FAPI-engineered nanoprobe induces dual-cell ferroptosis therapy and enables light-up magnetic resonance imaging for gastric cancer peritoneal metastasis

Abstract

Gastric cancer with peritoneal metastasis (GC PM) is associated with poor prognosis and limited therapeutic options. Cancer-associated fibroblasts (CAFs), as critical components of the tumor microenvironment (TME), drive tumor progression and confer treatment resistance. Single-cell RNA-seq analysis of gastric cancer specimens revealed that GPX4 expression was elevated in CAFs, indicating their intrinsic resistance to ferroptosis. To overcome this stromal barrier, we developed a multifunctional theranostic nanoprobe, integrating ferroptosis induction, CAF-targeted delivery, and magnetic resonance imaging (MRI). The nanoprobe co-delivers cisplatin (CDDP) and mesoporous superparamagnetic iron oxide nanoparticles (MSPIONs) to enhance lipid peroxidation and is further modified with fibroblast activation protein inhibitor (FAPI) to achieve selective accumulation in CAFs. A pH-responsive gadolinium-based contrast agent Gd-PASP (GP) allows T₂-to-T₁ signal conversion, generating a distinct light-up effect under MRI. After integrating all these components, the final formulation is referred to as CDDP2@MSPION@GP3/FAPI. In vivo evaluations in CAF-enriched subcutaneous and peritoneal metastasis models confirmed efficient CAF targeting, dual-cell ferroptosis activation, stromal disruption, tumor inhibition, and enhanced imaging contrast. This study presents a smart, integrated strategy for ferroptosis-based nanotheranostics, offering a precise and effective approach to treating GC PM.

Graphical abstract

Highlights

• •CDDP2@MSPION@GP3/FAPI integrates ferroptosis, CAF targeting, and MRI imaging.
• •GP enables pH-triggered T₂-to-T₁ MRI signal conversion for tumor visualization.
• •Dual-cell ferroptosis disrupts CAF barrier and enhances GC PM therapy.

1. Introduction

Gastric cancer (GC) is the fifth most commonly diagnosed malignancy and the fifth leading cause of cancer-related death worldwide [1]. Despite advances in surgery, chemotherapy, and targeted therapies in recent years, the treatment of advanced GC remains challenging, with peritoneal metastasis (PM, hereafter referred to as GC PM) representing one of the most critical obstacles [2]. PM is characterized by a high incidence, rapid progression, and extremely poor prognosis, and is considered a major contributor to GC-related mortality [3,4].

The tumor microenvironment (TME) plays a pivotal regulatory role in the occurrence and progression of PM [[5], [6], [7]]. Among its components, cancer-associated fibroblasts (CAFs) are a predominant stromal cell population. CAFs not only promote tumor adhesion, migration, and invasion by secreting cytokines such as Transforming Growth Factor-beta (TGF-β) [8] and Interleukin-6 (IL-6) [9] and remodeling the extracellular matrix (ECM), but also induce immunosuppression and enhance chemoresistance, thereby accelerating tumor progression [[10], [11], [12]]. Notably, CAFs are widely distributed around tumor masses and can form a "protective barrier" surrounding neovasculature, impeding drug penetration and supporting tumor cell survival [13,14]. Therefore, targeting CAFs in combination with tumor cell inhibition may serve as an effective strategy to dismantle the TME support system from its source.

In recent years, ferroptosis has attracted increasing attention as a form of programmed cell death that depends on iron ions and is driven by lipid peroxidation [[15], [16], [17]]. In cancer therapy, ferroptosis shows great promise as a precision treatment strategy because it can bypass conventional apoptotic pathways [[18], [19], [20]]. However, current studies on ferroptosis have mainly focused on tumor cells themselves [21,22], while key components of the TME, particularly CAFs, have not been systematically investigated in this context. Previous study has shown that CAFs can inhibit ferroptosis in GC cells by suppressing the accumulation of lipid reactive oxygen species (lipid-ROS) [23]. To explore whether CAFs themselves exhibit ferroptosis resistance, we analyzed single-cell RNA sequencing (scRNA-seq) data from GC patients available in public databases. The analysis revealed that the ferroptosis-negative regulator Glutathione Peroxidase 4 (GPX4) was significantly upregulated in CAFs compared to normal fibroblasts (NFs), which may help CAFs resist ferroptosis within the TME and sustain their tumor-promoting activity.

Delivering exogenous iron to tumors has been widely recognized as an effective strategy to trigger ferroptosis [[24], [25], [26]], primarily by increasing the intracellular labile iron pool, promoting Fenton reactions that generate hydroxyl radicals (·OH), and driving lethal lipid peroxidation (LPO). Recent study has demonstrated that mesoporous superparamagnetic iron oxide nanoparticles (MSPIONs) are pH-sensitive and degrade under acidic TME conditions, releasing Fe²⁺/Fe³⁺ ions [27]. These ions sustain reactive oxygen species (ROS) generation via Fenton-like reactions and subsequently induce ferroptosis through excessive lipid peroxidation [[28], [29], [30], [31]].

Meanwhile, cisplatin (CDDP), a first-line chemotherapeutic agent for GC [32], has also been shown to sensitize cells to ferroptosis by promoting ROS accumulation [33], depleting intracellular glutathione (GSH), and inactivating GPX4 [34]. Notably, previous studies have reported that CDDP can enhance Fe³⁺ to Fe²⁺ recycling and accelerate lipid peroxide accumulation in tumors, thereby synergizing with iron-based nanomaterials to amplify ferroptotic responses [27,35].

Based on these mechanisms, we propose that the combination of MSPION and CDDP synergistically induces ferroptosis in both tumor cells and CAFs. MSPION serves as an iron reservoir and catalytic platform for ROS generation, while CDDP disrupts antioxidant defenses and promotes lipid oxidation. Together, they act on two key ferroptotic pathways, potentially overcoming the stromal barrier posed by CAFs and achieving enhanced dual-cell ferroptosis-mediated therapeutic efficacy [15,36].

In this study, we developed a smart theranostic nanoplatform capable of precisely remodeling the TME and enabling integrated diagnosis and therapy for GC PM. To achieve active targeting of CAFs, we employed a fibroblast activation protein inhibitor (FAPI) as a surface ligand. The resulting multifunctional nanoprobe, CDDP2@MSPION@GP3/FAPI, was constructed by encapsulating CDDP within MSPIONs, integrating the T₁-weighted MRI contrast agent (CA) Gd-PASP (GP) [37], and functionalizing the surface with FAPI (Scheme 1A). We selected MSPIONs as the core material of our nanoplatform due to their multiple structural and functional advantages. First, their unique mesoporous architecture significantly enhances drug-loading efficiency, enabling the co-encapsulation of both hydrophilic and hydrophobic agents and providing superior delivery performance compared to non-mesoporous SPIONs [28]. Second, MSPIONs can gradually degrade under the acidic conditions of the TME, releasing Fe²⁺ ions that trigger Fenton reactions to generate hydroxyl radicals, thereby inducing lipid peroxidation and ferroptosis. Moreover, a one-pot, surfactant-free green synthesis strategy was employed, offering potential for large-scale production and facilitating future clinical translation [28]. MSPIONs also exhibit favorable particle size distribution and excellent biocompatibility in vivo, which helps avoid rapid clearance by the reticuloendothelial system (RES) and ensures desirable biodistribution [38,39]. Importantly, MSPIONs possess high r₂ relaxivity, resulting in excellent T₂-weighted MRI performance. Upon exposure to acidic environments, the release of GP enables a transition from quenched T₂ signals to enhanced T₁ signals, thereby improving imaging contrast and spatial resolution [40,41].

Scheme 1.

Schematic illustration of CDDP2@MSPION@GP3/FAPI nanoparticles for dual-cell ferroptosis and MRI. A) Synthetic process of CDDP2@MSPION@GP3/FAPI. B) Mechanism of the dual-cell ferroptosis strategy for tumor therapy. C) Illustration for the light-up effect of tumor MRI.

Building on this rational design, the nanoprobe is activated in the acidic TME, where it releases Fe²⁺/Fe³⁺ and CDDP to synergistically induce ferroptosis in both tumor cells and CAFs (Scheme 1B). This dual-cell ferroptosis activation not only directly promotes cell death but also disrupts the supportive stromal architecture, thereby enhancing intratumoral drug penetration and amplifying therapeutic efficacy. The ability to modulate both malignant and stromal components makes this strategy particularly well-suited for addressing the complex biological barriers present in GC PM.

To support real-time monitoring of the treatment process, the system incorporates an intelligent T₂-to-T₁ signal conversion mechanism. Upon exposure to acidic conditions, MSPIONs degrade, releasing the embedded GP and inducing a transition from quenched T₂-weighted to enhanced T₁-weighted signals. This shift generates a distinctive “light-up” effect on MRI, offering improved imaging contrast and spatial resolution while minimizing false-positive signals (Scheme 1C). The dynamic imaging capability enables precise spatiotemporal visualization of nanoprobe accumulation and therapeutic progress. By comparison, conventional small-molecule MRI CAs often suffer from poor specificity, rapid clearance, and limited diagnostic windows [40,42]. In contrast, our nanoplatform integrates structural versatility, selective targeting, and imaging functionality into a single system, enabling TME-responsive precision theranostics.

To comprehensively evaluate its diagnostic and therapeutic efficacy, we established clinically relevant, patient-derived models by isolating primary CAFs from surgical GC specimens and co-injecting them with MKN45 GC cells. Two in vivo models were constructed: (1) a CAF-enriched subcutaneous tumor model, used to assess nanoprobe accumulation and therapeutic effects; and (2) a peritoneal metastasis model, which mimics intraperitoneal dissemination of tumor cells and evaluates the efficacy of the dual-cell ferroptosis strategy and its MRI visualization performance in complex TME.

In summary, this study presents the first integration of CAF targeting, dual-cell ferroptosis, and MRI functionality into a single nanoplatform for smart theranostics of GC PM. This strategy effectively remodels the TME and enhances both treatment efficiency and imaging precision, offering new insights for ferroptosis-based interventions and clinical translation of nanomedicine.

2. Results and discussion

2.1. Isolation and validation of primary fibroblasts

To investigate the phenotypic differences between CAFs and NFs, primary fibroblasts were successfully isolated from fresh GC specimens using a combined enzymatic digestion method involving trypsin and collagenase.

qRT-PCR analysis (Fig. S1, Supporting Information) revealed that the mRNA expression levels of α-Smooth Muscle Actin (α-SMA, encoded by ACTA2), fibroblast activation protein (FAP), and IL-6 were significantly elevated in primary CAFs compared to NFs. α-SMA is a marker of CAF activation, reflecting their myofibroblast-like phenotype [41]; FAP is a well-established CAF marker, frequently upregulated during tumor-associated fibrosis and stromal remodeling [43]; IL-6, a pro-inflammatory cytokine, plays a role in promoting inflammatory responses and tumor cell proliferation within the TME [9,44].

Immunofluorescence staining further confirmed the qRT-PCR findings, showing markedly higher protein expression levels of α-SMA and FAP in CAFs than in NFs (Fig. S2, Supporting Information), consistent with the expected phenotype. In addition, Western blot analysis corroborated these results, demonstrating significantly increased expression of α-SMA and FAP proteins in primary CAFs compared to NFs (Fig. S3, Supporting Information).

