediated fibrosis [[56]10–[57]12]. Recent studies have detected a specific lipid-associated macrophage subset in obese adipose tissue that participates in lipid metabolism reprogramming to accommodate energy excess by enhancing the phagocytic activity of lipid droplets and the expression of lysosomal lipase [[58]13]. Therefore, elucidating the molecular targets that affect the phenotypic transformation of macrophages during obesity progression may provide a novel strategy for the treatment of obesity-related disorders. In recent years, the integration of multi-omics technologies with artificial intelligence (AI) methodologies has catalyzed a paradigm shift in studies investigating disease mechanisms. The combination of transcriptomic analyses and single-cell sequencing enables the systematically elucidation of disease-specific molecular events [[59]14]. At the transcriptional level, genome-wide differential expression analysis facilitates the identification of core signaling pathways associated with obesity [[60]15]. At the cellular interaction level, single-cell sequencing provides valuable insights into the spatial distribution and intercellular communication among various cell types within the adipose tissue microenvironment [[61]16, [62]17]. However, the high dimensionality and heterogeneity of multi-omics data pose challenges for traditional analysis methods to effectively extract key driving factors. In this context, AI-based algorithms, such as random forest and LASSO regression, have demonstrated unique advantages in filtering core gene sets with significant associations from the massive gene expression data, while mitigating multicollinearity to improve model interpretability [[63]18]. In summary, the research framework based on “multi-omics data, AI-driven mining, and experimental verification” is anticipated to offer robust methodological support for precisely identifying key regulatory targets in the pathological process of adipose tissue, thereby advancing research on precision diagnostics and therapeutics for obesity-related diseases. Materials and methods Data collection Obesity-related transcriptomic datasets ([64]GSE142187 and [65]GSE132706) were retrieved from GEO database ([66]https://www.ncbi.nlm.nih.gov/geo). Both datasets contain information from mice subjected to either a high-fat diet (HFD) or a normal diet (ND) for 8 and 20 weeks. RNA-seq was performed on RNA extracted from the white adipose tissue of mice at each time point. The [67]GSE214982 dataset examined mesenchymal vascular cells isolated from omental fat and epididymal fat using single-cell RNA sequencing technology, aiming to analyze differences in cell expression between lean and obese individuals. The [68]GSE210014, [69]GSE280875 and [70]GSE214618 datasets were used to validate the expression of TREM2 across different stages of obesity in mouse models. The [71]GSE205668 dataset was used to validate the expression of TREM2 in white adipose tissue of lean and obese groups (Table [72]1). Table 1. Determining features of the different analyzed datasets GEO dataset Public time Sample Sample source Platform Sample organism [73]GSE142187 2019 24(6-ND8w/6-HFD8w/ 6-ND20w/6-HFD20w) Perigonadal white adipose tissue [74]GPL19057 Mus musculus [75]GSE132706 2020 24(6-ND8w/6-HFD8w/ 6-ND20w/6-HFD20w) Epididymal adipose tissue [76]GPL17021 Mus musculus [77]GSE214982 2023 9(4-lean/5-obese); 4(ND2-HFD2) Omental fat; epididymal fat [78]GPL24676; [79]GPL24247 Homo sapiens; Mus musculus [80]GSE210014 2022 6(3-ND12w/3-HFD12w) Visceral adipose tissue [81]GPL17021 Mus musculus [82]GSE280875 2025 6(3-ND12w/3-HFD12w) Gonadal white adipose tissue [83]GPL24247 Mus musculus [84]GSE214618 2022 13(6-ND20w/7-HFD20w) Epididymal adipose tissue [85]GPL24247 Mus musculus [86]GSE205668 2022 61(35-lean/26-obese) Subcutaneous adipose tissue [87]GPL16791 Homo sapiens [88]Open in a new tab Screening of obesity-related differentially expressed genes Datasets acquired from the GEO database underwent pre-processing, including ID conversion, data cleaning, quality control, and normalization. For genes with multiple probes, the average expression value was calculated to ensure accurate representation. DEGs between the HFD and ND groups at 8 and 20 weeks were analyzed using the Limma R package. DEGs were selected based on the criteria of an adjusted P-value < 0.05 and |Log2FC| >1. The results were visualized via volcano plots, Venn diagrams, and PPI networks constructed using specialized online tools ([89]https://cloud.oebiotech.com/, [90]http://www.bioinformatics.com.cn/, and [91]https://hiplot.com.cn). GO, KEGG and Wiki pathways enrichment analyses Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes(KEGG), and Wiki Pathways enrichment analyses of the intersecting genes were performed using the ClusterProfiler package. The results were visualized on the online platform Wei Sheng Xin ([92]http://www.bioinformatics.com.cn/). Enriched GO categories and KEGG pathways with a P-value < 0.05 were identified and visualized to