Hypoxic migrasomes drive colorectal cancer liver metastasis by mediating CD5L+ macrophage efferocytosis via NRP2/PROX1 axis Authors: Yuqi Luo, Yiwen Huang, Yaoxing Huang, Longzhen Chen Published: 2025 ABSTRACT ObjectiveMigrasomes are a newly discovered class of extracellular vesicles, yet their roles in colorectal cancer (CRC) metastasis remain poorly understood. This study aimed to investigate the functional significance of CRC-derived migrasomes, particularly under hypoxic conditions, in promoting liver metastasis and modulating the tumor immune microenvironment.MethodsMigrasomes in CRC tissues and cells were characterized using transmission electron microscopy or immunofluorescence. A mouse liver metastasis model and single-cell RNA sequencing (scRNA-seq) were employed to explore functional outcomes and cellular interactions.ResultsMigrasome structures were observed in both primary CRC and metastatic liver tissues, and live-cell imaging revealed hypoxic CRC cells released increased numbers of migrasomes. In vivo imaging demonstrated hepatic accumulation of hypoxic migrasomes and enhanced liver metastasis in mice. ScRNA-seq of liver metastases revealed that hypoxic migrasomes reprogrammed the tumor microenvironment, notably expanding a Tmem45a⁺ fibroblast subset with myofibroblast features and promoting CD5L⁺ macrophage differentiation with elevated efferocytic capacity. Mechanistically, NRP2, enriched in migrasomes derived from hypoxic CRC cells, was transferred to macrophages, binding with PROX1 to drive CD5L expression and upregulate of efferocytosis receptors. NRP2 knockdown in CRC cells abrogated migrasome-induced CD5L⁺ macrophage polarization and impaired apoptotic tumor cell clearance.ConclusionThese findings demonstrate that hypoxic CRC-derived migrasomes facilitate liver metastasis by reprogramming stromal and immune compartments, particularly through NRP2/PROX1-mediated education of macrophages toward a pro-efferocytic CD5L⁺ phenotype. Our study reveals a previously unrecognized intercellular communication axis involving migrasomes in CRC progression and provides a potential therapeutic target for metastatic disease.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12967-025-07485-0. Introduction Liver metastasis is a major cause of colorectal cancer (CRC)-related mortality, accounting for 70% of all metastatic CRC cases. Despite significant advances in diagnostic and therapeutic technologies, the prognosis for patients with CRC liver metastasis remains poor, with a 5-year survival rate as low as 30% [1]. Recent studies have revealed that the successful dissemination of CRC cells to the liver is not merely a mechanical process, but is also driven by specific crosstalk events between cancer cells and immune cells in the microenvironment [2]. Therefore, identifying key mediators that drive the communication between cancer cells and immune cells, and exploring the mechanisms by which these interactions facilitate CRC liver metastasis, is crucial for developing new therapeutic strategies and improving the prognosis of patients with CRC liver metastasis. Migrasomes are a recently discovered type of extracellular vesicle, providing a novel mechanism for intercellular communication and the lateral transfer of cellular contents [3, 4]. Unlike other extracellular vesicles, migrasomes exhibit unique biogenesis, cargo packaging, transport, and effects on recipient cells. The formation of migrasomes is dependent on cell migration and occurs at the tips or intersections of contractile fibers, with a diameter ranging from 0.5 to 3 μm, exhibiting a pomegranate-like shape. Cellular contents, including mitochondria, mRNA, and proteins, can be actively transported into migrasomes; as the contractile fibers break, migrasomes are released from the donor cell and subsequently taken up by recipient cells [5]. Thus, migrasomes mediate intercellular communication and material transfer. Several migrasome markers have been identified, including TSPAN4, integrins, NDST1, PIGK, CPQ, and EOGT [6]. Increasing evidence suggests that migrasomes play roles in mitochondrial quality control [7], angiogenesis [8], and organogenesis [9], and are involved in various pathological processes, including cardiovascular diseases, neurodegenerative disorders, and cancers. For instance, neutrophil migrasomes are enriched with adhesion molecules and coagulation factors, allowing them to preferentially accumulate at sites of injury, triggering platelet activation and clot formation [10]. In