A magnetic resonance imaging-guided drug delivery system for premetastatic niche theranostics and colorectal cancer liver metastasis intervention Authors: Wangping Luo, Xiaoyan Yang, Yining Wang, Jianxia Xu, Yijie Lou, Jiayi Chen, Jue Hou, Xiaojie Wang, Yuanfei Lu, Jin Mao, Guping Tang, Hongzhen Bai, Risheng Yu Published: 2026 ABSTRACT Colorectal cancer liver metastasis (CRLM) poses a major challenge in cancer theranostics, in which hepatic stellate cell (HSC) activation participates in fibrosis progression and colorectal cancer (CRC) cell recruitment, catalyzing the formation of fibrosis-concomitant premetastatic niche (PMN). Targeting the crosstalk between activated HSCs (aHSCs) and CRC cells, we develop a bioengineered, mesoporous superparamagnetic iron oxide nanoparticle (mSPIONP)-based theranostic nano-system (mSPIONP-Cur@eCM). Camouflaged with CXC chemokine receptor type 4 (CXCR4)-overexpressing CRC cytomembranes, this nano-system exhibits a dual PMN-targeting function, exerting a homologous recognition to CRC cells and anchors to aHSCs with a high secretion of stromal cell-derived factor-1 (SDF-1). Its mSPIONP-Cur core affords not only T2-weighted magnetic resonance imaging (MRI) but also curcumin (Cur) delivery, initiating liver fibrosis detection and metastasis foci delineation, simultaneously blocking the HSC-CRC cell crosstalk by amplifying Cur’s therapeutic effect. Verified in fibrosis and CRLM models, mSPIONP-Cur@eCM exhibits great potential for suppressing the fibrosis-concomitant PMN and intervening the CRLM progression.Graphical abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12951-026-04725-3. Introduction Metastasis is regarded as the ultimate manifestation of cancer, in which liver metastasis is detected in a wide range of malignancies, particularly gastrointestinal ones, such as colorectal cancer (CRC) [1]. For the cases of CRC liver metastasis (CRLM), radical resection and chemotherapy remain the standard treatment modalities in clinic [2]. However, only 10%-20% of patients are eligible for liver resection with a 5-year survival rate of approximatively 30%, as there is a dilemma that resections are appropriate in early CRLM settings with a small quantity of lesions, but the therapy window is usually narrowed by the limited accuracy of early CRLM diagnosis [3, 4]. To this end, most patients with unresectable metastases receive systemic therapies including chemodrugs, molecular targeted therapeutics and immunotherapeutic agents, but the prognosis can be even worse due to poor drug bioavailability and acquired drug resistances [5]. In this scenario, early detection of liver metastasis combining with tailored implementation of therapeutic intervention is still a formidable challenge in the field of CRLM management. Recent understandings of key cellular components in the formation of pre-metastatic niche (PMN) provide mechanistic pathways to address this challenge [6, 7]. Hepatic stellate cells (HSCs), as a kind of non-parenchymal liver pericytes, play important roles in liver homeostasis and CRLM, owing to their phenotypic plasticity [8]. Once reprogrammed by disseminated CRC cells, HSCs undergo continuous activation, contributing to PMN by depositing extracellular matrix (ECM), recruiting CRC cells and mediating immunosuppressive orientation [9, 10]. On the other hand, activated HSCs (aHSCs) also potentiate fibrocellular signatures through the autocrine secretion of fibronectin and profibrogenic signals, triggering hepatic fibrosis [11–13]. This pathological condition not only harbors cirrhosis risk, but also promotes metastatic organotropism, simultaneously posing biological barriers to the infiltration of therapeutic and diagnostic agents [10, 14–17]. Our follow-up magnetic resonance imaging (MRI) examination and clinical diagnosis data for 104 patients receiving colorectal tumor resection further evidence the link between fibrosis and metastasis, indicating that the metastasis could be catalyzed by hepatic parenchymal fibrosis, meanwhile it exacerbates the fibrosis condition (Fig. 1A, S1). In this scenario, aHSCs have been identified to comprise different subtypes with collaborative contributions to liver fibrosis and metastasis. Specifically, HSCs can be partially transdifferentiated into cancer-associated fibroblasts (CAFs) with fibroblast activation protein (FAP) marker (Fig. S2), producing profibrogenic factors such as transforming growth factor-β (TGF-β) and connective tissue growth factor (CTGF) [8, 18–20]. Meanwhile, aHSC phenotype characterized by α-smooth muscle actin (α-SMA) expression serves as a major source of stromal cell-derived factor-1 (SDF-1), augmenting the colonization of CXC chemokine receptor type 4 (CXCR4)-overexpressed CRC cells via SDF-1/CXCR4 signaling axis. Reprogrammed by colonized CRC cells, HSC activation is further aggravated, initiating a feedback loop to foster the PMN [10, 21–24]. The central role of aHSCs in linking fibrosis and PMN is also confirmed by our transcriptome analysis (Fig. 1B, S3). Differential gene expression profiles of FAP, matrix metalloproteinases (MMPs) and collagens highlight aHSCs engaging pathways regulating fibrosis [18, 25, 26], while upregulation of PI3K/Akt, JAK/STAT and SDF-1/CXCR4 signaling axis-involving genes, such as S100A4, STAT3, DPP4, PDGFR, TWIST, evidences aHSCs promoting metastasis (Fig. 1C) [27–31]. As confirmed in the immunofluorescence analysis in a preclinical hepatic fibrosis model, the fibrosis progression potentiates HSC activation, augmenting the expression levels of α-SMA and SDF-1 in the fibrotic septa and their peripheries (Fig. 1D). With this expression profile, aHSCs would trigger the colonization of CRC cells in the fibrotic regions, and the interaction between aHSCs and CRC cells consequently accelerates the formation CRC-receptive, fibrosis-concomitant PMN. Therefore, the reciprocal cell-cell crosstalk between CRC cells and aHSCs can be leveraged as a potential target for PMN inhibition, which opens up exciting opportunities fighting against CRLM. Except for therapeutic intervention, imaging-guided malignancy characterization also has substantial implications for CRLM management, including disease staging, therapy implementing and outcome monitoring. Superparamagnetic iron oxide nanoparticles (SPIONPs), as contrast agents in MRI, have optimized the diagnostic accuracy of MRI techniques for CRLM detection, holding potentials for the fabrication of targeting theranostics systems [32]. To integrate imaging functionality with drug loading capacity for SPIONPs, mesoporous structuralization has been developed, facilitating MRI-guided drug, gene, photosensitizer delivery, or magnetic hyperthermia [33–37]. Although SPIONP-based ablation therapies have been investigated for unresectable CRLM settings [38, 39], drug interventions should be an optimal therapeutic approach for the PMN prevention and the choice of medication requires caution as the PMN formation is a very early stage of CRLM. Curcumin (Cur), sourced from Curcuma longa, has emerged as a potential therapeutic agent in cancer treatment due to its anti-inflammatory, antioxidant, and antitumor effects [40–43], which has recently been demonstrated to inhibit the activation and proliferation of HSCs via metabolism regulation [44–46]. Our molecular docking simulation confirms that curcumin serves as an antagonist for some key proteins involved in HSC signaling, potentially suppressing HSC activation (Fig. S4). These features render curcumin as a promising candidate for fibrosis treatment and PMN inhibition. Regarding to targeting delivery, strategies combining polymer coating with targeting molecule conjugation have been introduced in SPIONPs according to expression profiles of different cancer cells [47, 48], whereas this intention may be insulted by the spatiotemporal heterogeneity of target expression during PMN progression [49, 50]. Considering the mainstay of HSCs in the recruitment of CRC cells, CRC-cell membrane camouflage can be a biomimetic targeting strategy for drug delivery, which leverages the HSC-CRC cell crosstalk to improve the PMN-specificity for the early CRLM theranostics. Herein, we develop a bioengineered, SPIONP-based theranostics nano-system for the detection and inhibition of PMN to intervene CRLM progression. MRI-guided drug delivery core is fabricated with mesoporous SPIONPs (mSPIONPs) and Cur, in which mSPIONPs are synthesized via one-step hydrothermal method, then loaded with Cur with drug loading rate of 17.1% to 18.7%. The theranostics core is camouflaged with bioengineered CRC-cell membranes (eCMs) for a dual targeting function to the PMN. The eCMs were isolated from transfected CRC cells with augmented levels of CXCR4 expression, which not only mimics the CRC colonization to access aHSCs, but also exerts a homologous recognition to CRC cells. Through integrated construction, the core-shell nano-system (mSPIONP-Cur@eCM) affords prolonged systemic circulation and enhanced PMN-specificity, initiating precise MRI detection and Cur delivery for the theranostics of PMN. Mechanistically, mSPIONP-Cur@eCM exerts the dual targeting function to visualize the fibrosis-concomitant PMN and CRLM progression. Releasing Cur, it suppresses aHSC-derived signaling pathways and induces apoptosis upon aHSCs and CRC cells, which blocks the HSC-CRC cell crosstalk within the PMN, thereby intervening CRLM progression (Fig. 1E). Fig. 1Verification data at the human, animal, cellular levels and schematic diagrams of mSPIONP-Cur@eCM from synthesis to biomedical applications. (A) Longitudinal MRI and histopathological conditions in three patients with CRLM, illustrating the association between metastasis and liver fibrosis. This figure presents multimodal examination data from three CRC patients (Patient #1, #2, and #3) with curative resection. For each patient, imaging and pathological conditions are shown at baseline (following primary tumor resection) and at follow-up, including MRI, histopathological sections with hematoxylin and eosin (HE) staining (original magnification, ×2) and c-MET staining (a non-specific metastasis marker, original magnification, ×2). Patient #1 (79-year-old male): At baseline, post-operative MRI showed a normal liver morphology (outlined by a solid orange line). A