scRNA‐Seq in Gastric Cancer Reveals Macrophage‐Derived C1q as a Diagnostic Marker

ABSTRACT

Tumor‐associated macrophages (TAMs) and the complement system play pivotal roles in reshaping the chronic inflammatory microenvironment and driving malignant transformation. Although C1QA⁺ TAMs have been identified in multiple tumor types, their role in gastric cancer (GC) remains unexplored. Single‐cell RNA sequencing (scRNA‐seq) was performed on paired tumor tissues and adjacent tissues from three treatment‐naïve patients to map macrophage heterogeneities. To evaluate the clinical and translational value of C1q, immunohistochemistry (IHC) was conducted on 100 GC samples, and C1q concentrations were measured via ELISA in plasma from a multi‐center cohort of 240 patients and 100 GC tissue. Additionally, the effects of C1q‐mediated THP‐1‐derived macrophages on the malignant behaviors of GC cells were investigated using in vitro migration. Transcriptomic analysis of 56,151 single cells successfully identified a distinct pro‐inflammatory subset: C1QA⁺ TAMs. IHC staining demonstrated that C1q and CD163 were highly co‐expressed in compromised tissues, and their elevated expression significantly correlated with poor patient prognosis. Multi‐center plasma analysis revealed that elevated circulating C1q levels robustly enhanced clinical diagnostic accuracy. In vitro functional assays confirmed that C1q significantly primed M2 macrophage polarization, thereby accelerating the migration of GC cells. We first identified and functionally validated the C1QA⁺ TAM subset as a key driver of microenvironmental remodeling. C1q serves as a crucial mediator of macrophage polarization and holds strong promise as a novel, non‐invasive biomarker for the early diagnosis.

Single‐cell sequencing of GC analyzed the tumor microenvironment and discovered the C1QA⁺ TAMs subpopulation with immunosuppressive function. The role of C1q in the prognosis and diagnosis of GC was demonstrated through plasma and tumor tissue from the GC cohort.

Abbreviations

GC

Gastric cancer

IHC

Immunohistochemistry

scRNA‐seq

Single‐cell RNA sequencing

TAMs

Tumor‐associated macrophages

TME

Tumor microenvironment

1. Introduction

Gastric cancer (GC) remains a prevalent and deadly malignancy worldwide, ranking fourth in incidence and fifth in mortality globally [1]. In China, a substantial proportion of GC patients are diagnosed at an advanced stage, leading to poor survival outcomes, with median overall survival often falling below 1 year [2, 3]. Despite progress in early diagnosis and treatment, the molecular mechanisms driving GC initiation and progression are not fully understood. The persistently low 5‐year survival rate for patients with advanced‐stage disease, below 5% [4, 5, 6], highlights the urgent need for a deeper understanding of the disease's biological basis and the identification of novel therapeutic targets.

The tumor microenvironment (TME), encompassing immune cells (e.g., lymphocytes, myeloid cells), stromal cells, and endothelial cells, among others [7, 8], plays a critical role in GC progression and the development of treatment resistance. Tumor‐associated macrophages (TAMs), a key component of the TME, have been strongly linked to poor patient prognosis and treatment resistance [9, 10]. TAMs promote GC invasion and metastasis through a variety of mechanisms, including mediating immunosuppression, remodeling the extracellular matrix, secreting pro‐tumorigenic factors, and inducing metabolic reprogramming. Single‐cell RNA sequencing (scRNA‐seq) technology provides unprecedented resolution for dissecting the cellular and molecular heterogeneity within the TME, allowing for the profiling of gene expression and metabolic characteristics at the single‐cell level [11]. Therefore, a systematic characterization of the transcriptional profiles of key immune cells, particularly TAMs, within the GC TME is essential for understanding the limitations of immunotherapy and for developing novel TME‐targeting therapeutic strategies to combat GC progression.

Here, we performed high‐precision scRNA‐seq on paired tumor and non‐tumor adjacent tissues (NATs) obtained intraoperatively from three treatment‐naïve GC patients. Through detailed analysis of the TME's cellular composition, we identified a key TAM subset at the single‐cell level: C1QA⁺ TAMs. Our research indicates that C1q can enhance the expression of immunosuppressive factors in macrophages, and that C1q is highly expressed in tumor tissues and patient plasma. Notably, patients with high C1q expression exhibit a lower 3‐year survival rate. Furthermore, C1q may potentially serve as a biomarker for the diagnosis of GC.

2. Materials and Methods

2.1. Patients and Samples

The tumor tissues and plasma samples of gastric cancer were obtained from Zhongda Hospital Affiliated to Southeast University and Wuwei Tumor Hospital. The study protocol was approved by the Ethics Committee of Zhongda Hospital, Southeast University and Wuwei Tumor Hospital. We enrolled three treatment‐naïve GC patients whose diagnoses were pathologically confirmed. Patients had not received prior chemotherapy, radiotherapy, or other anticancer drug therapies. After review by two independent pathologists, the baseline pathological features of the 3 paired NATs samples used for single‐cell sequencing (all collected from the distal margins 5 cm away from the tumor edge) are as follows. All samples were confirmed to have no infiltration of any malignant tumor cells (ensuring the histological non‐tumor nature). Patient 1 (C1): Chronic non‐atrophic gastritis with mild lymphoplasmacytic infiltration; Helicobacter pylori ( H. pylori ) negative; No obvious intestinal metaplasia or atypical hyperplasia. Patient 2 (C2): Chronic atrophic gastritis with focal mild intestinal metaplasia; H. pylori positive. Patient 3 (C3): Chronic non‐atrophic gastritis; H. pylori negative; No atrophy or intestinal metaplasia. The primary tumor tissues and normal tissues were collected by gastroscopy. Detailed patient characteristics, including age and gender, are presented in Tables S1–S4.

