Ethyl caffeate reprograms tumor-associated macrophages to enhance CD8⁺ T-cell immunity in colorectal cancer

Abstract

Background

The colorectal cancer (CRC) tumor microenvironment contains diverse myeloid populations that critically regulate antitumor immunity and therapeutic response. Tumor-associated macrophages are key regulators of macrophage–T-cell crosstalk within the CRC microenvironment. This study aimed to investigate the antitumor effects of Ethyl caffeate (EC) and determine whether EC modulates macrophage-associated immune remodeling in CRC.

Methods

Public single-cell RNA sequencing data from CRC tissues were analyzed to characterize myeloid infiltration, macrophage heterogeneity, Cell-cell communication, pseudotime trajectories, and macrophage-associated prognostic signatures. RAW264.7 cells and bone marrow-derived macrophages (BMDMs) were used to evaluate the effects of EC on macrophage viability and polarization in vitro. An immunogenic syngeneic MC38 CRC model was established in C57BL/6 mice to assess the antitumor activity of EC. Flow cytometry was used to quantify tumor infiltrating macrophages, CD86⁺ M1-like macrophages, CD8⁺ T cells, Granzyme B expression, and PD-1 expression. Clodronate liposome-mediated macrophage depletion was performed to determine whether the antitumor effect of EC was macrophage dependent.

Results

Single-cell analysis revealed extensive myeloid infiltration and marked TAM heterogeneity in CRC. State-specific analyses identified CXCL9/CXCL10 IFN-responsive TAMs as a transitional immune activating population closely associated with CXCL10-CD8⁺ T-cell immune signatures. Macrophage-derived gene signatures were associated with overall survival in the TCGA-COAD/READ cohort. EC showed limited cytotoxicity at 30 and 60 µM and promoted macrophage polarization toward an M1-like antitumor phenotype in RAW264.7 cells and BMDMs. In the immunogenic MC38 syngeneic CRC model, EC treatment significantly suppressed tumor growth, reducing tumor volume by 31.3% ± 10.0% at 25 mg/kg and 46.6% ± 13.5% at 50 mg/kg. EC increased CD86⁺ inflammatory macrophage-associated activation, enhanced CD8⁺ T-cell infiltration, increased Granzyme B expression, and reduced PD-1 expression on tumor infiltrating CD8⁺ T cells, suggesting enhanced CD8⁺ T-cell cytotoxic activation. Macrophage depletion largely attenuated the antitumor effect of EC.

Conclusions

EC suppresses colorectal tumor growth at least partly through macrophage-dependent immune remodeling. These findings identify EC as a potential macrophage-reprogramming compound that enhances CD8⁺ T-cell antitumor immunity in CRC.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12885-026-16469-2.

Introduction

Colorectal cancer (CRC) is a leading cause of cancer-related morbidity and mortality worldwide [1]. Based on microsatellite status, CRC is broadly classified into microsatellite-stable (MSS) and microsatellite instability-high (MSI-H) subtypes, with MSS CRC accounting for the majority of cases and showing limited responsiveness to immune checkpoint blockade (ICB) [2]. The immunosuppressive tumor microenvironment remains a major barrier to effective immunotherapy in CRC. Among the various immune cells in the tumor microenvironment (TME), macrophages are typically among the most abundant immune cell populations in solid tumors, where they support stromal structures, remodel the vasculature, and regulate local immune responses [3]. These tumor infiltrating macrophages are collectively referred to as tumor-associated macrophages (TAMs) [4, 5]. However, TAMs do not represent a single cell type; rather, they constitute a dynamic and highly heterogeneous cell population whose composition and functional states continuously evolve during tumor initiation, progression, metastasis, and therapeutic response [6]. Researchers typically categorize these cells into two subsets: classical M1-like macrophages and alternative M2-like macrophages [7]. Classically activated M1-like macrophages exert antitumor effects by killing tumor cells through nitric oxide (NO) and reactive oxygen species (ROS) production, secreting pro-inflammatory cytokines, and mediating antigen presentation [6]. M2-like macrophages are both a key participant in tumor development and progression, as well as an important barrier to the effectiveness of cancer treatments such as immune checkpoint blockade and Chimeric antigen receptor T-cell therapy (CAR-T) [8].

Ethyl caffeate (EC), a naturally occurring Phenolic Compound found in Bidens pilosa, has attracted attention for its diverse pharmacological activities, including anticancer, anti-inflammatory, antifungal, antiviral, and anti-fibrotic effects [9–14]. However, whether EC exerts antitumor effects in CRC by modulating the tumor immune microenvironment remains unclear.

In this study, we combined public single-cell analysis, macrophage assays, and an immunogenic MC38 syngeneic CRC model to investigate whether EC modulates macrophage-associated immune remodeling and CD8⁺ T-cell activation.