liver biopsy obtained during the primary resection confirmed normal hepatic parenchyma. In follow-up data, MRI revealed the emergence of three new liver metastasis (indicated by yellow arrowheads). The HE staining further confirmed the metastatic lesions (highlighted by an orange star) surrounding by fibrotic stroma (indicated by purple triangles). c-MET immunohistochemistry demonstrated a strong positivity within the metastatic foci and the periphery of fibrotic stroma. Patient #2 (60-year-old male): At baseline, MRI revealed features suggestive of hepatic fibrosis, including widened fissures and an irregular liver contour (indicated by white triangles). A concurrent liver biopsy confirmed existing fibrotic bands disrupting the normal lobular architecture. Follow-up data indicated that the patient developed rapid and extensive CRLM. MRI displayed aggressive disease progression with significant mass effect and architectural distortion. HE staining of the resected specimen revealed abundant fibrotic stroma surrounding the metastatic deposits. c-MET expression was confirmed in both the metastasis foci and the adjacent fibrous regions. Patient #3 (82-year-old male): At baseline, both MRI and HE staining confirmed a solitary liver metastasis without evidence of peritumoral fibrosis. By contrast, histopathological analysis displayed obvious fibrotic features in metastatic lesions with c-MET deposition, suggesting the concomitance of metastasis and fibrotic stroma. (B) GSEA displaying the differences in gene sets related to liver fibrosis and tumor formation between activated (in co-incubation with CRC cell supernatant) and quiescent LX-2 groups. (C) Volcano plot visualization of the differentially-expressed genes between activated and quiescent LX-2 groups. The top 5 upregulated and downregulated genes, as well as multiple genes related to the SDF-1/CXCR4 axis, are highlighted. (D) Representative immunofluorescence staining and quantitative analysis of SDF-1 in fibrotic mouse livers at different time points, along with images from the metastasis model showing SDF-1 deposition around metastatic and fibrotic areas. The white dotted line shows the liver metastatic lesions. (E) Schematic diagram of the synthesis of mSPIONP-Cur@eCM and its targeted delivery to the PMN for exerting diagnostic and therapeutic functions Verification data at the human, animal, cellular levels and schematic diagrams of mSPIONP-Cur@eCM from synthesis to biomedical applications. (A) Longitudinal MRI and histopathological conditions in three patients with CRLM, illustrating the association between metastasis and liver fibrosis. This figure presents multimodal examination data from three CRC patients (Patient #1, #2, and #3) with curative resection. For each patient, imaging and pathological conditions are shown at baseline (following primary tumor resection) and at follow-up, including MRI, histopathological sections with hematoxylin and eosin (HE) staining (original magnification, ×2) and c-MET staining (a non-specific metastasis marker, original magnification, ×2). Patient #1 (79-year-old male): At baseline, post-operative MRI showed a normal liver morphology (outlined by a solid orange line). A liver biopsy obtained during the primary resection confirmed normal hepatic parenchyma. In follow-up data, MRI revealed the emergence of three new liver metastasis (indicated by yellow arrowheads). The HE staining further confirmed the metastatic lesions (highlighted by an orange star) surrounding by fibrotic stroma (indicated by purple triangles). c-MET immunohistochemistry demonstrated a strong positivity within the metastatic foci and the periphery of fibrotic stroma. Patient #2 (60-year-old male): At baseline, MRI revealed features suggestive of hepatic fibrosis, including widened fissures and an irregular liver contour (indicated by white triangles). A concurrent liver biopsy confirmed existing fibrotic bands disrupting the normal lobular architecture. Follow-up data indicated that the patient developed rapid and extensive CRLM. MRI displayed aggressive disease progression with significant mass effect and architectural distortion. HE staining of the resected specimen revealed abundant fibrotic stroma surrounding the metastatic deposits. c-MET expression was confirmed in both the metastasis foci and the adjacent fibrous regions. Patient #3 (82-year-old male): At baseline, both MRI and HE staining confirmed a solitary liver metastasis without evidence of peritumoral fibrosis. By contrast, histopathological analysis displayed obvious fibrotic features in metastatic lesions with c-MET deposition, suggesting the concomitance of metastasis and fibrotic stroma. (B) GSEA displaying the differences in gene sets related to liver fibrosis and tumor formation between activated (in co-incubation with CRC cell supernatant) and quiescent LX-2 groups. (C) Volcano plot visualization of the differentially-expressed genes between activated and quiescent LX-2 groups. The top 5 upregulated and downregulated genes, as well as multiple genes related to the SDF-1/CXCR4 axis, are highlighted. (D) Representative immunofluorescence staining and quantitative analysis of SDF-1 in fibrotic mouse livers at different time points, along with images from the metastasis model showing SDF-1 deposition around metastatic and fibrotic areas. The white dotted line shows the liver metastatic lesions. (E) Schematic diagram of the synthesis of mSPIONP-Cur@eCM and its targeted delivery to the PMN for exerting diagnostic and therapeutic functions Results Construction of mSPIONP-Cur@eCM We first synthesized monodisperse SPIONPs via a one-step hydrothermal method (Fig. 2A). This approach involved the mild and direct reaction of FeCl3·6H2O with sodium acrylate in an ethylene glycol (EG) solution, in which sodium acrylate facilitated the coordination of carboxylate groups to iron cations, as evidenced by carboxylate vibration peaked at 1600 cm− 1 and 1405 cm− 1 in Fourier transform infrared (FTIR) spectra, enabling the batch production of Fe3O4 structures with high dispersibility (Fig. S5). X-ray diffraction (XRD) spectra confirmed the good crystallinity of the obtained SPIONPs, consistent with the standard profile of magnetite Fe3O4 (PDF# 03-065-3107) (Fig. S6A). X-ray photoelectron spectroscopy (XPS) peaked at 711.7 eV and 725.2 eV indicated Fe 2p core-level spectra, reflecting the iron oxidation states in SPIONP (Fig. 2B, S6B-D), which was further verified by energy-dispersive spectroscopy (EDS) mapping results (Fig. S7A). On the basis of the Scherrer equation from the XRD pattern, the average crystallite size of the synthesized SPIONP was approximately 5 nm, implying their high saturation magnetization, as iron oxide nanoparticles (NPs) less than 20 nm in diameter were known to be superparamagnetic [51]. Scanning electron microscopy (SEM) imaging revealed that the product consists of smaller secondary ferric oxide nanocrystals featured by a rough surface and a pseudo-mesoporous architecture (Fig. 2A, S7B-C). The mesoporous property was characterized by nitrogen adsorption‒desorption experiments, illustrating the specific surface area of 88.1 m2/g with pore size of 4.0 nm (Fig. 2C). The above experiments suggested the successful synthesis of SPIONPs with high saturation magnetization and mesoporous microstructures (mSPIONPs), and these properties rendered these mSPIONPs as a versatile nanocarrier for drug delivery. The mSPIONPs exhibited a potent Cur carrying capacity, yielding Cur-loaded nano-systems (mSPIONP-Cur) with drug loading rates ranging from 17.1% to 18.7%, as determined by ultraviolet-visible (UV-Vis) spectroscopy and thermogravimetric analysis (TGA) (Fig. S8). XRD and XPS analysis of mSPIONP-Cur confirmed that the incorporation of Cur induced negligible alterations on the valence state of mSPIONPs (Fig. S6), highlighting the single crystal stacking robustness of this inorganic nanocarrier. The mSPIONP-Cur cores were then encapsulated by CRC cell membranes to mimic HSC-recruited CRC cells colonizing into PMN. To potentiate this bionic effect, we conducted the transfection upon CRC cells using appropriate plasmids to augment the expression of CXCR4 membrane protein. Immunofluorescence and western blot assays corroborated that the overexpression efficiency peaked within 24 to 48 h after transfection (Fig. 2D, S9). Subsequently, we isolated engineered cell membranes (eCM) for the biomimetic camouflage of mSPIONP-Cur cores and obtained uniform coating of the hazy cell membrane in the mSPIONP-Cur@eCM, as visualized by SEM imaging and SDS-PAGE (Fig. 2A and E, and S10). With Cur loading and eCM camouflaging, the mSPIONPs-based nano-system underwent a dimension increment and a zeta potential variation with particle size increasing from 164 nm to 324 nm and zeta potential changing from neutral to negative (Fig. 2F, S11). These results together suggested the successful construction of mSPIONP-based nano-formulation with an inorganic core and a biomimetic shell. Release profiles of Cur and iron from the mSPIONP-Cur@eCM nano-system were evaluated using UV-Vis and inductively coupled plasma‒mass spectrometry (ICP-MS). The concentration of Cur in mSPIONP-Cur peaked after 6–8 h incubation in PBS, whereas the Cur release in mSPIONP-Cur@eCM was prolonged over 48 h (Fig. 2G). As acidity accelerated the dissociation of coordination between carboxylate groups and iron cations, the release rate of Cur significantly increased in an acidic environment (pH 5.5). In comparison with Cur release, iron leak from SPIONP-based formulations displayed slower processes under a neutral condition, as determined by ICP-MS (Fig. 2H). In an acidic solution, both mSPIONP-Cur and mSPIONP-Cur@eCM exhibited accelerated iron release, contributing to the rapid Cur unloading. This observation confirmed the biological stability of mSPIONP-Cur@eCM, further demonstrating the pH-responsive release of mSPIONP-Cur@eCM under acidic conditions. To evaluate the magnetic response of the mSPIONP-based nano-systems, the magnetic properties of mSPIONPs under an external magnetic field were characterized using a superconducting quantum interference device (SQUID) magnetometer. The mSPIONPs exhibited ferromagnetic behavior, with a saturation magnetization of approximately 75.8 emu/g and a