CXCR4-engineered fibroblast membrane nanovesicles for Photothermal-enhanced ferroptotic therapy through chemokine-navigated tumor homing

Abstract

Background

Cell membrane-camouflaged nanoparticles (NPs) have attracted increasing attention for cancer-targeted drug delivery, but their clinical translation faces critical challenges due to the biosafety concerns and scalability issues. Dermal fibroblasts are an abundant, clinically accessible membrane source, and genetic programmability offers a route to active homing.

Results

We engineered fibroblast membrane nanovesicles from dermal fibroblasts overexpressing CXCR4 to actively home toward CXCL12-enriched tumor microenvironments. These CXCR4-engineered membrane nanovesicles demonstrated tumor-selective accumulation in multiple preclinical models with high CXCL12 secretion. The near-infrared (NIR) photosensitizer IR780 and the ferroptosis inducer RSL3 were co-loaded to form FbM@IR/RSL3 NPs for photothermal-controlled ferroptosis therapy. Upon NIR irradiation, FbM@IR/RSL3 NPs generated localized photothermal heating and simultaneously triggered RSL3 release. The combination of glutathione peroxidase 4 (GPX4) inhibition and iron-catalyzed lipid peroxidation amplified ferroptotic tumor cell death. In vivo studies reveal enhanced tumor suppression across heterogeneous carcinoma models compared to monotherapy approaches, with systemic biocompatibility confirmed by comprehensive hematological and histopathological analyses.

Conclusions

Fibroblast membrane engineering, combined with chemokine-gradient navigation and photothermally controlled therapeutic activation, represents the translational potential of nanomedicine in clinical by developing precision nanomedicines that coordinate biological recognition with stimulus-responsive bioactivity.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-025-03951-5.

Background

Nanoparticles have been widely used in tumor diagnosis and treatment because of their small size, high surface-to-volume ratios, and modifiable surfaces. These carriers reach tumor sites either via the passive enhanced permeation and retention (EPR) effect or through active targeting [1]. However, the indispensable toxicity and immune evasion of synthetic nanoparticles and the complexity of surface modification techniques for effective targeting significantly restrict their clinical translation [2]. To solve these problems, a “top-down” strategy has been developed by fabricating nanoparticles with cell membranes. This method leverages the natural surface antigens of the source cells, which shield the nanoparticles from immune recognition while simultaneously prolonging their circulation in the bloodstream [3].

Common types of cell membranes used in constructing biomimetic nanoparticles include those derived from red blood cells, immune cells, cancer cells, platelets and stem cells [3, 4]. In cancer therapy, tumor cell membrane-coated nanoparticles are often highlighted for their homologous targeting properties, which depend on the cell adhesion molecules present on the tumor cell membranes [5, 6]. Moreover, the overexpression of CD47 on certain tumor cell membranes confers resistance to macrophage-mediated phagocytosis [7]. While these features of tumor cell membranes enhance drug delivery specificity and reduce immunogenicity, they also introduce several technical and ethical challenges that restrict their clinical use. The first is that, obtaining autologous and immuno-compatible tumor cell membranes requires patients to undergo invasive surgery to excise tumor tissue, which is not only complex but also poses potential risks of tumor metastasis. Additionally, tumor tissues contain a variety of cell types beyond tumor cells, including immune cells, stromal cells, and endothelial cells [8]. Therefore, isolating pure tumor cells and cultivating them in vitro to provide sufficient tumor cell membranes for nanoparticle fabrication may require a substantial amount of tumor tissue. Furthermore, the potential transfer of oncogenic proteins and genetic elements during cell membrane preparation may raise safety concerns in patients. In response to these challenges, dermal fibroblasts would be potential sources for membrane-based cancer therapies. These cells are located in the skin’s dermis and play a pivotal role in skin development and wound healing due to their proliferative capabilities [9]. Dermal fibroblasts are safe and can be harvested from patients with non-invasive procedures, providing a practical advantage for clinical applications. Moreover, their ability for rapid proliferation facilitates scalable production. These characteristics make dermal fibroblasts an excellent choice for deriving cell membranes through in vitro cultivation, offering promising clinical translation value in cancer therapy.

By leveraging the principle of ligand-receptor interactions, modifying cell surface proteins to “identify” aberrantly expressed molecules within the tumor microenvironment, thereby facilitating the tumor-specific delivery of biomimetic membrane nanoparticles. The stromal cell-derived factor 1 (SDF-1 or CXCL12) binds to the transmembrane receptor CXCR4, initiating hematopoietic stem cell homing [10]. CXCL12 upregulation has been observed in various cancers, including breast cancer, pancreatic cancer and esophageal squamous cell carcinomas [11–13]. Based on the above, we propose to engineer dermal fibroblast to overexpress CXCR4. This modification might enable fibroblast membrane-coated nanoparticles to bind CXCL12 within the tumor microenvironment, thereby enhancing the specificity of tumor diagnosis and drug delivery.

Traditional cancer therapies, such as radiotherapy and chemotherapy, primarily inhibit tumor growth by inducing apoptosis. However, the efficacy of these treatments might be compromised by the overexpression of BCL-2 family proteins, which reduce tumor cell sensitivity to apoptotic signals and contribute to drug resistance [14]. Ferroptosis is featured by the accumulation of intracellular lipid peroxides, thereby disrupting the integrity of cell membranes and leading to cell death [15]. Distinct from apoptosis, ferroptosis is independent of BCL-2 regulation, making it a promising strategy for cancer therapy. Commonly used ferroptosis inducer RSL3 inhibits the activity of GPX4, disrupting its ability to catalyze the reduction of oxidized glutathione (GSH), which leads to the accumulation of lipid reactive oxygen species (ROS) [16]. Despite RSL3 demonstrating significant anti-tumor activity in vitro, its hydrophobic nature and poor solubility limit its clinical application [15]. Therefore, incorporating RSL3 into biomimetic nanoparticles could potentially enhance its safety and bioavailability. Studies have shown that photothermal therapy could enhance GSH depletion in tumor cells [17]. Moreover, the high temperature from photothermal effect could accelerate the Fenton reaction, promoting the accumulation of ROS and further exacerbating oxidative stress in cells [18]. These findings indicate that combining photothermal therapy with RSL3 may synergistically enhance ferroptosis in tumors, potentially providing a more effective treatment strategy for tumor therapy.

Herein, we aimed to develop a dermal fibroblast membrane-coated nanoparticle system, FbM@IR/RSL3 NPs. This system was consisted of the fibroblast membrane shell overexpressing CXCR4 (FbM) and IR780-conjugated lipid nanoparticle core encapsulated with ferroptosis inducer RSL3 (IR/RSL3 NPs). We sought to investigate the potential of these genetically modified biomimetic nanoparticles for specific tumor targeting. In the meanwhile, we proposed to determine whether the combination of FbM@IR/RSL3 NPs with photothermal therapy could effectively enhance ferroptosis. Finally, we planned to explore the overall antitumor efficacy of this combined therapeutic approach both in vitro and in vivo, aiming to establish FbM@IR/RSL3 NPs as a promising strategy for targeted cancer treatment.

Results

Characterization of CXCR4-OE fibroblasts

Dermal fibroblasts were selected as the cellular platform due to their clinical accessibility and robust in vitro expansion capacity. Primary fibroblasts isolated from neonatal mice exhibited characteristic spindle-shaped morphology, as confirmed by microscopy (Figure. S1). Lentiviral-mediated genetic modification successfully enhanced CXCR4 expression, with RT-qPCR analysis revealing a 199-fold increase in mRNA levels compared to unmodified controls (Fig. 1A). Western blotting corroborated this finding at the protein level, demonstrating clear CXCR4 upregulation in engineered fibroblasts (Fig. 1B). Consistent with the known membrane localization of chemokine receptors [19], immunofluorescence staining confirmed predominant CXCR4 expression on the fibroblast surface (Fig. 1C-D). Functional validation through transwell assays demonstrated that CXCR4-overexpressing fibroblasts exhibited significantly enhanced migration toward CXCL12 gradients compared to control cells (Fig. 1E-F). This enhanced chemotaxis was effectively abolished by pretreatment with LIT-927, a specific CXCL12 inhibitor, confirming the receptor-ligand dependency of the observed effect. Collectively, these results verify the successful generation of CXCR4-OE fibroblasts with preserved receptor functionality and CXCL12-responsive migration capacity.

Fig. 1.