highlight the most significant enrichments. Weighted gene coexpression network analysis (WGCNA) Use the sva R package in OmicsTools software to eliminate batch effects, integrate the two datasets ([93]GSE142187 and [94]GSE132706) and perform WGCNA [[95]19]. Genes exhibiting similar expression patterns were clustered based on their gene association across samples. Correlations between gene modules and the high-fat diet feeding cycle in mice were then analyzed. Hub modules and genes most strongly correlated with obesity progression were identified based on correlation coefficients. These hub genes were subsequently subjected to downstream enrichment analyses. Identification of macrophage-associated signature genes The results from the Limma differential analysis and core genes identified in the WGCNA analysis were intersected, and differential enrichment analysis was performed using the Metascape database ([96]http://metascape.org) to uncover potential molecular mechanisms underlying obesity. Subsequently, macrophage-related genes were retrieved from the Human Protein Atlas database ([97]https://www.proteinatlas.org). The overlapping of these macrophage-related genes with the genes from the Limma-WGCNA analysis was then analyzed to identify DEGs specific to obesity-associated macrophages. Screening and validation of diagnostic markers Gene interactions were visualized using Cytoscape ([98]https://cytoscape.org/), and genes were ranked based on their degree values. The median degree value was used as the cut-off, and genes with degree values exceeding this threshold were selected for further analysis. To validate the most relevant DEGs associated with obesity, random forest analysis and Lasso regression modeling were performed. Through these analyses, two key genes impacted by macrophage changes in obesity were identified. Immune cell infiltration analysis Genes intersecting between DEGs and weighted co-expressed core modules were selected for further analysis. To explore the immune infiltration landscape, the CIBERSORT tool was employed to estimate the proportions of various immune cell types in adipose tissue. The resulting visualizations were generated using the Hiplot online platform, providing comprehensive gene expression profiles of immune cell subtypes. Single-cell transcriptome analysis An obesity-related single-cell dataset of stromal vascular cells isolated from human and mouse adipose tissue was downloaded from the GEO database ([99]GSE241982). RAW data were downloaded and re-analyzed using OmicsTools software to examine single-cell transcriptomic profiles in clinical patients and obese mouse models. The results illustrated the distribution and expression of cells and the cellular localization of related genes across different groups. Animals Healthy male C57BL/6 mice (7 weeks, 20–22 g) were obtained from Beijing Huafukang Technology Co., Ltd. [License No: SCXK (Beijing) 2024-0003]. All mice were housed in the Experimental Animal Center of Beijing Hospital of Traditional Chinese Medicine, Capital Medical University. A total of 40 mice were used for this study, maintained at 25℃ ± 1℃, a 12 h light/dark cycle with 65% ± 5% humidity, and given food and water freely. The mice were then randomly assigned to either the normal diet (ND, ^#SPF-F02-001) or high-fat diet (HFD, ^#D12492) group. The experimental protocols received approval from the Ethics Committee for Animal Experimentation at the Beijing Institute of Traditional Chinese Medicine (approval no. BJTCM-M-2024-06-02). ELISA Mouse blood samples were collected, and serum was isolated by centrifugation. The serum levels of circulating soluble TREM2 (sTREM2) were measured by ELISA according to the manufacturer’s protocol. Hematoxylin-eosin (H&E) staining Following euthanasia by cervical dislocation, peritesticular white adipose tissue was dissected and fixed in adipose tissue-specific fixative for 24 h. After fixation, samples were dehydrated, embedded, sectioned, and baked for 2 h. Afterward, they were stained with H&E using standard protocols, mounted with neutral gum, and examined and imaged under a microscope. Immunofluorescent staining After fixation, embedding, dehydration, and sectioning, adipose tissues were baked at 60℃ for 1 h, dewaxed, dehydrated with gradient alcohol, subjected to antigen repair for 20 min, blocked with serum, and incubated with primary antibodies (TREM2 and F4/80) for 4℃ overnight. The following day, the sections were rewarmed at room temperature for 30 min, washed with PBS, and incubated with secondary antibodies at 37℃ for 1 h. After mounting, the sections were observed under a microscope, and images were captured. Data and statistical analysis Statistical analyses and data visualization were conducted using GraphPad Prism 10.1.2, and the results were expressed as mean ± standard Inline graphic deviation. Group comparisons were performed using one-way ANOVA. A P-value less than 0.05 was considered statistically significant. Results Identification of obesity-related DEGs at