2.2. Single‐Cell Suspension Processing and Library Preparation

Single‐cell suspensions passing quality control (viability > 85%) were washed, resuspended to a concentration of 700–1200 cells/μL, and loaded onto the 10× Genomics Chromium system. This system generates Gel Beads‐in‐Emulsions (GEMs) to achieve single‐cell partitioning. Cell barcoding was accomplished via reverse transcription (RT) of GEMs in a PCR thermal cycler. Subsequently, GEM emulsions were broken, and the barcoded first‐strand cDNA was purified and enriched using magnetic beads. Following cDNA amplification and successful QC, Next‐Generation Sequencing (NGS) libraries were constructed. Library preparation involved fragmentation, adapter ligation, and sample indexing via PCR. Final libraries underwent quantification and QC before being sequenced on either the Novaseq Xplus or DNBSEQ‐T7 platform using a paired‐end 150 bp (PE150) configuration. A targeted sequencing depth of ≥ 20,000 reads per cell was achieved. We excluded clusters with high mitochondrial gene content (> [e.g., 20%]) and low feature counts, which typically indicate dead cells or empty droplets (Table S5). We used [DoubletFinder] to identify and remove potential doublets. Additionally, we manually inspected and removed clusters co‐expressing lineage‐specific marker genes (putative doublets).

2.3. Analysis of Single Cell Sequencing Data

Raw sequencing reads underwent initial quality assessment using Fastp, evaluating base composition (A/T/G/C) distribution and error rate profiles, followed by comprehensive processing through the 10× Genomics Cell Ranger pipeline which executed: STAR alignment to the reference genome; quantification of high‐quality cells, genes per cell, and alignment rates; cell identification via barcode assignment; and gene expression quantification via UMI counting. To address inherent technical artifacts (empty droplets, multiplets, aberrant gene counts, elevated mitochondrial content), additional quality filtering was applied post‐expression matrix generation using outlier removal criteria: genes per cell (nGene), UMIs per cell (nUMI), mitochondrial/ribosomal gene expression, and DoubletFinder‐derived doublet detection, with systematic exclusion of cells exceeding ±2 standard deviations from parameter means to ensure downstream analytical robustness.

2.4. Dimension Reduction and Clustering

Following expression matrix filtration, Seurat implements default “LogNormalize” normalization: gene counts per cell are divided by total cellular transcripts (library size) to derive relative abundances, scaled by a factor of 10,000, and log‐transformed. Post‐normalization, technical variations are regressed out through linear modeling. Dimensionality reduction is performed using UMAP, while cell clustering identifies distinct subtypes based on shared transcriptional profiles. Resulting embeddings visually resolve cellular distributions.

2.5. Cell Type and Marker Gene Identification

Cell type annotation integrates marker gene analysis with computational tools (e.g., SingleR) through two primary approaches: (1) Supervised/semi‐supervised marker‐based identification leverages known cell‐type‐specific markers for manual/algorithmic annotation—straightforward yet susceptible to subjective bias and marker database limitations, (2) Unsupervised reference‐based identification classifies cells by similarity to well‐annotated expression datasets, minimizing human bias though constrained to broad categorizations due to reference scope limitations. To enhance reliability, these complementary methodologies are typically integrated.

2.6. Signature Score Analysis

To infer the functional status of macrophage subpopulations, we collected a list of gene signatures including M1, M2, angiogenesis, and phagocytosis‐related gene signatures from a recent study by Cheng et al. Signature scores for these gene signatures and pathways were calculated for each macrophage subpopulation using the ssGSEA method in the GSVA package.

2.7. Differential Gene Expression Analysis of Cellular Subpopulations

Following clustering analysis, cells were partitioned into distinct subclusters. Using Seurat, we performed comparative analysis between each subcluster and all remaining subclusters to identify differentially expressed genes (DEGs). The FindAllMarkers algorithm [12] was employed to detect cluster‐specific upregulated genes with the following thresholds: Expressed in ≥ 20% of cells within the cluster; Average log fold‐change > 1.5; Adjusted p‐value (FDR) < 0.05.

2.8. Gene Set Variation Analysis (GSVA)

We conducted GSVA [13, 14] using KEGG pathways and Gene Ontology gene sets from the Molecular Signatures Database (MSigDB v7.0). Pathway activity scores were calculated for each cell cluster using GSVA (v1.34.0), with results visualized in bar plots representing cluster‐wise average expression levels.

2.9. Cell Communication Analysis

We inferred intercellular interactions between immune cells using CellChat (v0.0.1) [15]. Interaction strengths were calculated based on expression levels of ligand‐receptor pairs. Significantly enriched ligand‐receptor pairs (p < 0.05) were extracted for further analysis.