Characterization of CXCR4-OE fibroblasts. (A) RT-qPCR analysis of CXCR4 expression in murine dermal fibroblasts. (B) Western blotting analysis of CXCR4 in fibroblasts and fibroblast membranes. C and D. Representative fluorescence images of fibroblasts stained with FITC-labeled CXCR4 antibody (C), and the relative FITC intensity in cells (D) (n = 3). E and F. Representative transwell invasion images of fibroblasts towards different treatments (Control, CXCL12, CXCL12 + LIT-927, CXCL12 + CTR-FbM@IR NPs and CXCL12 + FbM@IR NPs) (E), and the number of invaded cells per field (F). Data are displayed as mean ± SEM. p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). CXCR4-OE, CXCR4-overexpressed fibroblasts; CTR, control fibroblasts

Synthesis and characterization of FbM@IR/RSL3 NPs

The synthesis scheme of FbM@IR/RSL3 NPs was depicted in Fig. 2A. IR780-DSPE-PEG was synthesized according to a previously reported substitution reaction [6]. The hydrophobic ferroptosis inducer RSL3 was co-assembled with IR780-DSPE-PEG and thermal-sensitive phospholipid DPPC using solvent diffusion, resulting in drug-loaded IR NPs. After the drug loading process, the nanoparticle size increased from 67.54 ± 0.93 nm to 98.72 ± 6.65 nm, indicating successful drug encapsulation (Fig. 2B; Table 1). High-performance liquid chromatography (HPLC) analysis demonstrated that RSL3 encapsulation efficiency in IR NPs was 66.90 ± 7.06% and its drug loading efficiency was 15.92 ± 1.43%. FbM vesicles extracted from CXCR4-OE murine fibroblasts were then fused onto the nanoparticle surface via controlled sonication. Western blot analysis verified the preservation of CXCR4 in the membrane coating (Fig. 1B). TEM imaging revealed intact core-shell nanostructures with continuous membrane encapsulation (Fig. 2D), and the membrane shell thickness was estimated at approximately 8 nm, consistent with reported values [20]. Following membrane coating, the average nanoparticle size of FbM@IR NPs and FbM@IR/RSL3 NPs increased to 150.50 ± 3.84 nm and 163.93 ± 3.95 nm, respectively, and their zeta potentials became more negative (Fig. 2B, 2 C and Table 1). These negatively charged nanoparticles can remain in circulation for longer periods due to minimal protein absorption in the bloodstream [21]. Notably, the engineered nanoparticles displayed excellent stability with minimal size variation in serum over four days (Figure. S2). The cytotoxicity of the drug delivery system was evaluated by incubating tumor cells and normal cells with various concentrations of drug-unloaded FbM@IR NPs. No significant cytotoxic effects were observed on either cell type, even at high concentrations (Fig. 2E). Collectively, FbM@IR NPs and FbM@IR/RSL3 NPs were successfully synthesized and possessed high stability, prolonged blood circulation and safety.

Fig. 2.

Synthesis and characterization of FbM@IR/RSL3 NPs. A. Schematic diagram of the preparation of FbM@IR/RSL3 NPs. B and C. Average particle sizes (B) and zeta potentials (C) of IR NPs, IR/RSL3 NPs, FbM, FbM@IR NPs and FbM@IR/RSL3 NPs. D. Representative TEM images of FbM@IR NPs and FbM@IR/RSL3 NPs (scale bar = 200 nm). Red arrows indicate the membrane shell. E. Cell viability of tumor cell lines (MDA-MB-468 and KYSE-30) and normal fibroblast cell line NIH-3T3 in response to different concentrations of FbM@IR NPs. Data are presented as mean ± SEM

Table 1.

Sizes and zeta potentials of IR NPs, IR/RSL3 NPs. FbM, FbM@IR NPs and FbM@IR/RSL3 NPs

Formulation	Size (nm)	Zeta Potential (mV)	Polydispersity Index
IR NPs	67.54 ± 0.93	6.17 ± 1.60	0.07 ± 0.03
IR/RSL3 NPs	98.72 ± 6.65	−15.60 ± 1.53	0.15 ± 0.01
FbM	152.53 ± 5.76	−28.90 ± 0.61	0.13 ± 0.02
FbM@IR NPs	150.50 ± 3.84	−21.35 ± 0.76	0.08 ± 0.03
FbM@IR/RSL3 NPs	163.93 ± 3.95	−22.94 ± 0.70	0.16 ± 0.01

Cellular internalization and CXCR4-CXCL12 axis-mediated targeting of FbM@IR NPs in tumor spheroids

The binding ability of FbM@IR NPs to CXCL12 was validated through a transwell invasion assay (Fig. 1E). In this assay, the lower chamber contained CXCL12 along with either FbM@IR NPs or CTR-FbM@IR NPs. The presence of FbM@IR NPs in the lower chamber significantly inhibited the invasion of CXCR4-OE fibroblasts compared to CTR-FbM@IR NPs, indicating the specific binding interaction between FbM@IR NPs and CXCL12. We further observed a time-dependent increase in the cellular internalization of FbM@IR NPs by tumor cells (Fig. 3A-D). The fluorescence intensity of FITC-labeled membrane-coated NPs within the cells was significantly higher at 9 h compared to 3 h, confirming the increased uptake of these nanoparticles over time. To investigate the targeting of FbM@IR NPs to CXCL12 in vitro, we first detected CXCL12 secretion by murine fibroblasts. ELISA analysis of the cell culture supernatant demonstrated that co-culturing fibroblasts with tumor cells significantly promoted their CXCL12 release (Fig. 3E). This suggested that tumor cells could stimulate fibroblasts to secret CXCL12, which was in good agreement with the characteristics of cancer-associated fibroblasts as previously reported paper [22]. To better mimic the tumor microenvironment in vivo, we established a 3D tumor spheroid model by mixing KYSE-30 tumor cells and fibroblasts at a 1:1 ratio. The FbM@IR NPs exhibited a stronger signal within the tumor spheroid compared to the control, while this effect was diminished upon blocking CXCR4 with its antagonist (Fig. 3F and Figure. S3). These findings are likely due to the pronounced CXCR4-CXCL12 interaction in the 3D environment. The cell density and microenvironment in the spheroid model was more closely resemble in vivo conditions, where the secretion and diffusion of CXCL12 created a significant concentration gradient. This allows CXCR4-postive NPs to effectively bind to CXCL12 around the tumor spheroid and be internalized by tumor cells. In summary, FbM@IR NPs effectively target the tumor microenvironment through the CXCR4-CXCL12 interaction in vitro.

Fig. 3.

Cellular internalization and CXCR4-CXCL12 axis-mediated targeting of FbM@IR NPs in tumor spheroids. A and B. Representative CLSM images showing the cellular internalization of FITC-labeled CTR-FbM@IR NPs and FbM@IR NPs into KYSE-30 cells (A) and MDA-MB-468 cells (B) after 3-hour and 9-hour incubation. C and D. The relative FITC intensity in KYSE-30 (C) and MDA-MB-468 cells (D). E. ELISA of CXCL12 levels in different cell culture supernatants. F. 3D-merged CLSM images showing the uptake of FITC-labeled FbM@IR NPs, FbM@IR NPs + plerixafor, and CTR-FbM@IR NPs in KYSE-30 spheroids after 24 h of incubation. Data are indicated as mean ± SEM. p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***)

Enhanced in vivo tumor targeting of FbM@IR NPs via CXCR4-CXCL12 interaction

Based on the in vitro findings that showed effective targeting of FbM@IR NPs to CXCL12, we next evaluated their targeting ability in vivo. Breast cancer and esophageal carcinoma xenograft models were respectively established by subcutaneously injecting MDA-MB-468 and KYSE-30 tumor cells into nude mice. Immunofluorescence analysis confirmed abundant CXCL12 expression within tumor stroma of both models (Figure. S4), establishing a molecular basis for targeted delivery. Subsequently, mice with tumors received intravenous injections of FbM@IR NPs. Fluorescence distribution was monitored at 2, 6, 24, 30 and 48 h post-injection using an in vivo imaging system. In both tumor models, FbM@IR NPs showed significant accumulation in the tumor regions compared to the control NPs (Fig. 4A-B and Figure. S5A-B), indicating enhanced targeting specificity due to CXCR4 overexpression. Upon inhibition of either CXCR4 or CXCL12, a notable reduction in the tumor accumulation of FbM@IR NPs was observed (Fig. 4A-B and Figure. S5A-B), confirming that their retention in the tumor region was specifically mediated through CXCR4-CXCL12 binding. To eliminate the potential skin autofluorescence interference, ex vivo fluorescence imaging of harvested organs and tumors was performed 48 h post-injection. The ex vivo imaging and quantitative analysis of dissected tumors supported the in vivo findings, further validating the active tumor-targeting capability of FbM@IR NPs (Fig. 4C-F and Fig. S6A-B). Additionally, minimal retention of FbM@IR NPs in the liver and kidneys indicated good biocompatibility and safety (Fig. 4C-D and Figure. S6A-B). These findings demonstrate the potential of FbM@IR NPs for tumor diagnosis through active homing to tumors via CXCR4-CXCL12 interaction.