different time points from the GEO database To investigate changes in gene expression during the pathological progression of obesity, transcriptomic data related to obesity at different time points were retrieved from the GEO database. OmicsTools software was employed for quality control and normalization. The normalized gene expression data were subsequently subjected to principal component analysis (PCA) to assess the relationships between samples from different perspectives (Supplementary Fig. [100]1). Following this, the R programming language and the limma package were used to identify DEGs between groups, with the results visualized using volcano plots. Volcano plot analysis of the [101]GSE142187 dataset revealed 2,807 DEGs between the HFD and ND groups after 8 weeks, including 1,913 upregulated and 894 downregulated genes. At 20 weeks, a total of 3,693 DEGs were identified, consisting of 2,341 upregulated and 1,352 downregulated genes (Fig. [102]1A-B). Similarly, analysis of the [103]GSE132706 dataset yielded 2,836 DEGs between the HFD and ND groups at 8 weeks, comprising 1,428 upregulated and 1,408 downregulated genes. At 20 weeks, 2,534 DEGs were detected, with 1,298 upregulated and 1,236 downregulated genes (Fig. [104]1C-D). Fig. 1. [105]Fig. 1 [106]Open in a new tab DEGs associated with obesity at various time points. A, B Volcano plot of DEGs in the [107]GSE142187 dataset between the HFD and ND groups after 8 and 20 weeks. C, D Volcano plot of DEGs in the [108]GSE132706 dataset between the HFD and ND groups after 8 and 20 weeks. E Venn diagram showing 535 intersecting key DEGs between the 8-week and 20-week datasets. F PPI network delineating the interconnectivity among the 535 key DEGs To further identify key DEGs involved in the pathological progression of obesity, the DEGs from the 8-week and 20-week datasets were intersected, resulting in the identification of 535 key differential genes, as illustrated in the Venn diagram (Fig. [109]1E). These 535 genes were then analyzed using the STRING database to assess their interactions, and the resulting network diagram is presented in Fig. [110]1F. Enrichment analysis of key DEGs through GO, KEGG, and WikiPathways in the pathological progression of obesity To further investigate the potential mechanisms underlying the pathological progression of obesity, GO, KEGG, and WikiPathways enrichment analyses were carried out on the 535 DEGs. Interestingly, GO analysis revealed that these DEGs were primarily enriched in biological processes (BP) such as myeloid leukocyte activation, leukocyte migration, and immune response-regulating signaling pathways, as well as in cellular components (CC) related to secretory granule and tertiary granule membranes, and molecular functions (MF) involving nicotinamide adenine dinucleotide (NAD(+)) nucleosidase activity and integrin binding (Fig. [111]2A, B). Fig. 2. [112]Fig. 2 [113]Open in a new tab Enrichment analysis of key DEGs. A GO enrichment analysis of the intersecting genes, categorized into BP, CC, and MF. B GO enrichment analysis of the intersecting genes. C KEGG pathway enrichment analysis of the intersecting genes, with each color representing a different category. D WikiPathways over-representation analysis of the intersecting genes KEGG pathway enrichment analysis categorized the results into five groups: metabolism, environmental information processing, cellular processes, organismal systems, and human diseases. Notably, the organismal systems category represented the largest proportion and included several immune inflammation-related pathways, such as the B cell receptor signaling pathway, complement and coagulation cascades, natural killer cell-mediated cytotoxicity, high-affinity IgE receptor (Fc epsilon RI) signaling pathway, leukocyte transendothelial migration, and chemokine signaling pathway (Fig. [114]2C). Furthermore, the results of the WikiPathways over-representation analysis revealed that the DEGs were predominantly enriched in pathways related to microglial pathogen phagocytosis, the TYROBP causal network in microglia, extrafollicular and follicular B cell activation by SARS-CoV-2, and B cell receptor signaling, among others (Fig. [115]2D). In summary, enrichment analysis of the 535 key DEGs highlighted a significant involvement in immune and inflammatory processes, particularly related to B cells, natural killer cells, white blood cells, and microglia/macrophages. The top 5 pathways of enrichment analysis of 535 key DEGs is shown in Supplementary Table [116]1. WGCNA-based analysis of temporal gene expression profiles in obese mice To explore the association between gene networks and the obese phenotype, as well as feeding duration, core genes within the network were identified. Data from the [117]GSE142187 and [118]GSE132706 datasets were integrated, and WGCNA was performed to identify the most co-expressed gene modules associated with the obesity phenotype. PCA was performed on the combined datasets, unveiling that samples from the same group clustered