2.10. Cell Trajectory Analysis

To determine potential linear differentiation between macrophage subpopulations, Monocle2 [16](version 2.26.0) was used and trajectories were inferred with cell subpopulation marker genes. A CellDataSet object was created using the normalized count data. Infer trajectories using default parameters after downscaling and cell sorting [17].

2.11. TCGA Data Analysis

To validate the presence of C1QA⁺ TAMs, NRP1⁺ TAMs, and HLA‐DPA1⁺ TAMs in primary GAC in a large primary GAC cohort with available expression and clinical data, we performed ecotype deconvolution analyses using a bulk RNA sequencing dataset. We downloaded bulk RNA‐seq data normalized to primary GC (STAD) generated by The Cancer Genome Atlas (TCGA). The RNA‐seq data were processed and normalized by the TCGA using their transcriptome analysis pipeline for processing and normalization. Clinical annotations of the TCGA‐STAD cohort and molecular subtypes defined by the TCGA analysis working group were downloaded from the recent PanCanAtlas study. In addition, we downloaded GAC‐seq data from the Gene Expression Omnibus database (GEO, https://www.ncbi.nlm.nih.gov/geo/). Raw gene expression values from microarray experiments were preprocessed (background corrected and log2 transformed) and quantile normalized using the Robust Multi‐Array Averaging (RMA) algorithm. For datasets GSE84437 and GSE62254 (ACRG), their original study‐defined clinical, histopathological, and survival data and molecular subtypes were downloaded and used for correlation analysis.

2.12. Cell Lines and Culture

Human gastric cancer cells AGS and HGC‐27 were purchased from the American Type Culture Collection (ATCC, USA). Cells were cultured in RPMI 1640 medium (Gibco, USA). THP‐1 cells were induced to differentiate into M0 cells by treatment with 100 ng/mL phorbol 12‐myristate 13‐acetate (PMA; Sigma‐Aldrich, USA) for 48 h. M0 cells were polarized into M2 cells using 20 ng/mL IL‐4 and IL‐13.

2.13. Transwell Migration Assay

Cells in the logarithmic growth phase were harvested and seeded into the upper chamber of a transwell insert at a density of 1.5 × 10⁴ cells per well. Cells were treated with 10 μg/mL C1q. After 18 h of incubation, the culture medium was aspirated and discarded. Cells remaining in the upper chamber were then: Washed three times with phosphate‐buffered saline; Fixed with 4% paraformaldehyde for 10 min; Stained with crystal violet for 10 min. Finally, migrated cells on the lower membrane surface were counted and imaged.

2.14. ELISA Assay for C1q in Human Plasma

Collect the peripheral blood of the patient in a collection tube, then centrifuge at 2000 rpm for 15 min at 4°C, extract the supernatant, and store it in a −80 degree refrigerator. The concentration of C1q was detected using the Human Anti‐C1q Antibody (C1q) ELISA Kit (YIFEIXUE Biotechnology, Nanjing, China).

2.15. Immunohistochemical Staining

The formalin‐fixed paraffin‐embedded gastric cancer tissue samples were obtained from patients who underwent radical gastrectomy for gastric cancer at the Zhongda Hospital Affiliated to Southeast University from April 2017 to April 2019. The pathological diagnosis was confirmed by at least two pathologists. For IHC, xylene and continuously diluted ethanol are used for dewaxing and dehydration of paraffin‐embedded tissue sections. The sections are incubated for 5 min at 121°C in a pressure sterilizer for antigen retrieval, then incubated with the primary antibody, including C1q (OM242378, OmnimAbs), at 4°C overnight, and the bound antibodies are detected using the 3,3′‐diaminobenzidine kit and hematoxylin. Cell quantification is performed using ImageJ software.

2.16. Lentiviral Transduction of THP‐1 Cells

Following The C1QA knockdown and control lines were generated by HanHeng Biotechnology Co. Ltd. C1QA (C1QA^KO) knockdown and control (C1QA^WT) lines were generated using lentiviral vectors. Stably expressing cells were screened for 2 weeks using a concentration of 10 μg/mL.

2.17. Statistical Analysis of Data

Survival differences were calculated using the Kaplan–Meier method and compared by log‐rank test. For data with non‐normally distributed continuous variables, the Wilcoxon rank sum test was used for analysis. All data analyses were done using R software (v4.3.1) and all statistical tests were two‐sided, with p < 0.05 considered statistically significant.

3. Results

3.1. Single‐Cell Sequencing Identifies 11 Major Cellular Subsets in GC

Fresh tumor and adjacent non‐tumor tissues were collected from three treatment‐naïve patients with non‐metastatic GC. After rapid enzymatic dissociation into single‐cell suspensions, we performed single‐cell transcriptome sequencing using droplet‐based microfluidic technology (10× Genomics) (Figure 1A). Following stringent quality control, 56,151 high‐quality cells were retained for analysis.

FIGURE 1.

Single‐cell atlas of gastric tumor and non‐tumor gastric tissue. (A) Flowchart of single‐cell sequencing data processing (n = 6). (B) UMAP plots of 24 major cell types from GC and normal tissues, with 31,442 cells from normal tissues and 24,709 cells from GC tissues. (C) Heatmap of correlations between 24 cell subclusters. (D) The proportion of 24 cell subtypes in 6 samples. (E) Violin plot of the main marker expression distribution for 24 cell subtypes. (F) UMAP plot of the expression of classic marker genes for 24 cell subtypes. (G) UMAP shows 11 cell types of 56,151 cells. *p < 0.05, **p < 0.01, ***p < 0.001.

