3D-printed implantable CAR-macrophages for post-surgery cancer immunotherapy

Abstract

Chimeric antigen receptor macrophages (CAR-M) therapy holds significant clinical potential, while its application is severely compromised by poor in vivo colonization and limited persistence. Here, we report a 3D-printed implantable CAR-M (iCAR-M) system, composed of a gelatin methacrylate (GelMA) scaffold loaded with IL13Rα2-targeting CAR-M, and signal regulatory protein-α-overexpressing extracellular vesicles (SIRPα-EVs). Functioning as a bioactive reservoir, the porous hydrogel mitigates anoikis and promotes nutrient transport, markedly elevating CAR-M viability and increasing the frequency of activated macrophages in mice from ~ 5% to ~ 30% after 7 days. Mechanistically, the co-released SIRPα-EVs mask CD47 on tumor cells to abrogate the “don’t eat me” signal, synergizing with CAR-mediated phagocytosis. In a 4T1 post-surgery model, this locoregional delivery system significantly inhibited tumor recurrence and remodeled the immunosuppressive microenvironment, boosting CD80⁺ macrophage polarization ~ 24% and CD8⁺ T-cell activation ~ 26%. Collectively, the iCAR-M functions as an engineered immune niche, integrating biomaterial-assisted persistence with checkpoint blockade to overcome solid tumor resistance.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04360-y.

Introduction

While surgery remains the primary treatment option for solid tumors, residual subclinical lesions and occult micro-metastases frequently cause post-surgery recurrence [1, 2]. Macrophages possess innate advantages for clearing residual disease through phagocytosis [3]. Genetically engineered chimeric antigen receptor macrophages (CAR-M) are engineered macrophages that can recognize tumor-associated antigens (TAAs) or tumor-specific antigens (TSAs), further offering enhanced targeting specificity and robust antitumor efficacy [4–6]. CAR-M demonstrates significant advantages in phagocytosis, antigen presentation, and the induction of a pro-inflammatory niche [7–9]. These features are particularly crucial for identifying and eliminating occult micro-metastases after surgery, positioning CAR-M as a robust strategy for treating potentially recurrent solid tumors.

Despite the remarkable tumor-suppressive potential of CAR-M, its clinical translation is hindered severely by poor in vivo colonization and limited persistence [8, 10]. Unlike T cells, primary macrophages are terminally differentiated cells that lack intrinsic proliferative capacity, which limits their expansion potential and long-term durability following infusion [11]. Furthermore, the hostile tumor microenvironment (TME) often induces CAR-M exhaustion or repolarizes them into an immunosuppressive M2-like phenotype, thereby compromising their stability and antitumor efficacy [12]. Additionally, cancer cells can overexpress anti-phagocytic surface proteins to evade macrophage clearance [13]. The high expression of CD47 “don’t eat me” checkpoint on tumor cells binds to the signal regulatory protein-α (SIRPα) on macrophages, inducing immune evasion and inhibiting CAR-M phagocytosis [14, 15]. While multiple competitive anti-CD47 antibodies have shown promising clinical outcomes, their application is inevitably limited by severe systemic toxicity [16]. Therefore, it is necessary to design a safe and efficient delivery strategy that maintains CAR-M activity and persistent release while synergistically boosting its tumor-clearing capacity.

3D bioprinting has emerged as a transformative technology in augmented manufacturing for biomedical applications [17–19]. As an advanced fabrication technique, 3D printing offers a robust strategy to construct patient-specific tissue architectures with high repeatability, precise structural fidelity, and rapid production capabilities, which are difficult to achieve with traditional hydrogel casting methods [20–22]. Bioinks, composed of living cells encapsulated within specific biomaterials, serve as the fundamental “raw materials” for this process [23]. Among them, gelatin methacryloyl (GelMA) hydrogel serves as an ideal bioink due to its tunable photo-crosslinking properties and inherent arginine-glycine-aspartic acid (RGD) motifs that mimic the native extracellular matrix (ECM) to support cell viability and spreading [24–26]. Extracellular vesicles (EVs), acting as natural delivery systems, have demonstrated significant potential in cancer vaccines due to their superior biocompatibility, active tumor-targeting ability, and prolonged systemic circulation [27–32]. Notably, recent advances in biosynthetic surface display technology allow for the high-density presentation of therapeutic proteins on EV membranes [33, 34]. In our previous study, we engineered hybrid EVs displaying high-affinity SIRPα variants to block the CD47 “don’t eat me” signal [35]. These hybrid EVs effectively amplify macrophage-mediated phagocytosis and potentiate immune responses against cancer recurrence and metastasis. Building on this, we here engineered SIRPα-EVs derived from HEK-293T cells to achieve efficient immune activation.

Here, we developed an implantable CAR-M (iCAR-M) system using 3D printing technology to construct a personalized immune niche. This system integrates IL-13Rα2-targeting CAR-M and SIRPα-EVs within a hierarchical porous GelMA hydrogel scaffold (Fig. 1A). Unlike traditional bolus injections, the 3D-printed porous architecture mimics the native ECM, thereby mitigating anoikis and facilitating nutrient transport. Consequently, this design significantly improves CAR-M viability and functions as a bioactive reservoir for their sustained locoregional release (Fig. 1B). Concurrently, the co-delivered SIRPα-EVs competitively mask the CD47 on tumor cells. This checkpoint blockade synergizes with CAR-mediated specific recognition to amplify phagocytosis and reverse the immunosuppressive TME (Fig. 1C). In vivo experiments demonstrate that this bioink hydrogel effectively inhibits post-surgery recurrence. Ultimately, the iCAR-M system integrates biomaterial-assisted persistence with dual-target immunotherapy to overcome solid tumor resistance.