patients with cerebral amyloid angiopathy, the number of migrasomes produced by macrophages increases, with CD5L encapsulated in the migrasomes and tethered to the vasculature, weakening resistance to complement activation [11]. Migrasomes produced by glioblastoma carry autophagosomes and alleviate endoplasmic reticulum stress [12]. Importantly, an increasing number of studies have demonstrated that tumor-derived migrasomes or exosomes can promote tumor metastasis. For instance, tumor-derived exosomal miR-934 [13] and HSPC111 [14] promote liver metastasis of CRC. Migrasomes derived from breast cancer cells mediate interactions between tumor cells and osteoclasts, inducing osteoclast subtypes associated with bone metastasis [15]. CD151-enriched migrasomes mediate liver cancer invasion by promoting angiogenesis [16]. In our previous studies, we reported that extracellular vesicles, exosomes, can mediate natural killer cell regulation of immune evasion in CRC [17]. However, it remains unexplored whether migrasomes, as a novel type of extracellular vesicle, also participate in immune regulation and the process of liver metastasis in CRC. Efferocytosis is the process by which phagocytic cells precisely recognize and engulf neighboring apoptotic cells, playing a crucial role in immune suppression. Macrophages are key phagocytes that efficiently process ingested apoptotic cells through continuous efferocytosis [18]. Efferocytosis consists of three steps [19, 20]: Attraction, where apoptotic cells release “find-me” signals, such as CX3CL1, ATP, and RPS19, to induce chemotaxis of phagocytes; Recognition, where phagocytes directly (e.g., BAI1, TIM1) or indirectly (e.g., AXL, MerTK) recognize phosphatidylserine on the surface of apoptotic cells; Engulfment (eating) and lysosomal degradation (digestion). Increasing evidence suggests that macrophage efferocytosis induces an immunosuppressive tumor microenvironment, aiding cancer cells in evading immune surveillance. Macrophage efferocytosis can clear apoptotic breast cancer cells, enhancing the production of immunosuppressive cytokines and the accumulation of myeloid-derived suppressor cells, thereby inducing immune suppression and promoting tumor metastasis [21]. In CRC, efferocytosis provides a favorable microenvironment for M2 macrophages, driving immune suppression by inhibiting the STING/IFN-I signaling pathway and reducing the activity of cytotoxic T cells [22]. Selective blockade of the efferocytosis receptor MerTK can suppress tumor growth in a CRC liver metastasis model [23]. These findings suggest that macrophage efferocytosis is a key driver of immunosuppressive microenvironment formation and tumor metastasis. However, it remains unknown whether migrasomes can regulate macrophage efferocytosis and thereby mediate CRC liver metastasis. This study aimed to investigate whether CRC cells under hypoxic conditions promote liver metastasis by secreting migrasomes that selectively induce efferocytosis in heterogeneous macrophage subpopulations. To this end, we first identified the presence of migrasomes in both clinical CRC liver metastasis samples and CRC cell lines. By integrating single-cell RNA sequencing (scRNA-seq) data, we pinpointed CD5L⁺ macrophages as the primary recipient cell population of CRC-derived migrasomes. Subsequently, a series of in vivo and in vitro functional assays, along with mechanistic investigations, were conducted to elucidate the regulatory effects and underlying molecular mechanisms by which hypoxia-induced CRC migrasomes influence CD5L⁺ macrophage differentiation and efferocytotic activity. This research provides new insights into the role of tumor-derived migrasomes in shaping the metastatic immune microenvironment and suggest a previously unrecognized mechanism contributing to CRC liver metastasis. Materials and methods Tumor samples The primary CRC lesions (n = 3), liver metastatic CRC tissues (n = 3), and adjacent non-neoplastic tissues (n = 3) were collected from Shenzhen Longhua District Central Hospital, with approval from the Institutional Review Board (IRB) of Shenzhen Longhua District Central Hospital (2024-049-02). All patients provided written informed consent prior to participation in the study. Patient anonymity was maintained throughout the study, and all procedures were in accordance with the Declaration of Helsinki. The inclusion criteria for the study were: (1) patients diagnosed with CRC based on histopathological examination; (2) patients undergoing surgery for