Gene expression data were normalized and subjected to principal component analysis (PCA), followed by graph‐based clustering, which partitioned the cells into 24 transcriptionally distinct clusters (Figure 1B). To assess cluster relationships, we performed inter‐cluster correlation analysis based on the top differentially expressed genes (DEGs) (Figure 1C). Cellular composition across clusters varied significantly between samples (Figure 1D). Based on canonical marker gene expression patterns (Figure 1E,F), we annotated and consolidated the 24 clusters into 11 major cell types (Figure 1G), defined by their characteristic surface markers: B cells (CD79A, MS4A1), T cells (CD3D, CD3E), Natural killer (NK) cells (NCR1, NKG7), Epithelial cells (PGC, PGA3), Endothelial cells (RAMP2, PECAM1), Neutrophils (CSF3R, S100A8), Smooth muscle cells (ACTA2, TAGLN), Macrophages (MRC1, CD68), and Monocytes (CD14, FCN1). Tissue stem cells (CD44, TROP2), CMP (CD34, CD117).

3.2. Heterogeneity of Myeloid Cells in the GC Microenvironment

Single‐cell transcriptomic analysis identified 16 myeloid cell subpopulations (Figure 2A). Macrophages (Clusters 0, 1, 5, 6, 8, and 11) were characterized by high expression of CD86, CD163, and MRC1. Monocytes (Clusters 2, 3, and 4) specifically expressed FCN1 and S100A8. Dendritic cells (DCs: Clusters 7 and 9) showed low CD14 and high HLA‐DR expression. The remaining 5 clusters were unassigned due to low quality or doublet status. Significant differences in subpopulation distribution were observed between tumor and adjacent tissues (Figure 2B). Further monocyte analysis revealed that CD14⁺/CD16⁺ populations formed two major pan‐cancer clusters, indicating microenvironment‐independent differentiation (Figure 2C–E). While all monocytes expressed canonical markers (CD14, CD16, and FCN1), tumor‐infiltrating subsets upregulated HLA‐DR/HLA‐DRB1 and macrophage genes (CD86 and CD163), confirming tissue‐macrophage differentiation (Figure S1, Table S6). Tumor‐associated monocytes specifically overexpressed inflammatory factors (IL1B, CCL4, CXCL2, and CXCR4), tissue‐residency markers (NR4A1, NR4A2, and NLRP3), growth factors (AREG and EGR1), and NF‐κB pathway genes (NFKB1 and NFKBIA). Among DCs, LAMP3⁺ cDC2 cells exhibited a migratory activated phenotype, with upregulated migration/maturation regulators but downregulated TLR signaling (Figure S1). This conserved feature across cancers is consistent with HCC reports [11], indicating their role as mature DCs. Macrophages demonstrated significant heterogeneity [18]. Specifically, the complement gene‐high Mφ_4 subset displayed tissue‐resident properties and pro‐tumoral M2‐like characteristics, with signature genes significantly enriched in TCGA pan‐cancer data (Figure 2F,G).

FIGURE 2.

Single‐cell sequencing characterization of tumor‐infiltrating myeloid cells in GC. (A) UMAP of 16 myeloid cell subsets. (B) Proportion of myeloid cell subsets of tumors and normal tissues. (C) Characteristic gene expression bubble map of myeloid cell subsets. (D) Distribution of characteristic gene expression. (E) Differential gene analyses across subpopulations. (F, G) Levels of Mac_4 infiltration in TCGA STDA and expression levels of characterized genes. *p < 0.05, **p < 0.01, ***p < 0.001.

3.3. Heterogeneous Transcriptional Features of Tumor‐Associated Macrophages in GC

Re‐clustering of GC macrophages identified six functionally distinct subsets (Figure 3A): NRP1⁺ TAMs (Mac_1), FOS⁺ TAMs (Mac_3), MT‐CO3⁺ TAMs (Mac_2), MRC1⁺ TAMs (Mac_5), HLA‐DPA1⁺ TAMs (Mac_4), and C1QA⁺ TAMs (Mac_6). UMAP visualization revealed subset‐specific gene expression patterns (Figure 3B). Specifically, C1QA⁺ TAMs exhibited high expression of M2‐polarization markers and immunosuppressive genes (HAVCR2, SIRPA, LAIR1), along with enhanced phagocytic activity; NRP1⁺ TAMs were enriched for angiogenesis‐related genes (CD44, ITGAV); and HLA‐DPA1⁺ TAMs co‐expressed M2 genes, immune checkpoints, the cancer stem cell core protein FN1 (which maintains proliferative and metastatic capacities) [19], and immunosuppressive genes (SPP1, LAIR1, etc.) (Figure 3C,D).

FIGURE 3.

Molecular characterization of tumor‐associated macrophages in GC. (A) UMAP plot of six macrophage clusters. (B) UMAP dot of marker genes for different macrophage subpopulations. (C) Expression levels of M1‐like, M2‐like, angiogenic, phagocytic, and checkpoint signature genes in macrophages. (D) Expression levels of 6 gene signatures across macrophages cluster. (E) Trajectory analysis of macrophage subpopulations. *p < 0.05, **p < 0.01, ***p < 0.001.