Taken together, these results indicate that we successfully isolated primary fibroblasts with characteristic CAF features and verified their identity using multiple molecular and protein-level assays, providing a solid experimental foundation for subsequent functional studies.

2.2. High GPX4 expression in CAFs identified by scRNA-seq analysis

Previous studies have shown that CAFs can inhibit ferroptosis in cancer cells [23,45,46]. To investigate whether CAFs themselves exhibit ferroptosis resistance compared to NFs, we analyzed single-cell RNA sequencing (scRNA-seq) data from GC samples in the GSE183904 dataset. After quality control and clustering, a total of 18,505 cells were divided into 27 separate clusters (Figs. S4 and S5, Supporting Information). Based on the expression profiles of specific marker genes, these clusters were annotated into nine major cell types: LYMPHOID (T&NK cells), MACROPHAGE, PLASMA, FIBROBLAST, EPITHELIAL, ENDOTHELIAL, B-CELL, DENDRITIC, and MAST (Fig. S6, S7 and S8, Supporting Information). To investigate the gene expression differences between CAFs and NFs, fibroblasts were isolated for further analysis (Fig. S9, Supporting Information). The results revealed that GPX4 expression was significantly upregulated in CAFs compared to NFs (Fig. S10, Supporting Information). Given that tumor tissues are often characterized by hypoxia, acidity, and excessive accumulation of ROS, CAFs may enhance their antioxidant systems to maintain lipid homeostasis, thereby improving their survival under such harsh conditions within the TME. The elevated expression of GPX4 not only helps CAFs resist lipid peroxidation-induced damage, but also suggests an inherent tolerance to ferroptosis, which may allow them to sustain their pro-tumorigenic functions.

2.3. Preparation and characterization of CDDP2@MSPION@GP3/FAPI

MSPION was synthesized using the bubble template method as previously reported [28]. Subsequently, CDDP2@MSPION@GP3/FAPI was prepared by sequentially introducing CDDP and GP into the mesopores of MSPION, followed by surface modification with FAPI. MSPION is primarily composed of magnetite (Fe₃O₄) with abundant surface hydroxyl (–OH) groups, which render the nanoparticles negatively charged and provide active sites for the attachment of functional molecules. In step 1, CDDP was loaded into the mesopores of MSPION. CDDP predominantly exists as positively charged platinum complexes in aqueous solution, enabling strong electrostatic attraction to the negatively charged MSPION surface [47]. In addition, CDDP can penetrate the highly ordered porous channels of MSPION, where it is retained through spatial confinement and electrostatic interaction. This dual mechanism accounts for its high loading efficiency and stability. As shown in Table S2 (Supporting Information), the loading content (LC) of CDDP in CDDP2@MSPION is much higher than that of CDDP3-4@MSPION, and slightly lower than that of CDDP1@MSPION. Moreover, the loading efficiency (LE) of CDDP2@MSPION is significantly higher than that of CDDP1@MSPION. Therefore, CDDP2@MSPION was selected as the optimal sample for subsequent synthesis [27]. In step 2, GP was conjugated onto the MSPION surface. GP is primarily composed of poly(aspartic acid) (PASP), which contains abundant negatively charged carboxyl (–COOH) groups that can coordinate with iron ions on the MSPION surface [48,49], thereby firmly anchoring GP to the particle surface and enabling its role as a T₁-weighted MRI CA. Table S3 (Supporting Information) shows that increasing the amount of GP added elevates GP loading, while correspondingly decreasing CDDP loading, reflecting competition for available binding sites. Based on this balance, CDDP2@MSPION@GP3 was chosen as the optimal formulation [50]. In step 3, carboxyl-containing FAPI was conjugated to CDDP2@MSPION@GP3. FAPI contains carboxyl (–COOH) groups that can coordinate with exposed Fe³⁺/Fe²⁺ ions on the MSPION surface, thereby enhancing binding stability, and can also form covalent bonds with amino groups from CDDP molecules on the MSPION surface [48,49], achieving stronger and more stable attachment. As shown in Table S4 (Supporting Information), FAPI modification had minimal effect on the loading capacities of either GP or CDDP.

Fig. lA shows transmission electron microscope (TEM) images of fabricated MSPION, CDDP2@MSPION, CDDP2@MSPION@GP3, CDDP2@MSPION@GP3/FAPI and the magnified view of the picture CDDP2@MSPION@GP3/FAPI, and their particle sizes can be found in Fig. S11 (Supporting Information), indicating that the diameter of the CDDP2@MSPION@GP3/FAPI is 67.26 ± 0.99 nm. The TEM image reveals a distinct mesoporous structure in MSPION, which facilitates efficient drug loading through enhanced surface adsorption. To further verify the chemical composition of CDDP2@MSPION@GP3/FAPI, high-angle annular dark-feld scanning transmission electron microscopy (HAADF-STEM) was used to visualize the key elements. The elements of Fe (from MSPION), Pt (from CDDP), Gd (from GP), and F (from FAPI) can be observed. All of the aforementioned components can be observed in the Energy Dispersive Spectroscopy (EDS) spectrum (Fig. S12, Supporting Information), indicating the successful fabrication of CDDP2@MSPION@GP3/FAPI.

Fig. 1.

The characterization of CDDP2@MSPION@GP3/FAPI. A) TEM images of the MSPION (Ⅰ), CDDP2@MSPION (Ⅱ), CDDP2@MSPION@GP3 (Ⅲ), CDDP2@MSPION@GP3/FAPI (Ⅳ) and the magnified view of the picture Ⅳ (Ⅴ). Scale bar: 100 nm. B) Fe, Pt, Gd and F elemental mapping of CDDP2@MSPION@GP3/FAPI. Scale bar: 50 nm. C) Size distributions. D) Zeta potentials. ns, not significant. E) N₂ adsorption-desorption isotherms of CDDP2@MSPION@GP3/FAPI. F) XPS spectrum of CDDP2@MSPION@GP3/FAPI. G) XRD spectrum of MSPION. H) UV–vis spectra of H₂O₂ solutions reacting with CDDP2@MSPION@GP3/FAPI or TMB at pH 6.5 to confirm •OH generation. I-J) UV–vis spectra of the CDDP2@MSPION@GP3/FAPI solutions with TMB and H₂O₂ to measure the •OH generation at different pH values (5.5, 6.5, or 7.4) (I), or at pH 5.5 for different concentrations (J). K-N) T₁-weighted (K) and T₂-weighted (M) MR images of CDDP2@MSPION@GP3/FAPI solutions with various C_Gd (0–100 μM) incubated at different pH values (5.5, 6.5, or 7.4) for 24 h observed by a 3.0 T MRI system. T₁ relaxation rate (L) and T₂ relaxation rate (N) plotted as a function of C_Gd for CDDP2@MSPION@GP3/FAPI in magnetic field of 3.0 T.

Additionally, variations in hydrodynamic diameters (d_h) and zeta potentials, as measured by dynamic light scattering (DLS), further confirm the successful preparation of CDDP2@MSPION@GP3/FAPI. Specifically, Fig. 1C and S13 (Supporting Information) indicate that the d_h of MSPION is 77.32 nm, increasing to 88.47 nm for CDDP2@MSPION, 98.29 nm for CDDP2@MSPION@GP3, and further to 101.5 nm for CDDP2@MSPION@GP3/FAPI. It is worth noting that the hydrodynamic diameter measured by DLS (101.5 nm) is larger than the core size observed by TEM (67.26 ± 0.99 nm), which is expected due to the presence of a hydration layer and flexible surface modifications in aqueous conditions. [[51], [52], [53]] Such discrepancies are frequently reported in nanoparticle studies and primarily arise from the intrinsic differences between dry-state, number-based TEM measurements and solution-phase, intensity-weighted DLS measurements. [54,55] The zeta potentials of MSPION, CDDP2@MSPION, CDDP2@MSPION@GP3 and CDDP2@MSPION@GP3/FAPI were measured as -17.39 mV, -11.02 mV, -38.49 mV and -39.95 mV, respectively (Fig. 1D and Fig. S14, Supporting Information). This is due to the synthesis of MSPION nanoparticles via the polyol method. During the reaction, ethylene glycol and diethylene glycol were used to control the particle size and surface functionalities of the nanoparticles. The high hydroxyl content in ethylene glycol resulted in MSPION nanoparticles with a high density of hydrophilic hydroxyl groups on their surface. This strong negative charge enhances electrostatic repulsion, preventing nanoparticle aggregation in aqueous media and significantly improving their dispersibility. Additionally, the decrease in negative zeta potential of CDDP2@MSPION is attributed to the introduction of positively charged CDDP on its surface. [56] Consistent with the DLS results, FAPI conjugation exerts minimal influence on these physicochemical parameters, likely due to its low molecular weight.

Fig. 1E shows the N₂ adsorption-desorption isotherms used to examine the mesoporous structure of MSPION and CDDP2@MSPION@GP3/FAPI. The specific surface areas of MSPION and CDDP2@MSPION@GP3/FAPI are measured to be 183.04 m²/g and 38.07 m²/g, respectively. Compared to MSPION, the significant decrease in surface area of CDDP2@MSPION@GP3/FAPI primarily results from the loading of CDDP and GP into the mesopores of MSPION. Considering that FAPI, as a targeting ligand, was introduced after the loading of CDDP and GP, its impact on the pore structure is negligible; instead, FAPI is expected to be displayed on the nanoparticle surface to preserve its biological targeting capability. Fig. S15 (Supporting Information) shows the pore size distribution curves of MSPION and CDDP2@MSPION@GP3/FAPI, which indicates the presence of mesopores on the MSPION surface. The average pore size of MSPION and CDDP2@MSPION@GP3/FAPI are measured to be 22.97 nm and 12.88 nm, which is favorable for the efficient loading and sustained release of CDDP and GP.

Fig. 1F displays the X-ray photoelectron spectroscopy (XPS) spectra of MSPION and CDDP2@MSPION@GP3/FAPI. Specifically, the high-resolution XPS spectrum of CDDP2@MSPION@GP3/FAPI is shown in Fig. S16 (Supporting Information), where the peaks at 725.0 eV and 711.9 eV correspond to Fe³⁺ 2p₁/₂ and 2p₃/₂, respectively, while the peaks at 723.0 eV and 709.9 eV are attributed to Fe²⁺ (Fig. S16A, Supporting Information). These results confirm that MSPION consists of Fe₃O₄ rather than Fe₂O₃ [57]. In addition, the high-resolution XPS spectrum of Gd 4d exhibits two distinct peaks at 149.20 eV and 143.4 eV (Fig. S16B, Supporting Information), and the high-resolution XPS spectrum of Pt 4f shows two characteristic peaks at 75.68 eV and 72.40 eV (Fig. S16C, Supporting Information). The X-ray diffraction (XRD) pattern of MSPION (Fig. 1G) reveals characteristic diffraction peaks of Fe₃O₄ at (111), (220), (311), (222), (400), (422), (511), and (440), which are consistent with previous reports [57,58].