Fig. 1.

Schematic illustration of the 3D-printed iCAR-M system and its anti-tumor mechanism. (A) Fabrication of 3D printing iCAR-M. Gel was modified into GelMA and mixed with a photoinitiator, LAP, to form a GelMA solution. CAR-M and SIRPα-EVs were added to the solution and printed into an iCAR-M, which was implanted subcutaneously into mice. CAR-M and SIRPα-EVs were gradually released into the mice’s bodies. (B) Post-surgery application of the iCAR-M system. The iCAR-M system improves CAR-M viability and acts as a reservoir for their sustained release to induce cancer cell death. (C) The specific mechanism of the iCAR-M system. IL-13Rα2-targeting CAR-M binds tumor cells, and the co-delivered SIRPα-EVs synergistically block CD47 to reverse immune suppression. Finally, the activated CAR-M in the 3D hydrogel causes cancer cell death

Results

Construction and characteristics of the 3D-printed hydrogel

Hydrogels are widely used in biomedical fields [36]. GelMA as a photocrosslinkable hydrogel with excellent biocompatibility [37], represents a highly promising system for drug or cell loading and delivery. To confirm GelMA’s successful synthesis, it undergoes cross-linking, transitioning from a sol phase to a gel phase upon irradiation with 405 nm light (Fig. 2A). The presence of characteristic peaks corresponding to carbon-carbon double bonds (C═C) and methyl groups (–CH₃) in the ¹H NMR spectrum of GelMA indicates successful grafting of methacrylic anhydride onto the gelatin (Gel) backbone (Fig. 2B). To examine the microstructure of GelMA, photocured hydrogels were dried and imaged using a scanning electron microscope (SEM, Fig. 2H). The results reveal an internally porous architecture, which can effectively accommodate cells and drugs while facilitating the permeation and transport of other substances. Porosity measurements further demonstrated that GelMA exhibits a porosity exceeding 50% (Fig. 2C). This interconnected porous network is conducive to drug release and nutrient diffusion. Swelling tests indicated an equilibrium swelling ratio of over 1500%, with nearly 1500% swelling achieved within 6 h (Fig. 2D and Figure S1), suggesting the hydrogel’s ability to adapt to the hydrated physiological environment and efficiently retain nutrients in aqueous media. Compression testing revealed a decrease in mechanical strength after cell loading, indicating increased operational difficulty for 3D printing of cell-laden constructs (Fig. 2E). Rheological time-sweep analysis showed that GelMA photocured within 1 min under 405 nm light at a power density of 30 mW cm² (Fig. 2F), while its viscosity remained relatively stable over time (Fig. 2G).

Fig. 2.

Synthesis and characterization of the GelMA hydrogel. (A) GelMA’s sol-gel transition. (B) ¹H NMR spectra of GelMA hydrogel. (C) The porosity of the freeze-dried GelMA hydrogel. (D) The equilibrium swelling rate of the GelMA hydroge. (E) The compression of the GelMA and Macrophages@GelMA hydrogels. (F) The dynamic time-sweep rheology of GelMA hydrogel exposed to a dose of 30 mW cm² of 405 nm laser. (G) The viscosity-time curve of the GelMA hydrogel. (H) The SEM of GelMA hydrogel. Scale bar, 20 μm. (I) The SEM of the bioprinting macrophages hydrogel. Scale bar, 40 μm. All data are presented as the mean ± s.d. (n = 3)

Owing to its photocurable nature, GelMA was processed using projection micro-stereolithography (PµSL) to fabricate cell-laden hydrogel constructs. Based on computer-aided design (CAD) models, bioink could be printed into scaffolds of various shapes and dimensions (Figure S2). After 2 days of culture in cell medium, the printed cell-laden hydrogels maintained their structural integrity and supported high cell viability (Figure S2). Importantly, SEM observation of dried bioink hydrogels revealed that cells can be loaded into the pore structure of the hydrogel (Fig. 2I), confirming the favorable biocompatibility of the scaffold.

Construction of IL-13 CAR-M and phagocytosis in vitro

Comparing different types of tumor cells in humans, IL13Rα2 was found to be overexpressed in tumors such as glioma and prostate carcinoma, and overexpression of IL-13Rα2 in ovarian carcinoma and breast cancer is associated with advanced disease and poor prognosis [38–43]. Therefore, we constructed a CAR structure carrying IL-13 to target IL13Rα2 on the surface of tumor cells. Macrophages were transduced with a lentiviral vector to generate IL-13 CAR-M in vitro (Fig. 3A and C). Both green fluorescent protein (GFP) fluorescence imaging, western blotting, and flow cytometry confirmed high transduction efficiency of the lentiviral vector (Fig. 3B, Figure S3). Subsequently, LPS was used to polarize CAR-M toward the anti-tumor M1 phenotype, and differentiated cells displayed characteristic M1 macrophage marker expression (Fig. 3D and E). To further evaluate the targeting and phagocytic capacity of CAR-M, we established a 4T1-IL13Rα2 cancer cell line (Fig. 3A and Figure S4). The cells were co-cultured with M0, M1 macrophages, or CAR-M to assess targeted phagocytosis. Flow cytometry results showed that phagocytosis of 4T1-IL13Rα2 cells by M0 macrophages was not significant, whereas CAR-M enhanced phagocytosis to nearly 27% (Fig. 3F and Figure S5). Additionally, confocal microscopy confirmed clear phagocytosis and co-localization between CAR-M and 4T1-IL13Rα2 cells (Fig. 3G and Figure S6). These results demonstrate the successful construction of IL-13 CAR-M and verify their targeted phagocytic ability against IL13Rα2-expressing tumor cells in vitro.