Fig. 4.

Enhanced in vivo tumor targeting of FbM@IR NPs via CXCR4-CXCL12 interaction. A and B. In vivo IR780 distribution in MDA-MB-468 (A) and KYSE-30 (B) tumor-bearing mice at different time points after i.v. injection of FbM@IR NPs, FbM@IR NPs + Plerixafor, FbM@IR NPs + LIT-927 and CTR-FbM@IR NPs. (Green dashed circles indicate tumor sites). C and D. Ex vivo IR780 distribution in tissues (heart, liver, spleen, lung, kidneys, tumor and lymph nodes) harvested from mice bearing MDA-MB-468 (C) and KYSE-30 (D) tumor at 48 h after injection of FbM@IR NPs, FbM@IR NPs + Plerixafor, FbM@IR NPs + LIT-927 and CTR-FbM@IR NPs. E and F. Ex vivo IR780 signals of tumors from MDA-MB-468 tumor-bearing mice (E) and KYSE-30 tumor-bearing mice (F) (n = 3). Data are presented as mean ± SEM. p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***)

Evaluation of photothermal effect and thermo-responsive drug release property of FbM@IR/RSL3 NPs

IR780 exhibits dual functionality as an NIR fluorescent tracer and photothermal agent, enabling simultaneous imaging and photothermal therapy [23]. Ultraviolet-visible spectral analysis confirmed that IR780-DSPE-PEG, FbM@IR NPs, and FbM@IR/RSL3 NPs maintained characteristic absorption peaks at 780–800 nm (Fig. 5A), consistent with the optical properties of native IR780 [6]. This confirmed that our synthesized biomimetic membrane-coated NPs retained their NIR absorption capacity, which is essential for effective photothermal therapy. In vitro photothermal conversion efficiency was evaluated using a thermal camera. Specifically, FbM@IR NPs and FbM@IR/RSL3 NPs reached maximum temperatures of 50.4 °C and 50.9 °C, respectively, significantly higher than the temperature peaked at 31.9 °C in PBS under the same irradiation condition (Fig. 5C-D). We further validated their in vivo photothermal performance in MDA-MB-468 tumor-bearing mice. After intravenous administration of FbM@IR NPs or FbM@IR/RSL3 NPs, mice were subjected to NIR irradiation, and the temperature changes in the tumor area were monitored by a thermal camera. Remarkably, the tumor regions in mice treated with FbM@IR NPs and FbM@IR/RSL3 NPs respectively reached 52.3 °C and 52.5 °C within 3 min of NIR irradiation, and the temperature remained constant throughout the 5-minute exposure (Fig. 5E-F). In comparison, the control group injected with PBS only reached a peak temperature of 40.3 °C, which was insufficient for effective photothermal ablation. These results collectively demonstrated the substantial photothermal potential of FbM@IR NPs and FbM@IR/RSL3 NPs.

Fig. 5.

Evaluation of photothermal effect and thermo-responsive drug release property of FbM@IR/RSL3 NPs. (A) UV spectra of IR780-PEG-DSPE, FbM@IR NPs and FbM@IR/RSL3 NPs solution. (B) NIR-responsive RSL3 release of FbM@IR/RSL3 NPs during 48 h of incubation (n = 3). (C) Representative infrared thermal images of FbM@IR NPs, FbM@IR/RSL3 NPs and PBS solution after continuous NIR irradiation (808 nm, 1.0 W/cm²). (D) Temperature changes of FbM@IR NPs, FbM@IR/RSL3 NPs and PBS solution upon continuous NIR exposure (808 nm, 1.0 W/cm²) (n = 3). (E) Representative infrared thermographic images of MDA-MB-468 tumor-bearing mice injected with FbM@IR NPs, FbM@IR/RSL3 NPs and PBS solution and received a 5-minute NIR irradiation (808 nm, 0.8 W/cm²). White square frames indicate tumor regions. (F) Temperature changes of tumor region in different groups upon laser irradiation (n = 3)

After confirming the NIR-induced temperature increase capability of the membrane-coated NPs, we further investigated their thermal-responsive RSL3 release behavior by HPLC using a dialysis method. The primary thermal-sensitive component DPPC in FbM@IR/RSL3 NPs is responsible for controlled drug release. When the temperature reaches 42 °C, DPPC undergoes a gel-to-liquid phase transition, which increases the permeability of the NPs, thereby releasing the encapsulated drug molecules [24]. The cumulative drug release rates were calculated by measuring RSL3 leakage at different time points. As demonstrated in Fig. 5B, free RSL3 exhibited a rapid release, with 85.89 ± 5.95% released within 4 h, remaining constant thereafter. In contrast, RSL3 release from FbM@IR/RSL3 NPs was significantly slower, with only 18.47 ± 3.67% released at 4 h. This indicated that the biomimetic NPs’ structure effectively controlled drug release, which extended the therapeutic duration and enhanced treatment stability. However, upon NIR laser irradiation (0.8 W/cm², 3 min), the release rate of RSL3 from FbM@IR/RSL3 NPS increased rapidly, reaching 70.52 ± 5.68% at 48 h, compared to 51.56 ± 2.18% release without NIR laser, indicating the capability of thermal-controlled drug release.

Synergistic induction of ferroptosis by FbM@IR/RSL3 NPs and photothermal therapy

The therapeutic potential and underlying mechanisms of our biomimetic membrane-coated NPs were investigated in vitro. Initially, cytotoxicity assays were performed to evaluate the impact of FbM@IR/RSL3 NPs on tumor cell viability. MDA-MB-468 cells were incubated with different samples and exposed to NIR laser for 3 min (0.8 W/cm²). As shown in Fig. 6A, both RSL3 and FbM/IR/RSL3 NPs induced tumor cell death, consistent with the reported pro-ferroptotic properties of RSL3 [25]. There was no significant difference in cytotoxicity between RSL3 and RSL3(+), and between PBS and PBS(+), indicating the excellent biocompatibility of NIR light. Notably, FbM@IR NPs(+) and FbM@IR/RSL3 NPs(+) exhibited significantly enhanced toxicity to tumor cells compared to their non-irradiated counterparts, suggesting the potential photothermal effects conferred by IR780. Remarkably, FbM@IR/RSL3 NPs(+) demonstrated superior toxicity compared to FbM@IR NPs(+), which could be attributed to the controlled release of RSL3 upon laser irradiation. No noticeable damage was observed in the non-irradiated FbM@IR NPs-treated group, confirming the safety of our delivery system in vitro. Similar cytotoxicity results were observed in KYSE-30 cells (Fig. 6B).

Fig. 6.

