Epigenetic demethylation sensitizes cancer cells to CAR-M triggered pyroptosis and remodels inflammatory immune responses in the tumor microenvironment to potentiate CAR-M cell therapy

Abstract

Background

Tumor epigenetic reprogramming serves as a critical hallmark of cancer, fostering transcriptional plasticity of gene expression and driving alterations in the suppressive inflammatory responses within the tumor microenvironment, which pose significant challenges to the advancement of novel CAR-macrophage (CAR-M)-based antitumor therapies. However, it remains limited mechanistic investigations exploring how epigenetic reprogramming strategies impact the efficacy of CAR-M treatments.

Methods

Human epidermal growth factor receptor 2 (HER2)-targeted CAR-Ms containing either an α1β1 integrin-mediated Fc-gamma receptor I (FcγRI) or CD3ζ signaling domain were generated and compared for tumor-directed migration and antitumor activity. HER2-positive tumor cells were pretreated with decitabine (DAC). CAR-M function, tumor-cell death, transcriptomic and inflammatory pathway changes, tumor immune remodeling, and therapeutic efficacy were assessed using in vitro assays, RNA sequencing, and murine tumor and metastasis models. Full-dose and half-dose CAR-M regimens were evaluated.

Results

α1β1 integrin-FcγRI CAR-Ms displayed significantly greater migration toward HER2-positive tumor cells than CD3ζ CAR-Ms. DAC potentiated anti-HER2 CAR-M activity by reprogramming tumor-cell gene expression and shifting the dominant cell-death mode from apoptosis to gasdermin E (GSDME)-dependent pyroptosis. RNA sequencing demonstrated enrichment of inflammatory immune-response programs, including nuclear factor kappa B and Janus kinase-signal transducer and activator of transcription signaling, after DAC treatment. DAC also favorably remodeled the tumor immune microenvironment, thereby supporting CAR-M activity. In murine models, DAC plus CAR-M therapy enhanced inflammatory antitumor immunity and improved control of primary tumor growth and metastatic disease compared with CAR-M monotherapy. Notably, DAC pretreatment combined with half-dose CAR-Ms achieved therapeutic efficacy comparable to that of full-dose CAR-M treatment.

Conclusions

DAC-mediated DNA demethylation sensitizes HER2-positive tumors to CAR-M therapy through GSDME-dependent pyroptosis, activation of inflammatory signaling, and favorable remodeling of the tumor microenvironment. Epigenetic priming may improve CAR-M efficacy while reducing the required cell dose, supporting further translational development of this combination strategy for solid tumors.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• CAR-macrophage (CAR-M) cells can infiltrate solid tumors, phagocytose cancer cells, and modulate the tumor microenvironment.

• Their limited persistence and the immunosuppressive tumor microenvironment restrict efficacy.

• DNA methylation contributes to cancer-cell plasticity and immunotherapy resistance, but its effect on CAR-M therapy remains unclear.

WHAT THIS STUDY ADDS

• The α1β1 integrin-mediated Fc-gamma receptor I signaling domain improves CAR-M accumulation and phagocytosis compared with a CD3ζ-based construct.

• Decitabine (DAC)-mediated demethylation restores gasdermin E (GSDME) expression and converts CAR-M-induced apoptosis into caspase-3/GSDME-dependent pyroptosis.

• The combination activates inflammatory signaling, remodels the tumor immune microenvironment, and improves antitumor and antimetastatic efficacy.

• DAC pretreatment may allow half-dose CAR-M to achieve efficacy comparable to full-dose CAR-M alone.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• DAC preconditioning may improve CAR-M efficacy, reduce the required cell dose, and support CAR-M treatment for solid tumors.

• It provides a novel epigenetic combinatorial strategy to guide CAR-M research and advance clinical solid tumor immunotherapy optimization.

Introduction

Engineered living therapeutics represent a transformative frontier in oncology, harnessing the programmable capabilities of living cells to mount targeted, adaptive responses against malignant tumors with unprecedented precision and therapeutic potential.^{1 2} Despite sustained inputs of chemokines and immune cell infiltration within the malignant tumor microenvironment (TME), the tumor exhibits an overall functional output characterized by immune suppression and anti-inflammation, which promotes tumor proliferation, invasion, metastasis, and chemoresistance.³ Compared with CAR-T cells, CAR-macrophages (CAR-M) exhibit unique advantages for the treatment of solid tumors owing to their superior tumor infiltration capacity and intrinsic phagocytic activity.^4–6 Emerging clinical evidence for CAR-M cell therapy has further heightened the anticipation of its therapeutic potential in clinical settings.⁷ However, despite harnessing the innate immunological merits of macrophages, the induction and sustained maintenance of robust clinical responses in CAR-M-treated patients remain significantly challenged by the immunosuppressive TME of malignant solid tumors.⁸

Currently, accumulating evidence highlights the pivotal role of epigenetic heterogeneity in driving the inflammatory immune regulation of tumors. Recent research reveals that aberrant DNA methylation modulates cancer-immune cell crosstalk.³ Comprehensive hypomethylation correlates with high genomic instability, immune evasion, and immunotherapy resistance, and drives an immunosuppressive microenvironment in tumors, including melanoma, via programmed death-ligand 1 and cytokine upregulation.^{9 10} Interestingly, hypermethylation also induces immune checkpoint blockade resistance by repressing interferon and checkpoint signaling, with (de)methylation of checkpoint genes shaping tumor immune traits across cancer types.^11–14 These findings highlight that (de)methylation may serve as a core determinant underlying the diverse mechanistic regulation of tumor immune behavior across multiple cancer subtypes. Epigenetic modifications, particularly DNA methylation, have been widely implicated in the initiation and progression of numerous malignancies.^15–17 Cancer-specific promoter DNA methylation represents a key driver of malignant transformation. Specifically, hypermethylation-mediated silencing of tumor suppressor genes leads to aberrant cell cycle progression and promotes the initiation and progression of cancer.¹⁸ Furthermore, aberrant hypermethylation contributes to therapeutic resistance in malignant tumors by silencing tumor suppressor genes and disrupting normal gene regulatory networks, thereby fostering tumor heterogeneity and immune evasion.¹⁹ Collectively, these mechanisms underlie the difficulties encountered by single-pronged CAR-M therapy in counteracting high tumor burden. This represents a critical challenge. However, current mechanistic studies remain limited in characterizing how tumor epigenetic alterations modulate CAR-M therapy.

In fact, compared with the traditional tumor drug paradigm, the efficacy of live cell therapy depends on both internal and external factors. These internal factors depend on the self-structural design of CAR-M, with a particular focus on the activation domain. This domain essentially determines the activity and persistence of CAR-M, and fundamentally influences its application as a potent cancer immunotherapeutic strategy.²⁰ Despite favorable preclinical findings, 44% of human epidermal growth factor receptor 2 (HER2) 3⁺ tumors achieved only transient disease stabilization in the CT-0508 trial, with its suboptimal clinical efficacy attributed to limited tumor penetration and poor persistence of systemically infused CAR-M.⁷

More importantly, in terms of external factors, emerging evidence suggests that epigenetic reprogramming, as the hallmark of cancer, encompasses not only alterations that favor the phenotypic selection of hallmark capabilities, thereby driving the clonal expansion of cancer cells with increased proliferative fitness, but also broad changes driven by the aberrant TME.²¹

Building on this premise, the optimization of structural domains serves as a foundational prerequisite for the advancement of CAR-M. As pioneering work, we previously constructed CAR-M cells using an α1β1 integrin-mediated Fc-gamma receptor I (FcγRI) intracellular domain, to drive persistent activation and potent antitumor activity of CAR-M (named as active CAR-M (ACT CAR-M)).²⁰ Compared with the classical CD3ζ domain-based CAR-M,²² derived from the T-cell signal transduction complex, ACT CAR-M showed improved HER2 antigen-mediated activation and phagocytic capacity in vitro, with robust antitumor efficacy in vivo in antitumor responses against tumor-bearing mice.²⁰

Here, after further validating the role of the α1β1 integrin-mediated FcγRI domain in enhancing the proliferation and phagocytic efficiency of ACT CAR-M in vivo, we developed an epigenetically enhanced CAR-M cell therapy strategy. Decitabine (DAC), a DNA methyltransferase inhibitor, was employed as a prototypical agent to investigate the antitumor efficacy and mechanism of epigenetic modifiers in combination with anti-HER2 ACT CAR-M. Preclinical data show DAC remodels tumor cells in promoting pyroptosis and rewires immunosuppressive TME, yet it exhibits poor single-agent clinical efficacy.²³ This study elucidates the synergistic mechanisms underlying epigenetic reprogramming and CAR-M cellular immunotherapy, assessing its efficacy against solid tumors with a specific focus on modulating cancer cell death pathways and inflammatory immune regulation within TME. In this work, following DAC pretreatment, even a half dose of CAR-M cells maintained significant antitumor efficacy in HER2-positive tumor models. These results prompted further investigation into the molecular mechanisms responsible for the observed therapeutic activity. Together, our findings would support DAC-mediated demethylation as a promising clinical strategy for substantially enhancing CAR-M cellular immunotherapy.