3,3′,5,5′-Tetramethylbenzidine (TMB) is a widely used probe for •OH and is commonly employed to verify •OH generation, [59] which can oxidize TMB to produce a blue-green color. [60] Fig. 1H shows UV–vis spectra of H₂O₂ solutions reacting with CDDP2@MSPION@GP3/FAPI or TMB at pH 6.5, confirming the generation of •OH radicals. No UV–Vis absorption is observed for CDDP2@MSPION@GP3/FAPI in the presence of TMB or H₂O₂ alone. However, a distinct UV–Vis absorption peak at approximately 652 nm appears when CDDP2@MSPION@GP3/FAPI is combined with TMB and H₂O₂. Upon the release of Fe²⁺/Fe³⁺ ions from CDDP2@MSPION@GP3/FAPI, the Fenton reaction occurs immediately, as Fe²⁺ exhibits a rapid Fenton reaction rate constant (k) of 76 M⁻¹s⁻¹. [61] Fig. 1I and Fig. S17 (Supporting Information) demonstrate that the amount of •OH increases as the pH decreases, indicating the pH-dependent catalytic activity of MSPION in the Fenton reaction. This phenomenon is attributed to the more rapid release of Fe²⁺/Fe³⁺ from CDDP2@MSPION@GP3/FAPI in a more acidic environment. Furthermore, the concentration-dependent behavior of CDDP2@MSPION@GP3/FAPI at pH 5.5 can be observed in Fig. 1J and Fig. S18 (Supporting Information), showing that the absorbance at 652 nm increases as the nanoparticles concentration rises.

To further validate the acid-activated degradation behavior of CDDP2@MSPION@GP3/FAPI, TEM images of CDDP2@MSPION@GP3/FAPI incubated in PBS (pH 5.5) at different time points are presented in Fig. S19 (Supporting Information). The degradation is primarily attributed to its Fe₃O₄-based core (MSPION), which reacts with H⁺ under acidic conditions. After 48 h of incubation, the spherical structure of CDDP2@MSPION@GP3/FAPI visibly decomposes into fragments. This pH-dependent degradation facilitates the Fenton reaction by releasing abundant Fe²⁺/Fe³⁺ ions in the TME.

As the concentration of gadolinium ions increases, the T₂-MRI images of CDDP2@MSPION@GP3/FAPI gradually darken, while the T₁-MRI images become progressively brighter at 3.0 T MRI scanner. T₁-weighted (Fig. 1K) and T₂-weighted (Fig. 1M) MRI images illustrate the imaging results of CDDP2@MSPION@GP3/FAPI solution at various Gd concentrations (0–100 μM) after 24-h incubation at different pH values (5.5, 6.5, or 7.4) in a 3.0 T MRI system. The longitudinal relaxivity (r₁) and transverse relaxivity (r₂) of CDDP2@MSPION@GP3/FAPI measured at 3.0 T MRI scanner are respectively calculated to be 13.66 mM⁻¹s⁻¹ and 64.08 mM⁻¹s⁻¹ (Fig. 1L and N). The high r₂/r₁ ratio (4.69) and low r₁ value indicate relatively weak T₁-MRI imaging capability under neutral conditions. Notably, after incubation in an acidic medium (simulating TME), the T₁-weighted MRI signal intensity was significantly enhanced. Interestingly, the r₁ value increased to 20.57 mM⁻¹s⁻¹ in PBS at pH 6.5 and further increased to 26.15 mM⁻¹s⁻¹ in PBS at pH 5.5, while the r₂/r₁ ratio decreased to 1.6 and 0.83, respectively. These results suggest that CDDP2@MSPION@GP3/FAPI is an effective TME-responsive T₁-MRI CA with excellent imaging performance. The significant enhancement of T₁ imaging capability is mainly attributed to the degradation of MSPION and the release of GP under low pH conditions. These findings indicate that CDDP2@MSPION@GP3/FAPI functions as an efficient pH-responsive CA capable of dynamic T₂-to-T₁ signal conversion, generating a characteristic light-up effect and demonstrating strong potential for precise in vivo imaging.

2.4. Evaluation of cellular uptake efficiency of CDDP2@MSPION@GP3/FAPI

Our previous experiments confirmed that the isolated CAFs exhibited significantly higher FAP expression than NFs, providing a solid basis for employing FAPI as a targeting ligand to guide nanoparticles toward CAFs within the tumor stroma.

To further verify whether FAPI-mediated targeting could facilitate active tumor-specific delivery of nanoparticles, we evaluated cellular uptake in MKN45 cells using confocal laser scanning microscopy (CLSM) and flow cytometry. Rhodamine 6G (R6G), a red fluorescent dye, was used to label the nanoparticles for intracellular localization. As shown in Figs. S20 and S21 (Supporting Information), after 12 h of incubation, both CDDP2@MSPION@GP3@R6G and CDDP2@MSPION@GP3/FAPI@R6G treatments resulted in clear red fluorescence signals in MKN45 cells and NFs, indicating successful nanoparticle internalization. Quantitative analysis revealed no significant difference in mean fluorescence intensity (MFI) between the two groups in these cell types. In contrast, in CAFs, CLSM images showed markedly enhanced red fluorescence intensity in the CDDP2@MSPION@GP3/FAPI@R6G group compared to the non-FAPI-modified CDDP2@MSPION@GP3@R6G group, with quantitative analysis confirming a statistically significant difference.

Flow cytometry results were consistent with the CLSM observations: no significant difference in fluorescence intensity was observed between groups in MKN45 cells or NFs, while in CAFs, the CDDP2@MSPION@GP3/FAPI@R6G group exhibited significantly higher MFI than the non-FAPI group (Figs. S22 and S23, Supporting Information).

Taken together, these findings demonstrate that the engineered CDDP2@MSPION@GP3/FAPI nanoparticles not only facilitate efficient cellular uptake but also exhibit strong targeting specificity toward CAFs, supporting their potential for TME accumulation and in vivo biosafety.

2.5. In vitro cytotoxicity and anti-proliferative effects of CDDP2@MSPION@GP3/FAPI

Given the excellent Fenton catalytic activity of the synthesized nanoparticles, we evaluated their in vitro antitumor efficacy using MKN45 cells and CAFs as models through CCK-8 assays. Fig. 2A and B show that at all tested concentrations, drug-free MSPION maintained cell viability above 70 %, indicating good biocompatibility and its potential as an ideal drug delivery carrier. In contrast, CDDP, CDDP2@MSPION@GP3, and CDDP2@MSPION@GP3/FAPI exhibited concentration-dependent cytotoxicity. Notably, in CAFs, CDDP2@MSPION@GP3/FAPI induced the strongest cell-killing effect, which was significantly greater than that of CDDP2@MSPION@GP3, confirming the specific targeting capability of FAPI modification toward CAFs. Moreover, in NFs, cell viability after CDDP2@MSPION@GP3/FAPI treatment remained significantly higher than that in CAFs (Fig. 2C), supporting its selective targeting and cytotoxicity toward CAFs. To further substantiate the dual-cell targeting efficacy, we performed a direct comparison of CDDP2@MSPION@GP3/FAPI cytotoxicity between CAFs and MKN45 cells at identical CDDP concentrations. The results (Fig. S24, Supporting Information) show that, under nanoparticle treatment, the overall viability of CAFs was significantly lower than that of MKN45 cells at most tested concentrations. Two-way ANOVA further confirmed a significant main effect of cell type (P < 0.0001), indicating that the average viability differed between the two cell types and that CAFs are more susceptible to nanoparticle treatment.

Fig. 2.

Assessment of cytotoxicity and proliferation inhibition of CDDP2@MSPION@GP3/FAPI on MKN45 cells and CAFs. A–B) The viabilities of MKN45 cells (A) and CAFs (B) treated with PBS (I), MSPION (II), CDDP (III), CDDP2@MSPION@GP3 (IV) or CDDP2@MSPION@GP3/FAPI (V) for 24 h. C) The viabilities of NFs and CAFs treated with CDDP2@MSPION@GP3/FAPI for 24 h. D–F) Fluorescence images (D) and flow cytometry analysis (E–F) of MKN45 cells from different treatment groups after staining with Calcein-AM/PI or 7-AAD. G–I) Fluorescence images (G) and flow cytometry analysis (H–I) of CAFs from different treatment groups after staining with Calcein-AM/PI or 7-AAD. J–L) Fluorescence images (J) and flow cytometry analysis (K–L) of MKN45 cells from different treatment groups after staining with EdU. M−O) Fluorescence images (M) and flow cytometry analysis (N–O) of CAFs from different treatment groups after staining with EdU. Scale bar: 100 μm ∗P < 0.05. ∗∗P < 0.01. ∗∗∗P < 0.001. ∗∗∗∗P < 0.0001.

Live/dead cell staining was further employed to evaluate the in vitro therapeutic efficacy of the nanoparticles. In this assay, calcein acetoxymethyl ester (Calcein-AM) and propidium iodide (PI) were used to label viable cells (green fluorescence) and dead cells (red fluorescence), respectively, and observed via CLSM. As shown in Fig. 2D and G, MKN45 cells and CAFs in the PBS and MSPION groups primarily exhibited green fluorescence with minimal red signals, indicating good biocompatibility of the material itself. In contrast, enhanced red fluorescence was observed in the CDDP, CDDP2@MSPION@GP3, and CDDP2@MSPION@GP3/FAPI groups, indicating increased cell death. The quantification results (Figs. S25 and S26, Supporting Information) were consistent with the CCK-8 assay. In addition, 7-AAD staining combined with flow cytometry was performed to further evaluate cell death (Fig. 2E, F, H, and I). 7-AAD-positive cells indicate non-viable cells with compromised membrane integrity. The proportion of 7-AAD-positive cells showed a trend consistent with the live/dead staining results, further confirming the synergistic cytotoxicity of CDDP and MSPION in vitro.

In the proliferation assay, the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) was incorporated into the DNA of proliferating cells and visualized by green fluorescence [62]. As shown in Fig. 2J and M, CLSM images of EdU-stained MKN45 cells and CAFs demonstrated abundant EdU-positive nuclei in the PBS and MSPION groups, indicating active proliferation. In contrast, the number of EdU-positive cells markedly decreased in the CDDP, CDDP2@MSPION@GP3, and CDDP2@MSPION@GP3/FAPI groups, suggesting inhibited cell proliferation. Quantitative analysis showed no significant difference in the anti-proliferative effects of CDDP2@MSPION@GP3 and CDDP2@MSPION@GP3/FAPI in MKN45 cells (Fig. S27, Supporting Information). However, in CAFs, the FAPI-modified formulation exhibited significantly greater anti-proliferative efficacy, further confirming its CAF-specific targeting capability (Fig. S28, Supporting Information). Flow cytometry analysis supported these findings, showing a consistent trend with the fluorescence imaging results (Fig. 2K, L, N, and O).

In summary, the combination of MSPION and CDDP demonstrated a potent synergistic cytotoxic effect in vitro. Furthermore, the incorporation of the FAPI targeting ligand significantly enhanced the selective killing of CAFs, suggesting that the designed CDDP2@MSPION@GP3/FAPI nanoplatform holds great potential for tumor stromal targeting and may contribute to overcoming tumor-associated barriers.

2.6. Ferroptosis characteristic evaluated on MKN45 cells and CAFs

This study systematically evaluated the ferroptosis-inducing mechanism of the CDDP2@MSPION@GP3/FAPI nanoplatform and its potential in tumor therapy. Ferroptosis is a form of programmed cell death characterized by iron dependence and excessive accumulation of LPO [[15], [16], [17]].