Synergistic induction of ferroptosis by FbM@IR/RSL3 NPs and photothermal therapy. A and B. Cell viability of MDA-MB-468 cells (A) and KYSE-30 cells (B) after 24 h of incubation with different treatments (NC, RSL3, FbM@IR NPs and FbM@IR/RSL3 NPs) with or without NIR laser (n = 3). C and D. Lipid peroxidation quantification of MDA-MB-468 cells (C) and KYSE-30 cells (D) using BODIPY 581/591 C11 staining. Cells were pretreated with NC, RSL3, FbM@IR NPs and FbM@IR/RSL3 NPs (with or without laser) for 8 h (n = 3). E and F. Western blotting of GPX4 protein in MDA-MB-468 cells (E) and KYSE-30 cells (F) after different formulation treatments for 6 h. G. CLSM images of ferrous iron in MDA-MB-468 cells (upper) and KYSE-30 cells (lower) upon 6-hour incubation with different formulations (with or without laser). H and I. Relative intracellular ferrous iron fold change in MDA-MB-468 cells (H) and KYSE-30 cells (I) (n = 3). J and K. Intracellular GSH/GSSG ratio in MDA-MB-468 cells (J) and KYSE-30 cells (K) upon 6-hour treatment with NC, RSL3, FbM@IR NPs and FbM@IR/RSL3 NPs (with or without laser) (n = 3). RSL3 dose was 1 µg/mL for MDA-MB-468 cells and 0.8 µg/mL for KYSE-30 cells. IR780-DSPE-PEG concentration was 3.5 µg/mL for both cell lines. NIR irradiation parameter was 808 nm, 1.0 W/cm². All data were expressed as mean ± SEM. p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***)

To explore whether the observed cell death was due to ferroptosis, we performed BODIPY 581/591 C11 staining of lipid ROS, which is a hallmark of ferroptosis. This fluorescent fatty acid analog integrates into the cell membrane and shifts its emission spectrum from red to green upon oxidation [26]. As shown in Fig. 6C, the control group treated solely with NIR light displayed no significant lipid ROS generation, indicating its minimal ferroptotic induction capacity. Groups treated with RSL3, RSL3(+), and FbM@IR/RSL3 NPs showed elevated lipid ROS levels due to RSL3’s capacity to inhibit GPX-4 activity. Interestingly, cells treated with irradiated FbM@IR NPs(+) produced a greater quantity of lipid ROS compared to FbM@IR NPs, consistent with previous research suggesting that elevated temperatures from photothermal effects can augment ferroptotic cell death [17, 27, 28]. Strikingly, FbM@IR/RSL3 NPs(+) elicited the highest levels of intracellular lipid ROS among all groups, substantiating our hypothesis that RSL3 and photothermal effects can synergize to induce lipid peroxidation. KYSE-30 cells mirrored the lipid ROS trends observed in MDA-MB-468 cells (Fig. 6D).

Given that photothermal effects can deplete intracellular GSH levels and accelerate the Fenton reaction [17, 27, 28], we further evaluated the intracellular GSH, GPX4 and iron levels in both MDA-MB-468 and KYSE-30 cell lines following various treatments. In both cell lines, the intracellular GSH/GSSG ratio was notably reduced in groups treated with FbM@IR NPs(+) upon NIR-induced temperature elevation, with the FbM@IR/RSL3 NPs(+)-treated group exhibiting the lowest ratio (Fig. 6J-K). Concurrently, GPX4 protein levels decreased following photothermal therapy with FbM@IR NPs(+), and the group treated with FbM@IR/RSL3 NPs(+) demonstrated the most pronounced GPX4 reduction (Fig. 6E-F). Furthermore, an increase in intracellular ferrous iron levels was detected in both cell lines after photothermal therapy, with the combination of RSL3 and photothermal treatment eliciting the highest iron levels (Fig. 6G-I). Collectively, these findings illustrated that upon cellular uptake of FbM@IR/RSL3 NPs, the temperature increase induced by NIR irradiation could trigger the intracellular release of RSL3. RSL3 directly inhibits GPX4 activity, initiating lipid peroxidation, while the elevated temperature from photothermal effects enhances GSH depletion, synergistically augmenting ferroptosis through GPX4 inhibition. The high temperature also accelerates the Fenton reaction, leading to increased lipid ROS generation.

In vivo antitumor efficacy and biosafety of FbM@IR/RSL3 NPs in tumor-xenografted mice

We further investigated the antitumor efficacy of FbM@IR/RSL3 NPs combined with photothermal therapy in MDA-MB-468 and KYSE-30 tumor-xenografted mice, following the dosing regimen depicted in Fig. 7A. Tumor growth was monitored as tumor volume during drug administration and observation periods for both models. In MDA-MB-468 tumor xenografts, FbM@IR NPs(+) slightly inhibited tumor progression (Fig. 7B), which not only demonstrated the antitumor effect of photothermal therapy, but also supported the excellent CXCR4-guided tumor-specific accumulation ability of the membrane-coated NPs. Free RSL3 moderately induced an antitumor effect, which was consistent with its reported therapeutic potential in tumor [29]. FbM@IR/RSL3 NPs displayed superior tumor suppression effect compared to free RSL3, which was attributed to the prolonged drug circulation and enhanced tumor-targeted drug delivery of our membrane-coated NPs. Notably, FbM@IR/RSL3 NPs(+) induced the most remarkable antitumor effect, compared to photothermal therapy and RSL3 monotherapy. Furthermore, the NIR light exhibited negligible tumor inhibition effect in the absence of photosensitive materials, as evidenced by the similar tumor growth trends between PBS and PBS(+) groups, and between RSL3 and RSL3(+) groups. The corresponding tumor weight and tumor size results collected from mice at day 21 were consistent with the variation of tumor volume. The FbM@IR/RSL3 NPs(+)-treated group had the lowest average tumor weight and the smallest tumor size, with a tumor inhibition rate of 93.18% (Fig. 7C and E, Table S1), which was significantly higher than those observed in the other treatment groups. H&E staining of the dissected tumor samples further supported the antitumor effect of FbM@IR/RSL3 NPs(+) (Fig. 7G), demonstrating that the FbM@IR/RSL3 NPs(+) treatment induced the most tumor cell death. Similarly, in KYSE-30 xenograft mice, the trends in tumor growth inhibition were consistent with those observed in the MDA-MB-468 model (Figure. S7A). FbM@IR NPs(+) slightly inhibited KYSE-30 tumor growth, while FbM@IR/RSL3 NPs(+) produced the most significant tumor suppression effect among all treatment groups, with a tumor inhibition rate of 95.04% (Table S2). Tumor weight measurements and tumor size image on day 21 showed that the FbM@IR/RSL3 NPs(+)-treated group had the lowest tumor weight and the smallest tumor size (Figure. S7B and S7C), in consistence with the results from the MDA-MB-468 model.

Fig. 7.

In vivo antitumor efficacy and biosafety of FbM@IR/RSL3 NPs in tumor-xenografted mice. (A) Schematic illustration of the establishment of MDA-MB-468 and KYSE-30 tumor-bearing mouse models and drug administration schedule. (B) Tumor volume changes of MDA-MB-468-bearing mice during 21 days of treatment with different formulations (PBS, RSL3, FbM@IR NPs and FbM@IR/RSL3 NPs) with or without NIR irradiation (808 nm, 0.8 W/cm²). The treatment doses were 1.2 mg/kg for RSL3 and 1.6 mg/kg for IR780-DSPE-PEG, with five mice per group (n = 5). (C) Average tumor weights of MDA-MB-468 tumors after 21 days of treatment with different formulations (n = 5). (D) Intra-tumoral MDA levels in MDA-MB-468 tumors 48 h after the final NIR therapy (n = 3). (E) Representative images of excised MDA-MB-468 tumors collected at day 21 (n = 5). (F) Average body weights of MDA-MB-468 tumor-bearing mice during 21 days of treatment. (G) H&E staining of MDA-MB-468 tumor tissues after the observation period (scale bar = 200 μm). (H) IHC staining of GPX-4 in MDA-MB-468 tumors collected 48 h after the last NIR irradiation (scale bar = 100 μm, n = 3). Data are shown as mean ± SEM based on all the replicates. p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***)

The involvement of ferroptosis in therapeutic tumor cell death was further explored by detecting tumor tissue malondialdehyde (MDA) levels. MDA is an end product of lipid peroxidation and serves as a key marker of ferroptosis. As shown in Fig. 7D, both FbM@IR NPs(+) and FbM@IR/RSL3 NPs(+) treatments resulted in elevated intratumoral MDA levels (6.61 and 10.26 nM/mg, respectively), suggesting photothermal-induced ferroptosis. Notably, FbM@IR/RSL3 NPs induced a higher MDA concentration (6.66 nM/mg) compared to free RSL3 treatment (4.71 nM/mg), due to the increased RSL3 accumulation at the tumor site. The highest MDA level was observed in the FbM@IR/RSL3 NPs(+) group (10.26 nM/mg), indicating the synergistic ferroptotic effect of RSL3 and photothermal therapy. Moreover, the most pronounced downregulation of tumor tissue GPX-4 was observed in FbM@IR/RSL3 NPs(+) group (Fig. 7H), which further supported the activation of key regulated pathway of ferroptosis in vivo.