Materials and methods

Cell lines and cell culture

Cell lines HEK293T, J774A.1, THP-1, 4T1, MKN45, and SKOV3 were obtained from the American Type Culture Collection. Human CD14⁺ monocytes were isolated from peripheral blood mononuclear cells of healthy donors.⁴ Cell culture conditions were prepared as previously reported.^{7 20}

Plasmid construction, lentiviral preparation, transduction, and production of engineered cells

The HER2-targeting ACT CAR construct was designed as previously described.²⁰ Control vectors included a CD3ζ-based CAR construct (replacing the α1β1-FcγRI signaling domains with CD3ζ) and a CAR-like control lacking the intracellular signaling domains.

Lentiviral supernatants used for the transduction of macrophages were prepared as previously described.²⁰ J774A.1, THP-1 cells, or human CD14⁺ monocytes were transduced with lentiviral CAR vectors in the presence of polybrene (Sigma-Aldrich) in six-well plates. After 24 hours, the medium was replaced with high-glucose Dulbecco’s modified Eagle’s medium containing 2% fetal bovine serum. CAR-M were cultured for an additional 48 hours to ensure complete CAR expression before functional assays. Transduction efficiency was assessed by flow cytometry based on green fluorescent protein (GFP) expression using a BD FACSCelesta flow cytometer or by laser-scanning confocal microscope (Olympus) to measure GFP expression. HER2⁺-4T1 cells were engineered to overexpress human HER2 based on the above lentiviral transfection system.

Confocal microscopy imaging analysis

To visualize CAR expression and cellular localization, macrophages were transduced with ACT CAR lentivirus and cultured for 72 h. Cells were then labeled with 5 μM DiI for 20 min at 37°C and washed three times with phosphate-buffered saline (PBS). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (1 μg/mL) for 5 min. Images were acquired using a laser scanning confocal microscope (FV3000, Olympus, Japan) with a 60× oil-immersion objective, as previously described.²⁴

Decitabine treatment protocol

DAC stock solution (10 mM in dimethyl sulfoxide) was stored at −20℃. Tumor cells were treated with 0–5 μM DAC for 0–72 hours, with daily medium changes. For all CAR-M co-culture assays, tumor cells were pretreated with DAC at the corresponding optimized concentration for the designed duration, then adding CAR-M cells: 5 μM DAC for 72 hours for cytotoxicity assays and 1 μM DAC for 24 hours for mechanistic studies on DAC-mediated apoptosis-to-pyroptosis switch and related inflammatory responses.

In vitro cytotoxicity assay

Target cells (HER2⁺-4T1, MKN45, or SKOV3; 1×10⁴ cells/well) were seeded in 96-well plates, pretreated with DAC or left untreated, and co-cultured with CAR-M at effector-to-target (E:T) ratios (1:1) for 24 hours as previously.²⁵ To account for the metabolic contribution of CAR-M cells to the Cell Counting Kit-8 (CCK-8) signal, wells containing CAR-M cells alone (without tumor cells) at equivalent cell numbers and E:T ratios were used as background controls. Cytotoxicity was normalized by subtracting CAR-M-only optical density (OD) values: % cytotoxicity=(1–[OD_co-culture–OD_CAR-M alone]/[OD_tumor cell control])×100%. The CCK-8 assay was performed using a commercially available CCK-8 (DONGXISW, Xi’an, China; Cat. No. DXS1500), according to the manufacturer’s instructions.

Flow cytometry analysis

CAR expression and transduction efficiency were assessed as previously described in our laboratory.²⁰ Tumor tissues were mechanically dissociated and enzymatically digested with collagenase type IV (1 mg/mL), DNase I (0.1 mg/mL), and hyaluronidase (0.5 mg/mL) (Sigma-Aldrich) in Roswell Park Memorial Institute-1640 for 45 min at 37°C. Single-cell suspensions were obtained by passing through 70 μm cell strainers, followed by red blood cell (RBC) lysis. Cells were stained with fluorochrome-conjugated antibodies against GFP, CD45, F4/80, CD80, CD206, CD3, CD4, CD8a, CD19, and B220 for 30 min at 4°C and analyzed by flow cytometry.

In vivo phagocytosis assay

Tumor tissues from CAR-M-treated mice were harvested and processed into single-cell suspensions. Following phagocytosis of HER2⁺ tumor cells by CAR-M, internalized HER2 antigen was detected intracellularly using antihuman HER2 antibody after cell permeabilization. Phagocytic events were thus defined as GFP⁺ HER2⁺ double-positive events.

Apoptosis detection by flow cytometry

Tumor cells (HER2⁺-4T1 or MKN45; 2×10⁵ cells/well) were treated with: PBS control, 5 μM DAC for 72 hours, ACT CAR-M (E:T ratio 1:1) for 24 hours, or DAC pretreatment followed by CAR-M co-culture. Cell death was quantified using the fluorescein isothiocyanate-conjugated Annexin V/propidium iodide (PI) Apoptosis Detection Kit according to the manufacturer’s instructions and subsequently analyzed by flow cytometry. Annexin V⁺/PI⁻ populations represented early apoptotic cells, while Annexin V⁺/PI⁺ cells included late apoptotic and pyroptotic populations. GSDME cleavage was further detected to confirm pyroptosis.

Quantitative real-time PCR

Total RNA was extracted using TRIzol reagent according to the manufacturer’s instructions. Complementary DNA (cDNA) was synthesized using the RevertAid First Strand cDNA Synthesis Kit. Quantitative real-time PCR was performed using SYBR Green Master Mix on a StepOne Real-Time PCR machine (Life Technologies). The fold expression of gasdermin E (GSDME, also known as DFNA5) was calculated by the 2^−ΔΔCT method. The primer sequences used in this study are presented in online supplemental table S1.

Lactate dehydrogenase and ATP release assay

Extracellular lactate dehydrogenase (LDH) activity was measured using a colorimetric LDH assay kit and ATP detection kit previously described in our laboratory.²⁶ Cell culture supernatants were collected and mixed with LDH working solution. Absorbance was measured at 490 nm with background correction at 600 nm, and 20 µL supernatant was collected after centrifugation (400×g, 5 min) for LDH assay. For ATP measurement, harvested cells were lysed and then centrifuged at 12 000×g for 5 min. The resulting supernatant was incubated with luminescent reaction reagent to analyze ATP levels. Data represent mean±SEM of triplicate samples.

Pyroptosis morphological analysis

Pyroptotic cell morphology was assessed using an inverted brightfield microscope. Following treatment, cells were imaged at 2-hour intervals using a 20× objective lens. Pyroptotic cells were identified based on the following criteria: (1) pronounced cell body swelling accompanied by homogeneous cytoplasmic clearing, (2) progressive plasma membrane blistering culminating in rupture with subsequent release of cellular contents, and (3) optional formation of discrete 1–2 μm pyroptotic bodies.

Western blot analysis

Western blot analysis was performed as previously described.²⁷ Cells were lysed in radioimmunoprecipitation assay buffer containing proteinase inhibitors, and protein concentrations were quantified using a bicinchoninic acid assay kit. Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to nitrocellulose membranes, blocked with bovine serum albumin, and probed with primary antibodies (1:1000) overnight at 4°C: anti-GSDME, anticaspase-3, anticleaved caspase-3, and anti-β-actin. Membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies for 1 hour. Protein expression was detected by chemiluminescence (Thermo Fisher Scientific) and quantified using ImageJ. Data represent mean±SEM from triplicate experiments.

RNA sequencing and transcriptomic analysis

For in vitro studies, MKN45 cells were treated with PBS, DAC (1 μM, 72 hours), ACT CAR-M (E:T 1:1, 24 hours), or DAC+CAR-M combination. For CAR-M co-culture groups, high-quality RNA was extracted (purity >95%) to ensure transcriptomic analysis accurately captured gene expression changes. For in vivo studies, tumor tissues were snap-frozen and stored at −80°C. Total RNA was extracted using TRIzol reagent (Invitrogen) and RNA integrity (≥7.0) was confirmed by Agilent 2100 Bioanalyzer. RNA sequencing (RNA-seq) libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit (New England Biolabs) with poly(A) enrichment.