Fig. 3.

The construction of IL-13 CAR-M and phagocytosis in vitro. (A) The plasmid diagram in this study. (B) Flow cytometry analysis of CAR binding to macrophages. (C) The experimental flow of CAR-M recognizing and targeting tumor cells in vitro. (D-E) Flow cytometry analysis and quantification of CAR-M0 polarization towards M1-phenotype. (F) Quantitative analysis of macrophage phagocytosis of tumor cells in vitro. (G) Confocal Imaging of CAR-M (green)-mediated phagocytosis on 4T1-IL13Rα2 cells (red). Scale bar, 20 μm. All data are presented as the mean ± s.d. (n = 3). Statistical significance was assessed by the one-way ANOVA with Tukey’s multiple comparisons test. ns, no significance; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

3D GelMA scaffold promotes the survival of CAR-M

3D bioprinting technology has attained remarkable prominence in biomedical research and applications in recent years due to its intrinsic ability to fabricate living cell-laden bioinks [44]. To ensure the viability and immunocompetence of the encapsulated CAR-M, CAR-M were mixed with GelMA solution, and a CCK-8 assay performed after 6 days of culture revealed that the GelMA solution had a minimal impact on CAR-M viability (Fig. 4B). Furthermore, a 3D hydrogel was fabricated. Live/Dead staining on day 7 demonstrated that the number of viable CAR-M within the 3D hydrogel was significantly higher than in traditional hydrogel controls (Fig. 4A), indicating the superior capacity of the porous architecture to maintain high CAR-M activity.

Fig. 4.

The in vivo sustained release of CAR-M in a 3D hydrogel. (A) Live/Dead staining analysis of CAR-M in 3D hydrogel and traditional hydrogel. Scale bar, 100 μm. (B) The cell viability of CAR-M in bioprinting ink. (C-D) Quantitation of the proportion of MHC II⁺, CD80⁺, and CD86⁺ in CD11b⁺ F4/80⁺ cells on day 4 and day 7. (E-F) Flow cytometry analysis and quantitation of the percentage of GFP in CD11b⁺ F4/80⁺ cells on day 4 and day 7. All data are presented as the mean ± s.d. (n = 3). Statistical significance was assessed by the one-way ANOVA with Tukey’s multiple comparisons tests (C) or unpaired two-tailed t-test (D), or two-way ANOVA with Tukey’s multiple comparisons tests (F). ns, no significance; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

To evaluate the sustained release and anti-tumor effect of scaffold-encapsulated CAR-M in vivo, three formulations: hydrogel alone, CAR-M encapsulated in traditional hydrogel (CAR-M@Hydrogel), and CAR-M encapsulated in 3D hydrogel (CAR-M@3D Hydrogel) were subcutaneously implanted in a mouse post-surgery model. Tissue adjacent to the implants was harvested on days 4 and 7 for flow cytometric analysis. The results showed that the proportions of CD86⁺ and CD80⁺ macrophages in the CAR-M@3D Hydrogel group were higher than those in the CAR-M@Hydrogel group at both time points (Fig. 4C, D, and Figure S7). Correspondingly, the percentage of GFP-expressing CAR-M in the tissue was also higher in the CAR-M@3D Hydrogel group, verifying that 3D hydrogels can sustainably release CAR-M and enhance the immune capacity in vivo (Fig. 4E, F, and Figure S8).

SIRPα-EVs act as an immune assistant of CAR-M

Driven by their superior safety characteristics and robust stability, recent evidence implicates EVs in diverse physiological and pathological processes while highlighting their promising translational utility as immunotherapeutic modalities for cancer treatment [45, 46]. Enhancing the anti-tumor efficacy of CAR-M, we incorporated EVs displaying a mutant SIRPα as an immune assistant. By binding to ligands such as CD47, the SIRPα-EVs can attenuate the “don’t eat me” signal that otherwise inhibits macrophage phagocytosis (Fig. 5C). A lentiviral vector was used to transfect HEK-293T cells with a mutant SIRPα gene (Figure S9). Immunofluorescence (IF) staining and flow cytometry confirmed successful surface expression of SIRPα on the transfected cells (Fig. 5A and B). The constructed SIRPα-EVs had an average diameter of approximately 200 nm and a zeta potential of about − 20 mV (Fig. 5D and Figure S9). Western blotting analysis then verified the presence of SIRPα on 293T-derived SIRPα-EVs (Fig. 5E and Figure S10), and their cup-shaped vesicular morphology was confirmed by transmission electron microscopy (TEM, Fig. 5F). These results collectively demonstrate the successful preparation of SIRPα-EVs.