The safety of FbM@IR/RSL3 NPs(+) was validated in both MDA-MB-468 and KYSE-30 tumor-bearing mouse models. In both models, stable weight gain was observed in all treatment groups throughout the drug administration and observation periods, with no significant body weight differences between groups (Fig. 7F and Figure. S7D). To further assess the in vivo toxicity of FbM@IR/RSL3 NPs(+), we performed pathological section examination, blood routine tests and serum biochemistry analysis by the end of observation period. H&E staining of major organs from both MDA-MB-468 and KYSE-30 tumor-bearing mice showed no significant pathological abnormalities in FbM@IR/RSL3 NPs(+)-treated mice (Figure. S8 and S9). Routine blood parameters across all treatment groups were within the normal range and no statistically significant differences were observed between FbM@IR/RSL3 NPs(+) and PBS groups (Figure. S10A and S11A). Additionally, serum biochemical analysis showed no significant changes in liver and kidney function markers, including aspartate aminotransferase (AST), alanine aminotransferase (ALT), total serum bilirubin (TBIL), uric acid (UA), and creatinine (CREA) across both models (Figure. S10B and S11B).

Discussion

Although cell membrane-camouflaged nanoparticles provide a promising approach in cancer nanomedicine, those derived from tumor cells still face several translational issues, including oncogenic risk, limited scalability and the need for invasive tissue collection. To overcome these challenges, we developed a nanoparticle system using membranes derived from dermal fibroblasts, which are clinically safe, easily accessible and can be expanded and genetically modified in vitro. By engineering fibroblast membranes to overexpress the chemokine receptor CXCR4, we aimed to enhance tumor-specific accumulation by targeting the CXCL12-rich tumor microenvironment through receptor-ligand binding mechanism.

This nanoparticle system combines IR780, a photothermal agent responsive to NIR irradiation, with a thermosensitive lipid core encapsulating the ferroptosis inducer RSL3. This design allows externally controlled drug release and heat generation in a spatially and temporally precise manner. Mechanistically, localized photothermal heating not only contributes directly to tumor cell damage but also facilitates the rapid release of RSL3. The released RSL3 inhibits GPX4 activity, promoting lipid peroxidation, while hyperthermia further enhances ferroptotic cell death by depleting intracellular GSH and accelerating iron-dependent Fenton chemistry (Scheme 1). In multiple tumor xenograft models, this combinatorial approach resulted in significant tumor growth inhibition, achieving greater efficacy than either monotherapy alone.

While these results are encouraging, they were obtained in nude mice bearing human xenografts, which helps isolate chemokine-guided targeting and photothermally enhanced ferroptotic effects. Meanwhile, the biomimetic membrane coating provides a host-compatible interface that can attenuate immediate innate recognition and phagocytic uptake compared with bare particles [30–32]. Even so, innate biointeractions-protein corona formation, complement/opsonin deposition and macrophage uptake-still occur in nude mice [33]. By contrast, the absence of mature T cells limits T-dependent humoral responses and the development of immunological memory with repeat dosing [34]. Photothermal and ferroptotic injury may also trigger immunogenic cell death, increasing antigen release and cross-priming. These effects could contribute to durable antitumor memory but are not fully captured in this model. In the future, the platform should be evaluated in immunocompetent models (e.g. BALB/c, C57BL/6) to provide evidence that is more informative for translation.

In terms of clinical translation, using fibroblast-derived membranes helps avoid the oncogenic risks associated with tumor cell membranes and is more amenable to scale-up. CXCR4 levels can also be adjusted by genetic engineering to control receptor density. The NIR-triggered step allows on-demand activation at the tumor site, which may help limit exposure to normal tissues. Our hematology and histopathology data indicate good systemic tolerance. Overall, this platform links biological recognition via the CXCR4-CXCL12 axis with stimulus-responsive release, enabling targeted delivery with temporal control.

Conclusions

With clinical implementation in mind, we developed CXCR4-engineered fibroblast membrane nanovesicles that home to CXCL12-rich tumors. Under NIR irradiation, they generate local heating and trigger RSL3 release, which amplifies ferroptotic cell death. In multiple models, the combination outperformed either monotherapy while maintaining systemic tolerance in blood tests and histology. By pairing an accessible membrane source and programmable receptor for chemokine-guided targeting with external-gated control, this platform tackles two common gaps-insufficient tumor specificity and limited spatiotemporal control-and supports further translational development.

Methods

Materials: Triethylamine (TEA) and octadecylamine (ODA) were purchased from Aladdin Reagent Co., Ltd. (Shanghai, China). IR780 was obtained from Life Technologies (CA, USA). 1,2-Dipalmitoylsn-glycero-3-phosphocholine (DPPC) and PEG2000-DSPE-PEG were supplied by Ponsure Biotechnology (Shanghai, China). Fluorescein isothiocyanate (FITC) was purchased from Sigma-Aldrich. DMEM, fetal bovine serum (FBS), penicillin-streptomycin (PS), PBS, hank’s balanced salt solution (HBSS) and trypsin were sourced from Gibco. Primocin was provided by InvivoGen (CA, USA). (1 S,3R)-RSL3 was obtained from MedChemExpress, USA.

Cell preparation and culture: Murine dermal fibroblasts (MF) were isolated as previously described by Nicolas, et al. [35]. In brief, the back skin of neonatal mice was disinfected and carefully excised using sterile forceps. The skin was then cut into small pieces (around 0.5–1.5 cm³) and washed with PBS three times. The skin fragments were placed onto the bottom of a 10 cm culture dish, with 1 mL of culture medium (DMEM supplemented with 10% FBS) added to just cover the tissue. The dish was incubated at 37 °C with 5% CO₂ for 48 h to allow tissue adhesion. After that, 2 mL of culture medium was gently added, and the cultures were maintained for 5 days to facilitate fibroblast migration from the tissue fragments. Once fibroblasts reached 80% confluency, they were trypsinized and passaged at a 1:3 ratio for expansion. The cells were maintained in DMEM containing 10% FBS and 0.2% primocin (1 ml per 500 ml medium). Human breast adenocarcinoma (MDA-MB-468; catalogue number: CL-0290; RRID: CVCL_0419), esophageal squamous cell carcinoma (KYSE-30; catalogue number: CL-0577; RRID: CVCL_1351) and mouse embryonic fibroblasts (NIH-3T3; catalogue number: CL-0171; RRID: CVCL_0594) cell lines were purchased from Procell Life Science &Technology (Wuhan, China) in May 2022 and cultured in DMEM supplemented with 10% FBS and 1% PS. Cell cultures were incubated in a 37 °C humidified incubator with 5% CO₂. All cells were performed mycoplasma screening using MycoAlert^® Mycoplasma Detection Kit (Lonza, USA) and found to be negative.

Genetic modification of MF and membrane extraction: Murine dermal fibroblasts were genetically modified by lentiviral transfection. Lentiviral vectors carrying the murine CXCR4 gene or a control sequence were designed and constructed as “lv-CXCR4” or “lv-CTR” by GeneChem (Shanghai, China). Fibroblasts were seeded at a density of 8 × 10⁴ per well in 6-well culture plates and transfected with the lentiviral vectors. After 48 h of incubation at 37 °C with 5% CO2, the transfection medium was replaced with standard culture medium containing 2 µg/mL puromycin to select stable CXCR4-overexpressing (CXCR4-OE) fibroblasts or control (CTR) fibroblasts. The expression of CXCR4 was validated by RT-qPCR, western blotting and immunofluorescence.

To isolate the MF membrane fragments, the Membrane and Cytosol Protein Extraction Kit (Beyotime, China) was used according to the manufacturer’s protocol. Modified fibroblasts were harvested from culture plates using a cell scraper and centrifuged at 4 °C, 600 g for 5 min to obtain cell pellets. The cell pellets were then resuspended in membrane protein extraction reagent supplemented with 1% phenylmethanesulfonyl fluoride (PMSF) and subjected to three cycles of freeze-thawing between liquid nitrogen and room temperature to ensure complete cell disruption. Subsequently, the cell lysates were centrifuged at 4 °C, 600 g for 10 min and the supernatant was collected and further centrifuged at 4 °C, 14,000 g for 30 min. The membrane precipitates (FbM and CTR-FbM) were finally resuspended in PBS and stored at −80 °C for future use.