We first assessed the intracellular generation of ROS in MKN45 cells and CAFs using the ROS-specific fluorescent probe DCFH-DA. Upon oxidation by ROS, DCFH-DA is converted to highly fluorescent DCF, thereby reflecting the oxidative stress level within cells [63]. Flow cytometry analysis revealed negligible fluorescence signals in the PBS group for both MKN45 cells (Fig. 3A) and CAFs (Fig. 3C). In contrast, cells treated with MSPION, CDDP, CDDP2@MSPION@GP3, or CDDP2@MSPION@GP3/FAPI exhibited markedly enhanced green fluorescence, indicating elevated intracellular ROS levels. These results suggest that both MSPION and CDDP contribute to ROS generation. Further quantitative analysis (Fig. 3B and D) showed that among all treatment groups, CDDP2@MSPION@GP3 and CDDP2@MSPION@GP3/FAPI induced the highest ROS levels, with the latter demonstrating particularly pronounced effects in CAFs. This difference is likely attributed to the CAF-targeting capability conferred by FAPI modification, which enhances nanoprobe accumulation within CAFs. CLSM further supported these findings: both MKN45 cells (Fig. S29, Supporting Information) and CAFs (Fig. S31, Supporting Information) displayed significantly increased green fluorescence after treatment. Corresponding quantitative fluorescence intensity analyses (Figs. S30 and S32, Supporting Information) were consistent with the flow cytometry data.

Fig. 3.

Evaluation of ferroptosis characteristics induced by CDDP2@MSPION@GP3/FAPI in MKN45 cells and CAFs. A–B) Fluorescence distribution (A) and quantification of ROS levels (B) in MKN45 cells treated with PBS (I), MSPION (II), CDDP (III), CDDP2@MSPION@GP3 (IV) or CDDP2@MSPION@GP3/FAPI (V) for 24 h, detected by DCFH-DA staining. C–D) Fluorescence distribution (C) and quantification of ROS levels (D) in CAFs from different treatment groups, detected by DCFH-DA staining. E–G) GPX4 expression in MKN45 cells and CAFs from different treatment groups as determined by Western blot (E), with corresponding statistical analysis (F–G). H–J) CLSM images (H) and corresponding flow cytometry analysis (I–J) of MKN45 cells from different treatment groups, stained with JC-1 to assess mitochondrial membrane potential. Red fluorescence indicates JC-1 aggregates in healthy mitochondria, while green fluorescence represents JC-1 monomers in depolarized (damaged) mitochondria. K–M) CLSM images (K) and corresponding flow cytometry analysis (L–M) of JC-1 stained CAFs from different treatment groups. N-P) CLSM images (N), flow cytometry fluorescence distributions (O), and the quantitative analysis (P) of MKN45 cells after treatments of groups I–V and staining with DAPI and C11-BODIPY. Red fluorescence: reduced form for the decreased LPO. Green fluorescence: oxidized form for the increased LPO. Q–S) CLSM images (Q), flow cytometry fluorescence distributions (R), and the quantitative analysis (S) of CAFs after treatments of groups I–V and staining with DAPI and C11-BODIPY 581/591. Scale bar: 100 μm ∗P < 0.05. ∗∗P < 0.01. ∗∗∗P < 0.001. ∗∗∗∗P < 0.0001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

To further verify the occurrence of ferroptosis, we examined the expression of GPX4, a key negative regulator of ferroptosis, using Western blot and immunofluorescence analysis. GPX4 reduces LPO into non-toxic lipids, thereby suppressing ferroptosis. As shown in Fig. 3E–G, treatment with CDDP, CDDP2@MSPION@GP3, and CDDP2@MSPION@GP3/FAPI resulted in a significant downregulation of GPX4 expression, with the most pronounced suppression observed in CAFs treated with the FAPI-modified formulation. In contrast, the reduction of GPX4 expression in MKN45 cells was modest and showed no statistically significant difference between formulations with or without FAPI modification. This result aligns with the mechanism of FAPI-mediated targeting, as FAP is abundantly expressed in CAFs but minimally or not expressed in tumor epithelial cells such as MKN45 [64,65]. Consequently, nanoparticle uptake in MKN45 cells occurs predominantly via nonspecific endocytosis rather than receptor-mediated internalization [[66], [67], [68]], resulting in relatively limited GPX4 suppression. Accordingly, no significant difference in GPX4 levels was detected between FAPI-modified and unmodified nanoparticles in MKN45 cells. Immunofluorescence results and corresponding statistical analysis (Fig. S33–S36, Supporting Information) further confirmed this observation. The downregulation of GPX4 indicates successful activation of ferroptosis.

In addition to ROS and GPX4, mitochondrial dysfunction is also considered a crucial promoter of ferroptosis. To assess mitochondrial function, we used the JC-1 probe to detect mitochondrial membrane potential (MMP). Under normal MMP, JC-1 forms aggregates that emit red fluorescence, while MMP depolarization results in increased green fluorescence. As shown in Fig. 3H and K, cells in the PBS and MSPION groups exhibited strong red fluorescence, indicating intact mitochondrial function. In contrast, cells treated with CDDP, CDDP2@MSPION@GP3, or CDDP2@MSPION@GP3/FAPI showed a reduction in red fluorescence and an increase in green fluorescence in both MKN45 cells and CAFs, suggesting a significant drop in MMP. Corresponding statistical results (Fig. S37 and S38, Supporting Information) and flow cytometry analysis (Fig. 3I, J, L, and M) further validated these findings. Notably, the FAPI-modified group exhibited the most significant MMP reduction in CAFs, likely due to enhanced •OH generation, as hydroxyl radicals can directly disrupt the mitochondrial membrane, impair energy metabolism, and synergistically promote ferroptosis [69].

LPO accumulation is a direct indicator of ferroptosis. To evaluate LPO levels, we used the C11-BODIPY 581/591 probe [70], which shifts fluorescence from red to green in the presence of LPO, allowing visual assessment of membrane lipid oxidation. As shown in Fig. 3N and Q, treatment with CDDP, CDDP2@MSPION@GP3, and CDDP2@MSPION@GP3/FAPI led to a significant increase in green fluorescence in both MKN45 cells and CAFs, indicating elevated LPO levels. Quantitative analysis of the green-to-red fluorescence ratio revealed that LPO accumulation was more pronounced in CAFs treated with the FAPI-modified formulation, whereas no similar increase was observed in MKN45 cells, supporting the CAF-targeting specificity of FAPI (Figs. S39 and S40, Supporting Information). Flow cytometry analysis (Fig. 3O, P, R, and S) displayed consistent trends, indicating that the nanoprobe effectively induces ferroptosis by promoting LPO accumulation.

In conclusion, the CDDP2@MSPION@GP3/FAPI nanoplatform induces ferroptosis in MKN45 cells and CAFs by orchestrating multiple key ferroptosis-related events, including enhanced ROS generation, GPX4 downregulation, mitochondrial membrane depolarization, and LPO accumulation. Notably, in CAFs, the nanoplatform exhibited strong targeting specificity, robust oxidative stress induction, and pronounced lipid peroxidation, highlighting its potential to penetrate the tumor stromal barrier and achieve precise therapy. This study not only deepens our understanding of ferroptosis mechanisms in CAFs, but also provides a theoretical and experimental foundation for the translational application of novel nanomaterials in cancer therapy.

2.7. Crosstalk modulation between MKN45 cells and CAFs for metastasis inhibition

The crosstalk between CAFs and tumor cells plays a pivotal role in tumor progression by promoting the secretion of various pro-proliferative and pro-migratory cytokines, thereby enhancing tumor cell proliferation, migration, and invasion [11]. Based on this, we proposed a strategy utilizing FAPI-modified nanomaterials to selectively deplete CAFs and disrupt their interaction with tumor cells. This approach aims not only to directly eliminate CAFs but also to weaken their tumor-supportive functions, offering a potentially more effective anti-tumor strategy, particularly in inhibiting tumor cell dissemination.

To evaluate the efficacy of this strategy, we designed a series of experiments, as outlined in Fig. 4A. NFs and CAFs were cultured separately and exposed to the nanoplatform for 24 h. Following treatment, the drug-containing medium was replaced with fresh medium, and the cells were incubated for an additional 24 h to allow for sufficient secretion of soluble cytokines. The conditioned media (CM) from NFs (NFs-CM), untreated CAFs (CAFs-CM), and nanoplatform-treated CAFs (NP-CAFs-CM) were then collected and used to culture MKN45 cells. Subsequent functional assays included colony formation, wound healing, and ELISA.

Fig. 4.

Inhibition of metastasis by modulating the crosstalk between MKN45 cells and CAFs using CDDP2@MSPION@GP3/FAPI. (A) Schematic illustration of the experimental design to evaluate the anti-metastatic effect of CDDP2@MSPION@GP3/FAPI by assessing its impact on MKN45 cells under different CM. (B–C) Representative images of colony formation (B) and wound healing (C) assays of MKN45 cells treated with: PBS (I), NFs CM (II), CAFs CM (III) or CAFs CM after treatment with MSPION (IV), CDDP (V), CDDP2@MSPION@GP3 (VI), or CDDP2@MSPION@GP3/FAPI (VII). (D) Schematic illustration of the Transwell co-culture assay of MKN45 cells and CAFs. (E) Representative optical images of migration and invasion assays of MKN45 cells co-cultured with NFs (II) or CAFs (III), with the CAFs group treated with MSPION (IV), CDDP (V), CDDP2@MSPION@GP3 (VI), or CDDP2@MSPION@GP3/FAPI (VII). (F) Quantitative analysis of the colony formation assay. (G) Quantitative analysis of the wound healing assay. (H) Quantitative analysis of IL-6 levels by ELISA. (I) Quantitative analysis of the migration assay. (J) Quantitative analysis of the invasion assay. Scale bar: 200 μm ∗P < 0.05. ∗∗P < 0.01. ∗∗∗P < 0.001. ∗∗∗∗P < 0.0001.

As shown in Fig. 4B, the colony formation assay demonstrated that CAFs-CM significantly promoted colony formation in MKN45 cells, outperforming the effect of NFs-CM. Notably, this pro-proliferative effect was markedly reduced in the NP-CAFs-CM groups, with colony numbers significantly decreased to levels comparable to or even lower than those of the NFs-CM group (Fig. 4F). These results indicate that the nanoplatform effectively disrupts the tumor-promoting function of CAFs, thereby indirectly suppressing tumor cell proliferation.

To further assess the influence of CAF-derived factors on tumor cell migration, a wound healing assay was conducted. As shown in Fig. 4C and G, CAFs-CM markedly accelerated the migration of MKN45 cells, leading to faster wound closure. This indicates that CAF-secreted factors promote tumor cell motility, possibly through enhanced cell-matrix interactions or activation of chemokine-associated signaling pathways. However, NP-CAFs-CM failed to promote migration, as evidenced by significantly slower wound closure. These findings suggest that targeting CAFs not only causes direct cellular damage but also disrupts their migration-promoting effect on tumor cells.

To further elucidate the underlying mechanism, IL-6 levels were measured by ELISA, as IL-6 is a key cytokine closely associated with tumor proliferation and migration. As shown in Fig. 4H, IL-6 was significantly elevated in CAFs-CM compared to NFs-CM. However, its level was markedly reduced in the NP-CAFs-CM groups, particularly in the FAPI-modified formulation. These data are consistent with the results of the colony formation and wound healing assays, confirming that FAPI-modified nanomaterials suppress CAF-derived cytokines and their tumor-promoting effects.