Fig. 5.

SIRPα-EVs as an immune assistant of CAR-M. (A) IF images of pristine cells and engineered cells. Scale bar, 10 μm. (B) Flow cytometry analysis of SIRPα in pristine cells and engineered cells. (C) Diagram of SIRPα-EVs blocks signal of the “don’t eat me”. (D) Nanoparticle DLS Analysis. (E) Western blotting analysis of Na⁺/K⁺-ATPase and SIRPα in 293T cells. (F) TEM image of SIRPα-EVs. Scale bar, 100 nm. (G) IF images of SIRPα-EVs targeting tumor CD47. Scale bar, 10 μm. (H) Imaging of CAR-M (green)-mediated phagocytosis on 4T1-IL13Rα2 cells (red) due to the CD47 checkpoint blockade by SIRPα-EVs. Scale bar, 20 μm. All data are presented as the mean ± s.d. (n = 4)

To further confirm the tumor-targeting ability of SIRPα-EVs in vitro, DiD-labeled SIRPα-EVs were co-incubated with tumor cells. The IF labeling of surface CD47 on tumor cells showed co-localization with SIRPα-EVs, indicating specific targeting (Fig. 5G). Subsequently, to investigate whether SIRPα-EVs could disrupt the CD47-SIRPα axis and enhance the anti-tumor function of CAR-M, a co-culture system of CAR-M and 4T1-IL13Rα2 tumor cells was established to assess phagocytosis. Confocal imaging and flow cytometry revealed that the addition of SIRPα-EVs to the co-culture system enhanced macrophage-mediated phagocytosis of tumor cells, compared to the origin-EVs control (Fig. 5H, Figure S6, Figure S11, and Figure S12). Moreover, ELISA of the co-culture supernatants showed that SIRPα-EVs treatment upregulated inflammatory cytokines such as IL-1β, IL-6, and NOS (Figure S12). These findings indicate that SIRPα-EVs can function similarly to conventional CD47-blocking antibodies, augmenting macrophage phagocytosis and thereby improving the efficacy of immunotherapy.

In vivo anti-tumor recurrence by the iCAR-M system

To evaluate the in vivo performance of the iCAR-M system, we engineered a GelMA hydrogel that co-encapsulates IL-13Rα2-targeting CAR-M and SIRPα-EVs (Fig. 6E). Firstly, a post-surgery recurrence model was established by inoculating 4T1-IL13Rα2 tumor cells into mice, followed by tumor resection and hydrogel implantation 8 days later. The following groups were subcutaneously implanted at the resection site: CAR-M and SIRPα-EVs loaded 3D hydrogel (iCAR-M), CAR-M loaded 3D Hydrogel (CAR-M@3D Hydrogel), CAR-M loaded traditional hydrogel (CAR-M@Hydrogel), blank traditional hydrogel (Hydrogel), and a control group receiving a subcutaneous injection of PBS (Fig. 6A).

Fig. 6.

Inhibition of tumor recurrence after surgery by iCAR-M. (A) Schematic illustration of in vivo experimental design and group identifications. (B-D) Tumor weights at the end of experiments, tumor growth curves, and body weight curves. (E) Schematic diagram of mouse implantation with 3D-printed hydrogels. (F-G) The infiltration and mean fluorescence intensity (MFI) of macrophages in the tumor were immunostained by antibodies against CD86 and CD206 in all treatment groups. Scale bar: 100 μm. (H) Quantitation of the proportion of CD80⁺, CD206⁺ in CD11b⁺ F4/80⁺ cells and CD8⁺ in CD3⁺ T cells. Data are presented as the mean ± s.d. (n = 5 for B-D and n = 4 for G-H). Statistical significance was assessed by the one-way ANOVA (B, G, H) and the two-way ANOVA (C) with a Tukey’s test. ns, no significance; *p < 0.05; **p < 0.01; ****p < 0.0001

The results showed no significant difference in tumor growth between the PBS and blank hydrogel groups (Fig. 6C), indicating that the hydrogel alone did not significantly inhibit tumor growth. In contrast, tumor growth was markedly suppressed in the CAR-M@3D Hydrogel group, with the most pronounced inhibition observed in the iCAR-M group (Fig. 6B and C). These findings suggest that the iCAR-M system can effectively enhance the anti-tumor activity in vivo. Besides, body weight curves indicated no significant decline in any treatment group (Fig. 6D). Blood routine and blood biochemistry analysis showed no difference in the treatment group (Figures S13 and S14). Immunohistochemistry showed no obvious toxicity in the treatment group (Figure S15).

Additionally, IF staining of tumor sections revealed a marked increase in CD86⁺ macrophages and a decrease in CD206⁺ macrophages in both the CAR-M@3D Hydrogel and iCAR-M groups (Fig. 6F and G). To elucidate the underlying anti-tumor mechanism, tumor tissues were analyzed by flow cytometry (Figure S16 and S17). The results demonstrated that the proportion of CD80⁺ macrophages in the iCAR-M group was significantly increased (approximately three-fold higher than in the PBS control). Within the CD3⁺ T cell population, the percentage of CD8⁺ T cells (~ 25%) was also notably higher than in other groups (Fig. 6H). Collectively, these results indicate that the iCAR-M system represents a novel strategy for suppressing post-surgery tumor recurrence, capable of activating immune responses and effectively inhibiting tumor growth and relapse.