Methods

Cell lines

MC38 cells (Immocell Biotechnology, IM-M006) were cultured in DMEM supplemented with 10% fetal bovine serum (FBS). RAW264.7 cells (Procell Biotechnology, CL-0190) were cultured in RAW264.7 complete medium according to the manufacturer’s instructions. All cells were maintained at 37 °C in a humidified incubator with 5% CO₂. Cell lines were authenticated by short tandem repeat (STR) profiling and confirmed to be free of mycoplasma contamination. Bone marrow-derived macrophages (BMDMs) were isolated from C57BL/6 mice and cultured in macrophage medium supplemented with Macrophage Colony Stimulating Factor (M-CSF, UA BIOSCIENCE, UA040128) at a concentration of 50 ng/ml for 7 days. The adherent cells were then collected and used as BMDMs.

Cell viability assay

Cells (5 × 10³ cells per well) were seeded into 96-well plates, incubated with various doses of EC for 24 h, followed by addition of Cell Counting Kit-8 (CCK-8, SparkJade, CT0001-A) (10 µL/well) and incubation for 1 h at 37℃, after which optical density was measured to assess cell viability. The absorbance was measured at 450 nm using a microplate reader, and cell viability was calculated relative to the vehicle-treated control.

Macrophage polarization

For macrophage polarization, RAW 264.7 cells were seeded in 6-well plates at 5 × 10⁵ cells per well and treated for 24 h with DMSO (0.1% v/v), lipopolysaccharide (LPS, 100 ng/mL), 30 µM EC, or 60 µM EC, LPS + 30 µM EC, or LPS + 60 µM EC. For BMDMs, cells were seeded in 6-well plates at 5 × 10⁴ cells per well on day 7 of differentiation and then treated under the same conditions for 24 h.

Natural compound library and EC

A macrophage-based compound screening was performed using 73 compounds selected from a natural compound library (L2000, Selleckchem, Houston, TX, USA). RAW264.7 cells were used for the primary screening because of their reproducibility and suitability for macrophage-based compound evaluation. Cells were treated with each compound at a final concentration of 10 µM for 24 h. The final DMSO concentration was kept identical across all groups and did not exceed 0.1% (v/v). Vehicle-treated cells were used as negative controls, and LPS-treated cells were used as positive controls for macrophage activation.

Candidate hits were defined as compounds that showed no obvious cytotoxicity in RAW264.7 cells and were able to induce CD86 expression. Compounds meeting these criteria in RAW264.7 cells were further validated in mouse BMDMs. Validated hits were required to show no detectable cytotoxicity in both RAW264.7 cells and BMDMs while reproducibly increasing CD86 expression.

For animal experiments, EC was formulated in a vehicle consisting of 5% DMSO, 40% PEG300, 5% Tween-80, and 50% sterile water. Immunogenic MC38 syngeneic CRC model was treated with EC at 25 or 50 mg/kg by intraperitoneal injection once daily for 7 consecutive days, with an injection volume of 100 µL per mouse. Mice in the control group received the same volume of vehicle alone using the same administration schedule.

Animal models

Six- to eight-week-old male C57BL/6 mice were purchased from Lanzhou Veterinary Research Institute, Chinese Academy of Agricultural Sciences (Lanzhou, China). The male C57BL/6 mice were used in this initial proof-of-concept study to reduce sex hormone cycle-related variability and to maintain consistency with our preliminary experiments. Mice were maintained under specific pathogen-free conditions in plastic cages with free access to standard pellet food and water at 20 ± 2 °C under a 12 h light/dark cycle. The subcutaneous CRC models were used to evaluate the antitumor effect of EC in vivo. A total of 5 × 10⁵ MC38 cells in 100 µL of a mixture (40 µL PBS and 60 µL Matrigengel (ABW, 082704)) were subcutaneously injected into the dorsal flank of C57BL/6 mice. Mice were randomized into Control group (PBS, once a day, i.p.) and EC treatment groups (25 mg/kg or 50 mg/kg, i.p., once daily). Tumor volume was monitored after inoculation according to the following formula: volume = length × width² × 0.5.

To determine whether the antitumor effects of EC are macrophage-dependent, a macrophage depletion model was established using clodronate liposomes. Mice received an intraperitoneal injection of 150 µL of Clodronate liposomes (LipoCLO, Abisin, abs90559) or control PBS liposomes (LipoPBS, Abisin, abs90560) one day before tumor cell inoculation, with the same dose repeated every two days thereafter.

Preparation of tumor and splenic single-cell suspensions

On day 7 after EC treatment, tumor tissues were carefully collected and processed to evaluate the effects of EC on tumor growth and the TME. After euthanasia, tumors were immediately excised and placed in tissue storage solution (Abisin, abs9474) to maintain immune cell viability. The tumor tissues were then minced into approximately 1 mm³ fragments using sterile scissors and digested in PBS containing 0.5 mg/mL collagenase D (Sigma, 11088858001) and 0.1 mg/mL deoxyribonuclease type I (DNase I, Stemcell, 07900). The mixture was incubated at 37℃ with shaking for 30 min to facilitate tissue dissociation.

After enzymatic digestion, the resulting cell suspension was filtered through a 70-µm cell strainer (BD, 352350) to obtain a single-cell suspension. Cells were then counted, and 1 × 10⁶ cells were seeded into six-well plates and cultured at 37 ℃ in a humidified incubator with 5% CO₂ for subsequent experimental analyses.