To assess the impact of CAFs on MKN45 cell migration and invasion, Transwell assays were performed (Fig. 4D). As shown in Fig. 4E, co-culture with CAFs significantly enhanced the migratory and invasive abilities of MKN45 cells. Compared to the PBS and NFs groups, the number of cells that traversed the Transwell membrane was markedly increased, indicating that CAF-derived soluble factors promote tumor invasiveness. Quantitative analysis of the migration (Fig. 4I) and invasion (Fig. 4J) assays revealed that treatment of CAFs with the nanoplatform significantly reduced the number of migrating and invading MKN45 cells, indicating effective inhibition of CAF-mediated pro-invasive effects and partial reversal of the tumor-supportive microenvironment.

In summary, CAFs play a significant role in promoting tumor migration and invasion within the TME. The nanomaterials developed in this study effectively interfere with CAF-mediated tumor-promoting activities, offering a new theoretical support and strategic guidance for CAF-targeted therapies aimed at suppressing metastasis.

2.8. In vivo biosafety of CDDP2@MSPION@GP3/FAPI

To evaluate the biosafety and potential toxicity of the synthesized nanoparticles, we conducted a comprehensive assessment at the cellular level, within the circulatory system, and across major organs.

As shown in Fig. S41 (Supporting Information), the hemolysis assay demonstrated that the supernatants from CDDP2@MSPION@GP3/FAPI-treated groups at various concentrations remained clear and transparent, with hemolysis rates below 2 %, comparable to the PBS group (negative control) and significantly lower than the H₂O group (positive control). These results indicate negligible damage to red blood cells (RBC) and excellent hemocompatibility of the nanoparticles.

To further investigate in vivo toxicity, CDDP2@MSPION@GP3/FAPI was administered to mice via tail vein injection, and blood biochemistry and hematology analyses were performed on days 1, 7, and 14 post-injection. Serum biochemical parameters, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN), and creatinine (Cr), showed no significant differences between the treatment and control groups, suggesting no apparent hepatotoxicity or nephrotoxicity (Fig. S42, Supporting Information). Moreover, hematological indicators such as white blood cells (WBC), RBC, platelets (PLT), hemoglobin (HGB), mean corpuscular volume (MCV), and mean corpuscular hemoglobin (MCH) remained within normal ranges before and after injection (Fig. S43, Supporting Information), further supporting the blood safety of the nanoplatform.

Histopathological examinations also corroborated these findings. Hematoxylin and eosin (H&E) staining of major organs (heart, liver, spleen, lung, kidney, and brain) revealed no noticeable morphological abnormalities in the treated groups compared to controls (Fig. S44, Supporting Information), further confirming the excellent biocompatibility and in vivo safety of the CDDP2@MSPION@GP3/FAPI nanoplatform.

Importantly, compared with conventional CAF-targeting strategies such as FAP-CAR-T cell therapy, which can induce severe off-target toxicity by attacking FAP⁺ mesenchymal stem cells in the bone marrow and lead to cachexia and hematopoietic failure [71], our FAPI-engineered nanoplatform offers a significantly safer alternative. This advantage is supported by previous studies on radiolabeled FAPI probes that demonstrated low off-target uptake, minimal systemic toxicity, and excellent tolerability in clinical settings [[72], [73], [74]]. These findings collectively highlight the superior biosafety profile of the CDDP2@MSPION@GP3/FAPI system for potential clinical translation.

2.9. Pharmacokinetics, biodistribution, and in vivo T₁-weighted MRI in subcutaneous and peritoneal tumor-bearing mouse models

To systematically evaluate the pharmacokinetic behavior and biodistribution of CDDP2@MSPION@GP3/FAPI in vivo, we examined its blood circulation half-life and organ distribution in mice. As shown in Fig. S45 (Supporting Information), the concentration of Gd in blood gradually decreased over time, indicating good metabolic stability. The calculated half-life was approximately 7.95 h, suggesting prolonged systemic circulation, which is favorable for tumor accumulation via both FAPI-mediated active targeting and the enhanced permeability and retention (EPR) effect. Notably, the FAPI ligand employed in this study is a clinically validated small-molecule FAP inhibitor with nanomolar binding affinity, capable of rapid and selective accumulation in CAF-rich stromal regions [72]. Compared to antibody- or cell-based CAF-targeting strategies, which often suffer from poor tissue penetration or unintended off-target effects [71,75], the covalent conjugation of FAPI on the nanoparticle surface in our system enables a dual-targeting mechanism and multivalent binding interactions, thereby enhancing tumor accumulation and CAF specificity. These structural features, coupled with favorable pharmacokinetics, significantly improve the overall targeting efficiency of the theranostic nanoplatform.

Fig. S46 (Supporting Information) further illustrates the biodistribution of Gd in major organs at different time points (1, 6, 12, 24, and 48 h), as determined by inductively coupled plasma optical emission spectrometry (ICP-OES). The results show predominant accumulation in the liver and spleen, suggesting clearance primarily via the RES. In contrast, Gd concentration in tumor tissues gradually increased, peaking at 12 h post-injection, indicating efficient tumor-targeted accumulation and retention.

Building on its superior in vitro MRI performance and favorable in vivo pharmacokinetics, the MRI capability of CDDP2@MSPION@GP3/FAPI was further assessed for detecting subcutaneous tumors and PM nodules in vivo. Fig. 5A and B presents T₁-weighted MRI images of mice bearing subcutaneous and PM tumors, respectively, established via co-injecting MKN45 cells and CAFs. MRI scans were performed at multiple time points following intravenous administration of either the experimental nanoprobe or the commercial CA Magnevist at a gadolinium dose of 5.0 mg/kg.

Fig. 5.

A) Axial T₁-weighted MR images of subcutaneous tumor-bearing mice established with NKN45 cells and CAFs, before and after intravenous injection of CDDP2@MSPION@GP3/FAPI or Magnevist at different time points. B) Sagittal T₁-weighted MR images of peritoneal metastasis model mice established with NKN45 cells and CAFs, before and after intravenous injection of CDDP2@MSPION@GP3/FAPI or Magnevist at different time points. C–D) ΔSNR analysis was performed based on T₁-weighted MR images from the subcutaneous tumor model mice (C) or the peritoneal metastasis model mice (D). Gd dosage is 5.0 mg/kg ∗P < 0.05. ∗∗∗∗P < 0.0001.

In the Magnevist group, the T₁ signal in tumor regions rose rapidly, peaking at 30 min post-injection before sharply declining, indicating a short imaging window. In contrast, the CDDP2@MSPION@GP3/FAPI group showed an initial drop within 1 h, followed by a gradual increase peaking at 12 h and persisting for 24 h. This dark-to-bright shift, referred to as the light-up effect, is attributed to the accumulation of the probe in CAF-rich regions of the TME, where acidic conditions trigger MSPION degradation and release of the T₁ CA GP. Although MSPION degradation can also release Fe²⁺/Fe³⁺ ions, previous studies have shown that their intrinsic T₁ effect is weak [76,77], and noticeable enhancement generally requires chelation with paramagnetic ions such as gadolinium or manganese [78,79], or fabrication into ultrasmall iron oxide nanoparticles [80]. Therefore, in this study, the marked T₁ signal enhancement is primarily attributed to GP release. Importantly, this T₂-to-T₁ signal conversion not only enhances imaging contrast but also functions as a non-invasive, real-time biomarker linking MRI to both nanoprobe activation and ferroptosis induction within the TME. The dynamic MRI signal change directly reflects the pH-triggered degradation of MSPION and the subsequent GP release, thereby capturing both the spatial distribution and activation status of the therapeutic payload. This capability enables precise localization of CAF-rich regions and longitudinal monitoring of treatment response, effectively integrating diagnostic and therapeutic functions into a cohesive, precision-guided theranostic strategy.

To quantitatively assess imaging performance, the signal-to-noise ratio change (ΔSNR), as previously reported in the literature, was used for analysis (Fig. 5C and D). In the subcutaneous tumor model, the maximum ΔSNR value of the CDDP2@MSPION@GP3/FAPI group was 166 ± 13 % at 12 h, significantly higher than that of the Magnevist group (103 ± 7 % at 30 min). A similar trend was observed in the PM model (118 ± 15 % vs. 95 ± 11 %).

Notably, the nanoprobe maintained a high signal intensity up to 24 h post-injection, demonstrating a prolonged imaging window and enhanced diagnostic capability. Although the overall signal in the PM model was slightly reduced due to the plasma–peritoneal barrier, the ΔSNR values in the CDDP2@MSPION@GP3/FAPI group remained markedly higher than those in the Magnevist group. These results further validate the robustness and adaptability of our imaging strategy within the complex TME.

This performance difference can be primarily attributed to structural and pharmacokinetic differences between the two CAs. Magnevist, as a small-molecule CA, relies on passive diffusion to enter tumor tissues; however, its penetration is hindered by the dense CAF population and high interstitial fluid pressure in the TME. Additionally, it is rapidly cleared by renal filtration, resulting in a short-lived imaging signal. In contrast, the CDDP2@MSPION@GP3/FAPI nanoplatform integrates FAPI-mediated active targeting, optimal nanoparticle size, and a pH-responsive structure, which collectively enhance tumor accumulation and imaging intensity while extending systemic circulation time.

In conclusion, the CDDP2@MSPION@GP3/FAPI nanoprobe effectively combined CAF-specific targeting with acid-activated T₂-to-T₁ signal conversion, resulting in excellent tumor accumulation and prolonged signal retention. This system offers strong technical support and promising translational potential for accurate tumor imaging and real-time therapeutic monitoring.

2.10. In vivo the CDDP2@MSPION@GP3/FAPI enhanced anti-tumor efficacy in both subcutaneous tumor and peritoneal implant mice models

The high abundance of CAFs in tumor tissues is widely recognized as a major barrier to drug penetration and a key contributor to reduced therapeutic efficacy. [13,14] To address this challenge, we designed a multifunctional nanoprobe, CDDP2@MSPION@GP3/FAPI, which combines ferroptosis induction with FAPI-mediated CAF targeting. This strategy aims to selectively deplete CAFs, thereby weakening their physical and biochemical barrier functions, enhancing drug delivery efficiency, and ultimately improving the overall therapeutic outcome.

Building upon the demonstrated in vitro antitumor potential of the nanoprobe, we further evaluated its therapeutic efficacy in vivo. MKN45 cells and CAFs were co-injected at a 1:1 ratio to establish two CAF-enriched tumor models: a subcutaneous tumor model and a PM model. Mice were randomly divided into five groups (n = 5) and treated with PBS (I), MSPION (II), CDDP (III), CDDP2@MSPION@GP3 (IV), or CDDP2@MSPION@GP3/FAPI (V), according to the treatment protocols shown in Fig. 6, Fig. 7A. During the treatment period, mouse body weight and tumor volume were recorded regularly. On day 14, mice were sacrificed, and tumor tissues were collected for volume and weight measurements to assess therapeutic outcomes.

Fig. 6.