Conclusion

In this study, we engineered a 3D-printed implantable CAR-M (iCAR-M) system, serving as an engineered immune niche to overcome the bottlenecks of poor colonization and limited persistence inherent to adoptive cell therapy. By harnessing the hierarchical porous architecture and biocompatibility of the GelMA scaffold, this platform functioned as a bioactive reservoir. It not only mitigated anoikis to significantly boost CAR-M viability but also facilitated the sustained locoregional release of functional macrophages. Notably, we integrated IL-13Rα2-targeting capabilities with checkpoint blockade within this 3D matrix. The co-delivered SIRPα-EVs effectively masked CD47 on tumor cells, abrogating the “don’t eat me” signal, directly enhancing CAR-M phagocytic activity against tumor cells. This blockade synergized with CAR-mediated recognition, thereby maximizing macrophage-mediated phagocytosis against antigen-positive solid tumors. Besides, SIRPα inhibition can activate inflammatory pathways and the cGAS-STING signaling cascade in macrophages, leading to increased production of proinflammatory cytokines and reactive oxygen species, thereby amplifying their antitumor efficacy and inducing a profound remodeling of the immunosuppressive TME [47].

After recognizing and phagocytosing tumor cells, CAR-M can process engulfed whole-tumor antigens into antigenic peptides and cross-present them to CD8⁺ T cells via MHC class I molecules, thereby priming tumor-specific adaptive immune responses. Through this cascade of “phagocytosis-presentation-activation,” CAR-M effectively bridges innate and adaptive immunity. Research has demonstrated that macrophages upregulate co-stimulatory molecules like CD80 and CD86 to enhance their capacity as antigen-presenting cells [48]. And polarized M1-like macrophages secrete pro-inflammatory cytokines to recruit and reactivate exhausted CD8⁺ T cells within the tumor microenvironment, reversing their functional dysfunction [33]. We observed a significant repolarization of tumor-associated macrophages toward an inflammatory M1 phenotype and the active recruitment of cytotoxic CD8⁺ T cells, triggering a robust adaptive immune response. Demonstrating the efficacy in a post-surgery 4T1 model revealed that this locoregional delivery strategy significantly suppressed tumor regrowth compared to CAR-M alone.

Although 3D-printed bioinks are widely utilized [49–53], our work presents the first integration of 3D bioprinting with CAR-macrophage therapy. By constructing a biomimetic scaffold for the sustained co-delivery of CAR-M and SIRPα-EVs, we combine biomaterial-assisted persistence with dual-target immunotherapy. Unlike conventional implantable immunotherapy systems that primarily serve as passive delivery vehicles for single immune modulators, our iCAR-M platform integrates multiple synergistic functionalities within a single bioactive construct. The 3D-printed GelMA scaffold enables sustained, locoregional co-delivery of living CAR-M and SIRPα-EVs, creating a persistent “immune niche”. Furthermore, the localized delivery of SIRPα-EVs provides effective CD47 blockade without the systemic toxicity associated with conventional antibody therapy. This multifaceted approach-combining living cell therapy, immune checkpoint modulation, and microenvironment remodeling-distinguishes our system from existing postoperative immunotherapy strategies. This iCAR-M system represents a novel, robust, and translatable strategy for preventing post-surgery recurrence and overcoming solid tumor resistance.

Currently, the applicability of this proposed cell delivery strategy to other resectable human cancers, such as pancreatic cancer and melanoma, remains to be verified. To further enhance the translational potential of this delivery method, the release kinetics and controllability of encapsulated CAR-M could be optimized by tuning the crosslinking density of the hydrogel network or introducing internal/external stimuli-responsive elements [54]. Additionally, further investigation is warranted to demonstrate the beneficial effects of this strategy in metastatic and unresectable tumor models. Furthermore, the hydrogel reservoir is highly versatile and can be engineered to incorporate other therapeutic bioparticles, thereby establishing a local “immune cell factory” to eliminate residual tumor cells. Ultimately, this paradigm establishes a new frontier in personalized postoperative care, offering a transformative solution to eradicate residual micro-metastases and achieve durable disease remission.

Methods

Reagents and materials

Chemical reagents were all purchased from Sigma-Aldrich unless otherwise stated. Fetal bovine serum (FBS), penicillin-streptomycin, phosphate buffer solution (PBS; 1×), trypsin, and Dulbecco’s Modified Eagle Medium (DMEM) were purchased from Gibco. 4’,6-diamidino-2-phenylindole (DAPI), 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindodicarbocyanine (DiL), 4-chlorobenzenesulfonate salt (DID), cell proliferation dye eFluorTM 670 (Invitrogen, 65-0840-85), and eBioscienceTM CFSE (Invitrogen, 65-0850-84) were purchased from Thermo Fisher Scientific.

Mice

BALB/c (female, 6–10 weeks) were obtained from Gempharmatech Co., Ltd. The animal experiment has been approved by the Institutional Review Board of Shenzhen Bay Laboratory and was by accordance with the principles of animal experiment protection guidelines.

Cell

The murine cell lines of 4T1 mammary carcinoma, HEK-293T cells, and Raw 264.7 macrophage were all obtained from the American Type Culture Collection (ATCC) and cultured under the guidelines offered by the ATCC.