Differential expression analysis (|log₂FC| > 2, FDR < 0.05) showed significant upregulation of 1089 genes in NRP1⁺ TAMs (Figure S2A). GSVA pathway enrichment demonstrated activation of ECM‐receptor interaction and PPARG signaling in NRP1⁺ TAMs, and upregulated MHC class II antigen presentation in HLA‐DPA1⁺ TAMs (Figure S2B,C). Prognostic analysis using TCGA‐STAD data confirmed that high expression of signature genes from both NRP1⁺ and HLA‐DPA1⁺ TAMs predicted poorer survival (Figure S2D). Pseudotime trajectory analysis suggested that NRP1⁺ and C1QA⁺ TAMs likely originate from MRC1⁺ TAMs (Figure 3E), indicating macrophage reprogramming toward pro‐tumorigenic phenotypes within the tumor microenvironment. This functional heterogeneity extends beyond classical M1/M2 classification paradigms. *p < 0.05, **p < 0.01, ***p < 0.001.

3.4. Characterization of C1QA ⁺ TAM‐Mediated Cell Communication Networks

As previously observed, C1QA⁺ TAMs were predominantly expressed in M2 macrophages and were also detected in other cancers, including BRCA, KIRC, and NSCLC, based on scRNA‐seq data (Figure S3A,B). The increased abundance of C1QA⁺ TAMs in immunotherapy‐treated KIRC suggests their potential role in cancer immunotherapy response. Ligand‐receptor interaction analysis using CellChat revealed significantly enhanced intercellular communication within the GC microenvironment compared to adjacent normal tissues (Figure 4A), with 2370 interacting pairs detected in tumor tissues versus only 1469 in paracancerous tissues (Figure 4B,C). Functioning as a central hub, C1QA⁺ TAMs established extensive connections with fibroblasts, stromal cells, endothelial cells, and immune cells, particularly Tregs and exhausted CD8⁺ T cells, possibly dynamically transmitting pro‐tumorigenic signals through ligand‐receptor pairs such as CXCL12‐CXCR4, ANXA1‐FPR1, and CCL3‐CCR1 (Figure 4D).

FIGURE 4.

Inference of cellular communication from GC single‐cell transcriptomes suggests a central role for C1QA⁺ TAMs. (A) Circle diagram showing the number of interactions between various immune cells in the tumor. (B) Analysis of ligand‐receptor numbers between tumor and normal tissue cells. (C) The bar graph represents the number of ligand‐receptors in GC tumor and normal samples. (D) Demonstration of cellular communication patterns. (E) Dot plots show predicted ligand‐receptor interactions between C1QA⁺ TAMs and various immune cells. (F, G) Analysis of ligand‐receptor differences in the FN1 and GALECTIN signaling pathways in tumor and normal samples. *p < 0.05, **p < 0.01, ***p < 0.001.

Further analysis of communication pattern clustering (Figure 4E) demonstrated high activity of C1QA⁺ TAMs in pro‐tumorigenic pathways, including SPP1, FN1, and GALECTIN signaling (Figure 4F,G). Notably, elevated SPP1 and FN1 expression not only promoted tumor angiogenesis but also significantly correlated with poor patient prognosis (Figure S4A,B). Collectively, this interaction network suggests that tumor cells may recruit macrophages to collaboratively establish an immunosuppressive microenvironment. Of course, CellChat, based on transcriptome prediction, cannot display the actual protein levels or functional intercellular communication. It does provide a predictive model, rather than direct spatial or functional evidence of protein‐level interactions. More discoveries should be verified spatially using multiple immunofluorescence or spatial transcriptomics to be more convincing. *p < 0.05, **p < 0.01, ***p < 0.001.

3.5. Prognostic Role of C1QA ⁺ TAMs in GC

To clarify the functional role of C1QA⁺ tumor‐associated macrophages (TAMs), we used UMAP plots to visualize the distribution of characteristic genes in GC (Figure 5A). This analysis revealed that C1QA⁺ TAMs highly express complement genes (C1QA, C1QB, C1QC) and immunosuppressive genes (MRC1, NRP1, MARCO, TREM2) (Table S6). Furthermore, C1QA⁺ TAMs were significantly more abundant in tumor tissues compared to non‐adjacent normal tissues (Figure 5B). We also evaluated the proportion of C1QA⁺ TAMs in each sample (T1: 5.59%, T2: 5.49%, T3: 12.31%, C1: 5.67%, C2: 4.58%, C3: 1.27%). KEGG pathway enrichment analysis indicated pro‐tumorigenic characteristics of C1QA⁺ TAMs, with enrichment in the VEGF signaling pathway and energy metabolism‐related pathways, such as oxidative phosphorylation and the PPAR signaling pathway (Figure 5C). This metabolic profile is consistent with the known properties of M2 macrophages, which primarily depend on oxidative phosphorylation for energy production.

FIGURE 5.

Impact of C1QA⁺ TAMs signature genes in survival prognosis of GC. (A) Distribution of C1QA⁺ TAMs signature gene expression in GC. (B) Abundance of C1QA⁺ TAMs in GC and normal tissues. (C) KEGG pathway enrichment of highly expressed genes in C1QA⁺ TAMs. (D, E) Kaplan–Meier curve of TCGA (Up) and GEO (Down) GC patients with high or low C1QA⁺ TAMs. (F, G) C1QA⁺ TAMs and clinical characteristics in TCGA (Up) and GEO (Down) with univariate, multivariate cox risk analysis. *p < 0.05, **p < 0.01, ***p < 0.001.