RT-qPCR: Total RNA of MF (CXCR4-OE and CTR) was isolated using the TaKaRa MiniBEST Universal RNA Extraction Kit (Takara Bio Inc., Japan). The concentration and purity of isolated RNA were assessed using a Nanodrop spectrophotometer. RNA samples were reverse-transcribed to cDNA using the qScript cDNA SuperMix (Quanta Biosciences, USA). Quantitative Reverse Transcription PCR (RT-qPCR) was performed on a Bio-Rad CFX Opus 96 System using SYBR green master mix (Applied Biosystems, USA). The primers are listed as follows: CXCR4 forward: 5’-GAAGTGGGGTCTGGAGACTAT-3’, CXCR4 reverse: 5’-TTGCCGACTATGCCAGTCAAG-3’, ACTB forward: 5’-GGCTGTATTCCCCTCCATCG-3’, ACTB reverse: 5’-CCAGTTGGTAACAATGCCATGT-3’. The expression levels of CXCR4 were normalized to the housekeeping gene ACTB. Relative gene expression was calculated using the 2^−ΔΔCt method.

Western blotting: Membrane protein samples were prepared as described above. Whole-cell proteins were extracted using RIPA lysis buffer (supplemented with 1% PMSF) by incubating the cells on ice for 15 min, followed by centrifugation at 12,000 g for 20 min at 4 °C to remove cell debris. Protein concentrations were determined using the bicinchoninic acid (BCA) assay (KeyGEN BioTECH, China). Equal amounts of protein from each sample were loaded onto SDS-PAGE gels for separation. The separated proteins were then transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, USA). These membranes were subsequently blocked with 5% non-fat dried milk for 1 h at room temperature to prevent nonspecific binding. The membranes were then incubated with primary antibodies at 4 °C overnight, followed by incubation with HRP-conjugated secondary antibodies at room temperature for 1 h. The primary antibodies used were anti-CXCR4 (11073-2-AP, Proteintech, 1:1000), anti-GPX4 (67763-1-Ig, Proteintech, 1:1000) and anti-β-actin (#4970, Cell Signaling Technology, 1:2000). The secondary antibodies included anti-rabbit IgG (7074, Cell Signaling Technology, 1:2000) and anti-mouse IgG (7076, Cell Signaling Technology, 1:3000). The membranes were developed using SuperSignal™ West Pico PLUS Chemiluminescent Substrate (ThermoFisher) and the protein bands were detected and captured using ChemiDoc MP imaging system (BioRad).

Immunofluorescence staining: CXCR4-OE and CTR fibroblasts were seeded at a density of 1 × 10⁵ cells per well in 24-well plates containing coverslips and incubated overnight at 37 °C in a 5% CO2 incubator. The cells were then washed with PBS and fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature. The fixed cells were permeabilized with 0.2% Triton X-100 for 5 minutes and blocked with 3% bovine serum albumin (BSA) for 1 h at room temperature. Subsequently, the cells were incubated with a rabbit anti-CXCR4 primary antibody (11073-2-AP, Proteintech, 1:200) overnight at 4 °C. After three washes with PBS, the cells were incubated with an Alexa Fluor 488-conjugated anti-rabbit antibody (#4412, Cell Signaling Technology, 1:200) for 1 h at room temperature. Following three times of PBS washing, the cell nuclei were stained with DAPI for 10 min at room temperature. Finally, the coverslips were carefully transferred onto glass slides using fine-tipped forceps and imaged using a Zeiss confocal laser scanning microscope (LSM 900). The acquired images were processed and analyzed using Zeiss ZEN Blue software.

ELISA: The impact of tumor cells on the secretion of CXCL12 by MF was evaluated by an enzyme-linked immunosorbent assay (ELISA) kit (FineTest, Wuhan, China). Briefly, MF, MDA-MB-468 and KYSE-30 cells were individually seeded in 6-well plates at a density of 5 × 10⁵ cells/well in 2 ml of DMEM supplemented with 10% FBS. Concurrently, co-cultures of MDA-MB-468 or KYSE-30 with MF were established at a 1:1 ratio, totaling 1 × 10⁶ cells/well in 2 ml of the same medium. After 72 h of incubation, the supernatants were collected for CXCL12 quantification following the manufacturer’s protocol. Absorbance was measured at 450 nm using a BioTek microplate reader (SYNERGY H1) to determine the CXCL12 levels.

Synthesis of FbM@IR NPs and FbM@IR/RSL3 NPs: The synthesis of the IR780-PEG-DSPE conjugate was performed according to a previously reported method [6]. IR780-modified lipid nanoparticles were synthesized using a solvent diffusion method [36]. In brief, a mixture of 14 mg DPPC, 4 mg IR780-PEG-DSPE, 1 mg cholesterol and 1 mg DSPE-PEG in 2 mL ethanol was stirred in a 60 °C water bath for 1 min. The ethanol solution was then rapidly dispersed into 20 mL F68 aqueous solution (1 mg/mL) while stirring at 60 °C for 8 min. After the reaction, the solution was cooled to room temperature to allow the formation of IR NPs. To prepare RSL3-loaded IR NPs (IR/RSL3 NPs), 4 mg RSL3, 11 mg DPPC, 4 mg IR780-PEG-DSPE and 1 mg cholesterol were dissolved in 2 mL of ethanol and stirred at 60 °C at 400 rpm for 1 min. The RSL3-containing ethanol solution was then dispersed into 20 mL of F68 solution (1 mg/mL) under the same conditions. After 8 min of stirring at 400 rpm in a 60 °C water bath, the solution was equilibrated to room temperature to obtain IR/RSL3 NPs.

The RSL3-loading capacity was analyzed using an HPLC system (Agilent 1260 Infinity II, Germany) with a C18 column. The mobile phase consisted of 1% acetic acid and acetonitrile (30:70, V/V) and the flow rate was maintained at 1 mL/min at 25 °C. The absorption wavelength for detecting RSL3 was set at 290 nm. A standard curve of RSL3 was generated by loading a series of concentrations of RSL3 (5 to 1000 µg/mL) under the same conditions to quantify the amount of RSL3 encapsulated in the NPs. The drug encapsulation efficiency (EE%) and drug loading efficiency (DL%) were calculated using the following formulas: EE% = * 100%, DL% = * 100%.

The fibroblast membrane-coated NPs (FbM@IR NPs and FbM@IR/RSL3 NPs) were synthesized through sonication. Initially, murine dermal fibroblast membrane fragments were sequentially extruded through polycarbonate membranes with pore sizes of 400 nm and 200 nm. The extruded membrane solution was then mixed with IR/NPs or IR/RSL3 NPs at a mass ratio of 2:1 (relative to the mass of IR780-PEG-DSPE). The mixture was subjected to sonication at 20% amplitude for 1 min (120 W, 20 kHz) to obtain the FbM@IR NPs and FbM@IR/RSL3 NPs.

Transwell invasion assay: Transwell invasion assays were performed by coating transwell inserts (8 μm pore size) with Matrigel solution (diluted 1:30 with serum-free DMEM). The upper chambers were seeded with genetically modified MF (5 × 10⁴ cells in 150 µL serum-free DMEM). The lower chambers contained 600 µL of DMEM supplemented with 5% FBS and various treatments: CXCL12 at 10 ng/mL, CXCL12 with LIT-927 (MedChemExpress, USA) at 90 ng/mL, CXCL12 with control membrane-coated nanoparticles (CTR-FbM@IR NPs at 25 µg/mL), and CXCL12 with FbM@IR NPs at 25 µg/mL. After 30 h of incubation, cells on the underside of the inserts were fixed with 4% PFA for 20 min and stained with 0.1% crystal violet for 30 min. Non-invaded fibroblasts were removed from the upper side of the inserts using a cotton swab, and the number of invaded cells on the lower side was quantified under a microscope.

Characterization of FbM@IR NPs and FbM@IR/RSL3 NPs: The particle size and zeta potential of FbM@IR NPs and FbM@IR/RSL3 NPs were determined by Zetasizer (3000HS, Malvern, UK). Stability was evaluated by monitoring the particle size in full serum at room temperature for up to 4 days. For morphological analysis, the samples were placed onto transmission electron microscopy (TEM) copper grids covered with carbon films and were subsequently left to air-dry at room temperature overnight, followed by uranyl acetate staining. The morphology of the samples was visualized by TEM microscopy (HT7700, HITACH, Japan). Membrane shell thickness was estimated on scale-calibrated TEM micrographs in ImageJ by measuring the perpendicular width of the membrane at annotated locations.