Preparation of GelMA hydrogel

GelMA was synthesized according to a previously reported method. Briefly, 10.0 g of Gel was dissolved in 100 mL of PBS under stirring at 55 °C to obtain a gelatin solution. Separately, 8.5 mL of methacrylic anhydride (MA) was mixed with 20 mL of PBS. This mixture was then added dropwise into the Gel solution with continuous stirring, and the reaction was allowed to proceed for 3 h at 55 °C in the dark, while maintaining the pH at approximately 8.5. The reaction was terminated by adding 600 mL of deionized water. The resulting solution was dialyzed against deionized water for 3 days, followed by lyophilization to obtain GelMA. The GelMA was dissolved in PBS containing 0.5% (w/v) photoinitiator lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) to prepare a 10% (w/v) solution. Upon exposure to 405 nm blue light for 1 min, the solution underwent photopolymerization, forming a cross-linked GelMA hydrogel.

Characterization of GelMA hydrogel

The methacryloyl modification of GelMA was verified by ¹H NMR spectroscopy (Bruker 400 MHz Advance, Switzerland) using D2O. Its photocuring behavior and viscosity of hydrogel were characterized by rheometry (TA Instruments DHR-2, USA) with a 20 mm parallel-plate geometry at 37°C under 405 nm light (30 mW cm²) (OmniCure Series 2000). The gelation point was defined as the time when the storage modulus (G’) exceeded the loss modulus (G’’). The mechanical properties of the hydrogel were characterized by uniaxial compression testing using an AGS-V universal testing machine. Cylindrical specimens (10 mm in diameter, 3 mm in height) were prepared via photocuring and compressed at a constant rate of 0.5–1 mm/min until either specimen fracture or a stress variation exceeding 1% was observed. The microstructure of the lyophilized hydrogel was observed by SEM.

The porosity of freeze-dried samples was determined using a gas pycnometer (true density meter). The apparent density (ρₐ) was calculated from the measured mass and geometric volume. The true density (ρₜ) was measured directly via gas displacement. Porosity was calculated as:

Swelling behavior was characterized by immersing freeze-dried hydrogels in PBS at 37 °C for 24 h. The swollen weight (W_S) was recorded, and the swelling ratio was calculated as: where W_d is the dry weight after lyophilization.

Construction and characterization of CAR-M

To develop CAR-M, Raw 264.7 cells were transfected with the CAR plasmid with IL-13 and GFP with Lipofectamine 8000 (Invitrogen). To establish the stable cells, CAR-M were transfected and further selected with Flow cytometry through green fluorescence. The characterization of CAR-M was monitored under a confocal laser scanning microscopy (CLSM, ZEISS LSM980) and flow cytometry (CytExpert, Beckman Coulter). CAR-M was polarized with LPS (20 ng/mL each) for 24 h. The resulting CAR-M was washed three times with PBS (500 g for each 10 min) and was detected for CD80 and CD86 through Flow cytometry.

Preparation and characterization of SIRPα-EVs

The construction of SIRPα-293T cells with mCherry proteins is also carried out using a similar method mentioned above. To obtain single genetically engineered EVs, SIRPα variant-engineered 293T cells were subjected to sonication for 2 min, then treated with DNase and RNase (Invitrogen), and centrifuged at 4000 g for 5 min. The enriched supernatants were extracted and sonicated for an additional 5 min, further centrifuged at 20,000 g for 30 min and 100,000 g for 1.5 h, respectively. After that, the EVs were collected and finally extruded stepwise through 400 nm nanopore polycarbonate membranes on a mini extruder (Avanti Polar Lipids). The hydrodynamic diameter and zeta potential of SIRPα-EVs were characterized with a dynamic light scatter (DLS, Nano-Zen 3600, UK). Total protein concentration of the EV lysates was quantified using a bicinchoninic acid assay (BCA, Thermo) according to the manufacturer’s protocol, with bovine serum albumin (BSA) as the standard. Besides, the morphology of the SIRPα-EVs was observed using TEM.

Western blotting

For sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), Raw 264.7 cells and CAR-M were added to the protein extraction buffer, and the protein contents were measured with a bicinchoninic acid (BCA) kit. The samples were heated at 95 °C for 5 min, and 20 µg of each sample was loaded into 10% SDS-polyacrylamide gel. The samples were run at 120 V for 2 h, and the segregated proteins were transferred onto polyvinylidene fluoride (PVDF) membranes, blocked with 5% (w/v) skimmed milk at 25 °C for 1 h, and incubated with GFP primary antibodies (Abcam) at 4 °C overnight, and further incubated with HRP-conjugated secondary antibody (Thermo Fisher Scientific). The blots were developed by using a West Pico PLUS Chemiluminescent Substrate kit (Thermo Fisher Scientific). The verification of SIRPα-EVs is also carried out using a similar method.

Phagocytosis assay in vitro

Raw 264.7 cells and CAR-M were stained with eBioscienceTM CFSE. They were subsequently seeded into 4-chamber confocal culture dishes and cultured overnight in DMEM. 4T1 cells (4T1-IL13Rα2) were stained with cell proliferation dye eFluorTM 670, washed twice with PBS, and co-cultured with Raw 264.7 cells and CAR-M in a 4-chamber confocal culture dish at a density of 2 × 10⁵ cells/well for 6 h. After the cells were washed twice with PBS, they were then fixed with 4% PFA and subjected to imaging. For the SIRPα-EVs group, the method used was the same. Before the 4T1-IL13Rα2 cells were added, the EVs were pre-incubated in 4T1-IL13Rα2 cells at 4 °C for 2 h.