RNA-seq data were aligned using HISAT2 (V.2.0.4). Differential gene expression analysis was performed using DESeq2. Differentially expressed genes (DEGs) with |log₂ fold change|≥1 and false discovery rate ≤0.05 were visualized by volcano plots. Functional enrichment analysis was conducted using clusterProfiler to identify significantly enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Heatmaps were generated by Z-score normalization, and Gene Set Enrichment Analysis (GSEA) was performed on the dataset using GSEA software V.4.2.3 with Hallmark gene sets from MSigDB. To validate DAC-mediated demethylation effects, in vitro assays were performed in MKN45 cells with azacitidine (AZA), a structurally distinct DNMT methyltransferase‌ inhibitor. Transcriptomic profiling of AZA-treated and untreated cells was conducted for comparison. In heatmap visualizations, individual sample labels (eg, PBS1–3, DAC1–3, AZA1–3) denote independent biological replicates within each treatment group.

Animal models and in vivo therapeutic studies

Female BALB/c mice (aged 4–6 weeks), female nude mice, and NCG mice (aged 4–6 weeks) were obtained from Beijing HFK Bio for humanized models. All mouse studies were performed in accordance with national guidelines for the treatment of animals, as previously described.²⁸ The scheme of the xenograft model used is shown in detail in the first panel of each correlogram.

For subcutaneous models, BALB/c mice received 1×10⁶ HER2⁺-4T1 cells, while NCG mice received 5×10⁶ MKN45 in 100 μL PBS via right flank injection. For the peritoneal metastasis model, nude mice received 5×10⁶ SKOV3-luciferase cells via intraperitoneal injection. Tumor growth was monitored by caliper measurements every 2–3 days (volume=length×width²/2) or bioluminescence imaging using an IVIS Spectrum system (PerkinElmer) with D-luciferin substrate (150 mg/kg) every 5 days. Mice were defined as dead at a fluorescence radiance threshold of ≥1×10¹⁰ photons/s/cm²/sr. At experimental endpoints, tumors were harvested and processed for analysis as previously described.²⁰

Cytokine and biochemical assays

In mouse models of cytokine release syndrome (CRS) and tumor lysis syndrome (TLS), HER2⁺-4T1 tumor-bearing BALB/c mice were treated with CAR-M cells. Serum samples were collected on days 1, 2, 3, and 5 post-treatments. For CRS-related cytokine profiling, serum levels of murine interleukin (IL)-6, interferon-γ, tumor necrosis factor-α, and IL-1β were quantified with commercial ELISA kits (Fankew, Shanghai) in accordance with the manufacturer’s protocols. For TLS-associated biochemical profiling, serum biochemical parameters including uric acid, creatinine, LDH, K⁺, inorganic phosphate, Ca²⁺, aspartate aminotransferase, alanine aminotransferase, and urea were measured by Servicebio using standard biochemical testing protocols.

Biosafety evaluation

At experimental endpoints, whole blood was collected from NCG mice for complete blood count analysis (RBC, hemoglobin, white blood cell, platelet). Major organs (heart, liver, spleen, lung, kidney) were harvested, fixed in 4% paraformaldehyde, paraffin-embedded, and sectioned for H&E staining.

Quantification and statistical analysis

All data were analyzed using GraphPad Prism 9.0. Differences between two groups were assessed using Student’s t-test, while one-way analysis of variance was used for multiple comparisons. A p value <0.05 was considered statistically significant.

Results

α1β1 integrin-mediated FcγRI domain drives CAR-M to possess enhanced motility in vivo

Our previous work has revealed that the α1β1 integrin-mediated FcγRI domain drives ACT CAR-M to show excellent tumor suppression ability compared with CD3ζ CAR-M in solid tumor models.²⁰ To further explore the mechanisms involved, we employed HER2-targeted CARs and assembled different domain combinations to respectively obtain ACT CAR, clinically approved CD3ζ CAR, or CAR-like M, which has single-chain variable fragment (scFv) but no intracellular domains (figure 1A). To mimic the in vivo tumor environment accurately, we engineered anti-HER2 CAR-M using mouse macrophages J774A.1, and meanwhile used an engineered cancer cell line, HER2⁺-4T1, to construct a standard model of HER2⁺ tumor for all related in vivo studies. Recombinant lentiviruses expressing these CARs were generated and used to obtain efficient expression of ACT CAR on macrophages (figure 1B,C). For experiments performed in J774A.1 cells, only transduction products with ≥80% GFP-positive cells were used for downstream experiments. Empty vector, CAR-like, and CD3ζ CAR also displayed robust surface CAR expression by analysis of GFP⁺ populations (online supplemental figure S1A,B).

Figure 1. α1β1 integrin-mediated FcγRI domain drives CAR-M to possess stronger motivation in vivo. (A) Schematic design of the lentiviral vector for expression of the CARs. The ACT CAR-M constructs contain HER2 scFv, α1β1 integrin-mediated FcγRI domain, and GFP. Empty plasmid, CAR-like, and clinically approved HER2-specific CD3ζ CAR were used for comparison. (B) Confocal images of ACT CAR-M (bar=10 µm). (C) Flow cytometry assay for transfection efficiency of ACT CAR using GFP labeling to represent transfected cells. (D) Schematic of the HER2⁺-4T1 breast cancer model and treatment process. Female BALB/c mice (aged 4–6 weeks) received subcutaneous injection of 1×10⁶ HER2⁺-4T1 cells on day −10. J774A.1-derived CAR-M cells (3.5×10⁶ cells, i.v.) were injected on day 0. Tumor-infiltrating CAR-M populations were analyzed by flow cytometry at days 0, 5, and 10 postinjections. Treatment groups: G1, PBS; G2, Empty; G3, CAR-like; G4, CD3ζ CAR-M; G5, ACT CAR-M (n=5). (E) Flow cytometric analysis and statistical results for CAR-M populations of different groups in tumors. (F) Flow cytometric analysis and statistical results for CAR-M phagocytosis of different groups in tumors. HER2 of tumor cells was detected intracellularly following cell permeabilization, while GFP⁺HER2⁺ double-positive events therefore reflect phagocytic internalization of HER2⁺ tumor cell material. Data are mean±SEM (n=3); *p<0.05, **p<0.01, ***p<0.001. CAR, chimeric antigen receptor; CAR-M, chimeric antigen receptor macrophage; FACS, fluorescence-activated cell sorting; FcγRI, Fc gamma receptor I; GFP, green fluorescent protein; HER2, human epidermal growth factor receptor 2; i.v., intravenous; ns, not significant; PBS, phosphate-buffered saline; s.c., subcutaneous; scFv, single-chain variable fragment; TM, transmembrane; UTD-M, untransduced macrophages.

Next, we tested how the α1β1 integrin-mediated FcγRI domain affected CAR-M motility in vivo by promoting activity and phagocytosis of ACT CAR-M in tumor tissues. To identify which aspects of CAR-M development are regulated by the α1β1 integrin-mediated FcγRI domain or the CD3ζ domain, we assessed the CAR-M cell populations in the tumor tissues among ACT CAR-M, CD3ζ CAR-M, and CAR-like M groups at 0, 5, and 10 days following intravenous injection of CAR-M cells (3.5×10⁶ cells per mouse; 10 days after tumor inoculation), using the PBS and Empty groups used as controls (figure 1D). The results revealed a significantly increased CAR-M cell ratio in the α1β1 integrin-mediated FcγRI domain group than those of the CD3ζ domain, CAR-like M, Empty, and PBS groups (figure 1E, online supplemental figure S1C). Additionally, the proportion in the Empty group was lower than in other CAR-M treatment groups (figure 1E), suggesting that the extracellular scFv antibody arm contributes to the accumulation of CAR-M within the tumor. Compared with the Empty group, the series of CAR-M groups showed a similar increase on day 5 and day 10 postinjection (figure 1E). Importantly, the α1β1 integrin-mediated FcγRI domain group showed significantly greater proliferation than those of the CD3ζ domain group on day 5 and day 10 postinjection (figure 1E). Interestingly, by day 10 postinjection, there were more CD3ζ CAR-M cells compared with CAR-like M cells within the tumor. These patterns suggested that involvement of the CAR intracellular domain plays a decisive role in the in vivo activation of CAR-M therapy.