Spleen tissues were processed using a similar procedure. Briefly, spleens were mechanically dissociated, and red blood cells were lysed with 1 mL of RBC lysis buffer (BD FACS™ Lysing Solution, 349202) at room temperature for 5 min. The lysis reaction was quenched by adding 5 mL of RPMI-1640 medium supplemented with 10% FBS to preserve cell viability. After filtration through a 70-µm cell strainer, the cell suspension was counted, and 1 × 10⁶ cells were seeded into six-well plates and cultured under the same conditions for further experimental evaluation.

Flow cytometry analysis

To evaluate macrophage polarization, single-cell suspensions were adjusted to a concentration of 1 × 10⁶ cells/mL and stained with fluorescently labeled antibodies against macrophage surface markers. Cells were incubated for 30 min at 4 °C in the dark, washed with PBS, and then analyzed by flow cytometry.

To evaluate lymphocyte function, single-cell suspensions were stimulated with Leukocyte Activation Cocktail (BD GolgiPlug™, 550583) for 4 h at 37 °C in a humidified incubator with 5% CO₂ to promote intracellular cytokine accumulation. After stimulation, cells were harvested and stained with fluorochrome-conjugated antibodies against surface markers for 30 min at 4 °C in the dark. Cells were then washed twice with PBS and processed for intracellular staining. For intracellular cytokine detection, cells were fixed and permeabilized using BD Cytofix/Cytoperm™ Fixation/Permeabilization Kit (BD, 554714) for 20 min at room temperature, followed by washing with Perm/Wash™ buffer (BD, 554723) Fluorochrome-conjugated antibodies against intracellular cytokines were then added, and cells were incubated for 30 min at room temperature in the dark. After washing, the stained cells were analyzed by flow cytometry. Details of the antibodies and reagents used for flow cytometry, including fluorochrome, catalog number and dilution are provided in Table 1.

Table 1.

Antibodies used in this study

Protein name	Product number
Zombie Yellow™Fixable Viability Kit	Biolegend 423,104
anti-mouse CD45-BV510	Biolegend 103,138
anti-mouse CD3-FITC	BD 553,061
anti-mouse CD4-APC-Cy7	BD 552,051
anti-mouse CD8-PerCP-Cy5.5	BD 551,162
anti-mouse PD-1-BV421	BD 562,584
anti-mouse Granzyme B-Alexa 647	Biolegend 515,406
Cell Activation Cocktail (with Brefeldin A)	Biolegend 423,304
Brilliant Stain Buffer	BD 566,349
CD11B-FITC	BD 557,396
F4/80-BV421	BD 565,411
CD86-APC	BD 558,703
CD206-PE	BD 568,273
BD Cytofix/Cytoperm™ Fixation/Permeabilization Kit	BD Biosciences
Purified Rat Anti-Mouse CD16/CD32 (Mouse BD Fc Block) (2.4G2)	BD Pharmingen 553,141

Statistical analysis

Data are presented as mean ± SD. Statistical analyses were performed using GraphPad Prism version 10.1.2. Comparisons between two groups were performed using an unpaired two-tailed Student’s T-test or Welch’s T-test when variances were unequal. For multiple-group comparisons, one-way ANOVA followed by Tukey’s multiple-comparison test was used. Tumor growth curves were analyzed using two-way repeated-measures ANOVA. When significant treatment effects or interaction effects were detected, post hoc multiple comparisons were performed with appropriate correction. Flow cytometry data were analyzed based on independent biological replicates. Kaplan–Meier survival curves were compared using the log-rank test, and Cox/LASSO-Cox regression analyses were performed for prognostic modeling. Differential expression analysis of single-cell RNA-seq data was performed using the Wilcoxon rank-sum test in Seurat. P < 0.05 was considered statistically significant.

Single-cell RNA-seq data processing and analysis

Public single-cell RNA-seq data for CRC were obtained from GEO dataset GSE132465. This reanalysis was based on the processed expression matrix and accompanying annotation files released by the original authors. According to the sample annotation, the dataset included 23 patients, comprising 23 tumor samples and 10 matched normal samples. The processed matrix was imported into Seurat v5 for downstream analysis. Additional quality control was performed by retaining cells with nFeature_RNA > 200, nFeature_RNA < 6000, and mitochondrial gene percentage (percent.mt) < 20%. Data were normalized, scaled, and visualized using principal component analysis and UMAP. Major cell types were annotated according to canonical marker genes and the original annotation information. Because the analysis was based on the processed matrix and curated annotations, no additional DoubletFinder/Scrublet-based doublet removal, global batch correction, or CNV-based malignant epithelial cell inference was performed in our main workflow. Epithelial cells were therefore classified as normal-derived or tumor-derived epithelial cells according to tissue source and available annotations, rather than being directly defined as malignant epithelial cells. Myeloid cells were extracted and reclustered to identify macrophages, monocytes, conventional dendritic cells, and plasmacytoid dendritic cells. Macrophages were first classified into M1-like and M2-like states using Seurat AddModuleScore based on established polarization-associated gene sets. To further resolve TAM heterogeneity, macrophages were reclustered and annotated into five TAM states according to representative marker genes: FCN1⁺ monocyte-like TAMs, CXCL9/CXCL10 IFN-responsive TAMs, C1QC/APOE TAMs, SPP1/TREM2 TAMs, and HLA-DRA APC-like TAMs. Cell-cell communication was inferred using CellChat based on normalized expression data and cell-type annotations. Pseudotime trajectory analysis of monocytes and macrophages was performed using monocle3, with monocyte-like cells set as the putative root state. To evaluate the clinical relevance of macrophage-associated gene signatures, bulk RNA-seq and survival data from TCGA-COAD/READ were analyzed using univariable Cox regression, LASSO-Cox regression, Kaplan–Meier survival analysis, and multivariate Cox regression.