tumor resection; (3) patients with no history of other malignancies or systemic diseases affecting immune function. Exclusion criteria included: (1) patients receiving preoperative chemotherapy or radiotherapy; (2) patients with any autoimmune disorders or severe infections; (3) patients with incomplete clinical or pathological data. Tissue samples were obtained during surgery and immediately frozen in liquid nitrogen or formalin-fixed for subsequent analyses. Clinical characteristics of all samples are shown in Supplemental Table 1. Patient demographic characteristics, including age and sex, were well balanced across the study groups. Isolation and culture of primary CRC cells from tumor specimens Primary CRC cells were isolated from freshly resected human CRC tumor tissues using Tumor Dissociation Kit, Human (#130-095-929, Miltenyi). Briefly, Fresh CRC tumor tissues were rinsed with cold PBS and finely minced into 2–4 mm fragments. Enzymatic digestion was performed by incubating the tissue fragments in 5 mL of enzyme solution prepared by Tumor Dissociation Kit in 4.4 mL RPMI 1640 medium. The mixture was incubated at 37 °C in a shaking incubator (75 rpm, 5% CO₂) for 30 min. After digestion, the suspension was gently pipetted up and down 10 times using a wide-bore pipette tip, followed by filtration through a 70 μm nylon mesh filter. The filtrate was centrifuged at 250 × g for 3 min at 4 °C, and the cell pellet was resuspended in complete culture medium (RPMI 1640 supplemented with 10% FBS and 1% penicillin-streptomycin). Transmission electron microscopy (TEM) TEM was used to observe the ultrastructure of migrasomes. In briefly, tissues were immediately fixed in glutaraldehyde followed by a secondary fixation with osmium tetroxide. After fixation, the sample is dehydrated through a graded ethanol series (30%, 50%, 70%, 90%, and 100% ethanol), followed by infiltration with a resin. The resin-embedded sample was cut into 50–100 nm thick sections using an ultramicrotome. The thin sections were mounted onto a copper grid and then stained with phosphotungstic acid. Finally, the stained sections on grids were placed into the TEM chamber, and images were captured by TEM (JEM-1200EX, Japan Electronics Co., LTD). Immunofluorescence (IF) Tissue sections or cultured cells were fixed with 4% paraformaldehyde and permeabilized with Triton X-100 followed by blocking in 5% BSA. Samples were incubated with primary antibodies, CD5L (1:500, 17224-1-AP, Proteintech), CD163 (1:200, ab182422, Abcam), NRP2 (1:250, #3366, CST), PROX1(1:250, sc-81983, Santa cruz), MERTK (200 µg/mL, sc-365499, Santa Cruz). After washing off unbound primary antibodies, the sample was incubated with a fluorophore-conjugated secondary antibody. Excess unbound antibodies were removed by washing and then sample was mounted onto glass slides with a mounting medium that contained DAPI. Finally, sample was examined under a fluorescence microscope. Cell culture, hypoxic treatment, and transfection The human CRC cell lines, including P53R (#iCell-h169) and HCT116 (#iCell-h071) cells were purchased from iCell Bioscience Inc, and the murine CRC cell lines, including CT26 (#iCell-m014) and MC38 (#iCell-m032) were also purchased from iCell Bioscience Inc. HCT116 cells were maintained in McCoy’s 5 A medium. CT26 cells were cultured in RPMI-1640 medium. P53R and MC38 cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM). All medium were supplemented with 10% fetal bovine serum (FBS, 10099-141, GIBCO), 1% penicillin-streptomycin (P/S, E607011, Sangon), 1% non-essential amino acid (iCell-01000, iCell), 1% sodium pyruvate (iCell-01100, iCell), 1% HEPES (iCell-01200, iCell) and 1% GlutaMAX-1 (iCell-0900, iCell). To evaluate the effects of hypoxic stimulation on CRC cells, cells were cultured under hypoxic conditions (1% O₂, 5% CO₂, 94% N₂) at 37 °C using a tri-gas incubator (Thermo Scientific) for 24 h. Normoxic controls were cultured in a standard incubator at 5% CO₂ and 37 °C for the same duration. Murine bone marrow-derived macrophages (BMDMs) were utilized in this study. The isolation procedure was as follows: femurs and tibias were harvested and thoroughly cleaned. The proximal and distal ends of each bone were cut using sterile scissors. A pre-cooled high-glucose DMEM was drawn into a 1–2 mL syringe and used to flush the bone marrow out of the femurs and tibias. The flushing was repeated at least three times until the marrow cavity appeared pale and free of visible red coloration. The resulting cell