We further investigated alterations in the phagocytic ability of CAR-M within the tumor tissues using flow cytometry. This increased phagocytic ability of CAR-M is critical because it is likely to facilitate antigen presentation to T cells, thereby enhancing the antitumor immune response. The α1β1 integrin-mediated FcγRI domain group significantly enhanced the phagocytic ability of CAR-M compared with other groups. Furthermore, both the α1β1 integrin-mediated FcγRI domain group and the CD3ζ domain group significantly upregulated phagocytic ability toward HER2⁺ tumor cells in tumor tissues by day 10 postinjection (figure 1F), which may be attributed to the intracellular domain design of CAR-M. However, it is important to note that although the domain design induces high phagocytic ability, different domain designs play key antitumor roles not only by increasing the CAR-M population beneficial for CAR-M activation and proliferation but also by enhancing the antitumor capacity of CAR-M through promoting the innate immune potential of macrophages and directly acting on tumor cells.

Methylation epigenetic modifications increase the efficiency of CAR-M therapy

DNA methylation acts as a crucial epigenetic mechanism for gene expression regulation in various cancers. It is hypothesized that these epigenetic changes may influence the therapeutic effect of CAR-M. Therefore, DAC was employed as a pretreatment agent for cancer cells to systematically evaluate its potential contribution to ACT CAR-M therapy, both under controlled in vitro conditions and in tumor-bearing murine models.

First, a low-dose/long-term protocol (0–5 μM DAC for 0–72 hours) was evaluated in cancer cells (online supplemental figure S2). The addition of DAC at a dose of 5 μM for 72 hours during HER2⁺-4T1 cells culture had little effect on cell viability (online supplemental figure S2A). To assess whether epigenetic pretreatment might affect cancer cells’ susceptibility to cytotoxic killing by ACT CAR-M, HER2⁺-4T1 cells were pretreated with DAC or PBS as control and subsequently incubated with ACT CAR-M cells. Strikingly, DAC-pretreated HER2⁺-4T1 cells showed enhanced susceptibility to cancer cell killing compared with ACT CAR-M treatment alone, with significantly increased cytolytic activity at different time points following low-dose DAC treatment (figure 2A).

Figure 2. DAC epigenetic pretreatment of HER2⁺-4T1 contributes to ACT CAR-M cytotoxicity and antitumor effects. (A) Cell viability of HER2⁺-4T1 cells following various treatments in vitro. Minor technical fluctuations relative to the untreated control, which was set as the 100% viability baseline. (B) Schematic of the HER2⁺-4T1 breast cancer model and treatment group. Female BALB/c mice were inoculated subcutaneously with 1×10⁶ HER2⁺-4T1 cells on day −7. DAC (0.5 mg/kg, i.v.) was administered on days 0, 2, and 4. J774A.1-derived ACT CAR-M cells (3.5×10⁶, i.v.) were injected on day 5. Treatment groups: G1, PBS; G2, ACT CAR-M; G3, DAC+ACT CAR-M (n=6). (C) Tumor images at the experimental endpoint after different treatments. (D) Quantitative analysis of tumor weight at the experimental endpoint after different treatments. (E) Tumor volume of mice in different groups. (F) Body weights in different groups. Data are mean±SEM (n=6); *p<0.05, **p<0.01, ***p<0.001. CAR-M, chimeric antigen receptor macrophage; DAC, decitabine; HER2, human epidermal growth factor receptor 2; i.v., intravenous; ns, not significant; PBS, phosphate-buffered saline; s.c., subcutaneous.

We further assessed the antitumor effect of ACT CAR-M combined with DAC administration in vivo using the HER2⁺-4T1 BALB/c mouse model. ACT CAR-M cells derived from J774A.1 were administered intravenously at a dose of 3.5×10⁶ cells per mouse, for proof-of-concept validation to confirm whether DAC pretreatment enhances the antitumor response (figure 2B). As expected, the cohort of mice receiving ACT CAR-M combined with DAC administration showed significant tumor control compared with the ACT CAR-M monotherapy group (figure 2C–E). Additionally, DAC-mediated ACT CAR-M treatment showed no significant decrease in body weight of mice (figure 2F), preliminarily indicating that the combination therapy had no obvious toxicity in mice.

DAC demethylation-mediated ACT CAR-M alters transcriptomic landscapes of cancer cells

Besides the constructed HER2⁺-4T1 cell lines, human classic HER2⁺ cancer cell lines (such as human gastric adenocarcinoma cell line MKN45 and human ovarian cancer cell line SKOV3) were carried out to verify the contribution of DAC treatment to enhancing CAR-M therapy with different E:T ratios. Similar results to those in online supplemental figure S2A were observed that the premise of DAC pretreatment had little effect on cell viability of MKN45 and SKOV3 (online supplemental figure S2B,C). Meanwhile, after DAC pretreatment at a dose of 5 μM for 72 hours, the consistency of the results clearly demonstrated that the premise of DAC pretreatment significantly enhanced ACT CAR-M cytotoxicity at different E:T ratios targeting MKN45 and SKOV3, as well as HER2⁺-4T1 (figure 3A,B and online supplemental figure S3).

Figure 3. DAC demethylation-mediated ACT CAR-M alters transcriptomic landscapes of in cancer cells. (A, B) ACT CAR-M cells were co-cultured with MKN45 or SKOV3 cells pretreated with or without DAC at different E:T ratios. Cytotoxicity was assessed by CCK-8. (C) Schematic of transcriptome sequencing. MKN45 cells received four distinct treatments: PBS, DAC, ACT CAR-M, DAC+ACT CAR-M. Total RNA from lysed cancer cells was purified, quality-checked, and preserved for library construction and sequencing. Comparisons were conducted between PBS versus DAC and ACT CAR-M versus DAC+ACT CAR-M. (D) Principal component analysis shows the closeness of sample clustering in each treatment group. The -1, -2, and -3 suffixes denote the three independent biological replicates for each treatment condition. Each replicate was derived from a separate cell culture experiment. (E) Volcano map analysis of upregulated and down-regulated genes of differential genes in DAC pretreatment group. (F) Key differential gene expression caused by DAC pretreatment in MKN45 cells. PBS1–3 and DAC1–3 represent three independent biological replicates per group (n=3). Each line represents a separate replicate experiment of the given condition. Y-axis values are log₂ fold changes calculated from raw sequencing data of differentially expressed genes (range: −2 to 2), with positive values for upregulated genes and negative values for downregulated genes. (G) Volcano map analysis of upregulated and downregulated genes of differential genes in DAC+ACT CAR-M group compared with ACT CAR-M group. (H) Heat map analysis of differential gene expression caused by combination therapy. Color scale represents log₂-transformed fold change values. (I) KEGG enrichment analysis showed that related pathways were significantly enriched in DAC+ACT CAR-M group compared with ACT CAR-M group. (J) GSEA of significantly enriched signaling pathways in DAC+ACT CAR-M group compared with ACT CAR-M group. Data are mean±SEM (n=3); *p<0.05, **p<0.01, ***p<0.001. CCK-8, Cell Counting Kit-8; DAC, decitabine; E:T, effector-to-target; ES, enrichment score; GSEA, Gene Set Enrichment Analysis; IL, interleukin; JAK-STAT, Janus kinase-signal transducer and activator of transcription; KEGG, Kyoto Encyclopedia of Genes and Genomes; NF, nuclear factor; PBS, phosphate-buffered saline; PPAR, peroxisome proliferator-activated receptor; TNF, tumor necrosis factor; VEGF, vascular endothelial growth factor.

We aimed to decipher how DAC-induced epigenetic reprogramming fundamentally alters the intrinsic cellular regulatory networks, thereby creating a molecular milieu that enhances susceptibility to CAR-M-mediated cytotoxicity. Human gastric cancer cell line MKN45 was selected as a template to conduct RNA sequencing analysis. Different groups, PBS, DAC, CAR-M, and DAC+CAR-M, were designed to elucidate the mechanisms by which CAR-M therapeutic efficacy was enhanced (figure 3C). First, the comparison between the DAC and PBS group was given attention because DAC pretreatment did not cause significant changes in cell viability. Interestingly, 1447 significant DEGs were identified in the DAC versus PBS group (figure 3D,E). We delved deeper into the transcriptional profiles and focused attention on DNA methylation regulatory-related DEGs. We observed that DAC treatment modulates pyroptosis-related genes such as GSDME and GSDMD, in MKN45 cancer cells (figure 3F). Meanwhile, a series of immune response-associated genes were identified following DAC treatment. Notably upregulated genes included CSF1R, CCL5, CXCL11, and CXCL16 (figure 3F), which are involved in immune cell recruitment and activation. Importantly, transcriptomic profiling of AZA-treated MKN45 cells was performed for in vitro validation. As a structurally distinct DNMT inhibitor, AZA regulated pyroptosis and immune recruitment genes in a pattern consistent with DAC, with GSDME being markedly downregulated (online supplemental figure S4A,B). In agreement with the findings of the present sequencing, TIMER2.0 analysis showed that low GSDME expression led to diminished immune cell infiltration in gastric cancer (online supplemental figure S4C,D).