Results

Single-cell analysis reveals extensive macrophage heterogeneity in CRC

After quality control, 63,689 cells from CRC tissues and matched normal tissues were retained for downstream analysis. Unsupervised clustering identified six major cell populations, including epithelial cells, T cells, myeloid cells, B cells, stromal cells, and mast cells (Fig. 1A). Compared with normal tissues, tumor tissues exhibited increased epithelial and myeloid cell infiltration, indicating substantial remodeling of the TME (Fig. 1B). Re-clustering of the myeloid compartment identified macrophages, monocytes, conventional dendritic cells (cDC2), and plasmacytoid dendritic cells (pDCs), with macrophages representing the dominant myeloid population (Fig. 1C–D). Using established polarization-associated gene signatures, macrophages were initially classified into M1-like and M2-like populations. M2-like macrophages were significantly enriched in tumor tissues, whereas M1-like macrophages were relatively enriched in normal tissues (Fig. 1E-F). However, the substantial overlap between M1-like and M2-like macrophages suggested that the conventional M1/M2 framework did not fully capture macrophage heterogeneity (Fig. 1G-I). Therefore, macrophage-specific re-clustering was performed and identified five transcriptionally distinct TAM states, including FCN1⁺ monocyte-like TAMs, CXCL9/CXCL10 IFN-responsive TAMs, C1QC/APOE TAMs, SPP1/TREM2 TAMs, and HLA-DRA APC-like TAMs (Fig. 1J). Marker-gene analysis confirmed the identity of each TAM state (Fig. 1K). Notably, tumor tissues were enriched in SPP1/TREM2 TAMs, HLA-DRA APC-like TAMs, and CXCL9/CXCL10 IFN-responsive TAMs, whereas C1QC/APOE TAMs predominated in normal tissues (Fig. 1L). These findings demonstrate substantial macrophage heterogeneity in CRC and support the use of state-specific TAM classification beyond the traditional M1/M2 paradigm.

Fig. 1.

A UMAP of major cell populations in CRC and adjacent normal tissues. B Cell-type composition in tumor and normal tissues. C-D Re-annotation of myeloid populations identifies macrophages, monocytes, cDC2, and pDCs. E-F Distribution of M1-like and M2-like macrophages in tumor and normal tissues. G-I UMAP visualization of conventional M1/M2 macrophage classification. J Macrophage-specific re-clustering identifies five TAM states. K Dot plot showing representative marker genes of each TAM state. L Relative abundance of TAM states in tumor and normal tissues

State-specific CellChat analysis reveals distinct communication programs among TAM subsets

CellChat analysis demonstrated extensive macrophage-centered communication within the CRC microenvironment, involving epithelial, stromal, and immune cell populations (Fig. 2A). Conventional M1-like macrophages exhibited broad interactions with epithelial cells and T cells through TNF, HLA-II, MIF, ICAM1, and SPP1 signaling pathways (Fig. 2B–C). Further analysis of TAM states revealed distinct communication patterns. CXCL9/CXCL10 IFN-responsive TAMs preferentially interacted with T-cell and NK-cell populations through HLA class I, CXCL16-CXCR6, ICAM1-SPN, and TNF-TNFRSF1B signaling pathways (Fig. 2D). C1QC/APOE TAMs displayed broad communication with immune, stromal, and epithelial cells (Fig. 2E), whereas HLA-DRA APC-like TAMs showed prominent interactions with CD4⁺ and CD8⁺ T cells (Fig. 2F). In contrast, SPP1/TREM2 TAMs primarily communicated with epithelial and stromal cells through SPP1- and integrin-associated pathways (Fig. 2G). Although conventional M1-like macrophages exhibited extensive communication with epithelial and T-cell populations, state-specific analysis resolved these broad interactions into distinct IFN-responsive, APC-like, C1QC/APOE, and SPP1/TREM2 TAM programs, highlighting the functional heterogeneity of CRC-associated macrophages beyond the traditional M1/M2 paradigm.

Fig. 2.