Evaluation of cellular uptake and spheroid penetration of FbM@IR NPs: To evaluate cellular uptake and penetration of FbM@IR NPs into multicellular spheroids, IR NPs were labeled with FITC prior to fibroblast membrane coating via sonication. FITC was conjugated to ODA to synthesize FITC-ODA, following a previously published method [37]. FITC-ODA (1 mg) was incorporated into IR NPs through the solvent diffusion method mentioned above, and the resulting nanoparticles were then coated with the membrane via sonication to obtain FITC-labeled FbM@IR NPs.

MDA-MB-468 cells (8 × 10⁵ cells/well) and KYSE-30 cells (5 × 10⁵ cells/well) were seeded in 6-well plates. Cells were incubated with FITC-labeled FbM@IR NPs or CTR-FbM@IR NPs (IR780-DSPE-PEG: 2.5 µg/mL) for 3–9 h. Cellular uptake of FbM@IR NPs was observed through confocal microscopy, and the intracellular FITC fluorescence intensity was analyzed using ZEN blue software.

For spheroid penetration studies, KYSE-30 cells and MF were labeled with CellTracker™ Deep Red dye (1 µM) and CellTracker™ Orange CMTMR dye (5 µM), respectively. The labeled cells were seeded at a 1:1 ratio (3 × 10³ cells per well) in Corning 96-well ultra-low attachment plates. After 72 h of co-culture, multicellular spheroids formed and were treated with FITC-labeled FbM@IR NPs or CTR-FbM@IR NPs (IR780-DSPE-PEG: 9.9 µg/mL) for 24 h. The spheroids were then washed with PBS, transferred to confocal dishes and analyzed using Z-stack scanning on a confocal microscope.

Confocal imaging and fluorescence quantification: Confocal immunofluorescence (IF) images were acquired using Zeiss ZEN software (blue edition) with identical microscope settings across groups. For each field, equal-sized square regions of interest (ROIs) were placed over areas with comparable cell density, and equal-sized background ROIs were drawn in cell-free areas of the same field. ZEN reported the mean gray value for each ROI; background-corrected intensity was calculated as relative IF intensity = Mean(cell ROI) - Mean(background ROI) [38]. One field-level value was obtained per image, and three images (fields) were analyzed for each group. For the FITC channel (488 nm excitation), background-corrected FITC intensity is reported without additional normalization. For FerroOrange (561 nm excitation), signals were processed using the same ROI-based background subtraction to obtain per-field IF intensity, then expressed as relative Fe(II) fold change by normalizing each treatment’s per-field IF intensity to the mean per-field IF intensity of the NC control (set to 1.0).

In vivo tumor targeting capacity of FbM@IR/RSL3 NPs: BALB/c nude mice (male and female, 4–6 weeks old) were purchased from Beijing Weitong Lihua Biotechnology to establish breast and esophageal cancer models. To establish the breast cancer mouse model, 1 × 10⁷ MDA-MB-468 cells suspended in 100 µL of a 1:1 mixture of PBS and Matrigel were injected into the left mammary gland of 6-week-old female BALB/c nude mice. For the esophageal carcinoma model, 8 × 10⁶ KYSE-30 cells (in a 1:1 PBS/Matrigel mixture) were subcutaneously injected into the left flank of 6-week-old male BALB/c nude mice. When tumors reached approximately 200 mm³, FbM@IR/RSL3 NPs or CTR-FbM@IR/RSL3 NPs (IR780-PEG-DSPE: 0.7 mg/kg) were intravenously administered via the tail vein. One hour prior to NPs injection, mice received an intraperitoneal injection of the CXCR4 antagonist Plerixafor (3.1 mg/kg, MedChemExpress, USA) or the CXCL12 inhibitor LIT-927 (3.3 mg/kg). In vivo fluorescence distribution of IR780-labeled NPs was monitored at 2, 6, 24, 30 and 48 h using an IVIS Lumina III imaging system (PerkinElmer, USA). After 48 h, the mice were euthanized, and their tumors along with major organs (heart, liver, spleen, lungs, kidneys and lymph node) were collected. Ex vivo fluorescence imaging of these tissues was performed using the same system. The excitation wavelength for fluorescence imaging was 780 nm and emission was captured between 760 and 845 nm.

Photothermal effect of FbM@IR NPs and FbM@IR/RSL3 NPs: The absorption spectra of FbM@IR NPs and FbM@ IR/RSL3 NPs were measured using a spectrophotometric microplate reader (BioTek SYNERGY H1), with the wavelength range scanned from 400 nm to 1000 nm. To evaluate the in vitro photothermal performance of IR780, solutions of PBS, FbM@IR NPs and FbM@IR/RSL3 NPs (equivalent IR780-DSPE-PEG concentration: 19.9 µg/mL) were exposed to NIR irradiation (808 nm, 1.0 W/cm²) for 3 min. The temperature changes were recorded using a FLIR thermal camera to monitor the heating efficiency of each sample. For in vivo photothermal evaluation, KYSE-30 tumor-bearing mice were intravenously injected with 250 µL of PBS, FbM@IR NPs or FbM@IR/RSL3 NPs (IR780-DSPE-PEG: 99.3 µg/mL). After 24 h, the tumors were irradiated with an NIR laser (808 nm, 0.8 W/cm²) for 5 min. Thermographic images were captured at various time points during the irradiation using the thermal camera.

NIR-triggered drug release evaluation: The NIR-triggered drug release profile of RSL3-loaded membrane-coated NPs was assessed using the dialysis method. Briefly, 2.0 mL of FbM@IR/RSL3 NPs or free RSL3 solution was placed into dialysis bags with a molecular weight cut-off (MWCO) of 3.5 kDa. The bags were then submerged in 10 mL PBS (pH 6.8) and incubated at 37 °C water bath with shaking. After 4 h of incubation, the dialysis bags were exposed to an 808 nm NIR laser (1.0 W/cm²) for 3 min. Samples of the incubation medium were collected at different time points (1 h, 2 h, 4 h, 6 h, 8 h, 12 h, 24 h and 48 h), and the RSL3 concentration was analyzed by HPLC.

Cytotoxicity assay: The in vitro safety of drug-free FbM@IR NPs was evaluated in both tumor cells (MDA-MB-468 and KYSE-30) and normal cells (NIH-3T3). MDA-MB-468, KYSE-30 and NIH-3T3 cells were seeded in 96-well plates at densities of 1 × 10⁴, 1 × 10⁴ and 8 × 10³ cells per well, respectively. Following overnight incubation, the cells were treated with different concentrations of FbM@IR NPs (20, 40, 60, 80, 100 and 120 µg/mL) for 48 h. After the incubation period, the medium was replaced with 100 µL of fresh medium containing 10 µL of Cell Counting Kit-8 (CCK-8) solution (Meilun Biotechnology, Dalian, China), followed by an additional 4 hours of CCK-8 metabolism. The absorbance of each well was measured at 450 nm using a microplate reader to assess cell viability.

For detecting the cytotoxicity of FbM@IR/RSL3 NPs, MDA-MB-468 and KYSE-30 cells were seeded in 96-well plates at a density of 1 × 10⁴ cells per well, respectively. After 24 h of incubation, the cell culture medium was replaced by different formulations, including negative control (NC), free RSL3, FbM@IR NPs, FbM@IR/RSL3 NPs. The NC group received fresh culture medium without any treatment. For MDA-MB-468 cells, RSL3 was administered at a concentration of 1 µg/mL for both the free RSL3 and FbM@IR/RSL3 NPs groups, while the IR780-DSPE-PEG concentration was fixed at 3.5 µg/mL for FbM@IR NPs and FbM@IR/RSL3 NPs. In KYSE-30 cells, the concentrations of RSL3 and IR780-DSPE-PEG were set at 0.8 µg/mL and 3.5 µg/mL, respectively. After 4 hours of incubation with the respective treatments, all groups were exposed to NIR laser irradiation (808 nm, 1.0 W/cm²) for 3 min. Cells were then incubated for an additional 20 h. Cytotoxicity was assessed using the CCK8 assay, and absorbance was detected using a microplate reader.