Evaluation of the anti-tumor efficacy of CDDP2@MSPION@GP3/FAPI in the subcutaneous tumor model. A) Schematic illustration of the establishment and subsequent treatment of a GC subcutaneous tumor model constructed with MKN45 cells and CAFs. B) Relative body weight curves of tumor-bearing mice. C) Relative tumor volume. D) Photographs of tumors isolated from mice in different groups after 14 days of treatment. E) Average tumor weight. F) Representative microscope images of tumor tissue sections stained with H&E, GPX4, Ki-67, α-SMA and TUNEL. G–J) Quantitative analysis of the results shown in (F). Scale bar: 100 μm ∗P < 0.05. ∗∗P < 0.01. ∗∗∗P < 0.001. ∗∗∗∗P < 0.0001.

Fig. 7.

Evaluation of the anti-tumor efficacy of CDDP2@MSPION@GP3/FAPI in the PM model. A) Schematic illustration of the establishment and subsequent treatment of the PM model constructed with MKN45 cells and CAFs. B) Relative body weight curves of tumor-bearing mice. C) Photographs of peritoneal metastases in mice from different treatment groups. D) Number of peritoneal nodules. E) Average weight of peritoneal nodules. F) Representative microscope images of tumor tissue sections stained with H&E, GPX4, Ki-67, α-SMA and TUNEL. G–J) Quantitative analysis of the results shown in (F). Scale bar: 100 μm ∗P < 0.05. ∗∗P < 0.01. ∗∗∗P < 0.001. ∗∗∗∗P < 0.0001.

In the subcutaneous tumor model (Fig. 6C), treatment with CDDP, CDDP2@MSPION@GP3, and CDDP2@MSPION@GP3/FAPI significantly suppressed tumor growth. Among them, the FAPI-modified group showed the most pronounced anti-tumor effect, with the lowest tumor volume (Fig. 6D) and weight (Fig. 6E), both showing statistically significant differences. Notably, no significant changes in body weight were observed among the groups during the treatment period (Fig. 6B), suggesting good biocompatibility of the therapeutic regimens.

Similar results were observed in the PM model. As shown in Fig. 7C, numerous tumor nodules were found on the peritoneal surface in the PBS group, while the number (Fig. 7C and D) and weight (Fig. 7E) of nodules were markedly reduced in the CDDP, CDDP2@MSPION@GP3, and especially the CDDP2@MSPION@GP3/FAPI groups. Meanwhile, body weights remained stable across all groups (Fig. 7B), further supporting the good tolerability of the nanoplatform. These effects are largely attributed to the FAPI modification, which enabled selective accumulation in CAF-rich regions, disrupted tumor-stroma interactions, alleviated physical barriers, and enhanced intratumoral drug penetration. Notably, the nanoscale size of CDDP2@MSPION@GP3/FAPI, approximately 101.5 nm, allows it to efficiently penetrate the dense collagen-rich stromal matrix, an ability that is often lacking in cell-based therapies such as FAP-CAR-T cells. These cells typically reach diameters of 11–12 μm upon activation and exhibit increased rigidity and viscosity, which hinders migration through fibrotic tumor regions [71,81]. In contrast, FAPI ligands have been shown to rapidly accumulate in CAF-rich tumors with high tumor-to-background contrast in both preclinical and clinical studies, reflecting their favorable tissue penetration [72].

To further elucidate the underlying anti-tumor mechanisms, we performed histological and immunohistochemical analyses, including H&E, GPX4, Ki-67, α-SMA and TUNEL (Fig. 6, Fig. 7F). H&E and TUNEL staining revealed that CDDP2@MSPION@GP3/FAPI treatment induced significant apoptosis, as evidenced by increased TUNEL-positive signals, indicating DNA fragmentation. Ki-67 immunostaining revealed a substantial reduction in tumor cell proliferation. Furthermore, GPX4 staining showed a marked downregulation of its expression, implying disruption of iron homeostasis and LPO accumulation, consistent with ferroptosis induction. Importantly, α-SMA staining revealed a significant reduction in activated fibroblast abundance within the tumor stroma, directly supporting the concept of stromal barrier disruption and confirming the CAF-targeting capability of the nanoprobe. These observations were further supported by quantitative analyses (Fig. 6, Fig. 7G–J).

In summary, CDDP2@MSPION@GP3/FAPI demonstrated significant in vivo antitumor efficacy. Its therapeutic mechanism is primarily based on the synergistic effects of iron supplementation from MSPION, CDDP-induced cytotoxicity, and CAF-targeted delivery via FAPI. This combinatorial strategy effectively disrupted tumor stromal barriers, enhanced drug accumulation, and promoted tumor cell death. Collectively, these findings offer both theoretical rationale and experimental validation for the development of ferroptosis-based precision cancer therapies.

3. Conclusion

In this study, we successfully developed an intelligent nanoprobe, CDDP2@MSPION@GP3/FAPI, which integrates CAF-targeted delivery, dual-cell ferroptosis induction, and MRI functionality, and systematically evaluated its theranostic potential in a GC PM model. Single-cell transcriptomic analysis of GC patient samples revealed elevated GPX4 expression in CAFs, suggesting their intrinsic resistance to ferroptosis and providing a theoretical basis for CAF-targeted ferroptosis therapy. Building on this insight, the designed nanoplatform not only achieved precise accumulation in CAFs but also synergistically induced ferroptosis in both tumor cells and CAFs, thereby disrupting the tumor-supportive stroma and significantly enhancing therapeutic efficacy. Meanwhile, the system enables T₂-to-T₁ MRI signal conversion in response to acidic TME conditions, generating a characteristic light-up effect, which improves imaging contrast and extends the visualization window. This provides real-time imaging feedback during ferroptosis-based therapy and supports the integration of diagnosis and treatment. In vivo studies confirmed the nanoprobe's excellent biocompatibility, imaging performance, and therapeutic effect, laying a solid foundation for clinical translation. In summary, this work proposes a novel integrated strategy that combines CAF-specific targeting, dual-cell ferroptosis induction and MRI-guided imaging. It expands the application of ferroptosis within complex multicellular systems and offers a new precision therapeutic approach for GC PM. It also provides meaningful insights and theoretical support for the clinical translation of nanomedicine in integrated cancer theranostics.

4. Materials and methods

4.1. Materials and reagents

NH₄HCO₃, Ferric trichloride hexahydrate (FeCl₃·6H₂O), ethylene glycol (EG), hydrogen peroxide (H₂O₂, 30 %), CDDP and R6G were purchased from Sigma-Aldrich (USA). Phosphate-buffered saline (PBS), Dulbecco's Modified Eagle Medium (DMEM), and Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F-12) were purchased from Gibco Laboratories (NY, USA). Fetal bovine serum (FBS) and 0.25 % trypsin-EDTA were also acquired from Gibco. The human gastric cancer cell line MKN45 was purchased from Procell Life Science & Technology Co., Ltd. (Cat. No. CL-0292). 4′,6-diamidino-2-phenylindole (DAPI) was purchased from Beijing Solarbio Science & Technology (Beijing, China). TMB was bought from Shanghai Macklin Biochemical Co., Ltd. The BCA Protein Assay Kit (P0012), One-Step TUNEL Apoptosis Assay Kit (Green Fluorescence), 2′,7′-Dichlorofluorescin diacetate (DCFH-DA), BeyoClick™ EdU-488 proliferation kit, Calcein acetoxymethyl ester (Calcein-AM) and PI Assay Kit and 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide (JC-1) Assay Kit were purchased from Beyotime Biotechnology Co. (Shanghai, China). C11 BODIPY^581/591-C11 was purchased from Shanghai Mao Kang Biotechnology Co., Ltd. Cell Counting Kit-8 (CCK-8) was purchased from GLPBIO. Human IL-6 ELISA Kit was purchased from ELK Biotechnology. Anti-GAPDH antibody and Anti-FAP antibody were purchased from Abcam. GPX4 antibody was bought from Beijing Solarbio Science & Technology Company (China). Anti-α-smooth muscle actin antibody (anti-α-SMA antibody) was purchased from proteintech.

4.2. Isolation of primary fibroblasts and cells culture

We collected fibroblasts from the tumor tissue and normal tissue distant from the tumor to isolate CAFs and NFs from patients undergoing surgery at Nanfang Hospital, Southern Medical University. All procedures involving human samples were conducted in accordance with the ethical standards of the institutional review board and with informed consent obtained from all patients. Sampling and transportation were conducted under strict aseptic conditions, with samples kept on ice to minimize cellular contamination. Upon receiving the gastric cancer tumor specimens in the laboratory, all operations were conducted in a laminar flow hood or biosafety cabinet. First, the samples were rinsed three times with PBS, then cut into small blocks approximately 1 mm³ in size and transferred to a 5 mL centrifuge tube. An appropriate amount of collagenase and deoxyribonuclease (DNase I) was added, and the samples were digested at 37 °C for 2 h. The digested samples were filtered through a 40 μm cell strainer, and the filtrate was centrifuged (1000×g, 3 min) to collect fibroblasts. The fibroblast pellet was resuspended in an appropriate amount of cell culture medium and seeded into culture dishes or flasks. The fibroblast culture medium consisted of F12/DMEM supplemented with 10 % FBS. Besides, MKN45 cells were cultured in Dulbecco's modified Eagle medium DMEM supplemented with 10 % FBS and 1 % penicillin-streptomycin, maintained at 5 % CO2 and 37 °C incubator.

4.3. Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted using TRIzol reagent (Invitrogen, Waltham, MA, USA), and reverse transcription was performed with the cDNA Synthesis SuperMix Kit (Yeasen, China) to obtain complementary DNA (cDNA). qRT-PCR was then carried out using SYBR Green Master Mix (Yeasen, China) on the QuantStudio™ 5 system (Thermo Fisher Scientific, USA), following the manufacturer's instructions. Primer sequences used for qRT-PCR are listed in Table S1 (Supporting Information). GAPDH was used as the internal reference gene for normalization.

4.4. Western blot

The cells were lysed using RIPA buffer (Beyotime, China) supplemented with phosphatase and protease inhibitors for protein extraction. The protein concentration was determined using a BCA Protein Assay Kit (CWBIO, China). Lysates were mixed with 4 × loading buffer and boiled at 100 °C for 10 min. Equal amounts of protein were separated by SDS-PAGE (Beyotime, China) and transferred onto PVDF membranes (Millipore, USA) using a Bio-Rad electrophoresis system (Bio-Rad, USA). The membranes were blocked with 5 % bovine serum albumin (BSA) for 2 h at room temperature and incubated overnight at 4 °C with primary antibodies against GAPDH, α-SMA, FAP, and GPX4. After washing twice with 2 % PBST, the membranes were incubated with HRP-conjugated goat anti-mouse or anti-rabbit IgG (H&L) for 1 h at room temperature. Protein bands were then visualized using enhanced chemiluminescence (ECL) reagents and detected with an ECL imaging system.

4.5. Immunofuorescence assays

MKN45 cells, NFs, and CAFs were fixed with 4 % paraformaldehyde for 30 min, followed by three washes with PBST. The cells were then permeabilized with 0.1 % Triton X-100 for 15 min at room temperature. To block non-specific binding, the samples were incubated with goat serum at room temperature. Primary antibodies against α-SMA and FAP (1:200 dilution) were applied and incubated overnight at 4 °C. The following day, cells were incubated with dye-conjugated secondary antibodies (Cell Signaling Technology, USA; 1:400) for 1 h at room temperature. After additional PBST washes, the cells were counterstained and mounted with a DAPI antifade mounting medium (Solarbio, China), and fluorescence images were captured using a confocal microscope (Nikon, Japan). GPX4 expression in nanoparticle-treated cells was assessed using the same immunofluorescence protocol described above.