predominantly at both 8 and 20 weeks. The minimal sample dispersion across the two datasets suggested the absence of significant batch effects following data integration (Supplementary Fig. [119]2A). Boxplot analyses are displayed in Supplementary Fig. [120]2B. Next, the combined expression matrix was examined for outlier samples, and the connectivity threshold was set at -2 to exclude discrete samples. The results of the sample clustering analysis and network connectivity threshold are displayed in Supplementary Fig. [121]2C, D. The similarity matrix showed optimal mean connectivity at a power value of 8 (Supplementary Fig. [122]2E), and the WGCNA network was constructed using the Pearson correlation method (Fig. [123]3A). The eigengene adjacency heatmap is shown in Supplementary Fig. [124]2F. A total of 12 co-expression modules, each represented by a distinct color, were identified through average hierarchical clustering and dynamic merging. Based on the correlation coefficient between modules and the phenotype, the module of interest was selected for further analysis (Fig. [125]3B). Fig. 3. [126]Fig. 3 [127]Open in a new tab Identification of genes associated with the obese phenotype and feeding duration using WGCNA. A Hierarchical clustering dendrogram of co-expressed genes, with color bands representing identified modules. B Module-trait relationship heatmap. C Scatterplot of the turquoise module and gene significance Comparative analysis revealed that the turquoise module demonstrated the most significant correlation with obesity-related pathologies. Genes within this module were significantly negatively correlated with the ND group and positively correlated with the HFD group, with time-dependent changes observed. Therefore, genes in the turquoise module were considered hub genes that influence obesity progression and were subsequently analyzed. The turquoise module contained 3,891 trait-related genes, with scatter plot visualizations presented in Fig. [128]3C. The results of the enrichment analysis are delineated in Supplementary Fig. [129]3A, B. Enrichment analysis of overlapping targets between DEGs and hub genes In the aforementioned experiments, the Limma package was employed to analyze gene expression data across different samples and identify DEGs. Subsequently, WGCNA was employed to investigate gene co-expression patterns and identify core genes influencing the obesity phenotype. Intersecting the identified DEGs with the core genes yielded 425 genes exhibiting significant involvement in the pathological progression of obesity and displaying highly synergistic changes. These results were visualized using a Venn diagram (Fig. [130]4A). Fig. 4. [131]Fig. 4 [132]Open in a new tab Enrichment analysis of overlapping targets between DEGs and hub genes. A Venn diagram illustrating the overlap between DEGs and turquoise module genes from WGCNA. B Top enriched biological pathways identified through Metascape analysis. C, D Enrichment analysis of the 425 DEGs. E Ten computationally identified MCODE within the interaction network. F Immune cell infiltration and correlation analysis Next, functional annotation and enrichment analysis of the 425 overlapping genes were performed using the Metascape database. The findings demonstrated that these genes were predominantly enriched in pathways associated with the positive regulation of immune response, cell activation, cytokine production, microglial pathogen phagocytosis, neutrophil degranulation, hemostasis, and other inflammation-related processes (Fig. [133]4B-D). These results highlighted that pathways involving these 425 genes were most closely associated with the positive activation of immune responses, suggesting that the progression of obesity-related pathology is strongly linked to inflammatory responses, potentially mediated by microglial/macrophage phagocytosis. Furthermore, the Molecular Complex Detection (MCODE) algorithm was applied to identify densely connected network components. The MCODE networks generated from the 425 genes are presented in Fig. [134]4E, with MCODE1 being the most prominent module, containing key genes such as PLCG2, SYK, TREM2, C3AR1, and PLK1. Immune infiltration analysis was performed on the expression matrix of the 425 overlapping genes using the CIBERSORT algorithm, which generated infiltration scores for 22 immune cell types. The analysis indicated significant alterations in immune cell populations, including memory B cells, regulatory T cells (Tregs), dendritic cells, and macrophages, in the HFD group compared to the ND group (Fig. [135]4F). To further investigate cell-specific changes associated with obesity, the [136]GSE214982 dataset was retrieved from the GEO database to conduct single-cell RNA sequencing on clinical samples. The results showed a significantly higher proportion of macrophages in the adipose tissue of obese patients compared to lean individuals. Based on these findings, subsequent analyses were focused on macrophage-associated DEGs (Supplementary