A Overall CellChat communication network. B M1-like macrophage - Epithelial cell communication. C M1-like macrophage - T-cell communication. D Communication network of CXCL9/CXCL10 IFN-responsive TAMs. E Communication network of C1QC/APOE TAMs. F Communication network of HLA-DRA APC-like TAMs. G Communication network of SPP1/TREM2 TAMs

TAM-state classification provides greater biological resolution than the conventional M1/M2 framework

Differential expression analysis identified distinct transcriptional programs between M1-like and M2-like macrophages. M1-like macrophages preferentially expressed inflammatory genes, including IL1B, IL1RN, PTGS2, TNFAIP6, and FCN1, whereas M2-like macrophages were enriched for APOE, APOC1, CCL18, FOLR2, and SLC40A1 (Fig. 3A). To evaluate their prognostic significance, LASSO Cox regression models were constructed based on M1-like and M2-like macrophage signatures (Fig. 3B, D). Kaplan-Meier analysis demonstrated that both M1 and M2 risk scores significantly stratified overall survival in CRC patients (both log-rank p < 0.0001; Fig. 3C, E). To further capture macrophage heterogeneity beyond the conventional M1/M2 framework, a TAM-state marker–based prognostic model was established using LASSO Cox regression (Fig. 3F). The resulting TAM-state risk score also significantly stratified patient survival (Fig. 3G). Correlation analysis revealed marked differences among TAM states. IFN-responsive TAM signatures showed the strongest positive correlations with CXCL10, CXCL9, IFNG, CD8A, and GZMB, whereas SPP1/TREM2 TAMs exhibited substantially weaker associations with these immune activation markers (Fig. 3H). C1QC/APOE TAMs displayed intermediate correlations.

Fig. 3.

A Differential gene expression analysis between M1-like and M2-like macrophages. B-E LASSO Cox regression and Kaplan–Meier survival analyses based on M1-like and M2-like macrophage signatures. F-G Construction and validation of a TAM-state–based prognostic model. H Correlation analysis of TAM-state signatures with CXCL9, CXCL10, IFNG, CD8A, and GZMB. I Multivariate Cox regression analysis of TAM-state signatures after adjustment for clinicopathological variables

Multivariate Cox regression analysis adjusted for age, sex, and tumor stage. Tumor stage remained the dominant prognostic factor, whereas individual TAM-state signatures were not independently associated with survival (Fig. 3I). Collectively, these findings indicate that although conventional M1/M2 signatures retain prognostic value, TAM-state classification provides greater biological resolution and identifies IFN-responsive TAMs as the macrophage population most closely associated with the CXCL10-CD8 immune axis in CRC.

Pseudotime analysis reveals a continuous monocyte-to-macrophage differentiation trajectory in CRC

To investigate macrophage-state dynamics in CRC, Monocle3 pseudotime analysis was first performed using the conventional M1/M2 classification framework. UMAP visualization revealed substantial overlaps between M1-like and M2-like macrophages, suggesting continuous transcriptional transitions rather than discrete polarization states (Fig. 4A). Trajectory analysis demonstrated a progressive differentiation process from monocytes to M1-like macrophages and subsequently to M2-like macrophages (Fig. 4B-D). Consistently, pseudotime distribution analysis showed enrichment of monocytes at early stages, M1-like macrophages at intermediate stages, and M2-like macrophages at later stages of the trajectory (Fig. 4E).

Fig. 4.

A UMAP visualization of macrophages classified by the conventional M1/M2 framework. B-D Monocle3 trajectory analysis showing the inferred monocyte-to-M1-like-to-M2-like macrophage continuum. E Pseudotime distribution of monocytes, M1-like macrophages, and M2-like macrophages. F Identification of five TAM states. G TAM-state trajectory. H Full TAM-state continuum. I TAM-state pseudotime violin

Dynamic gene-expression analysis further revealed distinct temporal patterns along pseudotime. Early-stage macrophages preferentially expressed inflammatory genes, including IL1B and TNF, whereas CXCL9, CXCL10, and CD86 were enriched during intermediate stages. In contrast, APOE, APOC1, C1QA, C1QB, C1QC, and CD74 progressively accumulated at later stages of the trajectory (Figure S1A-B). These findings suggested that the conventional M1/M2 framework captures only part of the macrophage differentiation continuum and does not adequately explain the emergence of multiple terminal TAM programs.

To achieve higher-resolution characterization of macrophage heterogeneity, TAM-state analysis was subsequently performed. Five transcriptionally distinct TAM populations were identified, including FCN1⁺ monocyte-like TAMs, CXCL9/CXCL10 IFN-responsive TAMs, HLA-DRA APC-like TAMs, SPP1/TREM2 TAMs, and C1QC/APOE TAMs (Fig. 4F). Pseudotime analysis revealed a continuous trajectory originating from FCN1⁺ monocyte-like TAMs and progressing through CXCL9/CXCL10 IFN-responsive TAMs toward HLA-DRA APC-like, SPP1/TREM2, and C1QC/APOE TAM states (Fig. 4G-H). State-level pseudotime distribution further showed that FCN1⁺ monocyte-like TAMs and CXCL9/CXCL10 IFN-responsive TAMs were mainly enriched at early-to-intermediate stages, whereas C1QC/APOE TAMs were predominantly distributed at later stages; SPP1/TREM2 TAMs and HLA-DRA APC-like TAMs showed broader intermediate-to-late distributions (Fig. 4I).

Analysis of dynamic gene-expression patterns demonstrated transient upregulation of CXCL9, CXCL10, STAT1, and IRF1 during the intermediate stage corresponding to IFN-responsive TAMs, whereas HLA-DRA, CD74, APOE, C1QC, SPP1, TREM2, and LGALS3 were enriched at later stages of the trajectory (Figure S2A-B). Collectively, these findings support a monocyte-derived continuum of TAM states in CRC and identify CXCL9/CXCL10 IFN-responsive TAMs as a transitional immune activating state that is not fully captured by the conventional M1/M2 classification framework.