remanent magnetization of approximately 7.9 emu/g—values that are relatively high for ferromagnetic NPs (Fig. 2I). The final formulation exhibited an r2 relaxivity of 51.4 mM− 1 s− 1 (Fig. S12A-B), indicating outstanding T2-weighted MRI performance and highlighting its potential for theranostics use. While mSPIONPs might theoretically breakdown into smaller crystals for T1-weighted MRI, experiments indicated that their T1 contrast was weaker than that of their T2 contrast (Fig. S12C). This was probably because that T1 relaxation relied on energy exchange between protons and contrast agents, and the large magnetic moment of the mSPIONPs obstructed effective energy transfer [52]. Fig. 2Structural and compositional characterization of mSPIONP-Cur@eCM. (A) Schematic of the synthesis and SEM images of mSPIONP, mSPIONP-Cur, and mSPIONP-Cur@eCM (scale bar, 100 nm). (B) Fe 2p XPS spectra of mSPIONP. (C) Nitrogen adsorption-desorption isotherms and corresponding pore size distribution curves of mSPIONP. (D) Representative confocal laser scanning microscopy (CLSM) images showing CXCR4 (green), F-actin (red), and nuclei (DAPI, blue) in CRC cells at different transfection time points (0 h, 12 h, 24 h, 48 h) (scale bar, 50 μm). The mean fluorescence intensity (MFI) of CXCR4 (green) was quantified to reflect its expression level. (E) SDS-PAGE analysis of protein profiles: L1, protein marker; L2, mSPIONP; L3, CRC cell membrane (CM); L4, mSPIONP@CM; L5, bioengineered CRC cell membrane (eCM); L6, mSPIONP-Cur@eCM. (F) DLS size distribution of mSPIONP, mSPIONP-Cur, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM. (G-H) Cumulative release profiles of curcumin (G) and iron (H) from the nanoparticles. Quantification data are shown as mean ± SD (n = 3). (I) Magnetic hysteresis curves of mSPIONP and mSPIONP-Cur@eCM Structural and compositional characterization of mSPIONP-Cur@eCM. (A) Schematic of the synthesis and SEM images of mSPIONP, mSPIONP-Cur, and mSPIONP-Cur@eCM (scale bar, 100 nm). (B) Fe 2p XPS spectra of mSPIONP. (C) Nitrogen adsorption-desorption isotherms and corresponding pore size distribution curves of mSPIONP. (D) Representative confocal laser scanning microscopy (CLSM) images showing CXCR4 (green), F-actin (red), and nuclei (DAPI, blue) in CRC cells at different transfection time points (0 h, 12 h, 24 h, 48 h) (scale bar, 50 μm). The mean fluorescence intensity (MFI) of CXCR4 (green) was quantified to reflect its expression level. (E) SDS-PAGE analysis of protein profiles: L1, protein marker; L2, mSPIONP; L3, CRC cell membrane (CM); L4, mSPIONP@CM; L5, bioengineered CRC cell membrane (eCM); L6, mSPIONP-Cur@eCM. (F) DLS size distribution of mSPIONP, mSPIONP-Cur, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM. (G-H) Cumulative release profiles of curcumin (G) and iron (H) from the nanoparticles. Quantification data are shown as mean ± SD (n = 3). (I) Magnetic hysteresis curves of mSPIONP and mSPIONP-Cur@eCM Cellular internalization of mSPIONP-Cur@eCM We then investigated the dual targeting function of mSPIONP-Cur@eCM to HSCs (human HSCs, LX-2 cells) and CRC cells (HCT116 cells). Activated LX-2 (aLX-2) cells via TGF-β stimulation, serving as a simplified in vitro system to mimic the physiological state of aHSCs, were characterized by contraction, proliferation, and migration, as visualized by microscopic observations (Fig. S13). When incubated with Cur, mSPIONP-Cur, mSPIONP-Cur@CM or mSPIONP-Cur@eCM, aLX-2 cells exhibited efficient phagocytosis abilities to nanosystems, as demonstrated by confocal laser scanning microscopy (CLSM) imaging and flow cytometry analysis (Fig. 3A and B). In comparison to other groups, mSPIONP-Cur@eCM treatment achieved the highest level of intracellular fluorescence with an approximatively 3-fold higher intensity than that of mSPIONP-Cur@CM (Fig. 3A). However, this enhanced internalization of mSPIONP-Cur@eCM relative to mSPIONP-Cur@CM was not observed in quiescent LX-2 cells. We speculated that such a specific affinity of mSPIONP-Cur@eCM to aLX-2 was driven by the recognition between SDF-1 and CXCR4. We thus blocked the CXCR4/SDF-1 axis in aLX-2 cells using the CXCR4 antagonist AMD3100, an anti-SDF-1 antibody, or an anti-CXCR4 antibody, all of which led to a substantial decrease in the endocytic uptake of mSPIONP-Cur@eCM (Fig. S14). These results suggested the specific internalization of mSPIONP-Cur@eCM by aHSCs originating from the ligand binding of CXCR4 to SDF-1, which confirmed the biomimetic function of bioengineered cytomembrane shell. The homologous targeting function of mSPIONP-Cur@eCM was next investigated in CRC cells (HCT116). CLSM imaging and flow cytometry analysis revealed that mSPIONP-Cur@CM and mSPIONP-Cur@eCM had higher affinities to CRC cells than other formulations (Fig. 3A and B). Together, the cellular internalization study verified the dual targeting function of mSPIONP-Cur@eCM, which exerted the homologous targeting derived from tumor cell membranes to enhance the CRC cell internalization and anchored the SDF-1/CXCR4 axis by CXCR4 overexpression to preferentially target aHSCs. Although SDF-1 from aHSCs dissolved in the pericellular region rather than locating on the cell membrane, this specific secretion generated a supportive and receptive microenvironment for CXCR4-overexpressed components [53], assisting the settlement of mSPIONP-Cur@eCM in the aHSC’s vicinity and facilitating its cell internalization. This dual-mechanism design lays a foundation for PMN-specific drug delivery. After cellular internalization, mSPIONP-Cur@eCM underwent the structural dissociation and fragmentation, displaying a dynamic degradation within 8 h, as imaged by bio-transmission electron microscopy (bio-TEM) (Fig. S15), which potentially facilitated the MRI detection and Cur release. HSC deactivation and antifibrosis of mSPIONP-Cur@eCM Next, we set out to investigate the deactivation and antifibrosis efficacy of mSPIONP-Cur@eCM by detecting key aHSC markers and profibrogenic factors. In consistence with our preliminary molecular docking analysis illustrating the binding of Cur to HSC activation-related markers (Fig. S4), the Western blot assay demonstrated that mSPIONP-Cur@eCM treatment reduced the expression levels of NOX4 and PTGS2 in LX-2 cells by 69% and 47%, respectively (Fig. 3C), verifying a potent HSC deactivation effect, as NOX4 and PTGS2 signaling pathways were implicated in the regulation of oxidative stress response and HSC polarization [54, 55]. Simultaneously, the phosphorylation of STAT in LX-2 cells was suppressed by 82% (Fig. 3C), potentially attenuating the profibrotic activity, since the elevated levels of STAT3-related inflammation in HSCs were correlated with the progression of hepatic fibrosis [56]. Consistently, the morphology of LX-2 cells after mSPIONP-Cur@eCM treatment showed few manifestations of cell activation (Fig. 3D). We also examined the key factors α-SMA and collagen type I (col I) involving in hepatic fibrosis [57, 58], confirming that mSPIONP-Cur@eCM treatment decreasing the expression levels of α-SMA and col I by 87% and 79%, respectively (Fig. 3E and F). In addition, our CLSM imaging displayed that the activation process gave rise to obvious mitochondrial fragmentation in aLX-2, resulting in the augmentation of mitochondrial curls and decreases in both the mitochondrial area and length/width ratio (Fig. S16). The mSPIONP-Cur@eCM treatment inhibited mitochondrial fission, thereby restoring mitochondrial morphology to a state resembling that of quiescent phenotype (Fig. 3G, S17). In contrast, mitochondrial fission was observed in the other groups, displaying pronounced blurring effect resulted from the accumulation and overlapping of mitochondria near the nucleus. Since inhibiting mitochondrial fission has been shown to deactivate and induce apoptosis toward HSCs [59], our finding underscored the substantial potential of mSPIONP-Cur@eCM in suppressing HSC polarization. These results together suggested that mSPIONP-Cur@eCM initiated deactivation and antifibrosis mechanisms in HSCs. Fig. 3In vitro cellular uptake of mSPIONP-Cur@eCM and its ability to reprogramme the HSCs. (A) Representative CLSM images of LX-2, aLX-2 and HCT116 cells after incubation with curcumin, mSPIONP-Cur, mSPIONP-Cur@CM, or mSPIONP-Cur@eCM for 4 h (scale bar, 20 μm). (B) Flow cytometry analysis of cell phagocytosis. (C) Western blot analysis validating the molecular docking-predicted targets of curcumin in LX-2 cells under various treatments for 24 h. The expression levels of NOX4, PTGS2, total STAT, and pSTAT were examined. β-Actin served as a loading control. (D) Representative CLSM images of α-SMA and NOX4 in LX-2 cells with various treatments (scale bar, 20 μm). (E) Western blotting analysis on the expression levels of α-SMA, col I, SDF-1 and β-Actin in LX-2 cells with various treatments. α-SMA and col I are established markers of hepatic fibrosis and HSC activation, and SDF-1 is a key pro-metastatic factor. (F) Representative CLSM images of α-SMA and col I in LX-2 cells with various treatments (scale bar, 20 μm). (G) The stimulated emission depletion (STED) imaging depicting changes in mitochondrial structure in various treatment groups (scale bar, 5 μm) In vitro cellular uptake of mSPIONP-Cur@eCM and its ability to reprogramme the HSCs. (A) Representative CLSM images of LX-2, aLX-2 and HCT116 cells after incubation with curcumin, mSPIONP-Cur, mSPIONP-Cur@CM, or mSPIONP-Cur@eCM for 4 h (scale bar, 20 μm). (B) Flow cytometry analysis of cell phagocytosis. (C) Western blot analysis validating the molecular docking-predicted targets of curcumin in LX-2 cells under various treatments for 24 h. The expression levels of NOX4, PTGS2, total STAT, and pSTAT were examined. β-Actin served as a loading control. (D) Representative CLSM images of α-SMA and NOX4 in LX-2 cells with various treatments (scale bar, 20 μm). (E) Western blotting analysis on the expression levels of α-SMA, col I, SDF-1 and β-Actin in LX-2 cells with various treatments. α-SMA and col I are established markers of hepatic fibrosis and HSC activation, and SDF-1 is a key pro-metastatic factor. (F) Representative CLSM images of α-SMA and col I in LX-2 cells with various treatments (scale bar, 20 μm). (G) The stimulated emission depletion (STED) imaging depicting