suspension, containing bone marrow cells, was pipetted repeatedly to disperse any cell clumps. The suspension was then passed through a 200-mesh cell strainer to remove debris and aggregates. The filtered cells were transferred to a 15 mL centrifuge tube and centrifuged at 300 × g for 5 min. The supernatant was discarded, and the cell pellet was resuspended in pre-cooled bone marrow macrophage culture medium for subsequent experiments. BMDMs were cultured in a specialized primary murine bone marrow-derived macrophage medium (iCell-i017-002 m, CytoSci Biosciences). To achieve stable NRP2 knockdown cell lines, MC38 cells were transduced with lentivirus particles carrying pSlenti-U6-shRNA (NRP2)-CMV-F2A-Puro-WRPE. Lentivirus was packaged in HEK293T cells using a standard three-plasmid system and harvested 48–72 h post-transfection. Viral supernatant was filtered (0.45 μm) and applied to MC38 cells in the presence of 8 µg/mL polybrene to enhance infection efficiency. At 48 h post-infection, cells were subjected to puromycin selection at 2 µg/mL for 3–5 days to eliminate non-transduced cells. To overexpress NRP2, the NRP2 coding sequence was cloned into the pcDNA3.1 vector, and the NRP2-overexpressing plasmid was transfected into BMDMs using Lipofectamine 2000 (#11668019, Invitrogen). For transient knockdown of NRP2, PROX1, and MERTK, the corresponding siRNAs were transfected into BMDMs using Lipofectamine 2000 according to the manufacturer’s instructions. The medium was replaced with fresh culture medium 6 h after transfection, and cells were harvested 24 h post-transfection. The sequences of all siRNAs are listed in Supplementary Table 1. Migrasomes isolation The isolation and purification of migrasomes were performed following the method described by Gu et al. [24]. Briefly, after 48 h of cell culture, cells were treated with 5 µg/mL cytochalasin B for 30 min to promote the detachment of migrasomes from the cells. The cells were then digested with trypsin to obtain a suspension, followed by centrifugation at 500 × g for 5 min. The supernatant was carefully transferred onto a Ficoll gradient (5%, 10%, 15%, and 25%) and centrifuged at 150,000 × g for 60 min. Finally, the interface between the 15% and 25% layers was carefully aspirated using a pipette to collect the migrasomes. Live-cell migrasome labeling and visualization To visualize migrasomes in live CRC cells, cells were incubated with Wheat Germ Agglutinin conjugated to Alexa Fluor 488 (WGA-488, Cat# W11261, Thermo Fisher Scientific) at a 1:500 dilution. Staining was performed at room temperature for 10 min in the dark, followed by immediate live-cell imaging using a fluorescence microscope. Moreover, to visualize migrasomes in live CRC cells, TSPAN4 was fused with GFP and cloned into an expression plasmid. Tumor cells were transfected with TSPAN4-GFP plasmid using Lipofectamine 2000. Live-cell imaging was performed using a fluorescence microscope to visualize GFP-labeled migrasomes. Western blot (WB) Tissues or cells were lysed to extract proteins and then protein concentration was determined by the BCA protein assay kit (Thermo scientific, USA) followed by separating with 10% SDS-PAGE gel. The separated proteins on the gel were transferred onto PVDF membranes and then blocked for nonspecific binding with 5% nonfat milk. After that, membranes were incubated with NRP2 (1:1000, Proteintech, 27193-1-AP), CALR (1:1000, Proteintech, 10292-1-AP), PIGK (1:1000, Proteintech, 15151-1-AP), CPQ (1:1000, Proteintech, 16601-1-AP), EOGT (1:1500, Proteintech, 27595-1-AP), CD63 (1:1000, Proteintech, 25682-1-AP), TSG101 (1: 5000, Abcam, ab125011), CD9 (1:3000, Proteintech, 20597-1-AP), PROX1 (1:500, Santa Cruz, sc-81983), and anti-GAPDH (1:1000, Proteintech, 60004-1-Lg) at 4 °C overnight. Membranes were incubated with incubated with HRP-conjugated Affinipure Goat Anti-Rabbit IgG(H + L) (1: 5000, Proteintech, SA00001-2) at room temperature for 1 h. Finally, after washing away unbound secondary antibody, the bound antibodies were detected and visualized by chemiluminescence apparatus (Qinxiang, Shanghai, China). Mice treatment Sixteen 4–6-week-old C57BL/6J mice were purchased from Junke Biology (CSXK(Zhe)2019-0004). The mice were housed under controlled conditions with a temperature of 24–26 °C, a humidity of 55–60%, and a 12-hour light/dark cycle, with ad libitum access to water. Mice were acclimatized to the housing environment for 7 days prior to the experiment. The animal experiments were approved by the Laboratory Animal