As expected, both CAR-M and DAC+CAR-M treatment groups triggered significant alterations in cancer cell transcripts. We focused on the potential mechanisms underlying the enhanced efficacy when DAC+CAR-M treatment was compared with CAR-M alone treatment. A total of 820 DEGs were obtained in the DAC+CAR-M versus CAR-M group (figure 3G). Similar gene expression patterns related to epigenetic regulation were still observed, such as YTHDF3, DNMT1, and WTAP, which shared similar expression changes with those observed in the DAC versus PBS group (figure 3H). Notably, key immunostimulatory cytokines and chemokines—including IL-15, CCL7, and CXCL8—demonstrated substantial upregulation in the DAC+CAR-M treatment group (figure 3H). Further KEGG pathway analysis revealed significant upregulation of multiple core inflammatory and immune response pathways in the DAC+CAR-M group, including the NF-κB and JAK-STAT signaling pathways (figure 3I). This pattern suggests that DAC-mediated reprogramming induces a TME that is immunologically sensitized and more susceptible to CAR-M-mediated attack. This enrichment suggests that DAC-induced reprogramming promotes an immunologically sensitized and therapeutically responsive TME, thereby facilitating the antitumor activity of CAR-M cells. Additionally, we observed that the DAC+CAR-M group presented a more prominent tumor proliferation-suppressing response compared with the CAR-M group, including downregulated pathways such as the cell cycle and VEGF signaling pathways. Using GSEA, the upregulation of hallmark immune and apoptotic pathways was further highlighted, including the IL-17 signaling, TNF signaling, and apoptosis-related gene sets (figure 3J and online supplemental figure S5). Together, these findings suggest a potential role for DAC in reprogramming both the intracellular signaling landscape and the tumor immune microenvironment, which may underlie the enhanced responsiveness to CAR-M immunotherapy.

DAC enhances ACT CAR-M-mediated tumor cell death by transforming apoptosis into pyroptosis

We have observed the alterations at the transcriptional level of MKN45 cancer cells caused by epigenetic modifications in the DAC versus PBS comparison, and also obtained the differential signaling pathways regulating the outcomes of the two treatment approaches in the DAC+CAR-M versus CAR-M comparison. We next considered what possible internal mechanisms are responsible for such treatment outcomes. GSDME-mediated pyroptosis represents an alternative immunogenic cell death that is conducive to immunotherapy.²⁹ Prior work reported that GSDME was epigenetically silenced in cancers owing to promoter hypermethylation.³⁰ We first verified whether DAC demethylation upregulates the mRNA expression of GSDME in MKN45, SKOV3, and HER2⁺-4T1 cells using a dose range of DAC treatment by quantitative PCR analysis. Dose-response analysis (0–5 μM DAC for 24 hours) demonstrated maximal GSDME induction at 1 μM (figure 4A), while time-course experiments (1 μM DAC for 0–120 hours) revealed peak expression at 24 hours (figure 4B). These data support that low-dose, short-duration DAC treatment (1 μM, 24 hours) optimally enhances GSDME expression (figure 4A,B). Thus, the convergence of optimal dose (1 μM) and duration (24 hours) defines a therapeutic window that maximizes GSDME upregulation efficiency.

Figure 4. DAC enhances ACT CAR-M-mediated tumor cell death by transforming apoptosis into pyroptosis. (A) Relative mRNA expression of GSDME in different groups. A dose-response was conducted with a fixed treatment duration of 24 hours across different DAC concentrations (0–5 μM). (B) Relative mRNA expression levels of GSDME following time-course treatment with a fixed 1 µM DAC concentration. (C–F) The analysis of ATP and LDH levels among different groups. Cell lysis was quantified via colorimetric LDH release assay and extracellular ATP was measured using bioluminescent detection. (G) Representative microscope images of MKN45 cells after different treatments; red arrows show pyroptosis. Scale bar: 100 μm. (H) Relative mRNA expression of caspase-3 was detected by qRT-PCR in different groups. (I, J) Indicated proteins for in different groups were analyzed by western blot analysis. N, N-terminus only protein. (K, L) Flow cytometric analysis and statistical results for apoptosis and pyroptosis in MKN45 and HER2⁺-4T1 cells after different treatments based on Annexin V-FITC and PI co-staining protocol. The upper row of plots corresponds to MKN45 cells and the lower row corresponds to HER2⁺-4T1 cells. Data are mean±SEM (n=3); *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. ACT CAR-M, CAR-M cells using an α1β1 integrin-mediated FcγRI intracellular domain, named as active CAR-M; CAR-M, chimeric antigen receptor macrophages; DAC, decitabine; FITC, fluorescein isothiocyanate; FL, full-length; FcγRI, Fc gamma receptor I; GSDME, gasdermin E; HER2, human epidermal growth factor receptor 2; LDH, lactate dehydrogenase; mRNA, messenger RNA; PI, propidium iodide; qRT-PCR, quantitative real-time PCR; UTD-M, untransduced macrophages.

Central to pyroptosis is the GSDME-mediated pore formation and membrane rupture, through which intracellular ATP and LDH are released, thereby supporting these molecules as ideal biomarkers for quantifying pyroptotic cell death. We examined the extracellular levels of ATP and LDH in MKN45 cells following various treatments, and significantly enhanced results were found in the DAC+CAR-M group (figure 4C,D). Consistent results were obtained in the HER2⁺-4T1 cell lines (figure 4E,F). Furthermore, MKN45 cells pretreated with DAC and treated with ACT CAR-M displayed the classic pyroptotic features including massive swelling, membrane blebbing, and plasma membrane rupture (figure 4G), confirming the successfully induced pyroptotic cell death.

HER2 blocking is known to activate caspase-3-mediated apoptosis.³¹ More recently, GSDME was identified as one of the pore-forming proteins leading to pyroptosis after its cleavage into the GSDME-N (N terminus of GSDME) by activated caspase-3.³² Since HER2-targeted CAR-M therapy significantly induces caspase-3 expression (figure 4H,I), which supports the activation of apoptotic signaling,³³ we further explored the underlying mechanisms of the DAC-mediated switch from apoptosis to pyroptosis in MKN45 cells. The expression and proteolytic processing of both caspase-3 and GSDME were systematically examined. Western blot analysis revealed that DAC treatment significantly upregulated the expression of GSDME protein, which subsequently underwent ACT CAR-M-induced cleavage to generate its active forms (figure 4I). Notably, neither did ACT CAR-M monotherapy enhance GSDME expression nor did DAC alone trigger GSDME cleavage (figure 4I,J), indicating that the synergistic action of both treatments is essential for pyroptosis induction. Flow cytometric analysis confirmed that the DAC+ACT CAR-M combination significantly increased the proportion of Annexin V⁺/PI⁺ double-positive cells in both MKN45 and HER2⁺-4T1 cell lines compared with control or monotherapy groups (figure 4K,L).

Collectively, these findings support that DAC-mediated GSDME upregulation triggers HER2 CAR-M-induced pyroptosis in cancer cells through a caspase-3-dependent mechanism, effectively converting the default apoptotic pathway into inflammatory pyroptotic cell death.