changes in mitochondrial structure in various treatment groups (scale bar, 5 μm) HSC-CRC cell crosstalk inhibition of mSPIONP-Cur@eCM Given the central role of HSC-CRC cell crosstalk in the formation of fibrosis-concomitant PMN, we then investigated the inhibitory effect of mSPIONP-Cur@eCM on this cell crosstalk using a co-incubation in vitro model. In the PMN formation, HSCs have been identified to offer an SDF-1/CXCR4 signaling axis-based driving force for CRC cell colonization, which was blocked by mSPIONP-Cur@eCM, as demonstrated by enzyme-linked immunosorbent assay (ELISA) assay. Despite TGF-β stimulation significantly increased SDF-1 secretion in aLX-2 cells, mSPIONP-Cur@eCM effectively restored SDF-1 expression to baseline levels (Fig. S18). Transwell invasion assays demonstrated that, in comparison with untreated group, the colonization of CRC cells in the mSPIONP-Cur@eCM treatment group was reduced by approximately 70% (Fig. 4A, S19A). In contrast, eCM vesicles alone exhibited only a weak inhibitory effect on CRC cell invasion (6.4%), whereas free Cur showed moderate inhibition (39.7%) (Fig. S19B-C). In consistence with the invasion inhibition, the scratch assay showed the mSPIONP-Cur@eCM treatment reducing the migration distance of CRC cells by approximately 42% comparing to the control (Fig. S20), confirming that mSPIONP-Cur@eCM inhibited the CRC cell-recruiting potential of HSCs. The SDF-1/CXCR4 signaling axis-dependency underlying this inhibition was validated by both AMD3100 and anti-CXCR4 antibody blocking experiments (Fig. S19-S21). These results illustrated that the invasion of CXCR4-overexpressing CRC cells was significantly suppressed by AMD3100 antagonism or anti-CXCR4 antibody pretreatment, substantiating that mSPIONP-Cur@eCM’s invasion inhibition originated from its interference with the SDF-1/CXCR4-mediated cell crosstalk. Negative results from the mSPIONPs group and suboptimal invasion inhibitions induced by other Cur-containing formulations indicated that the crosstalk inhibition was contributed by Cur cargos, which was amplified in the mSPIONP-Cur@eCM formulation due to its HSC-specific accessibility and biodegradability. In addition to the SDF-1/CXCR4 blocking, mSPIONP-Cur@eCM exhibited potent cytotoxic effects upon aLX-2 cells and CRC cells. MTT assay revealed that the cytotoxicity of mSPIONP-Cur@eCM in aLX-2 cells was higher than that in quiescent LX-2 cells, affording IC50 values for aLX-2 and LX-2 cells at ~ 7.4 µM and ~ 11.2 µM, respectively (Fig. 4B). Since the mSPIONPs without Cur cargos induced negligible cell death even when the concentration reaching 1000 ng/mL (Fig. S22), it could be confirmed that mSPIONP-Cur@eCM’s cytotoxicity originated from its encapsulated Cur. Live/dead staining further verified mSPIONP-Cur@eCM’s selective lethality to aLX-2 cells, inducing complete cell death at a concentration down to 15 ng/mL (Fig. S23). On the other hand, mSPIONP-Cur@eCM improved the cytotoxicity of Cur upon CRC cells, which was attributed to the CRC-targeting capacity of the eCM shell exerting homotypic binding effects to the source cells (Fig. 4B). These data indicated that the biomimetic behavior of mSPIONP-Cur@eCM could strengthen the lethal effects of Cur to target cells, offering another avenue to abrogate the crosstalk of aHSCs and CRC cells, two key cellular components in the PMN. Transcriptome sequencing was next conducted to study the killing mechanism of mSPIONP-Cur@eCM to aLX-2 cells, demonstrating that mSPIONP-Cur@eCM induced significant alterations in the expression of genes encoding proteins associated with the cell cycle and apoptosis, including members of the cellular communication network (CCN) and cyclin-dependent kinase (CDK) families [60], as well as FADD, FAS, and TP53 (Fig. 4C). As confirmed by flow cytometry analysis, mSPIONP-Cur@eCM treatment resulted in significant G2/M phase accumulations, accompanied by an approximately two-fold reduction in the G0/G1 phase relative to the control (Fig. 4D, S24). Massive arrest at the G2/M phase in turn triggered cell apoptosis, as verified by the augmented proportion of early and late apoptotic LX-2 cells in the mSPIONP-Cur@eCM group (Fig. 4E, S25A). It has been reported that Cur has an antineoplastic activity in both liver and colon cancers by inducing cell apoptosis [61–63]. This effect was potentiated by the mSPIONP-Cur@eCM formulation, as validated by flow cytometry analysis illustrating that the apoptotic CRC cells of mSPIONP-Cur@eCM group significantly increased relative to other groups, with the total apoptotic population reaching 45.1% (Fig. 4F, S25B). These experimental data demonstrated that mSPIONP-Cur@eCM suppressed aHSC-CRC cell crosstalk by not only blocking key signaling pathways but also potentiating Cur’s pro-apoptosis effect upon HSCs and CRC cells. Fig. 4mSPIONP-Cur@eCM inhibits HSC-CRC cell crosstalk. (A) Schematic diagram of the co-incubation in vitro model and results of the transwell invasion assay. HSCs (LX-2) in the lower chamber were pretreated with various formulations for 24 h, followed by TGF-β stimulation for 24 h. CRC cells (HCT116) were then seeded in the upper chamber and allowed to invade toward the lower chamber. Representative images of invaded CRC cells (stained with crystal violet) are shown. (B) Cytotoxicity of mSPIONP, Cur, mSPIONP-Cur, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM on LX-2, aLX-2, and HCT116 cells after an 8-hour treatment. Quantification data are shown as mean ± SD (n = 3). (C) Heatmap of gene expression profiles in LX-2 and aLX-2 cells associated with the cell cycle and apoptosis pathways. (D) Representative flow cytometry analysis of cell cycle status in LX-2 cells after 24 h of different treatments. (E-F) Representative flow cytometry analysis of apoptosis levels in LX-2 cells (E) and HCT116 cells (F) after 24 h of different treatments mSPIONP-Cur@eCM inhibits HSC-CRC cell crosstalk. (A) Schematic diagram of the co-incubation in vitro model and results of the transwell invasion assay. HSCs (LX-2) in the lower chamber were pretreated with various formulations for 24 h, followed by TGF-β stimulation for 24 h. CRC cells (HCT116) were then seeded in the upper chamber and allowed to invade toward the lower chamber. Representative images of invaded CRC cells (stained with crystal violet) are shown. (B) Cytotoxicity of mSPIONP, Cur, mSPIONP-Cur, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM on LX-2, aLX-2, and HCT116 cells after an 8-hour treatment. Quantification data are shown as mean ± SD (n = 3). (C) Heatmap of gene expression profiles in LX-2 and aLX-2 cells associated with the cell cycle and apoptosis pathways. (D) Representative flow cytometry analysis of cell cycle status in LX-2 cells after 24 h of different treatments. (E-F) Representative flow cytometry analysis of apoptosis levels in LX-2 cells (E) and HCT116 cells (F) after 24 h of different treatments Theranostics of mSPIONP-Cur@eCM in a hepatic fibrosis model Since the metastatic organotropism of liver is associated with hepatic fibrosis, we established a fibrosis animal model to investigate the in vivo theranostics behaviors of mSPIONP-Cur@eCM (Fig. 5A). The mice received intraperitoneal injection of CCl4 for 35 days (details in methods) [23, 64, 65], and the model mice were characterized by the presence of fibrotic lesions and high levels of aHSC infiltration in their livers (Fig. S26). We first conducted the pharmacokinetics analysis in the model mice, demonstrating that the membrane-coated formulations (mSPIONP-Cur@CM and mSPIONP-Cur@eCM) exhibited prolonged circulation and sustained Cur release in the bloodstream for more than 16 h after intravenous administration (Fig. 5B), due to the biological stability provided by cell membrane camouflage. Biodistribution was next investigated with fluorescence imaging system. After three hours post the administration of Cy5-labeled nano-formulations, ex vivo imaging was conducted for the excised major organs, displaying that hepatic fluorescence intensity significantly exceeded that in other organs, which was attributed to the capture effect of liver tissues [66]. Particularly, livers in mSPIONP-Cur@eCM group had approximately 5.87-fold and 1.97-fold greater fluorescence intensity than those in the mSPIONP-Cur and mSPIONP-Cur@CM groups, respectively (Fig. 5C, S27), indicating the effective and selective accumulation of CXCR4-overexpressing mSPIONP-Cur@eCM in fibrotic liver tissues. Meanwhile, considerable renal fluorescence was detected in the mSPIONP-Cur@CM and mSPIONP-Cur@eCM groups, which may be resulted from CCl4-induced renal fibrosis [67]. As illustrated by fluorescence imaging, the membrane-coated formulations exerted persistent retention in livers, by which the released iron ions could be trapped due to the high level of ferritin receptor expression in livers (Fig. S28), potentiating their efficacy as a negative T2 contrast agent and expended the window of MRI diagnosis [68]. With the progressive release of iron ions (Fig. 5B), T2 signal intensity in the hepatic parenchyma was decreased, which consequently enhanced the conspicuity of fibrous tissue. As the eCM shell exerted the biomimic anchoring to aHSCs, mSPIONP-Cur@eCM administration yielded a superior accumulation in the liver parenchyma relative to mSPIONP-Cur@CM administration, affording clear differentiation between normal liver parenchyma and fibrotic lesions (Fig. 5D, S29A). Imaging sequence analysis indicated that the signal-to-noise ratio (ΔSNR) of mSPIONP-Cur@eCM in the liver region progressively increased after injection and maintained high values over 3 h, followed by a gradual decrease due to metabolic clearance (Fig. S29B). As observed through the T2WI sequence, the MRI enhancement of mSPIONP-Cur@eCM accurately reflected the severity of liver fibrosis, which could be used for assessing the risk of liver PMN formation in CRC settings. We then investigated the fibrosis intervention and HSC deactivation of mSPIONP-Cur@eCM in the hepatic fibrosis model. Continuous monitoring of body weight for the model mice revealed that the fibrosis induced weight loss of more than 10% over the 21-day period, whereas this situation was relieved in the Cur-containing groups. Particularly, mSPIONP-Cur@eCM treatment significantly