Ethics Committee of Guangdong Medical University (GDY2402433). To investigate the distribution of migrasomes in mice, six mice were randomly divided into two groups. Each group received a tail vein injection of DiR-labeled normoxic migrasomes (Nor-Mig) or hypoxic migrasomes (Hyp-Mig), respectively. Fluorescence distribution was observed using in vivo imaging 3 h after injection. A CRC liver metastasis model was established to investigate the specific effects of migrasomes on CRC. Ten mice were randomly divided into two groups, weighed, and anesthetized with avertin. Following sterile skin disinfection, a longitudinal incision (~ 0.5 cm) was made below the left axillary line along the posterior rib margin to expose the spleen, and a 0.2 mL suspension of MC38 single cells was injected into the spleen. MC38 cell-derived migrasomes (50 µg) were administered via tail vein injection, the injections were repeated every 3 days. After 14 days, in vivo imaging was performed to observe liver metastases. Mice were euthanized using an excess of CO2, and liver tissues were excised for examination of tumor nodules. scRNA-seq analysis Tumor tissues from MC38 tumor-bearing mice treated with migrasomes derived from MC38 cells cultured under normoxic or hypoxic conditions were collected for scRNA-seq, with one sample per group. The cell concentration was adjusted to 1 × 106, and cells were loaded onto GemCode gel beads and chips within 15 min of preparing the single-cell suspension, following the manufacturer’s instructions. The loaded samples were processed in the Chromium™ Controller (Pleasanton, California) for cell lysis and RNA reverse transcription. Libraries were prepared using the 10x Genomics Library Kit (MobiDrop, China) and sequenced on the Illumina NextSeq500 platform, as per the manufacturer’s recommendations. A sequencing depth of over 100 million reads per library was achieved, corresponding to an average of approximately 30,000 reads per cell. Data analysis of scRNA-seq The samples were matched and counted using the official MobiVision software from MobiDrop. Quality control and filtering were performed using the Seurat R package v4.3.0.1, and cells were retained if they met the following criteria: UMI count > 500 and < 100,000; detected genes > 200 and < 9,000; mitochondrial gene percentage < 25%; and red blood cell percentage < 3%. Genes expressed in fewer than three cells were removed. Doublets were identified and removed using the DoubletFinder software v2.0.4. Batch correction for multiple samples was performed using the default methods in Seurat. Dimensionality reduction was then conducted sequentially using PCA and UMAP, ultimately projecting the cells onto a two-dimensional plane. Based on the results of PCA, cell clustering was performed using the shared nearest neighbor (SNN) algorithm. Cell annotation was performed using the SingleR package based on the BlueprintEncodeData dataset. The initial annotation results were then refined using known markers to obtain the final cell annotation. Differentially expressed genes (DEGs) for each cell type were identified using the FindAllMarkers function in the Seurat package, while DEGs between the two groups were identified using the FindMarkers function. Genes with |log2FC| >1 and an adjusted p value (Bonferroni correction) < 0.05 were considered statistically significant. Cell-cell communication between different cell types was analyzed and identified using the CellPhoneDB software. A total of 341 efferocytosis-related genes was compiled by integrating gene sets associated with efferocytosis pathways from the KEGG and MSigDB databases. Similarly, 23 migrasome-related genes were curated from published literature. The single-sample gene set enrichment analysis (ssGSEA) implemented in the GSVA package (version 1.46.0) was employed to compute efferocytosis scores and migrasome scores for each sample. Paired boxplots comparing the scores between the two cell groups were generated using the ggplot2 package, and statistical significance was assessed with the Wilcoxon test. Flow cytometry The CD5L+ macrophages were sorted by flow cytometry. Primary murine BMDMs were cultured in specialized macrophage medium (iCell-i017-002 m, CytoSci Biosciences) in the presence of 100 ng/mL recombinant M-CSF (91109ES10, Yeasen) for 6 days. On day 6, adherent macrophages were detached using 12 mM lidocaine (HY-B0185A, MCE) for 3 min, followed by an additional 3-minute digestion with trypsin. Cells were gently pipetted to create a