EC reprograms macrophages toward an antitumor phenotype

EC was selected for further study because it showed a favorable combination of low cytotoxicity and reproducible induction of CD86-associated macrophage activation in both RAW264.7 cells and BMDMs. In addition, EC was selected based on its biological and translational relevance as a natural compound, and because EC-mediated macrophage remodeling and CD8⁺ T-cell immune activation in CRC had not been systematically investigated. Dose-response validation was subsequently performed using EC at 10, 30, 60, 60, 90, 120, 150, and 180 µM. EC is a natural phenolic compound found in Bidens pilosa (Fig. 5A). CCK-8 analysis was performed to determine a non-cytotoxic concentration range of EC in RAW264.7 cells and BMDMs. After 24 h of treatment, EC showed estimated IC50 values of 117 µM and 112 µM, respectively, whereas 30 and 60 µM caused no detectable growth inhibition (Fig. 5B). These two concentrations were therefore used for subsequent assays. Flow cytometry further showed that EC promoted macrophage polarization toward an M1-like antitumor phenotype (Fig. 5C, D).

Fig. 5.

A Compound-screening workflow and chemical structure of EC, a major active component of Bidens pilosa. The Bidens pilosa image was adapted from “Bidens pilosa (4936846138) (cropped).jpg” by Dinesh Valke, available via Wikimedia Commons under the Creative Commons Attribution-ShareAlike 2.0 Generic license (CC BY-SA 2.0). Changes were made by incorporating the image into the figure layout. B Dose-response analysis of EC cytotoxicity. C EC-induced polarization in RAW264.7 cells. D EC-induced polarization in BMDMs. The experiment was carried out three times. ns, no significance; *, P < 0.05; **, P < 0.01; ***, P < 0.001

EC efficiently inhibited tumor growth and promoted antitumor immunity

In the immunogenic MC38 syngeneic CRC model, daily intraperitoneal administration of EC at 25 or 50 mg/kg significantly suppressed tumor growth compared with PBS treatment, as shown by reduced tumor volume and tumor weight. Tumor volume was reduced by 31.3% ± 10.0% in the 25 mg/kg group and by 46.6% ± 13.5% in the 50 mg/kg group (n = 6, P < 0.05; Fig. 6A, B). Flow cytometric analysis showed that EC did not significantly alter the overall proportion of F4/80⁺ CD11b⁺ macrophages but significantly increased CD86⁺ M1-like macrophages in tumor tissues (Fig. 6C, D). EC treatment also increased tumor infiltrating CD8⁺ T cells and enhanced Granzyme B expression, while reducing PD-1 expression on CD8⁺ T cells (Fig. 6E-G). EC suppressed MC38 tumor growth, and this effect was associated with macrophage-containing myeloid remodeling and enhanced CD8⁺ T-cell cytotoxic activation.

Fig. 6.

A Experimental design of EC treatment (n = 6). B EC treatment suppresses tumor growth in vivo (n = 6). C Quantification of F4/80⁺CD11b⁺ macrophages in tumor tissues (n = 6). D EC treatment promotes antitumor macrophage polarization (n = 6). E The proportion of CD8⁺ T cells in tumor tissues (n = 7). F The expression of Granzyme B on CD8⁺ T cells (n = 7). G The expression of PD-1 on tumor infiltrating CD8⁺ T cells. (n = 7)

Clodronate liposome-mediated depletion attenuates EC-induced tumor suppression

To determine whether the antitumor effect of EC was mediated by macrophages, we established a macrophage depletion model using intraperitoneal administration of clodronate liposomes in CRC-bearing mice (Fig. 7A). Although EC treatment significantly inhibited tumor growth in the LipoPBS + EC group compared with the LipoPBS group, this antitumor effect was largely abrogated after macrophage depletion, indicating that EC suppressed tumor growth at least partially through macrophage-dependent mechanisms (Fig. 7B). Flow cytometric analysis showed that the percentages of both CD3⁻CD19⁻ myeloid cells and F4/80⁺CD11b⁺ macrophages were significantly lower in the LipoCLO group than in the LipoPBS group, with reductions of 39.8% and 50.6%, respectively, confirming successful depletion of splenic myeloid cells and macrophages (Fig. 7C, D). In tumor tissues, the percentage of F4/80⁺CD11b⁺ macrophages were significantly reduced in the macrophage-depleted groups, LipoCLO and LipoCLO + EC, compared with the non-depleted groups, LipoPBS and LipoPBS + EC. No significant difference was observed between LipoPBS and LipoPBS + EC or between LipoCLO and LipoCLO + EC, indicating that EC treatment did not affect the efficiency of macrophage depletion (Fig. 7E).

Fig. 7.