attenuated the symptom, resulting in less than 3% weight loss (Fig. S30). We next examined the level of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in the blood of treated mice (Fig. S31), since these two factors could be primary indicators reflecting hepatic functional alterations [69]. Biochemical analysis confirmed that mSPIONP-Cur@eCM effectively relieved CCl4-induced liver injury, as verified by the restoration of ALT and AST to normal ranges (Fig. 5E, S32). As illustrated by gross observation of isolated livers and hematoxylin and eosin (HE) staining of whole liver sections (Fig. 5F and G), fibrotic livers from control group displayed extensive collagen deposition surrounding the central veins and hepatic sinusoids with blunt, rounded edges and fine granular protrusions on the surface (fibrotic area > 15%), whereas after mSPIONP-Cur@eCM treatment, the collagen deposition and fibrotic lesions were significantly reduced and the liver remained a normal bright brown-red color (fibrotic area < 1%). HE staining also showed that mSPIONP-Cur@eCM treatment mitigated fibrosis progression by ameliorating the CCl4-induced disorder of hepatocyte arrangement and reducing the infiltration of liver pseudolobules, fibrous connective tissue and inflammatory cells (Fig. 5G). Sirius red and Masson’s trichrome staining further confirmed the fibrosis intervention of mSPIONP-Cur@eCM (Fig. 5G, S33). The HSC deactivation of mSPIONP-Cur@eCM was substantiated by immunohistochemistry analysis, which showed that mSPIONP-Cur@eCM treatment effectively suppressed α-SMA and col I expressions within livers (Fig. 5H, S33), implying a reduction of activation phenotype [65]. Simultaneously, mSPIONP-Cur@eCM intervention alleviated oxidative stress-related inflammation responses by reducing NOX4 expression, which also contributed to the HSC deactivation (Fig. 5H, S33). Notably, by inhibiting the early activation of HSCs, mSPIONP-Cur@eCM treatment effectively downregulated the SDF-1 level from 27% in the control group to 2–3% (Fig. 5H, S33), potentially declining the pro-metastasis inclination of HSCs. Together, these data underscored the therapeutic potential of mSPIONP-Cur@eCM in suppressing HSC activation and fibrosis progression, laying a foundation for intervening the formation of fibrosis-concomitant PMN. Fig. 5Theranostics investigation of mSPIONP, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM in the hepatic fibrosis model. (A) Schematic diagram for the establishment and monitoring of hepatic fibrosis model. (B) Pharmacokinetics of curcumin and iron after the injection of Cur, mSPIONP-Cur, mSPIONP-Cur@CM, or mSPIONP-Cur@eCM. Quantification data are shown as mean ± SD (n = 3). (C) Ex vivo fluorescence images of organs harvested from hepatic fibrosis mice after intravenous administration of Cy5.5-labled mSPIONP, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM. (D) The pseudo-color images of T2-weighted MR images in fibrotic liver (indicated by white dashed line) and fibrous tissue (indicated by white arrow). TR/TE = 2000/80 ms. (E) Serum-level analysis of ALT and AST with different treatments. Quantification data are shown as mean ± SD (n = 3). Statistics were done using one-way analysis of ANOVA with Dunnett’s post-test, *** P < 0.001. (F) Comparison of liver tissues extracted from mice with different treatments. (G) Representative images of HE, Masson and sirius red stained liver tissue sections with different treatments. (H) Representative immunohistochemical staining of α-SMA, col I and NOX4, and immunofluorescence staining of SDF-1 in liver tissues with different treatments Theranostics investigation of mSPIONP, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM in the hepatic fibrosis model. (A) Schematic diagram for the establishment and monitoring of hepatic fibrosis model. (B) Pharmacokinetics of curcumin and iron after the injection of Cur, mSPIONP-Cur, mSPIONP-Cur@CM, or mSPIONP-Cur@eCM. Quantification data are shown as mean ± SD (n = 3). (C) Ex vivo fluorescence images of organs harvested from hepatic fibrosis mice after intravenous administration of Cy5.5-labled mSPIONP, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM. (D) The pseudo-color images of T2-weighted MR images in fibrotic liver (indicated by white dashed line) and fibrous tissue (indicated by white arrow). TR/TE = 2000/80 ms. (E) Serum-level analysis of ALT and AST with different treatments. Quantification data are shown as mean ± SD (n = 3). Statistics were done using one-way analysis of ANOVA with Dunnett’s post-test, *** P < 0.001. (F) Comparison of liver tissues extracted from mice with different treatments. (G) Representative images of HE, Masson and sirius red stained liver tissue sections with different treatments. (H) Representative immunohistochemical staining of α-SMA, col I and NOX4, and immunofluorescence staining of SDF-1 in liver tissues with different treatments Theranostics of mSPIONP-Cur@eCM in an early CRLM model The initiation of CRLM involves the CRC cell colonization and the metastatic foci formation. We thus conducted an early CRLM model through the splenic injection of luciferase-transfected CRC cells to induce micrometastasis lesions in fibrosis model mice (Fig. 6A) (details in methods) [70]. To characterize the progression of metastasis in this fibrosis‑associated CRLM model, we examined the metastatic burden at different time points after tumor cell inoculation. Notably, at one week post‑injection, the metastasis initiated, presenting micro-metastatic lesions without peritoneal or distant spread (Fig. 6B, S34), which was defined as the early‑stage CRLM for subsequent theranostics. MRI-based liver metastasis detection was then investigated with mSPIONP-based formulations, displaying that the membrane-camouflaged formulations had superior diagnoses for both primary tumors and liver metastases relative to mSPIONP, consistent with their performances in the fibrosis model (Fig. 6C). Particularly, mSPIONP-Cur@eCM exhibited pronounced abilities for T2-weighted contrast MRI, effectively delineating the lesion margins within 3 h post-administration (Fig. S35A), as manifested by marked signal regions within liver parenchyma. In the 8-hour MRI, a significant reduction in the T2 signal intensity of the metastases was observed, implying that mSPIONP-Cur@eCM gradually penetrated in the metastasis lesions, as visualized by the pseudocolor images (Fig. 6C). Quantitative analysis (Fig. S35B) further confirmed that mSPIONP-Cur@eCM exhibited the most pronounced and sustained T2 signal decrease over time among all formulations. Moreover, mSPIONP-Cur@eCM showed superior sensitivity in detecting small lesions (< 1.0 cm) compared to non-targeted controls (Fig. S35C), with an AUC of 0.9333 (Fig. S35D). This sequential “negative contrast imaging-to-metastasis penetration” process was driven by mSPIONP-Cur@eCM’s dual-targeting function and progressive dissociation, increasing the drug infiltration and facilitating the subsequent intervention. Collectively, our mSPIONP-based formulation enables early detection of liver micro-metastases by reducing the background liver signal, extends the imaging time window due to sustained lesion retention, and achieves targeted drug delivery for the intervention of liver metastasis. These capabilities of mSPIONP-Cur@eCM underscored its potential as a theranostic nanosystem for the treatment of early-stage CRLM. In vivo imaging at 8-hour after administration for the treated mice in supine and lateral positions further substantiated the CRLM-specificity of mSPIONP-Cur@eCM (Fig. 6D, S36-S37), as illustrated by a high degree of distribution consistency between Cy5-labeled mSPIONP-Cur@eCM and luciferase-positive malignant sites with a concordance rate reaching approximately 85–90% (Fig. 6D). This CRLM-specific infiltration opened up the therapeutic window for early CRLM. We next conducted the in vivo study via intravenous injection of different formulations once every two days for 24 days and the therapeutic efficacy on early CRLM was first evaluated with in vivo bioluminescence imaging on the 21th day after the establishment of liver metastasis. The results revealed that, compared to the control or Cur group with biofluorescence signals spreading in the abdominal cavity, the mSPIONP-Cur@eCM group displayed signals confining in spleen areas with slight distributions in livers (Fig. S38A-B). As for mSPIONP-Cur@CM and mSPIONP-Cur groups, considerable infiltrations of luciferase-positive CRC cells were observed in both spleens and livers. Combined with the MRI data, this comparison underscored that mSPIONP-Cur@eCM prevented the CRC cell colonization from the primary tumors in spleens, thus inhibiting the early CRLM progression. Consistently, ex vivo tissue imaging showed shrinking splenic primary tumors and nearly complete inhibition of liver metastasis with in mSPIONP-Cur@eCM group (Fig. 6E, S38C), which yielded a weight loss less than 7% (Fig. 6F) and a mortality down to 33% (Fig. 6G). Macroscopic dissection further confirmed the therapeutic effects of mSPIONP-Cur@eCM, in which the Ki67 and TUNEL staining for splenic tumor sections displayed the potent inhibition of mSPIONP-Cur@eCM upon CRC cells and the HE staining for liver sections confirmed its suppression on the formation of hepatic micro-metastases with significantly lower metastatic foci weights relative to that in the control group (Fig. 6H, S39). Notably, Masson’s trichrome staining visualized the fibrosis during micro-metastasis formation, showing metastatic foci were surrounded by prominent collagen fiber septa, with some fibrous structures irregularly distributed within the metastatic masses (Fig. 6H). With the mSPIONP-Cur@eCM treatment, the level of liver fibrosis was significantly attenuated and sporadic fibrosis spots were distributed in liver parenchyma. These data highlighted the intervention of mSPIONP-Cur@eCM on the fibrosis-concomitant CRLM. Underlying therapeutic mechanisms were studied with immunofluorescence staining. In addition to the apoptosis induction observed in Ki67 and TUNEL sections, mSPIONP-Cur@eCM exerted effective deactivation on HSCs in the liver, as verified by the significant downregulation of α-SMA, an indicator of the profibrotic HSC phenotype, in the section