4.6. scRNA-seq analysis

To compare GPX4 expression levels between fibroblasts derived from gastric cancer tissues (CAFs) and those from normal gastric tissues (NFs), we analyzed scRNA-seq data from the GSE183904 dataset using the “Seurat” package (v5.0.1) in R (v4.3.2), which included four tumor samples and four normal samples. For quality control, the minimum number of cells sharing each gene/feature was set to 3. Cells were retained if they expressed between 500 and 6000 genes, and if their mitochondrial RNA content was less than 20 %. The filtered data were then normalized and scaled, and the top 3000 highly variable genes were selected for further analysis. Principal component analysis (PCA) was performed, followed by batch effect correction across samples using the RunHarmony function. Cell clustering was conducted using the FindNeighbors function with a resolution of 1.25, and the results were visualized using the UMAP algorithm. Marker genes for each cluster were identified using the FindAllMarkers function. Based on the marker gene definitions described by Kumar et al., [44] we annotated nine major cell types in the gastric tissue samples: LYMPHOID (T&NK), MACROPHAGE, PLASMA, FIBROBLAST, EPITHELIAL, ENDOTHELIAL, B-CELL, DENDRITIC, and MAST. Fibroblasts annotated as “FIBROBLAST” were extracted for further analysis to investigate the gene expression differences between CAFs and NFs. GPX4 expression levels in CAFs and NFs were visualized using box plots, and statistical significance was assessed using the Wilcoxon rank-sum test.

4.7. Synthesis of mesoporous superparamagnetic iron oxide nanoparticles (MSPION)

MSPION was synthesized via a bubble template-assisted method. The reaction system consisted of FeCl₃ and NH₄HCO₃ as precursors, and a mixed solvent of diglycol and ethylene glycol in a volume ratio of 0.61. The synthetic mechanism is outlined as follows: At 200 °C, NH₄HCO₃ decomposes, releasing NH₃·H₂O (alkaline) and CO₂ gas bubbles (Eq. (1)). The generated NH₃·H₂O reacts with Fe³⁺ to form Fe(OH)₃ colloids (Eq. (2)). Subsequently, Fe(OH)₃ further reacts with ethylene glycol (HOCH₂CH₂OH), yielding a mixed colloidal system containing Fe(OH)₃ and Fe(OH)₂ (Eq. (3)). After continuous heating at 200 °C for 12 h, Fe(OH)₃ and Fe(OH)₂ are converted into Fe₃O₄ nanoparticles (Eq. (4)), referred to as SPIONs. These SPIONs were then subjected to a second thermal treatment at 250 °C for 12 h, during which Ostwald ripening occurred, resulting in the formation of MSPIONs with a well-defined mesoporous architecture [28].

$${\text{NH}}_4{\text{HCO}}_3\to {\text{CO}}_2+{\text{NH}}_3·{\mathrm{H}}_2\mathrm{O}$$ (1)

$${\text{Fe}}^{3+}+3{\text{NH}}_3·{\mathrm{H}}_2\mathrm{O}\to \text{Fe}{\left(\text{OH}\right)}_3+3{\text{NH}}_4^{+}$$ (2)

$${\text{HOCH}}_2{\text{CH}}_2\text{OH}+10\text{Fe}{\left(\text{OH}\right)}_3\to 10\text{Fe}{\left(\text{OH}\right)}_2+2{\mathrm{H}}_2{\text{CO}}_3+6{\mathrm{H}}_2\mathrm{O}$$ (3)

$$2\text{Fe}{\left(\text{OH}\right)}_3+\text{Fe}{\left(\text{OH}\right)}_2\to {\text{Fe}}_3{\mathrm{O}}_4+4{\mathrm{H}}_2\mathrm{O}$$ (4)

4.8. R6G labeling of nanoparticles

4.0 mL of CDDP2@MSPION@GP3 or CDDP2@MSPION@GP3/FAPI (C_Fe = 5.0 mM) were mixed with 1 mL of R6G (10 μM) under magnetic stirring at room temperature. After 24 h, the CDDP2@MSPION@GP3@R6G or CDDP2@MSPION@GP3/FAPI@R6G was obtained by centrifugation (15000×g, 10 min) and washing with pure water for further use.

4.9. Characterizations

TEM was used to observe the morphology and structure of the nanoparticles. The crystal structures of Nanoparticles were characterized by XRD. XPS was utilized to determine the valence states of elements for the nanoparticles. Malvern Zetasizer Nano ZS90 was applied to detect the zeta potential and dynamic light scattering of various samples. UV–vis absorbance spectra of the samples were determined by a UV-1601 spectrophotometer. The pore size distribution of the samples was determined using the Brunauer-Emmet-Teller (BET) method. Elemental concentrations of the nanoparticles were determined by ICP-OES.

4.10. Fenton reaction monitoring

A TMB colorimetric method is used to monitor the Fenton reaction. Briefly, CDDP2@MSPION@GP3/FAPI (2.0 mg/mL) was incubated with or without TMB (2 mg/mL) or H₂O₂ (4 mM) at pH 5.5, 6.5, or 7.4 for 90 min, respectively. Besides, various concentrations of CDDP2@MSPION@GP3/FAPI (25–1.56 mg/mL) were mixed with H₂O₂ (4 mM) and TMB (2 mg/mL) at pH = 5.5 for 90 min. After that, UV–vis spectroscopy was used to monitor the Fenton reaction at different conditions. To eliminate interference from the intrinsic UV–vis absorption of the CDDP2@MSPION@GP3/FAPI solution across the spectrum, the nanoparticles were removed by centrifugation under identical conditions prior to measuring the absorbance peak of the TMB solution.

4.11. In vitro MRI performance

The MRI contrast potential of CDDP2@MSPION@GP3/FAPI was explored on MRI scanner systems (3.0 T, Philips, Ingenia, NL). CDDP2@MSPION@GP3/FAPI was dispersed in PBS solutions (pH 7.4, 6.5, or 5.5) with different Gd concentrations (100, 50.0, 25.0, 12.5, 6.25 or 3.125 μM). The dispersions were incubated for 24 h at 37 °C, and then were placed in Eppendorf tubes for MRI test. The relaxivity value of r₁ or r₂ was obtained from the slope of the linear curve of relaxation rate (1/T₁ or 1/T₂, s⁻¹) versus Gd concentration (mM). For T₁ relaxation rates at magnetic field of 3.0 T: TR is 200 ms, TE is 8.2 ms. For T₂ relaxation rates at magnetic field of 3.0 T: TR is 5000 ms, and TE is 80 ms.

4.12. Cellular uptake, cytotoxicity, and proliferation assays

Nanoparticles effects on the proliferation and viability of MKN45 cells, NFs and CAFs were examined in the CCK-8, Live/Dead, 7-AAD, and EdU assays. For the CCK-8 assay, MKN45 cells, NFs and CAFs were plated in 96-well plates (5 × 10³ cells per well) and incubated in DMEM or DMEM/F12 medium containing 10 % FBS and 1 % antibiotics at 37 °C overnight. Subsequently, the cells were incubated with MSPION, CDDP, CDDP2@MSPION@GP3, CDDP2@MSPION@GP3/FAPI at different concentration gradients. The cell viability was assessed with the CCK-8 reagent after 24 h.

In addition, to evaluate the cellular uptake capacity, MKN45 cells, NFs and CAFs were seeded and treated with CDDP2@MSPION@GP3@R6G (C_CDDP = 10 μg/mL) or CDDP2@MSPION@GP3/FAPI@R6G (C_CDDP = 10 μg/mL) for 12 h, respectively. Then, the cells were imaged and photographed using confocal laser scanning microscopy (CLSM, Olympus) after staining with DAPI for 10 min, followed by statistical analysis with ImageJ. To further assess the cellular uptake capacity, collect the cells and perform flow cytometry to quantify cellular uptake efficiency and analyzed with Flowjo.

MKN45 cells and CAFs were treated with PBS, MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) and CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL). for 24 h, The fluorescence imaging was recorded after treated with live/dead staining assay and the percentage of dead cells were quantified by using ImageJ software. For the cell proliferation test, various treated cells were incubated with a nucleoside analog of thymidine, EdU for 6 h, then, followed by fixation and staining cells with the use of the Click-iT EdU Assay kit. After staining the nuclei with DAPI, the images and percentage of EdU-positive cells were obtained by the same methods as above. To further assess the degree of nanoparticles induced proliferation and apoptosis, MKN45 cells and CAFs were seeded and treated with MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) and CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL), respectively. Stain the cells with EDU and 7-AAD. Then, the proliferation and apoptosis rate of the treated cells was detected by flow cytometry using CytoFLEX (Beckman, USA) and analyzed with Flowjo.

4.13. Evaluation of intracellular ROS

ROS probe DCFH-DA was utilized to evaluate intracellular ROS levels. For CLSM analysis, MKN45 cells and CAFs were first seeded into 24-well plates (5.0 × 105 cells/well) and incubated overnight to achieve adherence. Next, the growth media were replaced with fresh one without or with MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) or CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL). After 12.0 h of incubation, the cells were washed with PBS, and incubated in the culture media containing DCFH-DA (10 μM). After further incubation for 30 min, the cells were washed with PBS and observed by CLSM. For flow cytometry analysis, MKN45 cells were first seeded into 6-well plates (2.5 × 10⁶ cells/well) and incubated overnight to achieve adherence. Then, the culture medium was replaced with fresh one without or with MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) or CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL). After 12.0 h of incubation, the cells were washed with PBS, and incubated in the culture media containing DCFH-DA (10 μM). After further incubation for 30 min, the cells were washed twice with cold PBS and harvested by trypsinization and centrifugation (1000×g, 5.0 min). The obtained cells were finally analyzed using flow cytometry.

4.14. Detection of mitochondrial membrane potential

The mitochondrial membrane potential was detected using JC-1 probe. Briefly, MKN45 cells and CAFs were seeded into 24-well plates (2.5 × 10⁵ cells/well) to achieve adherence. The culture medium was replaced with fresh one without or with MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) or CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL). After 24 h of incubation, the cells were washed with PBS, and stained with the JC-1 (10 μg/mL) probe for 30 min. Then cells were then rinsed thrice with PBS, and observed by CLSM. A green channel of excitation wavelength was utilized for JC-1 monomers, while a red channel of excitation wavelength was utilized for JC-1 aggregates. For flow cytometry analysis, MKN45 cells were first seeded into 6-well plates (2.5 × 10⁶ cells/well), and incubated overnight to achieve adherence. Then, the culture media were replaced with fresh one without or with MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) or CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL). After 24 h of incubation, the cells were then stained with JC-1 (10 μg/mL) for 30 min. The cells were washed twice with PBS, and harvested by trypsinization and centrifugation (1000×g, 5.0 min). The obtained cells were finally re-suspended in PBS, and analyzed using flow cytometry.