Fig. [137]4A-D). Identification of key genes implicated in obesity progression through multi-method analysis While long-term clinical studies have demonstrated that obesity is closely related to inflammation, the underlying mechanisms remain to be elucidated. The present findings suggest that the pathological progression of obesity is closely associated with inflammatory infiltration, particularly through the phagocytic activity of microglia/macrophages. Macrophages constitute the most abundant immune cell population in the adipose tissue of both genetically and diet-induced obese models. In the lean state, ATMs represent approximately 10% of the immune cells in adipose tissue; however, the proportion can increase to over 40% under obese conditions. To further investigate the genes involved in macrophage activation in adipose tissue following obesity, macrophage-related genes were retrieved from the Human Protein Atlas database and intersected with the 425 core genes identified earlier, resulting in the identification of 81 key DEGs related to macrophages (Fig. [138]5A). Fig. 5. [139]Fig. 5 [140]Open in a new tab Identification of key genes implicated in obesity progression through multi-method analysis. A Venn diagram demonstrating the 81 intersected genes between macrophage-related genes and 425 core genes. B PPI network of genes with a degree value greater than 18. C, D Feature importance ranking using the RF model. E, F LASSO coefficient profiles for the 24 selected core genes. G Venn diagram demonstrating the final two hub genes associated with obesity Subsequently, these 81 genes were input into the STRING database to generate an interaction network, which was visualized using Cytoscape. Network analysis was performed to determine the degree values for each gene. Genes with a degree value greater than 18 were selected as core genes, with a threshold of 18 representing the two-fold median (Table [141]2). The analysis yielded 23 core genes, which were visualized using the GeneMANIA database (Fig. [142]5B). To further identify key obesity-related genes, two machine learning techniques (LASSO regression and RF) were applied to the 24 selected core genes. The random forest algorithm identified 19 hub genes (Fig. [143]5C-D), while LASSO analysis identified 2 hub genes (Fig. [144]5E-F). Intersection of these results revealed that TREM2 and CXCR4 as the final hub genes associated with obesity (Fig. [145]5G). Table 2. The top 23 DEGs related to macrophages ranked by degree value Gene Degree Average shortest path length Betweenness centrality Closeness centrality Clustering coefficient TYROBP 45 1.4428571428571428 0.24238871797922606 0.6930693069306931 0.32626262626262625 TNF 41 1.5285714285714285 0.11090128759840741 0.6542056074766356 0.39146341463414636 ITGAM 36 1.6428571428571428 0.0521265947130502 0.6086956521739131 0.4666666666666667 FCER1G 35 1.6714285714285715 0.05572316170178386 0.5982905982905983 0.4336134453781513 CD68 33 1.6285714285714286 0.04784634429626021 0.6140350877192983 0.5 CTSS 32 1.6714285714285715 0.03844185727087682 0.5982905982905983 0.5120967741935484 SPI1 32 1.7285714285714286 0.029714496795291666 0.5785123966942148 0.5040322580645161 IL10 32 1.7428571428571429 0.025390374948044185 0.5737704918032787 0.5221774193548387 ITGAX 31 1.7571428571428571 0.017283963488778614 0.5691056910569106 0.5591397849462365 CYBB 28 1.7857142857142858 0.01945058524459422 0.5599999999999999 0.5634920634920635 SYK 27 1.8142857142857143 0.01680080181410049 0.5511811023622047 0.5356125356125356 CXCR4 25 1.8142857142857143 0.0322824251256326 0.5511811023622047 0.5533333333333333 TREM2 24 1.8142857142857143 0.04824623676351425 0.5511811023622047 0.5072463768115942 C3AR1 23 1.8571428571428572 0.007744673558727149 0.5384615384615384 0.6679841897233202 TLR7 23 1.8857142857142857 0.011530900134905821 0.5303030303030303 0.6521739130434783 TLR1 23 1.9 0.007493443345875121 0.5263157894736842 0.6521739130434783 CD44 22 1.8142857142857143 0.03197232990869283 0.5511811023622047 0.5714285714285714 CCR5 22 1.8857142857142857 0.021101681318752986 0.5303030303030303 0.6623376623376623 SIRPA 21 1.9 0.007017363806581904 0.5263157894736842 0.6476190476190476 NCF2 20 1.9285714285714286 0.00314387711002259 0.5185185185185185 0.7473684210526316 HAVCR2 19 1.9428571428571428 0.009027647202730719 0.5147058823529412 0.672514619883041 LAPTM5 19 1.9285714285714286 0.04468863191732004 0.5185185185185185 0.49122807017543857 C5AR1 19 1.957142857142857 0.002699711276273321 0.5109489051094891 0.7368421052631579 [146]Open in a new tab Finally, the expression of TREM2 and CXCR4 was validated in adipose tissue using the Human Protein Atlas database. The results suggested that TREM2 was specifically expressed on macrophages in adipose tissue (Supplementary Fig. [147]5A), whereas CXCR4 was abundantly expressed on T cells, macrophages, NK cells, leukocytes, plasma cells, and B cells in adipose tissue (Supplementary Fig. [148]5B). These findings collectively suggest that the significant increase in the proportion