analysis. Immunofluorescence assay further indicated that mSPIONP-Cur@eCM blocked the SDF-1/CXCR4 signaling-based HSC-CRC cell crosstalk by reducing the SDF-1 expression (Fig. 6I, S39). Consequently, the metastatic organotropism was inhibited, as confirmed by the downregulation of N-Cadherin (N-cad), a mesenchymal-type cadherin serving as a PMN biomarker associated with epithelial-mesenchymal transition (EMT) [71]. These results suggested that mSPIONP-Cur@eCM effectively inhibited HSCs activation, thereby mitigating downstream liver fibrosis and excessive extracellular matrix deposition. Simultaneously, the released Cur induced apoptosis to the CRC cells recruited in the liver, disrupting the metastatic cascade. Therefore, by targeting both aHSCs and CRC cells, mSPIONP-Cur@eCM suppressed the activity of pro-metastatic cells and their interactions in the liver PMN, intervening early CRLM. Fig. 6Theranostics investigation of mSPIONP, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM in the CRLM model. (A) Schematic diagram for the establishment and monitoring of early CRLM model. (B) Time‑course of liver metastasis and intrasplenic tumor growth following intrasplenic injection of luc-HCT116 cells. Excised livers and spleens from fibrosis model mice were photographed at 1, 2, and 3 weeks post‑injection. Minimal metastatic involvement was observed at week 1, presenting as micro-metastatic lesions without peritoneal or distant spread, which was defined as the early‑stage CRLM. Whereas scattered foci emerged by week 2, and multiple diffuse metastases were evident throughout the liver by week 3. (C) T2-weighted MR images of the liver and spleen tissues in CRLM model mice with different treatment and corresponding pseudo-color images in liver tissue. The dotted line indicates liver and spleen, and the arrow indicates the in situ and metastatic tumors. Pseudo-color images at 2–8 h post-treatment illustrate the progressive reduction in signal intensity within liver metastatic lesions. TR/TE = 2000/80 ms. (D) In vivo fluorescence, bioluminescence, and overlapping signal images of CRLM model mice after intravenous injection with Cy5-labeled formulations and luciferase substrate at different time points. Overlapping images were generated by merging Cy5 fluorescence and bioluminescence signals; Color gradient indicates the degree of signal colocalization. (E) Comparison of metastatic liver tissues at the end of experiment. Liver metastases appear as dense, punctate white nodules on the liver surface. Three representative livers are shown per group (n = 5 initially enrolled; two additional mice per group were used for histology). (F-G) Body weight changes (F) and survival rate changes (G) of CRLM model mice during different treatments (n = 5). (H) In situ tumor slice images stained with TUNEL and Ki67, and representative images of HE, Masson’s trichrome staining metastatic liver tissue sections. Masson’s trichrome staining revealed liver metastatic lesions (orange stars) and fibrotic structures (purple triangles). Fibrotic areas are seen in close proximity to metastatic foci. (I) Immunofluorescent staining of the metastatic liver after different treatments. The cell nuclei (DAPI), fibrosis-related protein (α-SMA), SDF-1, and tumor invasion-related protein (N-cad) were stained with blue, red, green, and pink colors Theranostics investigation of mSPIONP, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM in the CRLM model. (A) Schematic diagram for the establishment and monitoring of early CRLM model. (B) Time‑course of liver metastasis and intrasplenic tumor growth following intrasplenic injection of luc-HCT116 cells. Excised livers and spleens from fibrosis model mice were photographed at 1, 2, and 3 weeks post‑injection. Minimal metastatic involvement was observed at week 1, presenting as micro-metastatic lesions without peritoneal or distant spread, which was defined as the early‑stage CRLM. Whereas scattered foci emerged by week 2, and multiple diffuse metastases were evident throughout the liver by week 3. (C) T2-weighted MR images of the liver and spleen tissues in CRLM model mice with different treatment and corresponding pseudo-color images in liver tissue. The dotted line indicates liver and spleen, and the arrow indicates the in situ and metastatic tumors. Pseudo-color images at 2–8 h post-treatment illustrate the progressive reduction in signal intensity within liver metastatic lesions. TR/TE = 2000/80 ms. (D) In vivo fluorescence, bioluminescence, and overlapping signal images of CRLM model mice after intravenous injection with Cy5-labeled formulations and luciferase substrate at different time points. Overlapping images were generated by merging Cy5 fluorescence and bioluminescence signals; Color gradient indicates the degree of signal colocalization. (E) Comparison of metastatic liver tissues at the end of experiment. Liver metastases appear as dense, punctate white nodules on the liver surface. Three representative livers are shown per group (n = 5 initially enrolled; two additional mice per group were used for histology). (F-G) Body weight changes (F) and survival rate changes (G) of CRLM model mice during different treatments (n = 5). (H) In situ tumor slice images stained with TUNEL and Ki67, and representative images of HE, Masson’s trichrome staining metastatic liver tissue sections. Masson’s trichrome staining revealed liver metastatic lesions (orange stars) and fibrotic structures (purple triangles). Fibrotic areas are seen in close proximity to metastatic foci. (I) Immunofluorescent staining of the metastatic liver after different treatments. The cell nuclei (DAPI), fibrosis-related protein (α-SMA), SDF-1, and tumor invasion-related protein (N-cad) were stained with blue, red, green, and pink colors Discussion Three key factors contributing to the PMN have been identified, including primary tumor-derived components, tumor-mobilized bone-marrow-derived cells (BMDCs), and metastatic organotropism modeled by the local stromal components [6, 72]. For CRLM, HSCs play a central role in linking the three factors, firstly harnessing cancer cell-like metabolic pathways to meet the energy demands for activation and proliferation, then participating in CRC cell colonization, BMDC recruitment, ECM deposition and hepatic fibrosis [73–76]. Such a diversity inspired us to develop an HSC-targeting drug delivery strategy for the intervention of PMN and early CRLM. To materialize this concept, we design the mSPIONP-Cur@eCM that integrates the biomimetic cytomembrane camouflage with the mSPIONP-based theranostic core, aiming to initiate PMN-specific, MRI-guided drug delivery. Given the central role of plastic HSCs in the fibrosis-concomitant PMN, Cur is selected as the active pharmaceutical ingredient due to its deactivation effect on HSCs and antineoplastic activity in gastrointestinal cancer, and our hybrid fabrication improves Cur’s bioavailability, overcoming its limitations in hydrophilicity, biostability and cancer specificity [77–79]. With the biomimetic and biodegradable properties, mSPIONP-Cur@eCM can exert a dual targeting effect to the liver PMN and initiate MRI-based theranostics. In animal models, mSPIONP-Cur@eCM manifest advantages in the systemic circulation and biodistribution, enabling accurate depiction of fibrotic deposits and micro-metastatic lesions through the T2-weighted MRI. Simultaneously, it suppresses the activity of aHSCs and CRC cells and blocks their crosstalk in the liver PMN, thus intervening the CRLM progression. As the hybrid mSPIONP-Cur@eCM is fabricated with biocompatible ingredients, its administration shows excellent biosafety, inducing negligible hemolysis, hepatorenal abnormalities and tissue lesions (Fig. S40). Apart from these, our bioengineered, MRI-guided drug delivery system can further be leveraged in other aspects of CRLM management, such as directing the treatment approach selection, implementing the preoperative chemotherapy, and monitoring the therapeutic responses to systemic treatments, providing a personalized theranostic platform in this field. Methods Materials Ferric chloride hexahydrate (FeCl3·6H2O), sodium acrylate (CH2=CHCOONa), sodium acetate (CH3COONa, NaOAc), EG, and AMD3100 were obtained from Sigma-Aldrich (USA). Curcumin (Cur, > 98.0%) was purchased from MCE (Shanghai, China). The CXCR4 overexpression plasmid (ov-CXCR4) and the corresponding negative control plasmid (ov-NC) were synthesized with the lentiviral expression vector GV230 by Shanghai Jikai Co., Ltd. The transfection regent Lipo3000 was purchased from Invitrogen (L3000008). SDF-1 kits were purchased from Elabscience (Wuhan, China). Additional materials, including 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), Hoechst and 4’,6-diamidino-2-phenylindole (DAPI), were acquired from Beyotime (Shanghai, China). The PKMITO mitochondrial dye was provided by Nanjing Genvivo Biotech Co., Ltd. (Jiangsu, China). All other chemicals and solvents used in the experiments were of analytical or chromatographic grade. Cell lines and animals HCT116, HT29, SW480, SW620 and BJ cells were purchased from ATCC. HCT116 and HT29 cells were maintained in McCoy’s complete medium (Procell, Wuhan, China), and SW480 and SW620 cells were maintained in complete RPMI-1640 medium (Gibco, NY, USA). Based on screening experiments including CXCR4 baseline expression, transfection efficiency, and induction of LX-2 activation, HCT116 was selected as the model cell line for subsequent experiments. BJ cells were cultured in complete DMEM (Gibco, NY, USA). Human hepatic stellate LX-2 cells were purchased from Haixing Biosciences (Suzhou, Jiangsu, China) and cultured in RPMI 1640 medium (Gibco, NY, USA). The LX-2/TGF-β model represents an established in vitro system for HSCs activation studies, although its limitations in fully mimicking the in vivo microenvironment should be noted. All the cell cultures were maintained at 37 °C in a humidified atmosphere with 5% CO2 supplemented with FBS (10%) and penicillin/streptomycin (1%). Male NOD-SCID mice (6–8 weeks of age, ≈ 21 g) were purchased from Shanghai Silaike Laboratory Animal Co., Ltd. (Shanghai, China). All animal experiments adhered to the relevant ethical guidelines and were approved by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University