DAC reshapes the tumor immune microenvironment to improve ACT CAR-M cancer immunotherapy

Considering that a series of immune-related signaling pathways were identified in the analysis of DAC+CAR-M versus CAR-M, we next explored whether DAC epigenetic modifications could promote immune enhancement mediated by ACT CAR-M cell therapy. To characterize the initial efficacy study in figure 2, this experiment was designed to elucidate the TME immune remodeling mechanisms and to evaluate the therapeutic efficacy of DAC pretreatment with varying CAR-M doses. We constructed a BALB/c mouse tumor model with a functional immune system using HER2⁺-4T1 cells (1×10⁶), which were respectively treated with PBS, DAC, CAR-M, or DAC+CAR-M, with an additional DAC+a half-dose CAR-M group as a challenge group (figure 5A). ACT CAR-M cells derived from J774A.1 were administered intravenously at either 7×10⁶ cells (full dose, groups G3 and G5) or 3.5×10⁶ cells (half dose, group G4). Rapid tumor progression was observed in the PBS (G1) and DAC (G2) treatment groups, whereas the treatment with ACT CAR-M cells (G3) significantly enhanced tumor inhibition (figure 5B). Interestingly, an equivalent therapeutic effect was observed in the mice treated with DAC+a half-dose CAR-M (G4) compared with the ACT CAR-M (G3) group (figure 5B). As predicted, the DAC+CAR M (G5) treatment group achieved significantly greater tumor suppression than all the other groups (figure 5B,C and online supplemental figure S6A). No significant differences were observed in the body weight or organ tissue histology of mice treated with G2, G3, G4, and G5 compared with G1 (online supplemental figure S6B,C).

Figure 5. DAC reshapes the tumor immune microenvironment to improve ACT CAR-M cell-based cancer immunotherapy. (A) Schematic of the HER2⁺-4T1 breast cancer model and treatment group. Female BALB/c mice (aged 4–6 weeks) were inoculated s.c. with 1×10⁶ HER2⁺-4T1 cells. DAC (0.5 mg/kg, i.v.) was administered on days 0, 2, and 4. J774A.1-derived ACT CAR-M cells (7×10⁶ or 3.5×10⁶, i.v.) were injected on day 5. Treatment groups: G1, PBS; G2, DAC; G3, ACT CAR-M; G4, DAC+half dose ACT CAR-M; G5, DAC+ACT CAR-M (n=3). (B) Tumor volume of mice in different groups. (C) Quantitative analysis of tumor weight at day 37 experimental endpoint post-treatment. (D) Percentage of CAR-M in tumors of different groups. (E) Percentage of M1, M2 phenotypes on CAR-M in tumors of different groups. (F) Percentage of M1, M2 phenotypes on macrophages in tumor microenvironment of different groups. (G) Percentage of CD8⁺ T cells and CD4⁺ T cells phenotypes in tumors of different groups. (H) Percentage of B cells in tumors of different groups. Data are mean±SEM (n=3); *p<0.05, **p<0.01, ***p<0.001. CAR-M, chimeric antigen receptor macrophages; DAC, decitabine; HER2, human epidermal growth factor receptor 2; i.v., intravenous; ns, not significant; PBS, phosphate-buffered saline; s.c., subcutaneous.

We then investigated whether this improved therapeutic effect was attributed to the immune-related changes in the TME. Quantitative analysis of tumor weight was conducted at the day 37 experimental endpoint when tumor volume reached 1000 mm³ for flow cytometry-based mechanistic profiling on key immune cell populations. As a focused central role in the present study, we first quantified CAR-M enrichment within the TME, and a dose-dependent effect on CAR-M enrichment was observed when comparing DAC+a half-dose CAR-M with DAC+CAR-M (figure 5D). Moreover, the DAC+CAR-M group showed significantly increased expansion of CAR-M cells compared with those of ACT CAR-M group (figure 5D).

Importantly, although the M1 polarization of macrophages showed no statistically significant difference between the DAC+CAR-M group (G5) and the CAR-M monotherapy group (G3) (figure 5E and online supplemental figure S6), further analysis of macrophage subsets in the TME demonstrated that DAC-mediated epigenetic modification reshaped macrophage populations following CAR-M infusion, leading to a significant increase in M1-phenotype macrophages (figure 5F and online supplemental figure S6E). Consistently, quantitative analysis of the M1/M2 ratio supported these findings (online supplemental figure S6F-K). In addition to the comparative analysis of macrophage surface markers CD80, CD86, and CD206, immunofluorescence staining of the endogenous proteins inducible nitric oxide synthase and arginase 1 was further performed to characterize macrophage polarization (online supplemental figure S6L-N), which yielded consistent results.

Specifically, consistent with the enhanced antitumor efficacy, flow cytometric analysis revealed that DAC pretreatment significantly increased T-cell infiltration while DAC+CAR-M treatment resulted in significantly increased CD8⁺ T-cell infiltration compared with ACT CAR-M treatment alone (figure 5G). Meanwhile, the results on figure 5H showed the DAC+CAR-M treatment group (G5) exhibited greater B-cell infiltration relative to the CAR-M monotherapy group (G3). Collectively, the DAC-mediated epigenetic modification strategy significantly reshapes the tumor immune microenvironment, and the resultant phenotypic observations facilitate a more comprehensive interpretation of our signaling pathway analysis.

Although DAC preprocessing reshapes the TME to potentiate CAR-M function, its effects on systemic therapy-associated toxicities remain uncharacterized. Here, the systemic toxicities of DAC+CAR-M therapy were systematically evaluated by monitoring CRS and TLS risks. Serial blood samples were collected at days 1, 2, 3, and 5 post-CAR-M infusion (covering the peak period of CRS and TLS) to detect CRS-related pro-inflammatory cytokines, hepatic/renal injury markers, and TLS-associated metabolic and electrolyte indicators. Multitimepoint multivariate analysis excluding individual mouse variations further confirmed the favorable safety profile of single CAR-M treatment and demonstrated that DAC preconditioning did not exacerbate CRS or TLS during combined CAR-M therapy (online supplemental figure S7A-D).

Preliminary evaluation of the safety and efficacy of DAC demethylation-mediated ACT CAR-M therapy in immunodeficient mouse tumor models

In order to approach preclinical evaluation, we continued to use MKN45 to establish a NCG mouse subcutaneous tumor model for assessing the efficacy of the DAC+CAR-M group compared with the ACT CAR-M group (figure 6A). Treatment with DAC+CAR-M led to significantly enhanced tumor control compared with ACT CAR-M (figure 6B). No significant differences were observed in the body weight, peripheral blood parameters, or organ tissue histology of mice treated with DAC, ACT CAR-M, and DAC+ACT CAR-M compared with PBS (figure 6C,D and online supplemental figure S8), assessed at the experimental endpoint, day 23 post-treatment. We further performed RNA-seq analysis on tumor tissues. DEGs related to inflammatory immunity were identified between DAC+ACT CAR-M and ACT CAR-M (figure 6E). The transcriptome changes induced by DAC combined with ACT CAR-M were mainly enriched in a series of pathways related to inflammatory response (figure 6F), including the ‘NF-κB’ and ‘ETS2 signaling pathway’ according to GSEA (figure 6G). These data point to an important role for DAC demethylation in enhancing CAR-M antitumor activity even in the absence of a complete immune system. In vivo experiments revealed upregulated GSDME as well as its cleaved GSDME-N fragment in the DAC+CAR-M group compared with the CAR-M alone group (online supplemental figure S9). Collectively, these in vivo results substantiate that enhanced antitumor efficacy in the DAC+CAR-M group stems from DAC-induced pyroptosis conversion of CAR-M-triggered apoptotic cell death.

Figure 6. DAC-sensitized ACT CAR-M therapy in NCG mouse subcutaneous tumor model. (A) Schematic for subcutaneous injection MKN45 NCG tumor model and treatment (n=5). NCG mice were inoculated subcutaneously with 5×10⁶ MKN45 cells in 100 μL PBS on day −7. DAC (0.5 mg/kg, i.v.) was administered on days 0, 2, and 4. THP-1-derived ACT CAR-M cells (1.5×10⁶, i.v.) were injected on day 5. Treatment groups were: G1, PBS; G2, DAC; G3, ACT CAR-M; G4, DAC+ACT CAR-M (n=5). (B) Tumor volume of mice in different groups. (C) Body weights in different groups. (D) The levels of RBC, HGB, WBC, and PLT in different groups measured at the experimental endpoint (day 23 post-treatment). (E) Hierarchical clustering of the RNA-seq analysis results shows differentially expressed genes between ACT CAR-M cells and ACT CAR-M cells combined with DAC. (F) GO group analysis was performed for signaling pathways induced by both DAC and ACT CAR-M in tumors using the online tool DAVID. Only the GOs with p<0.05 are displayed. (G) Representative GSEA enrichment plot demonstrating the regulation of signaling pathways between ACT CAR-M cells and ACT CAR-M cells combined with DAC. Data are mean±SEM (n=5); **p<0.01. CAR-M, chimeric antigen receptor macrophages; DAC, decitabine; DAVID, Database for Annotation, Visualization and Integrated Discovery; ETS2, ETS proto-oncogene 2; GO, Gene Ontology; GSDME, gasdermin E; GSEA, Gene Set Enrichment Analysis; HGB, hemoglobin; IL, interleukin; i.v., intravenous; NF-κB, nuclear factor kappa B; PBS, phosphate-buffered saline; PLT, platelet count; RBC, red blood cell count; RNA-seq, RNA sequencing; s.c., subcutaneous; WBC, white blood cell count.