Analysis of C1QA⁺ TAM signature gene expression in TCGA‐STAD and GEO GC transcriptomic datasets confirmed that higher abundance of these macrophages correlates with poorer overall survival in GC patients, consistent with previous reports. Concurrently, ROC curves based on this signature demonstrated good predictive efficacy for 1‐year, 3‐year, and 5‐year survival rates (Figure 5D,E). Further univariate and multivariate cox regression analysis identified both C1QA⁺ TAM abundance and tumor grade as independent risk factors for GC (Figure 5F,G). Thus, these findings collectively suggest that C1QA⁺ TAMs significantly promote malignant progression in GC. *p < 0.05, **p < 0.01, ***p < 0.001.

3.6. C1q Represents a Promising Novel Biomarker for the Clinical Diagnosis of GC

Bioinformatic analysis of The Cancer Genome Atlas (TCGA) stomach adenocarcinoma (STAD) cohort revealed significant upregulation of C1QA mRNA in tumor tissues compared with adjacent non‐tumor tissues (Figure 6A,B). RT‐qPCR and transcriptome sequencing also indicated that the expression of C1QA was significantly higher in tumor tissues compared to adjacent non‐tumor tissues (Figure 6C,D). This finding was further validated in an independent, prospectively collected tissue cohort from Wuwei Tumor Hospital, where C1q protein expression was also significantly elevated in GC specimens relative to matched adjacent non‐tumor tissues (Figure 6E).

FIGURE 6.

C1q represents a promising novel biomarker for the clinical diagnosis of GC. (A) Expression of C1QA mRNA in GC (TCGA). (B) The expression of the C1QA gene in paired GC tumor was higher than that in the adjacent non‐tumor tissues (TCGA). (C) The expression of C1QA in tumor and adjacent non‐tumor tissues of patients from Wuwei Tumor Hospital was analyzed by RT‐qPCR (n = 34). (D) Transcriptome sequencing indicates that C1QA is highly expressed in GC tissues (n = 9). (E) The expression concentrations of C1q in paired adjacent non‐tumor tissues. (F) The expression of C1q in serum of healthy control (normal, n = 60), atrophic gastritis (n = 60), EGC patients (n = 62) and advance GC (n = 58). (G) The ROC curve is used to analyze the predictive power of C1q for EGC. Cut off, 1958.788 ng/mL. Sensitivity, 66. Specificity, 96.7. (H) The expression of C1q is significantly correlated with the tumor size. (I) Validation of the expression of serum C1q in healthy controls (n = 32) and in patients with early gastric cancer from the Zhongda Hospital cohort (n = 32). (J) ROC curve analysis was performed to assess the predictive power of C1q in EGC patients from the Zhongda hospital cohort. *p < 0.05, **p < 0.01, ***p < 0.001.

To assess the diagnostic utility of C1q, we quantified circulating C1q in plasma and found markedly elevated concentrations in GC patients compared to healthy controls (Figure 6F). To evaluate the diagnostic accuracy of C1q in EGC, we conducted a receiver operating characteristic (ROC) analysis, which showed an area under the curve (AUC) of 0.8735 (Figure 6G), indicating a strong ability to distinguish between EGC cases and controls. Furthermore, plasma C1q levels showed a statistically significant positive correlation with primary tumor size (Figure 6H). Furthermore, in the test set, the serum samples were collected from Zhongda Hospital, which was consistent with our findings in the discovery set. The serum C1q expression in EGC patients was significantly higher (Figure 6I), and the ROC curve showed an AUC of 0.8486 (Figure 6J). These results collectively indicate that serum C1q is a potential diagnostic marker for GC.*p < 0.05, **p < 0.01, ***p < 0.001.

3.7. C1q Is Positively Correlated With CD163 ⁺ M2 Macrophages in GC

To further validate the relationship between C1q and macrophage polarization, we observed a significant correlation between C1q expression and polarization markers (Figure 7A,B). Furthermore, C1QA expression was associated with poorer prognosis in gastric cancer patients (Figure 7C). To verify the relationship between C1q and macrophages in GC, we used IHC to examine the relationship between C1q and CD163 in our cohort (Table S3), finding that C1q was expressed in tumor‐infiltrating CD163⁺ macrophages (Figure 7D). Additionally, dividing C1q and CD163 expression into high and low groups revealed significant differences in overall survival (OS) and disease‐free survival (DFS) when considering only the group with high C1q expression and high CD163 density (Figure 7E–G). Multivariate Cox analysis showed that both C1q and CD163 were high‐risk factors for GC prognosis (Figure S5). In summary, analysis of clinical gastric cancer tissues revealed that C1q is expressed in tumor‐infiltrating M2 macrophages, and its expression correlates with the abundance of CD163⁺ macrophages. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 7.

C1q is positively correlated with M2 macrophage polarization in GC. (A, B) Analysis of the correlation between C1QA and M2 markers based on the TCGA database. (C) High expression of C1QA is associated with a poorer prognosis in GC patients, while the patients with high expression of both C1QA and CD163 have even worse survival. (D) Immunohistochemistry staining of C1q and CD163 in GC tissues, n = 100. (E–G) The OS and DFS of GC patients with high expression of C1q and CD163 in our cohort was significantly lower. *p < 0.05, **p < 0.01, ***p < 0.001.