3D bioprinting of iCAR-M

Before printing, 5 mg of LAP was added to 1 mL of GelMA solution, which was thoroughly mixed and incubated at 4 °C for 1 h. Subsequently, 5 × 10⁶ CAR-M and 1 mg SIRPα-EVs were incorporated into the LAP/GelMA solution and gently mixed by pipetting to prepare the iCAR-M bioink. For 3D bioprinting, a digital light processing (DLP) printer (nanoArch S240, BMF Material Technology Inc.) equipped with a 405 nm laser was employed, and the 3D models were designed using SolidWorks software, detailed schematic diagrams of 3D models and fluorescent images of cell-loaded constructs are shown in Fig. 4A and Figure S2. And table S1 displays the 3D bioprinting parameters. After printing, the constructions were initially examined using a digital microscope and then immediately placed in a sterile environment for culture.

To verify the survival of CAR-M in the GelMA solution and 3D hydrogel, the CCK-8 kit and Live/Dead staining were used for detection. Following a 7-day incubation period, the samples were stained with Calcein-AM and PI for 20 min and subsequently imaged using CLSM (ZEISS LSM980). Cell status within the scaffolds was also examined by SEM. CAR-M were mixed with hydrogel at a density of 5 × 10⁵ cells/mL. After adding the photoinitiator LAP, the mixture was immediately subjected to 3D bioprinting. Table S1 displays the 3D bioprinting parameters. The cell-laden scaffolds were subsequently cultured in inactivated serum medium for 3 days, harvested, and washed three times with PBS. The samples were then fixed with 2.5–3.0% glutaraldehyde for 2 h, followed by dehydration through a graded ethanol series (30%, 50%, 75%, 80%, 95%, and 100% twice, 15 min each step). After dehydration, the scaffolds were subjected to critical point drying using a Carbon Dioxide Critical Point Dryer (CO2 CPD, Quorum) for 30 min and finally observed under SEM.

Flow cytometric analysis of CAR-M sustained release

5 healthy BALB/c (female, 6 weeks) mice were assigned to each post-surgery treatment group (after tumor removal, hydrogel is implanted) (Hydrogel, CAR-M@Hydrogel, and CAR-M@3D Hydrogel). On day 4 and day 7, the mice were euthanized, and the tissues below the hydrogel scaffold were collected for flow cytometric analysis. The CAR-M were washed with PBS and seeded onto scaffolds. The flow cytometric analysis procedures are: the tumors were dissected, cut into pieces, and incubated with RPMI 1640 media containing 0.8 mg/mL collagenase D (Roche), 0.2 mg/mL DNases and 0.1 mg/mL hyaluronidase for 40 min at 37 °C. The homogenates were washed with PBS and passed through a 70 μm nylon mesh to acquire single-cell suspensions. Before labeling the immune cells, the dead cells in the single-cell suspensions were labeled with a Live/Dead dye using Zombie AquaTM fixable viability dye BV510 (BioLegend) for 10 min at 4 °C. Then, cells were blocked with rat anti-mouse CD16/CD32 (BD Biosciences) at 4 °C for 5 min and stained with the following fluorescence-labeled antibodies: CD45-PerpCP-Cy5.5 (clone HI30; BioLegend), CD11b-APC (clone M1/70; BioLegend), F4/80-APC-R700 (clone T45-2342; BD Biosciences), CD80-BV421 (clone 16-10A1; BioLegend), CD86-PE (clone GL-1; BioLegend), MHC-II-PE-Cy7 (clone M5/114.15.2; BioLegend). Then cells were washed with FACS buffer. Finally, 300 µL FACS buffer was added to each tube and analyzed by CytoFLEX flow cytometer (Beckman Coulter).

IF analysis of the tumor cells and macrophages

4T1-IL13Rα2 cells were seeded on 4-chamber confocal culture dishes for 24 h and treated with SIRPα-EVs stained by DID at 4 ℃ for 2 h. Afterwards, the cells were then washed with PBS twice and fixed in freshly prepared 4% PFA for 15 min at room temperature. Following twice washes with PBS, the cells were permeabilized and blocked with a blocking buffer (1× PBS, 5% BSA, 0.3% Triton X-100) for 1 h at room temperature. Cells were incubated with primary antibodies of CD47 (A11382; Abclonal) overnight at 4 °C. Subsequently, the cells were rinsed with PBS three times, followed by incubation with secondary antibodies (SA00003-2; Proteintech) at room temperature in the dark for 1 h. After the cells were washed with PBS three times, NucBlue was added to stain the nuclei according to the manufacturer’s instructions, and the cells were detected by imaging.

Cytokine level detection

To determine the cytokine binding capability, 0.5 × 10⁵ CAR-M were mixed with 0.5 × 10⁵ 4T1-IL13Rα2 cells (after the vesicle incubation). The cells were seeded in 12-well plates at a density of 1 × 10⁵ cells/well and allowed to adhere overnight. After centrifugation of collected supernatants at 15,000 g for 15 min, NOS, IL-6, and IL-1β concentrations were quantified using ELISA kits.