Lipid peroxidation detection: Lipid peroxidation was evaluated using BODIPY™ 581/591 C11 Lipid Peroxidation Sensor (D3861, Invitrogen), following the manufacturer’s instructions. Briefly, MDA-MB-468 and KYSE-30 cells were respectively seeded in 6-well plates at a density of 5 × 10⁵ cells per well and incubated overnight. Subsequently, cells were subjected to different treatments: NC, NC with NIR laser (NC(+)), RSL3, RSL3 with NIR laser (RSL3(+)), FbM@IR NPs, FbM@IR NPs with NIR laser (FbM@IR NPs(+)), FbM@IR/RSL3 NPs and FbM@IR/RSL3 NPs with NIR laser (FbM@IR/RSL3 NPs (+)). RSL3 concentrations were set at 1 µg/mL for MDA-MB-468 cells and 0.8 µg/mL for KYSE-30 cells, with IR780-DSPE-PEG fixed at 3.5 µg/mL for both cell lines. After 4 h of incubation, the NIR irradiation groups (NC(+), RSL3(+), FbM@IR NPs(+) and FbM@IR/RSL3 NPs(+)) were exposed to NIR laser (808 nm, 1.0 W/cm²) for 3 min. Following an additional 4-hour incubation period, the media were removed, and the cells were incubated with 5 µM BODIPY™ 581/591 C11 in fresh medium for 30 min at 37 °C. The cells were then harvested by trypsinization, resuspended in PBS, and analyzed for lipid peroxide levels using a flow cytometer (CytoFLEX, Beckman Coulter, CA, USA). The BODIPY™ 581/591 C11 fluorescence was excited with a 488 nm laser, and emissions were collected using the FL1 channel.

Intracellular GSH and GSSG detection: The levels of GSH and oxidized glutathione (GSSG) were detected using the GSH/GSSG-Glo™ Assay (Promega Corporation, USA). MDA-MB-468 and KYSE-30 cells were plated at a density of 1 × 10⁴ cells per well in white, opaque 96-well plates and incubated overnight. On the following day, the cell culture medium was replaced with the corresponding treatments: NC, NC(+), RSL3, RSL3(+), FbM@IR NPs, FbM@IR NPs(+), FbM@IR/RSL3 NPs and FbM@IR/RSL3 NPs(+). The concentration of RSL3 used was 1 µg/mL for MDA-MB-468 cells and 0.8 µg/mL for KYSE-30 cells, while the concentration of IR780-DSPE-PEG was 3.5 µg/mL for both cell lines. At the 4-hour mark of a 6-hour incubation, the four laser treatment groups were exposed to NIR irradiation (808 nm 1.0 W/cm²) for 3 min. Following the 6-hour incubation, the media were removed, and the cells were lysed with either the total GSH lysis buffer or the GSSG lysis buffer provided in the assay kit for 5 min. The cell lysates were then incubated with the luciferin generation reagent at room temperature for 30 min, followed by the addition of luciferin detection buffer and an additional incubation for 15 min at room temperature. The luminescence was measured using a microplate reader. The concentrations of total GSH and GSSG were determined from relative luminescence units (RLU) using a standard curve generated from a series of GSH concentration gradients. The GSH/GSSG ratio was calculated using the following formula: GSH/GSSG ratio = .

Intracellular ferrous iron detection: Fe(II) levels were detected using the Dojindo FerroOrange kit according to the manufacturer’s instructions. MDA-MB-468 cells and KYSE-30 cells were seeded in 35 mm confocal dishes at a density of 4 × 10⁵ cells per dish. The cells were then exposed to various formulations, including NC, RSL3, FbM@IR NPs, FbM@IR NPs(+), FbM@IR/RSL3 NPs and FbM@IR/RSL3 NPs(+). For MDA-MB-468 cells, the concentration of RSL3 was 1 µg/mL, while for KYSE-30 cells it was 0.8 µg/mL; the concentration of IR780-DSPE-PEG was 3.5 µg/mL for both cell lines. The treatment duration was 6 h, with the laser-treated groups receiving 3 min of NIR laser irradiation (1.0 W/cm²) at the fourth hour. After treatment, the cells were washed three times with HBSS and then incubated with 1 µM FerroOrange working solution at 37 °C for 30 min. Fluorescence images were captured using a confocal microscope with excitation at 561 nm and emission at 572 nm.

In vivo antitumor efficacy evaluation in tumor-bearing mouse models: The breast and esophageal tumor models were established following the previously described methodology. To investigate the anti-tumor effects of various agents, MDA-MB-468 and KYSE-30 tumor-bearing mice were assigned to eight groups: PBS, PBS(+), RSL3, RSL3(+), FbM@IR NPs, FbM@IR NPs(+), FbM@IR/RSL3 NPs and FbM@IR/RSL3 NPs(+). Each group consisted of five mice. Once tumors reached a volume of 80–100 mm³, the mice were intravenously administered the respective formulations. The dosing regimen was set at 1.2 mg/kg for RSL3 and 1.6 mg/kg for IR780-DSPE-PEG, administered every other day for a total of five treatments. The laser-treated groups received a 3-minute NIR laser irradiation (808 nm, 0.8 W/cm²) 24 h after each injection. Tumor volumes were measured every other day using a vernier caliper to assess treatment effectiveness, and body weight was monitored concurrently to evaluate treatment safety. Mice that achieved a complete response, defined as no measurable tumor by caliper, were retained in the analysis, and a tumor volume of 0 mm³ was assigned at all subsequent time points. The mice were euthanized 21 days after the initial administration, and their tumors were excised and weighed. The tumor tissues were fixed in 4% PFA, processed into paraffin sections, and stained with hematoxylin and eosin (H&E) to further evaluate treatment efficacy. Mice’s major organs (heart, liver, spleen, lungs, kidneys, lymph node) were collected and processed for H&E staining to assess potential side effects. Additionally, blood samples were obtained for routine blood testing, and serum separation was performed to evaluate liver and kidney functions.

In vivo tumor MDA quantification: MDA-MB-468 tumor tissues were collected 48 h after the final NIR irradiation treatment. The excised tumors were homogenized in IP lysis buffer at a ratio of 1 mg of tissue per 10 µL of buffer (w/v). Homogenized samples were centrifuged at 10,000 g for 15 min at 4 °C to collect the supernatant for further analysis. The concentration of MDA in the supernatants was measured using an MDA assay kit (Solarbio, Beijing, China). Protein concentrations in the same samples were determined by the BCA assay. MDA levels were normalized to the protein content and displayed as nanomoles per milligram of protein (nM/mg).

Immunohistochemical (IHC) examination: Tumors from the MDA-MB-468 mouse model were harvested 48 h following the last NIR therapy, fixed in 4% PFA, and processed into paraffin sections. Tissue sections were then deparaffinized and rehydrated using distilled water. Antigen retrieval was performed using citrate buffer (pH 6.0) under pressure boiling, followed by three times PBS-washing, each for 5 min. To block endogenous peroxidase activity, the sections were incubated in 3% hydrogen peroxide for 25 min at room temperature, followed by PBS washes. The sections were then blocked with 3% BSA for 30 min at room temperature to prevent non-specific binding. This was followed by overnight incubation at 4 °C with anti-GPX4 primary antibody (67763-1-Ig, Proteintech, 1:50 dilution). After primary antibody incubation, sections were washed in PBS and incubated with HRP-conjugated goat anti-mouse IgG (GB23301, Servicebio, 1:200 dilution) for 50 min at room temperature. Following PBS washes, diaminobenzidine (DAB) was applied to visualize the staining and hematoxylin was used for counterstaining of the nuclei. After dehydration and mounting under coverslips, the sections were examined using a bright-field microscope.

Statistical analysis: All statistical analyses were conducted using GraphPad Prism 10.0. Data were displayed as mean ± standard deviation (SD) or mean ± standard error of the mean (SEM). Student’s t-test was used for comparisons between two groups, while one-way ANOVA was applied for comparing the data across multiple groups. Statistical significance was considered as p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***).

Author contributions

LB conceived the study, performed the experiments, conducted data analyses and wrote the manuscript. LfZ, ML and JL assisted with experiments and data analyses. YmY, YzL, HfL and FmK provided related technical and methodical support for this study. XyG and YnT conceived, supervised the study and revised the manuscript. All authors have agreed with the manuscript for publication.

Funding

This work was supported by Shenzhen Medical Research Fund (A2303055). Also, this study was supported by Shenzhen Science and Technology Program (No. ZDSYS20210623091811035, JCYJ20220818103014030, JCYJ20220818103012025, KQTD20180411185028798), and Sanming Project of Medicine in Shenzhen (SZSM202211017). We also thank Clinical, Translational and Basic Research Laboratory, The University of Hong Kong-Shenzhen Hospital for technical support.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All animal experiments were conducted in accordance with the guidelines approved by the Ethical Committee of the University of Hong Kong-Shenzhen Hospital.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.