3.8. The C1q Is Associated With the Tumor‐Promoting Function of M2 Macrophages

Bioinformatic analyses indicated that elevated expression of C1QA⁺ TAMs correlates with a poorer prognosis in various cancers. Multiple immunofluorescence also indicated co‐expression of C1q and CD163 in tumor‐infiltrating macrophages (Figure 8A). Given that C1q may be involved in the polarization of M2 macrophages, we confirmed the role of C1q in M2 immunosuppression through in vitro experiments. THP‐1 cells were differentiated into M0 macrophages (THP‐1‐PMA cells) for 48 h in medium containing 100 ng/mL phorbol‐12‐myristate‐13‐acetate (PMA), followed by treatment with 20 ng/mL IL‐4 and IL‐13 for 24 h to promote M2 polarization. RT‐qPCR results showed that C1QA, CD206, and TGF‐β were highly expressed in M2 macrophages (Figure 8B). Furthermore, compared with treatment using M0 macrophage‐conditioned medium (CM), treatment using M2 macrophage‐conditioned medium enhanced the migratory capacity of GC cells (Figure 8C). Further, we pre‐treated THP‐1‐induced M0 macrophages with C1q. Subsequent RT‐qPCR analysis confirmed that C1q promotes macrophage polarization toward the M2 phenotype (Figure 8D). Furthermore, M0 macrophages treated with C1q enhanced the migration of GC cells (Figure 8E). Furthermore, we constructed C1QA‐knockdown THP‐1 cells (referred to as C1QA^KO) using lentivirus. Compared with the control group (C1QA^WT), the conditioned medium from C1QA^KO M2 significantly reduced the migration ability of GC cells (Figure S6). These results indicate that C1q treatment also significantly enhanced the tumor‐promoting capacity of M2 macrophages. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 8.

The C1q is associated with the tumor‐promoting function of M2 macrophages. (A) Representative images of patients with multiplex fluorescence immunohistochemistry (IHC) showed a high expression of C1q on CD163⁺ M2 macrophages in tumor stroma. (B) RT‐qPCR results indicated that C1QA, CD206, and TGF‐β were highly expressed in M2 macrophages. (C) The migration of GC cells was increased when treated with conditioned medium (CM) from M2 macrophages compared with treatment with M0 CM. (D) After treating THP‐derived M0 cells with C1q (10 ng/mL), the expressions of CD206 and TGF‐β significantly increased. (E) After treating M0 macrophages with C1q, it significantly promoted the migration of GC cells. *p < 0.05, **p < 0.01, ***p < 0.001.

4. Discussion

GC is a highly heterogeneous malignancy influenced by both genetic and environmental factors, posing significant challenges for diagnosis and treatment. Emerging single‐cell technologies have become powerful tools for resolving tumor heterogeneity, enabling in‐depth exploration of key biological processes including tumor progression stratification, immune tolerance, intercellular interactions, and tumor infiltration. These advances deepen our understanding of GC biology. This study comprehensively delineates the GC tumor landscape at single‐cell resolution [20, 21, 22].

Our analysis revealed pronounced heterogeneity within the GC tumor microenvironment (TME), particularly among myeloid cell populations. We provided detailed characterization of conventional dendritic cell (cDC) heterogeneity. Dendritic cells (DCs) are classically categorized into conventional (cDCs) and plasmacytoid (pDCs) subtypes, with cDCs further differentiating into CD141⁺ and CD1C⁺ subsets [23].

LAMP3⁺ cDCs, recognized markers of DC maturation typically induced by CD40L stimulation, have been reported to originate from both cDCs and pDCs [24, 25, 26].

We identified and annotated a mature LAMP3⁺ CCR7⁺ DC subset that elucidates myeloid cell functions in GC pathogenesis. These LAMP3⁺ DCs highly express chemokines (CCL19, CCL21) and co‐stimulatory ligands (CCL17, CCL22), recruiting CCR4‐expressing regulatory T cells (Tregs) to tumors and fostering an immunosuppressive microenvironment [27, 28]. Furthermore, tumor abundance of cDC‐derived LAMP3⁺ DCs correlated significantly with infiltration levels of exhausted CD8⁺ T cells and Tregs [11, 29]. Collectively, the LAMP3⁺ DC subset identified in GC mirrors populations observed across multiple cancers and likely participates in regulating T cell infiltration and functional exhaustion.

Tumor‐associated macrophages (TAMs) constitute a key TME component [30]. While historically simplified as pro‐inflammatory (M1) or anti‐inflammatory/pro‐tumorigenic (M2) phenotypes based on murine models, this dichotomy inadequately captures TAMs' complexity in human tumors [31, 32, 33]. Our scRNA‐seq analysis identified six TAMs subsets predominantly enriched in tumor tissue. Among these, NRP1⁺ TAMs and C1QA⁺ TAMs exhibited stronger M2‐like polarization. NRP1⁺ TAMs highly expressed pro‐angiogenic signature genes, and their enrichment in tumor‐promoting signaling pathways suggests roles in cancer progression and metastasis.