of macrophages in obesity may be primarily influenced by TREM2. To further investigate the expression patterns of different cell types and cell-expressed genes under both obese and lean conditions, immune infiltration and single-cell sequencing analyses were performed. Single-cell sequencing validation of cellular localization and expression of key genes in obesity onset To further elucidate changes in cellular composition and molecular characteristics in HFD- and ND-fed mice, the single-cell sequencing dataset of mouse gonadal adipose tissue vascular cells ([149]GSE214982) from the GEO database (Fig. [150]6A) was re-analyzed. Clustering analysis of scRNA-seq data was conducted and visualized using UMAP and t-SNE, revealing 22 distinct cell clusters, including endothelial cells, monocytes, smooth muscle cells, and macrophages (Fig. [151]6B-C). Significant heterogeneity was observed between cells from different samples, with marked alterations in the composition and interactions of immune cells in HFD-fed mice compared to ND-fed mice, particularly the infiltration of macrophages (Fig. [152]6D). In the adipose tissue of obese mice, macrophages were significantly activated, a finding consistent with previous immune infiltration and single-cell sequencing analyses of clinical samples. Fig. 6. [153]Fig. 6 [154]Open in a new tab The scRNA-seq analysis of [155]GSE214982. A-C Cluster analysis and cell type annotation results. D Heterogeneity analysis of cell types in HFD and ND groups. E Violin plots showing the expression of TREM2 and CXCR4 across cell types. F Box plots depicting the relative cell type abundance of TREM2 in the HFD and ND groups To verify the consistency between the two core genes identified through transcriptome analyses and the single-cell sequencing results, the expression of TREM2 and CXCR4 was examined in various cell types across the HFD and ND groups using violin plots. The results showed that TREM2 expression was significantly upregulated in macrophages in the HFD group, in agreement with previous findings. Although CXCR4 was expressed in adipose tissue B cells, monocytes, neutrophils, and T cells, no significant differences were observed between the HFD and ND groups (Fig. [156]6E). Box plots further illustrated the expression of TREM2 in specific cell populations in both groups (Fig. [157]6F). The results of high-throughput sequencing of clinical samples from [158]GSE205668 indicated that compared with lean individuals, the expression of TREM2 in obese individuals was significantly upregulated, with KEGG enrichment analysis of DEGs principally implicating metabolic pathways (Supplementary Fig. [159]6). In summary, these findings indicate that TREM2, as a macrophage-specific gene, may play a pivotal role in the pathological progression of obesity. Accordingly, we identified TREM2 as a key core gene and subsequently validated its expression pattern during obesity progression in follow-up animal experiments. Verify the expression of TREM2 through animal experiments and external datasets Subsequently, animal experiments were conducted to further analyze the expression of TREM2 in ATMs during obesity progression. High-fat diet was used to induce obesity in mouse models (8w–20w) over different time points, with the body size of different groups of mice is shown in Fig. [160]7A. With prolonged high-fat diet feeding, the body weight, waist circumference, and Lee’s index of mice progressively increased and subsequently plateaued at 16 weeks (Fig. [161]7B-E). H&E staining indicated that the volume of adipose tissue steadily increased with sustained obesity induction (Supplementary Fig. [162]7). Subsequent ELISA analysis revealed a time-dependent increase in sTREM2 levels during the feeding period, implying a positive correlation between sTREM2 expression and obesity progression (Fig. [163]7F). Fig. 7. [164]Fig. 7 [165]Open in a new tab Expression of TREM2 in ATMs across obesity progression. A Representative image of mouse body size across feeding groups. B−E Body weight, waistline, and Lee’s index of mice across groups. F Serum expression levels of sTREM2. G IF micrographs of TREM2 and F4/80 in adipose tissue at different obesity time points. Scale bars: 50 μm. H−J: Expression profiles of TREM2 across multiple independent datasets in the GEO database To further investigate the relationship between TREM2 and macrophages, IF staining was performed to assess their spatial distribution. During obesity progression, both TREM2 and the macrophage marker F4/80 exhibited time-dependent upregulation in adipose tissue. Notably, significant colocalization of TREM2 and F4/80 was noted, forming characteristic crown-like structures (Fig. [166]7G). This result further validates the indispensable role of TREM2 in the pathological process of obesity and suggests that it is closely related to macrophage activation. Systematic retrieval and analysis of obesity-related datasets in the GEO database consistently demonstrated significant upregulation of TREM2 expression in