DAC demethylation-mediated ACT CAR-M demonstrates efficacy in a mouse model of ovarian cancer peritoneal metastasis and preliminary validity verification with primary human monocytes in vitro

Next, we performed a tumor metastasis model to challenge the therapeutic effect of DAC demethylation-mediated ACT CAR-M. After SKOV3 luciferase-expressing tumor cells were injected intraperitoneally into athymic BALB/c nude mice and tracked with bioluminescent imaging, mice were regionally treated intravenously with either PBS, DAC, ACT CAR-M alone, or DAC+ACT CAR-M (figure 7A). Compared with the control, significant therapeutic responses were observed in ACT CAR-M alone or DAC+ACT CAR-M treated mice by quantitative analysis of luminescence intensity from intraperitoneal luciferase-labeled tumor xenografts, with the most significant antitumor effect obtained in the DAC+ACT CAR-M group. Interestingly, we still observed an equivalent clearance effect on SKOV3 cells for DAC+a half-dose CAR-M compared with ACT CAR-M (figure 7B–I). Kaplan-Meier survival analysis further confirmed prolonged humane-endpoint-free survival in CAR-M-treated groups compared with the PBS control group, with the most favorable survival outcome observed in the DAC+ACT CAR-M group (figure 7I). Meanwhile, no significant body weight loss was observed in any treatment group (figure 7J). In summary, our findings support the therapeutic potential of DAC to enhance the favorable efficacy of CAR-M cells in vivo while maintaining a safety profile, thereby highlighting the translational potential of this approach for clinical applications.

Figure 7. DAC demethylation-mediated ACT CAR-M demonstrates safety and efficacy in a mouse model of ovarian cancer peritoneal metastasis. (A) Schematic for i.p. SKOV3 tumor athymic nude mice model and treatment (n=5). SKOV3-luciferase cells (5×10⁶) were injected i.p. on day 0. DAC (0.5 mg/kg, i.v.) was administered on days 0, 2, and 4. THP-1-derived ACT CAR-M cells (0.75×10⁶ or 1.5×10⁶, i.v.) were injected on day 5. Bioluminescent imaging was performed every 5 days from day 0. Treatment groups: G1, PBS; G2, DAC; G3, ACT CAR-M; G4, DAC+half dose ACT CAR-M; G5, DAC+ACT CAR-M (n=5). (B) Representative bioluminescent flux imaging of tumor-bearing mice treated with PBS, DAC, ACT CAR-M, or DAC+ACT CAR-M combined with (n=5). (C–H) Tumor bioluminescence intensity growth kinetics in different groups. (I) Survival rates of mice with different treatments. (J) Body weights in different groups (n=5). (K, L) Cytotoxicity of primary human CD14⁺ monocyte-derived CAR-M against DAC-pretreated MKN45 (K) and SKOV3 (L) cells at standard and half doses, assessed by CCK-8 assay, (n=3); *p<0.05, **p<0.01, . CAR-M, chimeric antigen receptor macrophages; CCK-8, Cell Counting Kit-8; DAC, decitabine; i.p., intraperitoneal; i.v., intravenous; ns, not significant; PBS, phosphate-buffered saline.

To prepare for future clinical research, we performed additional experiments using CAR-M cells derived from primary human CD14⁺ monocytes rather than the THP-1 cell line. Primary CD14⁺ monocyte-derived CAR-M cells were co-cultured with DAC-pretreated MKN45 cells or SKOV3 cells, and cytotoxicity assays were performed. Consistent with figure 3A, DAC pretreatment significantly enhanced the cytotoxic efficacy of primary CAR-M cells against MKN45 and SKOV3 tumor cells (figure 7K, L). Remarkably, DAC-pretreated tumor cells co-cultured with a reduced CAR-M cell dose (50% of the standard dose) demonstrated cytotoxic responses comparable to those observed with the full-dose CAR-M alone, highlighting the critical role of DAC in augmenting CAR-M-mediated cytotoxicity and supporting the clinical translatability of this combinatorial approach.

Discussion

Solid tumor treatment remains one of the most formidable challenges in oncology, largely due to the complex interplay of tumor cell heterogeneity and the immunosuppressive tumor TME.³ Recent advances have highlighted the critical role of epigenetic modifications in driving transcriptional plasticity, enabling cancer cells to achieve proliferative expansion and malignant progression while evading immune surveillance.³⁴ These epigenetic alterations create a dynamic landscape where single-agent therapies, including CAR-based immunotherapies, struggle to maintain durable responses against the evolving tumor burden. Unlike CAR-T cells, which face significant barriers in solid tumor infiltration and immune exhaustion, CAR-M possesses inherent advantages, including superior tumor penetration capacity, phagocytic activity,³⁵ and the ability to reshape the TME. However, few studies have explored how epigenetic alterations in tumors impact the efficacy of CAR-M cell therapy.

Objectively, the in vivo proliferative potential of CAR-M is inferior to that of CAR-T cells. Researchers have shown that CAR-M is cleared from circulation within 4–48 hours but successfully traffics to tumor sites in which CAR-M was detected in the TME of 75% of patients at day 8 biopsy and 27% of patients at week 4 biopsy (USA CT-0508 ZI KAN2025).⁷ The relatively short in vivo persistence of CAR-M, as evidenced by these clinical data, underscores the critical need to optimize their domain(s) for enhanced cellular activity. In our previous pioneering work, after CD3ζ was confirmed to be hardly expressed on macrophages, we designed a CAR-M harboring the α1β1 integrin-mediated FcγRI signaling domain to supersede the conventional CD3ζ domain-based CAR-M. In the previous study, we constructed HER2-targeted ACT CAR-M and CD3ζ CAR-M, and found that ACT CAR-M treatment achieved 100% survival at 85 days in the HER2⁺-4T1 subcutaneous tumor model, compared with only 33.3% survival in the CD3ζ CAR-M group.²⁰ In the present study, using the established HER2⁺-4T1 tumor model, flow cytometric analysis at multiple time points demonstrated that the ACT CAR-M group exhibited significantly greater tumor infiltration than the CD3ζ domain group in a time-dependent pattern. This mechanism can be partially explained by our previously performed in vitro EdU proliferation assays,²⁰ in which α1β1 integrin mediated FcγRI engagement promoted the proliferation of CAR-M cells more effectively than that of the CD3ζ signaling domain following HER2 antigen stimulation. Meanwhile, the present study demonstrated that the α1β1 integrin-mediated FcγRI domain significantly enhanced the phagocytic capacity of CAR-M in tumors in vivo compared with the CD3ζ domain group as treatment duration increased. Beyond directly reducing tumor cell burden, the markedly enhanced phagocytic activity of CAR-M serves a pivotal role in promoting robust antigen presentation to T cells, thereby substantially amplifying systemic antitumor immune responses. A phase I clinical trial of CAR-M therapy for HER2-overexpressing advanced solid tumors demonstrated favorable safety and tolerability of CT-0508 (a CD3ζ-based CAR-M construct),⁷ which supports the potential clinical translatability of the current ACT CAR-M design.

Given the epigenetic changes within the TME, tumors develop primary/acquired drug resistance and exhibit reduced responsiveness to immunotherapies, partly mediated by epigenetic alterations.³⁶ The tumor responsiveness mechanisms underlying epigenetic reprogramming enhanced CAR-M cytotoxicity remain underexplored. Here, we identified DNA methylation as a potent entry point for immune enhancement strategies on CAR-M, demonstrating that demethylation induced by DAC promotes CAR-M-mediated tumor destruction. Existing work confirms DAC enhances CAR-T function, and low-dose DAC epigenetically remodels CAR-T cells to suppress exhaustion genes and prolong in vivo antitumor persistence.³⁷ Considering prospective clinical translation, we established a feasible bridging protocol consisting of DAC preconditioning before CAR-M cell infusion.