C1QA⁺ TAMs represent a tissue‐resident macrophage subset [34, 35] defined by core genes (C1Qs, APOE, TREM2, MRC1). C1q proteins enhance phagocytic function while suppressing pro‐inflammatory cytokine production [36, 37, 38], thereby exerting pro‐tumor effects. We further discovered that C1QA⁺ TAMs exhibit distinct lipid metabolic signatures, with established links to maintaining immunosuppressive capacity through fatty acid metabolism regulation [39, 40]. Cell communication analysis revealed extensive interactions between C1QA⁺ TAMs and multiple exhausted CD8⁺ T cell/Treg populations, underscoring their role in shaping the immunosuppressive TME. Of course, because CellChat predicts cell‐to‐cell communication based on mRNA expression levels, it lacks direct validation at the spatial or protein level. Future studies employing methods such as multiplex immunofluorescence or spatial transcriptomics will be crucial for verifying these predicted signaling networks.

We used multiple immunohistochemistry to prove that C1q is highly expressed in CD163⁺ macrophages. Experiments conducted in vitro have shown that C1q is associated with M2 macrophage polarization and enhances the malignant migration of GC cells. Of course, we still need more experiments (such as the recovery experiments after C1QA knockdown and the mouse gastric cancer model) to verify the mechanism of C1q's role in macrophage polarization. This will help clarify the specific functional role of C1q in the M2‐type polarization and the migration of GC cells mediated by macrophages. In our gastric cancer cohort, high expression of C1q was associated with a poorer prognosis. C1q was significantly elevated in tumor tissues and plasma, and the C1q present in plasma may be a potential biomarker for the diagnosis of GC. However, we should compare the diagnostic performance of C1q with traditional serum tumor markers (such as CEA, CA19‐9, and CA72‐4). At the same time, we should also investigate whether combining C1q with these markers can enhance diagnostic performance, thereby more fully demonstrating the clinical advantages of C1q. This is a deficiency in our current research, and we will optimize our research plan in future investigations. In summary, the graphical summary of this study is shown in Figure 9.

FIGURE 9.

Graphical table of contents. A graphical overview of the value of C1q in the diagnosis of GC based on single‐cell sequencing and GC cohort analysis.

Author Contributions

Saisai Gong: writing – original draft, writing – review and editing. Yupeng Zhao: conceptualization. Sheng Yang: methodology. Tong Liu: validation. Chengyun Li: visualization. Yun Zhou: formal analysis. Geyu Liang: writing – review and editing, funding acquisition.

Funding

The present study was supported by the National Natural Science Foundation of China (81972998), Postgraduate Research & Practice Innovation Program of Jiangsu Province (KYCX24_0483), Gansu Province Health Care Industry Research Program Projects (GSWSKY2024‐09).

Disclosure

The authors have nothing to report.

Ethics Statement

This study involves human subjects. The research design has been approved by the Ethics Committee of Zhongda Hospital Affiliated to Southeast University, with the approval number 2024ZDSYLL312‐P01. All human tumor tissue samples were collected in accordance with national and institutional ethical guidelines, and all participants signed the informed consent form before participating in the study.

Consent

Informed consent was obtained from all participants.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Violin plots depicting the expression profiles of signature genes across distinct cell populations. (A, B) Violin maps showing genes expressed differently between monocytes. (C) Gene set scoring in cDC. (D) Violin maps of differential gene expression in cDC.

Figure S2: Functional characteristics of macrophage subpopulations. (A) Volcano maps of NRP1⁺ TAMs and HLA‐DPA1⁺ TAMs differential genes. (B) Gene set variation analysis (GSVA) identified the differentially expressed signaling pathways enriched in NRP1⁺ TAMs and HLA‐DPA1⁺ TAMs GSVA scores for differential genes. (C) Gene‐set enrichment analysis was performed on gene sets of ECM receptor interaction. Positive NES indicate higher expression in NRP1⁺ TAMs. (D) Survival curves of two types of macrophages.

Figure S3: The functional characteristics of C1QA⁺ TAMs in various cancer tissues. (A) The presence of C1QA⁺TAMs was demonstrated by t‐distributed random neighbor embedding (t‐SNE) in stomach adenocarcinoma (STAD), acute myeloid leukemia (AML), non‐small cell lung cancer (NSCLC), colorectal cancer (CRC), liver hepatocellular carcinoma (LIHC), breast cancer (BRCA), and renal clear cell carcinoma (KIRC). (B) Gene expression of C1QA, C1QB and C1QC in different cellular subpopulations of STAD was analyzed using the Tumor ImmunoSingle Cell Centre (TISCH) database (http://tisch.comp‐genomics.org).

Figure S4: Survival analysis of SPP1 and FN1 expression across pan‐cancer cohorts. (A) Survival curves of SPP1 and FN1 (B) in various cancers.

Figure S5: Multivariate Cox analysis showed that both C1q and CD163 were high‐risk factors for gastric cancer prognosis.

Figure S6: The effect of C1QA knockdown on the migration ability of GC cells.

Table S1: Clinical information of patients undergoing single‐cell sequencing.

Table S2: Summary of the Clinical Features of Gastric Cancer Tissue.

Table S3: The correlation of C1q and CD163 Expression with Clinicopathological Factors.

Table S4: Summary of the Clinical Features of Gastric Cancer plasma.

Table S5: Cell quality assessment and filtration statistics table.

Table S6: Marker genes of myeloid cell populations.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.