School of Medicine (Approval Number: 2025 − 181). A hepatic fibrosis mouse model was established by the administration of carbon tetrachloride (CCl4; Sigma-Aldrich, USA). A mouse model of colorectal cancer liver metastasis was established using the spleen-preserving method. Retrospective analysis of the correlation between liver fibrosis and liver metastasis Ethical approval for the use of MRI and clinical data was granted by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University School (Approval Number: 2021 − 1029). A retrospective cohort of 104 patients who underwent orthotopic colorectal cancer tumor resection at the Second Affiliated Hospital of Zhejiang University School between 2024 and 2025 was included. Liver fibrosis was assessed using a step-up algorithm, with histology as the reference standard. Advanced fibrosis was defined as a METAVIR stage F ≥ 2 or Ishak stage ≥ 3. In cases where biopsy was unavailable, noninvasive tests were employed, as follows: APRI > 1.5, FIB-4 ≥ 3.25, or FibroScan ≥ 7.9 kPa (≥ 9.5 kPa for cirrhosis). Kaplan‒Meier analysis was used to compare the incidence of liver metastasis following colorectal cancer resection between the fibrotic and nonfibrotic groups. Liver stiffness measurements, acquired via transient elastography using FibroScan, were compared between patients with and without subsequent development of liver metastasis using an unpaired two-sample Student’s t-test. Coculture and transcriptomic analysis of LX-2 cells with CRC cells supernatants CRC cells (HT29, 620, SW480, and HCT116) and hepatic stellate cells (LX-2) were cultured separately. Once the cancer cells reached the logarithmic growth phase, their media were collected by centrifugation. The media were then used to replace the LX-2 media, and the LX-2 cells were cultured for 48 h. To assess morphological changes, LX-2 cells stimulated with conditioned media from each CRC cell line were fixed and stained with phalloidin (for F-actin) and DAPI. Cytoskeletal reorganization and cell morphology were observed using CLSM. Additionally, to evaluate the transition of LX-2 cells toward a CAF-like phenotype, cells were subjected to immunofluorescence staining for FAP, a canonical CAF marker. Images were acquired using CLSM to visualize the expression and distribution of FAP. Total RNA from the cultured LX-2 cells was extracted using TRIzol Reagent, and RNA sequencing was performed by Novogene Co., Ltd. Functional enrichment analyses, including Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), transcription factor, and protein‒protein interaction network analyses, were also performed. Synthesis of mSPIONPs Initially, sodium acrylate (0.445 g) was dissolved in 0.7 mL of deionized water and homogenized using ultrasonication. FeCl3·6H2O (1.35 g) and NaOAc (5.97 g) were dissolved in EG (20 mL). Subsequently, the aqueous sodium acrylate solution was slowly added to the EG solution with magnetic stirring. The resulting homogeneous yellow solution was transferred into a Teflon-lined stainless-steel autoclave, which was sealed and heated to 200 °C. After 10 h, the autoclave was allowed to cool to room temperature naturally. Finally, the obtained mSPIONPs were thoroughly washed with ethanol and water and subsequently dried under vacuum for 12 h. Construction of mSPIONP-Cur To incorporate curcumin molecules into mSPIONP, 10 mg of mSPIONP was dispersed in 2 mL of an acetone solution containing 20 mg of curcumin, followed by vigorous stirring overnight. The mSPIONP-Cur composite was subsequently collected using an external magnet over 1 h. The drug loading capacity and encapsulation efficiency of mSPIONP-Cur were evaluated using a UV‒vis spectrophotometer (U-4100, HITACHI, Japan) and confirmed by TGA (SDT Q600, TA Instruments, USA). Preparation of bioengineered CRC-cell membranes (eCM) and fabrication of mSPIONP-Cur@eCM CRC cells were transfected with plasmid DNA for CXCR4 overexpression using Lipofectamine 3000 following the manufacturer’s instructions. CXCR4 expression was assessed at 12,24,48, and 72 h post-transfection via immunofluorescence and Western blotting. Once cell lines stably overexpression CXCR4 were established, larger populations were transfected with polyethylenimine (PEI; Invitrogen). The membranes of the CXCR4-overexpressing cells were isolated using a membrane protein extraction kit (Beyotime, China) following the manufacturer’s protocol. Cells were lysed by freeze‒thaw treatment, and the lysate was centrifuged to remove intact cells and nuclei and then further centrifuged to isolate cell membranes. For the preparation of mSPIONP-Cur@eCM nanoparticles, mSPIONP-Cur was combined with CXCR4-overexpressing CRC cell membranes through sonication, followed by extrusion through a 200 nm polyethylene terephthalate membrane (LiposoFast, Avestin, Canada). The resulting mSPIONP-Cur@eCM nanoparticles were harvested by centrifugation and washed with water. Characterization SEM (GeminiSEM 300, Zeiss, Germany) was utilized to investigate the morphological characteristics of the nanoparticles, whereas EDS (XMAXn65 T, Oxford, UK) facilitated elemental mapping of C, O, Fe, and N. Nitrogen adsorption-desorption analyses were conducted using a Micromeritics Tristar II analyzer (Micromeritics, USA), and data interpretation was performed through application of the Brunauer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) models to determine the specific surface areas and pore size distributions. XRD (D/MAX-2550 PC, Rigaku, Japan) was employed with Cu Kα radiation (2θ = 10–80°) to elucidate the phase structure, whereas XPS (ESCALAB 250Xi, Thermo Fisher, USA) was used to determine the valence states of iron and to monitor changes in the valence state of iron following curcumin loading. FTIR spectroscopy (VECTOR22, Bruker, Germany) was conducted within the range of 400–4000 cm− 1. The particle size and zeta potential were measured via dynamic light scattering (Litesizer 500, Anton-Paar, Austria). The magnetic properties of the mSPIONP and mSPIONP-Cur@eCM powders were characterized using a SQUID magnetometer (Quantum Design MPMS3, San Diego, CA, USA) at 300 K, with magnetic field cycling between − 10 and 10 kOe. The content of iron in the supernatant was measured by ICP-MS (Thermo Fisher, USA). Cellular internalization First, LX-2 cells were activated with 10 ng/mL TGF-β for 24 h. Prior to treatment, both activated and nonactivated cells were serum-starved overnight and seeded onto 8-chamber slides. Upon reaching 70% confluence, the medium was replaced with fresh medium containing curcumin, mSPIONP-Cur, SPIONP-Cur@CM, and mSPIONP-Cur@eCM. After 4 h, the cells were washed with PBS, fixed with 4% paraformaldehyde for 30 min, and stained with actin-tracker and DAPI. Imaging was performed using a Nikon microscope (Nikon Eclipse Ti-S, USA). Quantitative analysis was conducted using flow cytometry analysis (NovoCyteTM, ACEA Biosciences, USA). Furthermore, cells exposed to different treatments were analyzed via bio-TEM. HSC deactivation LX-2 cells were seeded and cultured in an 8-chamber slide for 24 h. Five different treatments were subsequently administered: PBS, curcumin, mSPIONP-Cur, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM. After a 24-hour incubation period, the medium was replaced with fresh medium containing 10 ng/ml TGF-β, and the cells were incubated for an additional 24 h. The cells were subsequently washed with cold PBS and fixed with 4% paraformaldehyde for 30 min. Triton X-100 was then applied for 10 min to ensure adequate permeabilization. The cells were subsequently blocked with blocking buffer for 2 h, after which they were incubated overnight at 4 °C with primary antibodies against NOX4, α-SMA, and collagen I. After being washed with PBS, the cells were incubated with the appropriate secondary antibodies for 2 h at room temperature. DAPI and an actin tracker were used to label the nuclei and F-actin, respectively. Finally, fluorescence images were captured using CLSM. Following live cell staining after different treatments in accordance with the provided protocol, mitochondrial morphology was examined using a stimulated emission depletion (STED) microscope (Abberior Instruments, Germany) equipped with a 100×/1.4 NA oil immersion objective. Images were acquired with a pixel size of 20 nm and a pixel dwell time of 10 µs, using a 640 nm excitation laser and a 775 nm depletion laser. The acquired images were analyzed using ImageJ software (version 2.3.0). Western blotting Prior to electrophoresis, the total protein concentrations were quantified using the BCA assay to ensure uniform sample loading. The proteins were subsequently resolved by SDS-PAGE and transferred onto PVDF membranes. The membranes were incubated overnight with primary antibodies at the following dilutions: anti-β-actin (1:1000; Abcam), anti-CXCR4 (1:2000; Proteintech), anti-PTGS2 (1:1000; Proteintech), anti-NOX4 (1:2000; Proteintech), anti-collagen I (1:2000; Proteintech), anti-STAT (1:1000; Abcam), anti-p-STAT (1:1000; Abcam), and anti-SDF-1 (1:500; Proteintech). After being washed, the membranes were treated with horseradish peroxidase (HRP)-conjugated secondary antibodies. The protein bands were subsequently visualized via chemiluminescence using a ChemiDoc Touch Imaging System (Bio-Rad, USA). Transwell assay Cell invasion assays were conducted utilizing a transwell cell culture chamber coated with Matrigel (Corning Costar, Inc.) in accordance with the manufacturer’s protocol. First, LX-2 cells were seeded in the lower chamber and treated with 20 ng/ml nanoparticles for 24 h, followed by stimulation with 10 ng/ml TGF-β for an additional 24 h to induce their activation. The medium was subsequently replaced with complete medium containing 20% FBS. CRC cells were then seeded in the upper compartment and cultured in serum-free medium. After incubation at 37 °C with 5% CO2 for 24 h, the upper chamber was carefully cleaned with a cotton swab to remove the cells from the upper surface of the membrane. The cells that had migrated to the bottom of the membrane were fixed with 4% paraformaldehyde and stained with 1% crystal violet. Data analysis was carried out using ImageJ software (version 7.0). Wound healing assay CRC cells from each experimental group were seeded