Particularly striking is our observation that although DAC pretreatment does not directly induce cytotoxicity in cancer cells, a half-dose ACT CAR-M with DAC pretreatment can achieve cytotoxic efficacy comparable to full-dose CAR-M in vitro. Furthermore, the antitumor efficacy of ACT CAR-M was significantly enhanced following DAC pre-administration in the mouse HER2⁺-4T1 subcutaneous tumor model. This HER2⁺-4T1 BALB/c mouse model possesses an intact immune system, and thus these findings may be explained by two hypotheses.

First, the mechanistic basis for this synergistic interaction appears to stem from DAC-mediated sensitization of tumor cells to CAR-M cytotoxicity, creating a more amenable gene expression profile in cancer cells. Consistent with this finding, our comprehensive transcriptomic analyses reveal that DAC treatment induces widespread reprogramming of cancer cell gene expression to increase its sensitivity to ACT CAR-M therapy. Interestingly, although DAC treatment does not induce direct tumor cell killing, it induces a series of gene expression changes with demethylation effects. One of the key synergistic mechanisms is our identification of GSDME as a crucial mediator of the DAC and ACT CAR-M combination effect. GSDME, frequently silenced in cancers through promoter hypermethylation, is re-expressed following DAC treatment through demethylation mechanisms.^{32 38} This epigenetic reversal primes cancer cells for caspase-3-dependent pyroptotic cell death on CAR-M engagement, fundamentally altering the mode of tumor cell death from immunologically ‘quiet’ apoptosis to inflammatory pyroptosis. Pyroptosis releases damage-associated molecular patterns including ATP and pro-inflammatory cytokines, creating a cascade of immune activation that amplifies the initial CAR-M response. Our experimental validation confirmed this mechanistic switch. Canonical gasdermin family member gasdermin D drives pyroptosis and innate immune activation, eliciting potent antitumor immunity.^{23 39} Importantly, our work identifies GSDME as a critical mediator that contributes to CAR-M therapeutic efficacy.

Second, DAC treatment promotes favorable immune reprogramming of the TME, which is conducive to enhanced ACT CAR-M efficacy. Consistent with this hypothesis, high-throughput sequencing of tumors treated with DAC-pretreated CAR-M revealed not only specific alterations in epigenetic regulators (DNMT1,⁴⁰ YTHDF3,⁴¹ WTAP⁴²), but also distinct changes in immune-related genes (CXCL8, CXCL16, CCL7, IL-15, IL-18), along with upregulated activation of inflammatory signaling pathways. Few studies have explored the influence of epigenetics on CAR-M cell therapy. It was reported that DNA demethylation activated expression of C-X-C inflammatory chemokines including CXCL1, CXCL2, CXCL5, and CXCL8, and triggered neutrophil-dependent lung metastasis in preclinical models of renal cancer.⁴³ The present data indicate DAC does not directly trigger macrophage repolarization, and DAC pretreatment neither elevates the proportion of M1-phenotype cells nor compromises the temporal durability of the M1 phenotype. However, this study showed that epigenetic modifiers can reverse immunosuppressive programs when combined with CAR-M therapy. Accordingly, a pronounced responsiveness of the NF-κB and JAK-STAT pathways would be expected to favor CAR-M efficacy by reinforcing pro-inflammatory macrophage phenotypes, tumoricidal effector functions, and immune-amplifying cues that are central to CAR-M-mediated antitumor activity.⁴⁴ It is indicated that DNA demethylation-mediated immune regulation exerts a double-edged sword effect. In inflammatory cancers, it may accelerate tumor progression. Coincidentally, the present CAR-M possesses the ability to mediate recruit immune cells and reshape the TME toward a pro-inflammatory state that is beneficial for tumor treatment. The present immune profiling analysis of the TME revealed that DAC treatment promotes favorable immune reprogramming by sustaining the M1-like polarization of ACT CAR-M cells while preventing their conversion into immunosuppressive M2-like macrophages. This reprogramming was further evidenced by enhanced CD8⁺ T-cell infiltration, which collectively remodeled the TME into a more inflammatory and therapeutically responsive state, thereby overcoming the inherently immune-excluded tumor phenotype. These findings are consistent with our transcriptomic data, which suggest that the synergistic efficacy of DAC combined with ACT CAR-M therapy is mediated through the upregulation of immune-stimulatory pathways.

Furthermore, we validated DAC demethylation-mediated CAR-M strategy in NCG mouse models bearing human gastric adenocarcinoma. Significant efficacy profiles were observed, which the transcriptomic analysis of tumor tissues revealed similar inflammatory immune gene expression patterns and pathway enrichment (NF-κB and ETS2 signaling) obtained in vitro studies. No significant toxicity was observed in body weight, peripheral blood parameters, and organ histology, supporting further progress toward clinical research in the future. Our previous study demonstrated that ACT CAR-M exerted superior antitumor efficacy relative to CD3ζ CAR-M in an intraperitoneal carcinomatosis model.²⁰ Consistently, the SKOV3 cell-derived peritoneal metastasis model further verified the robust anti-metastatic capacity of DAC demethylation-primed CAR-M therapy, which yielded remarkable tumor suppression in this refractory metastatic scenario.

CAR-M eliminates tumors primarily through phagocytosis and is presumed to induce milder systemic inflammatory responses than cytotoxic CAR-T cells, which are prone to triggering severe CRS. We therefore systematically assessed the CRS and TLS risks of combined DAC and CAR-M therapy in immunocompetent tumor-bearing mice. The results revealed that standalone CAR-M exhibited a favorable toxicological profile, and DAC pretreatment did not exacerbate CRS or TLS following combinatorial treatment. Despite the inherent limitations of murine preclinical models, these safety findings provide credible preclinical evidence for the translational potential of DAC plus CAR-M combination therapy, rendering the present strategy a feasible alternative for future clinical practice.

Although we demonstrated that DAC demethylation-mediated CAR-M therapy significantly enhances the antitumor and antimetastasis effects on tumor in immunodeficient mouse models compared with single-agent CAR-M therapy, we also observed an equivalent antitumor effect between DAC+a half-dose CAR-M and a dose CAR-M in peritoneal metastasis mouse model. But immunodeficient mice lack a sound immune system, whether DAC demethylation-mediated CAR-M therapy could trigger human lymphocytes to obtain a powerful promoting immunotherapy needs to be explored in future studies. Besides, there are still several limitations in our present work. First, although we have posited DAC demethylation-mediated CAR-M therapy with the potential to induce pyroptosis-related genetic alterations and share the remodeling of the immune microenvironment, using RNA-seq enrichment analysis on DEGs should obtain other potential mechanisms to need be supported. Second, HER2 tumor antigen exhibits target-specific regulatory mechanism mediating apoptosis turns into pyroptosis. In this study, we did not fully explore the impact of differential tumor antigen targets on the DAC demethylation-mediated CAR-M therapy. More expansive research works are required to discover the role of differential tumor antigen in enhancing DAC demethylation-mediated CAR-M therapy efficacy. Finally, although we obtained significant antitumor effect with murine CAR-M model (constructed by J774A.1 strain) in tumor-bearing mice, further investigation is required to evaluate the efficacy validation of preclinical trials with primary mice or human macrophages constructed CAR-M for considering the limited survival time of primary macrophages in tumor-bearing mice. Meanwhile, determining the minimum effective epigenetic sensitization and safe dosage of DAC for human application remains one of the key focuses of future clinical studies. Building on our established human CAR-M platform,²⁰ future investigations will explore the capacity of clinically approved epigenetic drugs to potentiate CAR-M-based immunity—a strategy that holds substantial promise for maximizing therapeutic efficacy within the inherently constrained in vivo lifespan of primary cell-based CAR-M, or alternatively, for lowering the requisite cell therapy infusion dose without compromising acceptable therapeutic outcomes, ultimately broadening the clinical accessibility of CAR-M therapy.

Conclusions

In conclusion, based on α1β1 integrin-mediated FcγRI domain as the novel CAR-M, the integration of RNA-seq analysis, cell biology experiments, and animal studies highlights the role of DAC demethylation as a key mediator that potentiates CAR-M efficacy. DAC demethylation not only alters the gene expression profile of cancer cells to sensitize them to CAR-M therapy but also enhances redox and inflammatory immune responses related signaling pathways and reshapes the tumor immune microenvironment to create a favorable milieu for CAR-M therapy. Our mechanistic discoveries provide evidence supporting epigenetic demethylation as a promising target to enhance CAR-M efficacy, offering valuable insights to support the evolution of clinical CAR-M therapy.

The funders were not involved in the study design, data collection, analysis, report writing, or decision to submit the paper for publication.

Data availability statement

Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.