obese states across multiple independent datasets (Fig. [167]7H-J). These findings substantiate the pivotal role of TREM2 in obesity pathogenesis and highlight its clinical potential as a therapeutic target. Discussion Obesity constitutes a significant global public health issue, functioning not only as an independent medical condition but also as a crucial risk factor for multiple chronic diseases. Recent studies indicate that in 2022, the global population of obese individuals surpassed one billion, posing considerable threats to human health and generating substantial socioeconomic burdens [[168]20]. Consequently, an in-depth understanding of the pathological mechanisms underlying obesity is essential for developing novel preventive and therapeutic strategies. This study analyzed obesity-related datasets from the GEO database using Limma for DEG analysis and WGCNA. A total of 425 co-expressed genes closely associated with the pathological progression of obesity were identified. Importantly, enrichment analysis revealed that these genes were significantly involved in immune-inflammatory pathways, including the positive regulation of immune responses, cell activation, microglial pathogen phagocytosis, and immune effector processes. Subsequently, based on a deconvolution algorithm, the proportion of immune cells was inferred from a 425-gene expression matrix. Combined with single-cell analyses from clinical samples, the results demonstrated that macrophages were significantly activated in the adipose tissue of obese patients. In 2003, studies using both obese mouse models and human clinical samples reported that monocyte-derived macrophages specifically infiltrate adipose tissue under obese conditions, contributing to the formation of a pro-inflammatory microenvironment [[169]21, [170]22]. Notably, macrophage infiltration was consistently observed in both genetic (ob/ob mice) and diet-induced (high-fat diet) obesity models, highlighting its conserved role in obesity-induced immune remodeling [[171]21]. Under physiological conditions, macrophages constitute approximately 10% of adipose tissue-resident immune cells; however, their proportion increases significantly in obesity [[172]23]. Thus, comprehensively analyzing macrophage-associated regulatory genes is crucial for elucidating obesity-driven chronic inflammation and immune regulation in metabolic diseases. Based on the aforementioned findings and previous literature, key targets related to macrophages were identified from the 425 differentially co-expressed genes. Degree analysis, the random forest algorithm, and the LASSO regression model identified CXCR4 and TREM2 as core genes influencing the pathological progression of obesity. TREM2 is a type I transmembrane receptor belonging to the immunoglobulin superfamily, originally found to be expressed in monocyte-derived dendritic cells and macrophages [[173]24]. Cross-platform validation demonstrated macrophage-specific expression of TREM2 in adipose tissue according to the Human Protein Atlas, whilst high-throughput sequencing analysis of clinical samples signaled that the expression of TREM2 was significantly upregulated in the obese patients, suggesting that it may serve as a novel regulatory target for macrophage activation in obesity. CXCR4, a chemokine receptor constitutively expressed on multiple cell types (e.g., T cells, macrophages, NK cells, leukocytes, B cells) [[174]25], showed consistent expression patterns in the Human Protein Atlas and our scRNA-seq data. Although preliminary bioinformatic analyses suggested that CXCR4 may be a key gene in the pathogenesis of obesity, its expression patterns were inconsistent across datasets. For instance, RNA sequencing ([175]GSE142187 and [176]GSE132706) revealed a significant difference in CXCR4 expression between obese and control mice, whereas single-cell RNA sequencing ([177]GSE241982) data did not reach statistical significance. Moreover, CXCR4 is broadly expressed across multiple cell types, and this non-specific distribution may limit its reliability as a marker for macrophage-specific functions. In light of these analytical limitations and considering the robustness of the available evidence, we ultimately designated the macrophage-specific gene TREM2 as the core target for downstream functional investigations. In 2019, a seminal study published in Cell first identified the TREM2^+ lipid-associated macrophage subset and demonstrated that this subpopulation exerts metabolic protective effects in obesity through the formation of crown-like structures in adipose tissue, including suppression of adipocyte hypertrophy, maintenance of systemic lipid homeostasis, and improvement of glucose metabolism disorders [[178]13]. Bioinformatic analyses revealed a significant positive correlation between TREM2 expression in adipose tissue and body mass index (BMI), a finding subsequently validated in clinical samples using MARS-seq analysis [[179]13, [180]26]. Functionally, TREM2 regulates gene programs involved in