into 6-well plates and incubated for 24 h before the culture medium was replaced with the supernatant collected from LX-2 cells treated with various nanoparticles. A yellow pipette tip was used to generate a scratch in each well, and the cells were cultured for an additional 48 h. Photographs of the cells were taken at 0, 6, 24, and 48 h. The distance between the two edges of the wound surface at each time point was quantitatively assessed using ImageJ software (version 7.0). Cell cycle detection Cell cycle distribution was evaluated utilizing a cell cycle detection kit (Liankebio, Hangzhou, China). First, cells from each experimental group were fixed in 70% ethanol at 4 °C overnight. The cells were subsequently collected via centrifugation and resuspended in 500 µL of RNase A/PI working solution, followed by incubation at room temperature for 30 min in the dark. Flow cytometry analysis was performed using a CytoFLEX S flow cytometer (Beckman Coulter, USA), and a minimum of 10,000 events were acquired per sample. The DNA content was analyzed using the cell cycle platform in FlowJo software (V10.8.1), with aggregates excluded using the PI-A vs. PI-W gating strategy. Cell apoptosis detection Apoptosis was detected using an Annexin V-FITC/PI apoptosis detection kit (Solarbio, Beijing, China). After the cells in each group were digested, 500 µL of binding buffer, 5 µL of FITC-annexin V, and 5 µL of PI working solution were added. Then, the cells were incubated at room temperature in the dark for 15 min. Finally, the level of apoptosis was detected by flow cytometry analysis using a CytoFLEX S flow cytometer, and a minimum of 10,000 events were acquired per sample. The data were analyzed using FlowJo software (V10.8.1), with apoptotic cells quantified as the percentage of cells in the early apoptosis (Annexin V+/PI-) and late apoptosis (Annexin V+/PI+) quadrants. Blood compatibility test To assess hemocompatibility, murine red blood cells were exposed to increasing concentrations of mSPIONP@eCM (25, 50, 75, 100, 125, 150, 175, and 200 µg/mL) at 37 °C for 4 h. After incubation, the hemolytic effects were quantified by measuring the absorbance of the supernatants at 570 nm, with deionized water and PBS serving as positive and negative controls, respectively. Systemic toxicity was assessed in healthy NOD-SCID mice given intravenous mSPIONP-Cur@eCM at dosages of 5, 10, or 20 mg/kg every three days until day 18, followed by a drug-free period until day 35 to detect delayed toxicity. PBS-treated mice served as controls. Blood samples were periodically collected for analysis of serum biochemical parameters. Hepatic function was assessed by measuring total protein (TP), albumin (ALB), ALT, AST, and alkaline phosphatase (ALP), while renal function was evaluated by determining blood urea nitrogen (BUN), creatinine (CREA), and uric acid (UA) levels using clinical diagnostic methods. Hepatic fibrosis model To investigate the in vivo distribution and tumor-targeting efficacy of the nanoformulations, we established hepatic fibrosis model (n = 5 per group) to simulate liver fibrosis-concomitant PMN. The hepatic fibrosis model was established via intraperitoneal injection of a 20% (v/v) CCl4 olive oil solution at a dose of 1 µL/g body weight every three days for a total of 5 weeks. Age-matched mice receiving injections of equivalent volumes of olive oil alone served as the control group. The successful establishment of liver fibrosis was confirmed by histological analysis (HE and Masson’s trichrome staining) prior to subsequent experiments. CRLM model A spleen-preserving technique was utilized to establish CRLM. Luc-HCT116 cells were washed and resuspended in serum-free RPMI-1640 medium for further use. The skins of the mice were disinfected, and a horizontal incision approximately 1 cm long was made 1–2 cm below the left thoracic ribs. The spleen was gently removed through this incision. Suspended cells (1 × 106 per mouse) were injected into the splenic capsule, reaching one-third to two-thirds of the depth of the spleen. Finally, the muscle layer and skin were sutured. The same operator conducted all surgical procedures to maintain consistency and reproducibility. In vivo imaging Cy5-labeled mSPIONP, mSPIONP@CM, or mSPIONP@eCM (1 mg/kg) was administered intravenously. Dynamic monitoring of the Cy5- and D-Luciferin fluorescence signals was performed at predetermined intervals using an IVIS imaging system (Caliper, PerkinElmer, USA). For in vivo fluorescence imaging, mice were anesthetized with 2% isoflurane in oxygen, and images were acquired using the ICG filter set (excitation: 745 nm, emission: 800 nm) with an exposure time of 2 s. At 8 h post-injection, the mice were euthanized, and major organs were collected for ex vivo fluorescence imaging to assess nanoparticle biodistribution and retention. For MRI analysis, T2-weighted images were acquired using a 3.0 T MRI scanner (Discovery MR 750, GE, USA). The transverse relaxation rate (r2) was determined by plotting the reciprocal of the relaxation time (1/T2) against the corresponding iron concentration and performing linear regression. For the in vivo MRI evaluation, mice received intravenous injections of mSPIONP, mSPIONP-Cur@CM, or mSPIONP-Cur@eCM. During MRI acquisition, mice were anesthetized with 1.5-2% isoflurane in air and placed in a dedicated small-animal coil. Respiration rate was monitored continuously using a pressure sensor (SA Instruments, Inc., USA), and body temperature was maintained at 37 ± 0.5 °C using a circulating water heating system. Imaging was performed using a T2-weighted fast spin-echo sequence with a total scan time of approximately 8 min per animal. The key parameters were: repetition time (TR) = 2000 ms, echo time (TE) = 80 ms, field of view (FOV) = 60 × 60 mm², matrix = 256 × 192, slice thickness = 2 mm, and number of excitations (NEX) = 4. The therapeutic effects in hepatic fibrosis model Firstly, the hepatic fibrosis model was established by intraperitoneal injection of CCl4 using the above-mentioned method. After two weeks of treatment, the mice were randomly divided into five experimental groups (n = 5 per group). Each group received intravenous injections of PBS, curcumin, mSPIONP-Cur, mSPIONP-Cur@CM, or mSPIONP-Cur@eCM (5 mg/kg for nanoparticles) every two days. At the same time, the intraperitoneal injection of CCl4 was still performed. Throughout the treatment period, body weight was systematically monitored. On day 21 after treatment initiation, all the mice were euthanized, after which blood and the livers were collected and the liver index was determined. Serum samples were subjected to biochemical analyses of liver function markers. Liver tissues were dehydrated, embedded in paraffin, and subsequently sectioned into formalin-fixed, paraffin-embedded sections for histopathological evaluation. Sections were subjected to HE staining, Masson’s trichrome staining, Sirius red staining, immunohistochemical staining for collagen I, α-SMA and NOX4, and immunofluorescence staining for SDF-1 to assess the therapeutic efficacy of the nanoparticles in inhibiting liver fibrosis through the suppression of hepatic stellate cell activation. The therapeutic effects in early CRLM model First, NOD-SCID mice were used to establish a hepatic fibrosis model via intraperitoneal injection of 20% (v/v) CCl4 olive oil solution (1 µL/g) every three days for three weeks. After fibrosis induction, a CRLM model was established by intrasplenic injection of luc-HCT116 cells. To identify the appropriate timing for early intervention, mice were dissected at 1, 2, and 3 weeks post-tumor cell inoculation. At the first week, no macroscopic metastatic nodules were detectable on the liver surface. Only scattered heterogeneous cells or occasional micrometastatic foci were observed within the fibrotic background, without evidence of peritoneal or distant metastasis, representing the early stage CRLM model. In contrast, mice at 2–3 weeks exhibited substantial metastatic burden. Given that PMN formation occurs within a narrow temporal window during the first week post-injection, treatment was initiated at day 3 (allowing for postoperative recovery) and metastatic burden was evaluated at week 1–2 to specifically capture the early effects on PMN disruption. Mice were randomly assigned to five groups (n = 5 per group): PBS control, free curcumin, mSPIONP-Cur, mSPIONP-Cur@CM, and mSPIONP-Cur@eCM. All treatments were administered via tail vein injection at a dose of 5 mg/kg every two days until the end of the experiment. To monitor metastasis progression, MRI was performed at one and two weeks after tumor cell inoculation, and tumor burden, including the number and size of metastatic lesions, was assessed. Three weeks after metastasis induction, bioluminescence imaging was performed to evaluate liver metastasis and intrasplenic tumor distribution of luc-HCT116 cells. Then, mice were euthanized. Two mice from each group were used for histological analysis of liver sections, including HE staining, Masson’s trichrome staining, and immunohistochemistry. The remaining three mice of each group were examined for macroscopic metastatic nodules on the liver surface. For survival analysis, an additional cohort of 30 mice was subjected to the same treatment regimen, with body weight monitored every three days and drug administration continued until day 50 after the start of CCl4 administration (i.e., covering the 3-week fibrosis induction period and the subsequent 4-week metastasis period) to evaluate overall survival and body weight. Statistical analysis Data analysis was performed using SPSS software (v26.0; IBM Corp., USA) and GraphPad Prism (v10.0.4; GraphPad Software, USA). Quantitative data are presented as the mean ± standard deviation (SD). For comparisons between two groups, a two-tailed unpaired Student’s t test was applied. One-way analysis of variance (ANOVA) was used for comparisons among multiple groups. For time-dependent multigroup comparisons, two-way ANOVA was performed, followed by appropriate post hoc tests. A p value < 0.05 was considered to indicate statistical significance. Significance levels are denoted as *P < 0.05, ** P < 0.01, *** P < 0.001, and **** P < 0.0001. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1