A Schematic of clodronate liposome–mediated macrophage depletion in CRC-bearing mice (n = 5). B Macrophage depletion attenuates the antitumor effect of EC (n = 5). C Clodronate liposomes reduced splenic CD3-CD19-myeloid cells (n = 5). D Clodronate liposomes reduced splenic F4/80⁺CD11b⁺ macrophages (n = 5). E Clodronate liposomes effectively depleted tumor infiltrating macrophages (n = 5). F EC increased CD86⁺ macrophage-associated activation among residual macrophages (n = 5)

Notably, the percentage of CD86⁺ M1-like macrophages was significantly increased in both EC-treated groups, LipoPBS + EC and LipoCLO + EC, compared with their respective controls, LipoPBS and LipoCLO (Fig. 7F). Together, these findings indicate that the macrophage-containing phagocytic myeloid compartment contributes to EC-mediated tumor suppression. However, because clodronate liposomes are not macrophage-specific, these data should not be interpreted as definitive evidence of macrophage-specific causality.

Discussion

The TME represents the ecological niche in which tumor cells reside and is typically characterized by extensive cellular infiltration [15]. Tumor cells dynamically interact with surrounding stromal and immune cells, thereby reshaping the local microenvironment. These interactions promote tumor immune evasion, angiogenesis, therapeutic resistance, phenotypic plasticity, and coordinated adaptive evolution under environmental stress [16]. Based on the pattern of immune cell infiltration within the TME, tumors can be broadly classified into immune-inflamed, immune-excluded, and immune-desert phenotypes [17, 18]. Although immune checkpoint blockade (ICB) has achieved remarkable clinical success in several cancer types, a substantial proportion of patients fail to benefit from this treatment [19, 20]. Accumulating evidence indicates that the functional state of immune cells within the TME plays a critical role in determining tumor progression and clinical prognosis in solid tumors [21].

ICB has become a major therapeutic strategy in cancer immunotherapy; however, its therapeutic efficacy varies substantially across different tumor types [2, 22]. This limitation is particularly evident in CRC, which is typically characterized by a relatively low tumor mutational/neoantigen burden, limited effector CD8⁺ T-cell infiltration, and enrichment of immunosuppressive myeloid cells, resulting in limited responsiveness to ICB therapy [17, 23]. Therefore, converting immunologically “cold” tumors into “hot” tumors represents a key strategy for improving the efficacy of immunotherapy in CRC [19].

MSS CRC generally exhibits limited effector immune cell infiltration and limited responsiveness to ICB and is therefore often regarded as an immunologically “cold” tumor with immune-desert or immune-excluded features [24]. Promoting T-cell infiltration and enhancing tumor immunogenicity may help convert these tumors into an immune-inflamed phenotype, thereby improving ICB responsiveness [2]. Accordingly, therapeutic strategies that remodel the immunosuppressive TME, such as reprogramming TAMs toward an antitumor phenotype, reducing immunosuppressive myeloid cells or regulatory T cells, enhancing antigen presentation, and promoting chemokine-mediated recruitment of effector CD8⁺ T cells, may provide promising approaches to enhance immunotherapy efficacy in MSS CRC [25, 26].

As a key component of the TME, macrophages play context-dependent roles in cancer progression. In early tumor development, M1-like macrophages can promote antigen presentation, secrete pro-inflammatory mediators, and produce NO and ROS, thereby contributing to antitumor immunity [3]. They also release chemokines such as CXCL9, CXCL10, and CXCL11, which recruit activated T cells and NK cells to the tumor site. However, during tumor progression, macrophages are often reprogrammed toward an M2-like phenotype, which supports tumor growth, impairs CD8⁺ T-cell function, and promotes an immunosuppressive microenvironment [27–29]. Accordingly, reprogramming TAMs from an M2-like immunosuppressive state toward an M1-like antitumor phenotype represents a promising strategy for enhancing cancer immunotherapy [30, 31].

In the present study, we found that EC reshaped the CRC immune microenvironment by modulating the macrophage–T-cell immune axis. EC treatment increased CD86⁺ inflammatory macrophage-associated activation within the conventional M1/M2 framework and was accompanied by enhanced CD8⁺ T-cell infiltration and cytotoxic activation. Together with the single-cell analyses, these findings suggest that EC may promote macrophage-associated immune remodeling rather than simply inducing a binary M2-to-M1 conversion. Furthermore, macrophage depletion attenuated EC-mediated tumor suppression, supporting the macrophage-dependent nature of its antitumor activity and indicating that TAMs are important cellular targets of EC-mediated immune modulation.

Taken together, these findings suggest EC as a natural macrophage-reprogramming compound that may alleviate the immunosuppressive CRC microenvironment by promoting M1-like macrophage polarization and increased CD8⁺ T-cell infiltration, enhanced Granzyme B expression, and reduced PD-1 expression on tumor infiltrating CD8⁺ T cells, suggesting enhanced cytotoxic activation and a less inhibitory CD8⁺ T-cell phenotype. These findings provide a potential rationale for combining EC with ICB therapy to improve immune responsiveness in CRC.