phagocytosis, lipid catabolism, and energy metabolism, whereas its deficiency disrupts the downstream activation of the LAM molecular program [[181]27, [182]28]. Loss of TREM2 impairs macrophage recruitment to hypertrophic adipocytes and promotes pronounced adipocyte hypertrophy, systemic hypercholesterolemia, inflammation, and glucose intolerance, while TREM2^⁻/⁻ mice display more severe hepatic steatosis than their wild-type (WT) counterparts [[183]13, [184]29, [185]30]. Notably, TREM2 also mediates the reparative processes induced by bariatric surgery, mitigating the progression of metabolic-associated fatty liver disease by enhancing the clearance of dying hepatocytes and reducing macrophage-mediated inflammation [[186]31]. Collectively, these findings establish TREM2 as a pivotal regulator of tissue immune responses, coordinating conserved and protective immune cell programs that preserve metabolic homeostasis. Herein, prolonged obesity duration in mice led to sustained increases in body weight, waist circumference, Lee’s index, and sTREM2 levels, accompanied by significant adipose tissue expansion. It is worthwhile emphasizing that obesity progression was associated with marked macrophage activation in adipose tissue, with TREM2 expression showing co-localization with macrophages, thereby corroborating the findings of previous studies. As a core marker of lipid-associated macrophages, TREM2 upregulation represents an adaptive response whereby macrophages mitigate lipid overload through efferocytosis to maintain adipose tissue homeostasis [[187]29, [188]30]. However, chronic obesity may impair the functional ability of TREM2^+ macrophages, thereby exacerbating adipose tissue inflammation, insulin resistance, and systemic metabolic dysregulation. These findings suggest that elevated sTREM2 levels and TREM2^+ macrophage accumulation may not only serve as indicators of obesity-associated immune activation but also represent valuable biomarkers for predicting metabolic disorders and potential therapeutic targets. In this study, the expression of TREM2 in macrophages steadily increased with the duration of obesity in the epididymal white adipose tissue of obese mice induced by a high-fat diet for 8 to 20 weeks, which was in line with the results of bioinformatics analysis and previous articles. Recent studies have evinced that TREM2 enhances macrophage-mediated clearance of lipid-laden apoptotic hepatocytes during early steatosis induced by prolonged hypernutrition, thereby inhibiting liver inflammation and maintaining hepatic lipid homeostasis. Meanwhile, during the MASH stage, TREM2 cleavage to soluble sTREM2 disrupts cell clearance functions, thereby aggravating inflammatory infiltration and disease progression [[189]32]. Similarly, the expression of TREM2 in the adipose tissue of patients with pathological obesity was downregulated, and it was significantly negatively correlated with the severity of insulin resistance [[190]33]. This phenomenon suggests that TREM2 may undergo a dynamic transition from a compensatory protective role to a decompensatory dysfunction during the pathological progression of obesity, with its protective effects diminishing or even reversing in advanced stages. Therefore, it is essential to further elucidate the regulatory mechanisms by which TREM2 modulates macrophage function during specific pathological stages of obesity and its associated complications. In summary, the present study identified TREM2 as a potential adipose tissue-specific regulatory factor that participates in obesity-related immune remodeling. Through integrated bioinformatics, machine learning, and experimental validation, TREM2 was found to modulate lipid-associated macrophage function, contributing to immune homeostasis and metabolic regulation. These findings provide new insights into the mechanisms underlying obesity-associated inflammation and position TREM2 as a promising biomarker and therapeutic target for metabolic diseases. Nevertheless, some limitations of this study cannot be overlooked. To begin, the specific mechanisms and causal roles of TREM2 within adipose tissue macrophages (ATMs) remain unknown. While previous studies have documented that global knockout of TREM2 exacerbates obesity and metabolic disturbances, they did not distinguish the independent contributions of ATM-derived versus other cellular sources of TREM2. Additionally, the duration of high-fat diet intervention in this study did not encompass the entire dynamic course of obesity, and the relatively small sample size may limit the generalizability and robustness of the conclusions. Future studies employing macrophage-specific conditional knockout models or targeted adeno-associated virus-mediated interventions, extended intervention durations, and larger cohorts are required to precisely elucidate the ATM-TREM2 signaling axis, thus enhancing the reliability and translational relevance of the findings. Supplementary Information [191]Supplementary Material 1^ (2.8MB, docx) Acknowledgements