4.15. Evaluation of intracellular LPO level

For CLSM analysis, MKN45 cells and CAFs were seeded into 24-well plates (5.0 × 10⁵ cells/well) and incubated overnight to achieve adherence. Next, the culture media were replaced with fresh one without or with MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) or CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL) for 24 h. After that, cells were washed with PBS for twice times gently and mixed with BODIPY-C11 581/591 (10 μM) for 30 min at 37 °C. The cells were then fixed with 4.0 % paraformaldehyde for 10 min, permeabilized with 0.10 % Triton X-100 for 5.0 min and stained with DAPI for 10 min. Finally, the cells were observed by CLSM (Olympus). For flow cytometry analysis, MKN45 cells were first seeded into 6-well plates (2.5 × 10⁶ cells/well), and incubated overnight to achieve adherence. Then, the culture media were replaced with fresh one without or with MSPION (equivalent concentration of 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL) or CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL). After 24 h of incubation, the cells were then stained with BODIPY-C11 581/591 (10 μM) for 30 min. The cells were washed twice with PBS, and harvested by trypsinization and centrifugation (1000×g, 5.0 min). The obtained cells were finally re-suspended in PBS, and analyzed using flow cytometry.

4.16. IL-6 detection by ELISA

The NFs-CM, CAFs-CM and NP-CAFs-CM were collected for IL-6 detection. CAFs were treated for 24 h with MSPION (equivalent to 89.8 μg/mL), CDDP (C_CDDP = 10 μg/mL), CDDP2@MSPION@GP3 (C_CDDP = 10 μg/mL), or CDDP2@MSPION@GP3/FAPI (C_CDDP = 10 μg/mL). Following treatment, the culture supernatants were collected and centrifuged to remove cellular debris. The concentration of IL-6 in the resulting CM was quantified using an ELISA kit according to the manufacturer's instructions.

4.17. Colony formation assay

The NFs-CM, CAFs-CM and NP-CAFs-CM were collected and subsequently used to culture MKN45 cells. MKN45 cells were suspended in the respective CM and seeded into 6-well plates at a density of 1000 cells per well. After 24 h, the CM was replaced with fresh complete medium. Following a 14-day incubation, the cells were fixed with 4 % paraformaldehyde for 15 min and stained with crystal violet for 30 min. Colony formation was then photographed and documented.

4.18. Wound healing

The NFs-CM, CAFs-CM and NP-CAFs-CM were collected and used to culture MKN45 cells. MKN45 cells seeded in 6-well plates were scratched using sterile pipette tips to create uniform wounds. The cells were then incubated with the respective CM for 24 h, after which the media were replaced with fresh complete medium and cultured for an additional 24 h.

4.19. Transwell migration and invasion assays

The NFs-CM, CAFs-CM and NP-CAFs-CM were collected and used to culture MKN45 cells. For the migration assay, MKN45 cells were seeded into the upper chambers of Transwell inserts, which were then placed into 24-well plates pre-seeded with fibroblasts 24 h in advance. After incubation with nanoparticles for 24 h, the cells that had migrated through the membrane were fixed with 4 % paraformaldehyde and stained with crystal violet. Migrated cells were imaged under a microscope and quantified using ImageJ software. For the invasion assay, Matrigel was applied to the inner surface of the upper chamber 4 h before seeding MKN45 cells. The subsequent steps were the same as those used in the migration assay.

4.20. Hemolytic assay of CDDP2@MSPION@GP3/FAPI nanoparticles

To evaluate the blood compatibility, the blood samples were collected from healthy experimental animals, centrifuged, washed with PBS three times and gained the RBC pellet. Then, an equal volume of RBC suspensions was dissolved in the nanoparticles with different concentration gradients. In addition, diluted RBC suspensions were also added to PBS and ddH₂O which were respectively served as negative control and positive control. The mixture was incubated at 37 °C for 4 h, after that, the erythrocyte suspension was centrifuged at rotated speed of 400×g for 10 min, the absorbance at 540 nm was measured. Subsequently the absorbance of the supernatant was measured, and the hemolysis ratio was calculated. The percentage of hemolysis rate was determined as follows:

$$\text{Hemolysis}\text{rate}(\%)=({\mathrm{A}}_{\mathrm{e}}-{\mathrm{A}}_{\mathrm{n}})/({\mathrm{A}}_{\mathrm{p}}-{\mathrm{A}}_{\mathrm{n}})\times 100\%$$

where A_e, A_n and A_p were the absorbance values of experimental groups, negative control group and positive control group, respectively.

4.21. In vivo systemic biosafety test

To evaluate the biosafety of the synthesized CDDP2@MSPION@GP3/FAPI nanoplatform, blood samples were collected from mice (CDDP dose: 5 mg/kg, n = 3) on days 1, 7, and 14 post-injection. Blood was collected into heparin sodium anticoagulant tubes for routine hematological analysis, while serum was obtained by collecting blood into anticoagulant-free EP tubes, followed by centrifugation at 3000 rpm for 15 min to isolate the supernatant for biochemical testing.

Hematological parameters included WBC, RBC, PLT, HGB, MCV and MCH. Serum biochemical indicators included liver function markers (ALT and AST) and kidney function markers (BUN and Cr).

For further biological safety evaluations, the healthy mice were intravenously injected with 0.10 mL of CDDP2@MSPION@GP3/FAPI with 5 mg/kg of CDDP (n = 3). After 14 days, the mice were sacrificed, and major organs (heart, liver, spleen, lung, kidney and brain) were collected. Then, the H&E staining method was used to observe and analyze the sections of major organs.

4.22. Subeutaneous and intraperitoneal tumor models

To establish a stroma-rich gastric cancer mouse model, MKN45 cells and CAFs were mixed at a 1:1 ratio and injected into 6–8-week-old nude mice. The subcutaneous MKN45 tumor models were established by injecting 5 × 10⁶ MKN45 cells and 5 × 10⁶ CAFs into the dorsal right flank of mice. When the tumors reach a volume of 100 mm³, they will be used for subsequent experiments. For the establishment of a peritoneal gastric cancer xenograft model, the mice were injected with 5 × 10⁶ MKN45 cells and 5 × 10⁶ CAFs suspended in sterile PBS directly to the peritoneal cavity. To ensure the mice were not suffering from the later stages of tumor growth, the mice were taken for the next experiment when the tumor volume reached 100 mm³ or had enlarged abdomens due to the production of ascites. The general condition of each mouse was monitored every 2 days until noticeable ascites or nodules were detected, which appeared in approximately 2 weeks.

All animal experiments were performed in compliance with institutional and national guidelines. The protocol was reviewed and approved by the Animal Ethics Committee of Nanfang Hospital, Southern Medical University, under approval number IACUC-LAC-20240509-007.

4.23. In vivo pharmacokinetics

For the pharmacokinetic analysis, subcutaneous and intraperitoneal tumor models were used. Before injection of CDDP2@MSPION@GP3/FAPI nanoparticles, 0.1 mL of blood (as a control to eliminate the signal of gadolinium ions inside blood) was collected from each mouse, and an equal volume of saline solution was immediately injected via the tail vein. Subsequently, CDDP2@MSPION@GP3/FAPI (Gd dosage of 5.0 mg/kg) were intravenously injected via the tail vein. Following injection, venous blood samples (0.1 mL each time) were collected from the orbital region at different time points, and an equal volume of saline solution was immediately supplemented via the tail vein after each blood drawing. Finally, the samples were analyzed for Gd concentration using ICP-OES.

4.24. In vivo biodistribution

When the tumor volume of subcutaneous tumor-bearing mice grew up to 100–150 mm³ or intraperitoneal tumor mice had enlarged abdomens due to the production of ascites, the mice were randomly divided into five groups (n = 3 per group) and then intravenously injected with 0.10 mL of CDDP2@MSPION@GP3/FAPI at Gd dosage of 5.0 mg/kg. At 1.0, 6.0, 12, 24, or 48 h post-injection, the mice were sacrificed. Subsequently, their major organs (heart, liver, spleen, lung, and kidney), tumors and peritoneal nodules were collected and completely digested with concentrated nitric acid. Gd contents were determined by ICP-OES, and expressed as a percentage of the injected Gd dosage per gram of tissue.

4.25. In vivo MRI performance

The MRI measurements were performed with a clinical MRI scanner (3.0 T, Philips, Ingenia, NL). To evaluate the in vivo MRI imaging efficiency of our prepared CDDP2@MSPION@GP3/FAPI nanoparticles, subcutaneous and intraperitoneal tumor models were established in nude mice. The CA CDDP2@MSPION@GP3/FAPI or the commercial CA Magnevist (C_Gd = 5.0 mg/kg) was administered via tail vein injection. T₁-weighted images acquired before and at various time points after injection were analyzed by quantifying signal intensity using MicroDicom software. The signal-to-noise ratio (SNR) and ΔSNR values were calculated according to equations (1), (2).

$$\text{SNR}={\text{SI}}_{\text{mean}}/{\text{SD}}_{\text{noise}}$$ (1a)

$$\Delta \text{SNR}=({\text{SNR}}_{\text{post}}-{\text{SNR}}_{\text{pre}})/{\text{SNR}}_{\text{pre}}\times 100\%$$ (2a)

4.26. Tumor therapy performance

To evaluate the in vivo efficacy of CDDP2@MSPION@GP3/FAPI-mediated ferroptosis therapy, both subcutaneous and PM tumor-bearing mice were randomly divided into five groups (n = 5): PBS, MSPION (equivalent dose of 44.9 mg/kg), CDDP (5.0 mg/kg), CDDP2@MSPION@GP3 (CDDP dose of 5.0 mg/kg), and CDDP2@MSPION@GP3/FAPI (CDDP dose of 5.0 mg/kg). Each group received intravenous injections of 100 μL of the assigned formulation. In the subcutaneous tumor model, treatments were administered on days 0, 3, and 6; in the peritoneal metastasis model, injections were given on days 0, 2, 4, and 6. Body weights and tumor volumes were recorded every other day throughout the treatment period (with tumor volumes kept below 2000 mm³). Tumor volumes were measured using vernier calipers and calculated using the formula: Volume (mm³) = ½ × width² (mm²) × length (mm), where length is the longest dimension and width is the perpendicular diameter.

In a parallel study, mice were sacrificed on day 14 post-treatment. Tumor weights and the number of metastatic nodules were recorded, and tumor tissues were collected for histological analysis. Excised tumors were fixed in 4 % paraformaldehyde, embedded in paraffin, sectioned, and subjected to H&E staining. Additional sections were analyzed using the TUNEL apoptosis assay and immunohistochemical staining for GPX4, Ki-67 and α-SMA. All sections were imaged using CLSM.

4.27. Statistical analysis

All data are expressed as mean ± standard deviation (SD), with measurements obtained from independent samples. Statistical comparisons between two groups were performed using the Student's t-test, while one-way analysis of variance (ANOVA) was applied for comparisons among multiple groups. A P-value less than 0.05 was considered statistically significant and denoted as follows: ns, not significant (P ≥ 0.05), ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, or ∗∗∗∗P < 0.0001. All statistical analyses were performed using GraphPad Prism (GraphPad Software, USA) and Origin (OriginLab Corporation, USA).

CRediT authorship contribution statement

Cuiyin Zhao: Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Yingxin Ren: Writing – original draft, Validation, Methodology, Investigation, Formal analysis, Conceptualization. Jing Yang: Validation, Resources, Methodology, Formal analysis. Jiaqiang Jiang: Validation, Methodology, Investigation. Yudie Lu: Resources, Methodology. Xiang Yu: Resources, Methodology, Investigation. Jie Feng: Resources, Methodology, Formal analysis. Zheyu Shen: Writing – review & editing, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization. Yanfeng Hu: Writing – review & editing, Validation, Supervision, Resources, Project administration, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.