In vivo tumor models and treatments

For the personalized post-surgery recurrence prevention model, healthy BALB/c (female, 6 weeks) mice were selected. The mice were divided into 5 groups, each consisting of 5 mice. 6 × 10⁵ 4T1-IL13Rα2 tumor cells were first injected subcutaneously on the right flank of mice. On the 14th day after cell injection, the tumor was excised, and then the personalized scaffolds were 3D printed and transplanted at the excision site of the tumor. The first group of mice received a dorsal injection of 100 µL of PBS, and the second group received a non-porous scaffold hydrogel without cell implantation. The third group underwent dorsal implantation of a non-porous scaffold hydrogel loaded with CAR-M (1 × 10⁶ cells/mouse), the fourth group underwent dorsal implantation of a porous scaffold hydrogel loaded with CAR-M (1 × 10⁶ cells/mouse) and finally, the last group underwent dorsal implantation of a porous scaffold hydrogel loaded with CAR-M and SIRPα-EVs (1 × 10⁶ cells and 400 µg SIRPα-EVs/mouse). The mice’s body weight and tumor volume were measured every two days. Once the tumor volume reached approximately 500 mm³, it was utilized for subsequent experimental analysis.

Flow cytometry analyses after treatments

The tumor tissue was minced and suspended in a digestion solution (Serum-free 1640 medium with 25 U/mL DNase I, 0.2 mg/mL Collagenase, and 0.1 mg/mL hyaluronidase) in a gentleMACS (TM) C Tube, with 5 mL digestion solution per tube. The m-imp tumor-01-01 was run on an automated processor (GentleMACS Dissociator Instruments) for 37 s each time, and then shaken at 37 °C at 150 rpm for 40 min. After digestion, the cell suspension was filtered through a 70 μm mesh into a 50 mL centrifuge tube, and the supernatant was removed by centrifugation. 5 mL of red blood cell lysis solution (Solarbio) was added to each tube to remove red blood cells.

Isolated cells (1 × 106 cells in a 100 µL system) were first determined Live/Dead using Zombie AquaTM Dye (BioLegend) for 10 min at 4 °C. Then, cells were blocked with rat anti-mouse CD16/CD32 (clone 2.4G2; BD Biosciences) at 4 °C for 5 min and then stained with the following fluorescence-labeled antibodies: CD45-PerpCP-Cy5.5 (clone HI30; BioLegend), CD11b-FITC (clone M1/70; Biolegend), F4/80-PE-Cy7 (clone BM8; Biolegend), CD3-BV395 (clone SK7; Biolegend), CD4-BV650 (clone OKT4; Biolegend), CD8-PE (clone SK1; Biolegend), CD80-BV421 (clone 16-10A1; Biolegend), CD206-APC (clone C068C2; BioLegend). All antibodies were diluted according to the manufacturer’s instructions and incubated with the cells for 30 min at room temperature. Before initiating the multicolor analysis, a single-stained control with beads for each fluorochrome was made for compensation for each experiment. Finally, 300 µL FACS Buffer was added to each tube and analyzed by CytoFLEX flow cytometer (Beckman Coulter).

Immunofluorescence (IF) of tumor tissues

The tumor tissues were fixed in Bouin’s solution (Solarbio, China) overnight at room temperature. The sections of tumor tissues were used for an IF study, and the rabbit anti-mouse CD86 and rabbit anti-mouse CD206 (1:100 dilution; Abcam) were separately used to label the infiltration macrophages in tumors. Consistently, the secondary antibodies, goat anti-rabbit IgG H&L (Alexa Fluor 488), were separately used to bind the primary antibody, and the cell nucleus was stained with DAPI (Vector Laboratories). The sections were scanned by an Aperio ScanScope CS scanner (Vista).

Statistical analysis

All results are presented as mean ± standard deviation (S.D.). The unpaired two-tailed t-test was used for two-group comparisons, and ordinary one-way (or two-way) ANOVA with a Tukey’s test was used for multiple group comparisons. All statistical analyses were performed with the assistance of Prism 9.0 software (GraphPad).

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

L.R., and Q.H. supervised this project. D.N., Y.S., Y.X., L.R., and Q.H. conceived the project and designed the experiments. X.H., J.Q., and X.H. performed the construction and characterization of the 3D-printed hydrogels. D.N., Y.S., Q.M., and P.S. contributed to the *in vitro* experiments. D.N. performed the *in vivo* experiments. D.N., Y.S., Y.X., L.R., and Q.H. wrote the paper. All authors discussed the data and commented on the manuscript.

Funding

This work was supported by Shenzhen Medical Research Fund (Nos. B2502017, A2503047, and A2502029), National Natural Science Foundation of China (Nos. 82372106 and 82402456), and Guangdong Basic and Applied Basic Research Foundation (Nos. 2024A1515010267 and 2025A1515011290).

Data availability

All data are available if requested.

Declarations

Ethics approval and consent to participate

All animal experiments were conducted in accordance with the guidelines and approved by the Animal Ethics and Welfare Committee of Shenzhen Bay Laboratory (No. AERL202501).

Consent for publication

All authors agreed to publish the article.

Competing interests

The authors declare no competing interests.