Several limitations should be acknowledged. First, although EC reduced PD-1 expression and increased Granzyme B expression in tumor infiltrating CD8⁺ T cells, CD8⁺ T-cell exhaustion is a complex functional and transcriptional state that cannot be defined by PD-1 alone. Additional exhaustion-associated markers, such as TIM-3, LAG-3, TIGIT, TOX, and TCF1, together with functional restimulation assays, are required to determine whether EC truly reverses CD8⁺ T-cell exhaustion. Second, the present in vivo experiments were performed using a subcutaneous MC38 syngeneic model, which is immunogenic and does not fully recapitulate the immune-cold phenotype of MSS CRC or the colonic mucosal microenvironment. Therefore, our findings should be interpreted as proof-of-concept evidence in an immunogenic murine CRC model, and further validation in AOM/DSS-associated, orthotopic, genetically engineered MSS-like, or patient-derived models are needed. Third, clodronate liposomes are not macrophage-specific and may also affect monocytes, dendritic cells, and other phagocytic myeloid populations. Thus, our depletion data supports the involvement of the macrophage-containing phagocytic myeloid compartment but does not prove macrophage-specific causality. Fourth, the direct effect of EC on MC38 tumor-cell viability was not systematically evaluated, and endotoxin contamination was not directly assessed by Limulus amebocyte lysate testing or polymyxin B neutralization. These issues limit our ability to fully distinguish macrophage-mediated effects from potential tumor-cell-intrinsic or endotoxin-related effects. Future studies should include tumor-cell viability assays, endotoxin testing, polymyxin B controls, and more specific macrophage-targeting approaches to further clarify the mechanisms underlying EC-mediated immune remodeling. Nevertheless, additional studies are needed to determine whether EC can enhance the therapeutic response to anti-PD-1/PD-L1 treatment, particularly in CRC models. The present study did not further clarify the precise mechanism by which EC-reprogrammed macrophages regulate CD8⁺ T-cell recruitment and function. Further studies are needed to identify the key chemokines, cytokines, and ligand-receptor interactions involved in macrophage–T-cell crosstalk within the CRC microenvironment. Because the present findings were generated primarily in murine experimental systems, further validation in human-relevant models is required to evaluate the efficacy, safety, and translational potential of EC-based immunomodulatory therapy.

Conclusion

In conclusion, EC suppressed tumor growth at least partly through macrophage-associated immune remodeling, accompanied by increased CD86⁺ macrophage activation, enhanced CD8⁺ T-cell infiltration, increased Granzyme B expression, and reduced PD-1 expression. These findings suggest that EC may promote antitumor immunity in CRC by remodeling the macrophage-containing myeloid compartment.

Abbreviations

BMDMs

Bone marrow-derived macrophages

CAR-T

Chimeric antigen receptor T-cell therapy

CCK-8

Cell Counting Kit-8

CRC

Colorectal cancer

cDC2

Conventional dendritic cell type 2

DEGs

Differentially expressed genes

DMEM

Dulbecco’s modified Eagle medium

DMSO

Dimethyl sulfoxide

DNase I

Deoxyribonuclease I

EC

Ethyl caffeate

FBS

Fetal bovine serum

GEO

Gene Expression Omnibus

ICB

Immune checkpoint blockade

i.p.

Intraperitoneal injection

LipoCLO

Clodronate liposomes

LipoPBS

PBS control liposomes

LPS

Lipopolysaccharide

M-CSF

Macrophage colony-stimulating factor

MSI-H

Microsatellite instability-high

MSS

Microsatellite-stable

NO

Nitric oxide

PBS

Phosphate-buffered saline

PCA

Principal component analysis

pDC

Plasmacytoid dendritic cell

PD-1

Programmed cell death protein 1

PEG300

Polyethylene glycol 300

ROS

Reactive oxygen species

SD

Standard deviation

STR

Short tandem repeat

TAMs

Tumor-associated macrophages

TCGA-COAD/READ

The Cancer Genome Atlas colon adenocarcinoma/rectal adenocarcinoma cohort

TME

Tumor microenvironment

TNF

Tumor necrosis factor

Authors’ contributions

ZB and RJ conceived and designed the study, performed the experiments, analyzed the data, and drafted the manuscript. JG and X HC contributed to the design and preparation of the tables and figures. XM and WY performed the public database analysis. All authors reviewed and approved the final manuscript.

Funding

This study was supported by the Central Incentive Fund for Regional S&T Development (Grant No. 25ZYJA024), the Gansu Provincial Key Talent Program (Grant No. 2025RCXM081), Open Project of State Key Laboratory of Neurology and Oncology Drug Development (Grant No. SKLSIM-F-2025107) and the Central Universities Postgraduate Innovation Project Fund (lzujbky-2025-it35). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The public single-cell RNA-seq dataset analyzed in this study is available in the Gene Expression Omnibus repository under accession number GSE132465. The TCGA-COAD/READ RNA-seq and clinical survival data analyzed in this study are available from The Cancer Genome Atlas through the Genomic Data Commons (GDC) Data Portal. Other data generated or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All experimental protocols were approved by the Animal Ethics Committee of the First Hospital of Lanzhou University (Approval No.: LDYYLL2026-672). The care and use of laboratory animals followed the Guidelines for the Welfare and Ethics of Laboratory Animals (Ministry of Science and Technology of the People’s Republic of China). In addition, this study adhered to the ARRIVE guidelines (https://arriveguidelines.org) for reporting in vivo experiments to ensure transparency and reproducibility. All efforts were made to minimize animal use and suffering.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.