single-cell suspension, centrifuged at 300 × g for 3 min, and resuspended in 1 mL PBS. Cells were counted and adjusted to a density of 1 × 10⁷ cells/tube, then centrifuged at 350 × g for 5 min. The pellet was resuspended in 200 µL antibody dilution buffer (2% FBS in 1× PBS) and incubated with 0.5 µg CD5L antibody (BAF2834, R&D Systems) at 4 °C for 60 min in the dark. After washing once with PBS and centrifugation (350 × g, 5 min), cells were resuspended in 200 µL antibody dilution buffer and stained with 0.25 µL donkey anti-goat IgG H&L (488) secondary antibody (1:1000, ab150129, Abcam) for 30 min at 4 °C in the dark. Following incubation, cells were washed with 1 mL antibody dilution buffer and centrifuged at 200 × g for 3 min. The cell pellet was finally resuspended in sorting buffer (5 mM EDTA, 10 mM HEPES, 2% FBS in 1× PBS) and subjected to flow cytometric sorting using a FACSVerse™ flow cytometer (BD Biosciences). Assessment of macrophage phagocytic function To evaluate the impact of hypoxic migrasomes on macrophage phagocytic activity, BMDMs were incubated with either normoxic or hypoxic tumor-derived exosomes (100 µg/mL) for 48 h, with PBS treatment serving as the negative control. MC38 tumor cells were labeled with 1 µg/mL propidium iodide (PI, Cat#IP5030, Solarbio) to mark apoptotic or necrotic cells. Labeled cells were harvested by centrifugation (1000 × g, 5 min) and co-incubated with pre-treated BMDMs for 24 h. Following incubation, cells were fixed with 4% paraformaldehyde for 30 min at room temperature, permeabilized using 0.3% Triton X-100 for 15 min, and blocked with 3% BSA for 30 min. After washing with PBS, cells were incubated overnight at 4 °C with F4/80 antibody (1:200, 28463-1-AP, Proteintech), followed by incubation with fluorescently labeled secondary antibody (1:500, ab150077, Abcam) for 1 h at room temperature in the dark. Nuclei were counterstained with DAPI (10 min, dark), and fluorescence was observed using a fluorescence microscope. Migrasome labeling and colocalization of NRP2-GFP To investigate the colocalization of NRP2 with migrasomes, the NRP2 coding sequence was cloned into a GFP-tagged plasmid vector, and the resulting NRP2-GFP construct was transfected into CRC MC38 cells cultured under hypoxic conditions using Lipofectamine™ 2000 Transfection Reagent (Invitrogen, USA) according to the manufacturer’s protocol. After 6 h, the medium was replaced with fresh complete medium, and cells were continuously cultured for 48 h before migrasome isolation. Migrasomes were labeled with the red fluorescent dye DiI (Cat#C1036, Beyotime, China) following standard procedures. Briefly, isolated migrasomes were incubated with DiI and then washed thoroughly with PBS. To remove unbound dye, the samples were centrifuged at 20,000 × g for 1 h at 4 °C, and the pellet was resuspended in PBS. Subsequently, PI-labeled migrasomes were incubated with BMDMs overnight at 37 °C with gentle rocking to facilitate uptake. For nuclear staining, Hoechst 33,342 (Cat#C1029, Beyotime, China) was added and cells were incubated at room temperature for 10 min in the dark. Finally, cells were imaged using a fluorescence microscope. Co-immunoprecipitation (Co-IP) assay Macrophages were washed twice with cold PBS and collected by scraping, followed by centrifugation at 800 × g for 5 min at 4 °C. Cell pellets were lysed in 1 ml Pierce Lysis Buffer (87787, Thermo Fisher) supplemented with protease (78438, Thermo Fisher) and phosphatase inhibitors (78426, Thermo Fisher) on ice for 30 min. Lysates were cleared by centrifugation at 13,000 rpm for 15 min at 4 °C. RNA was digested with RNase A/T1 mix (EN0551, Thermo Fisher) at room temperature for 30 min. An aliquot of 1/10 of the lysate was saved as input. The remaining lysate was pre-cleared with 25 µl Protein A/G magnetic beads (88803, Thermo Fisher) at 4 °C for 1 h. Samples were then incubated with 2.5 µg NRP2 antibody (CST #3366) overnight at 4 °C, followed by capture with pre-washed Protein A/G magnetic beads for 4 h at 4 °C. Beads were washed three times with Co-IP washing buffer (10 mM Tris-HCl pH 7.5, 1 mM EDTA, 1 mM EGTA, 150 mM NaCl, 1% Triton X-100, 0.2 mM sodium orthovanadate) for 5 min each. Bound proteins were eluted with 1× loading buffer at 95 °C for 10 min and subjected to western blot. RT-qPCR analysis The RNA was reverse-transcribed into cDNA (#K1622, Thermo) followed by qPCR amplification using 2×Master Mix (Roche) on ABI Q6 Real-Time PCR System (Applied Biosystems). All primers were synthesized by Sangon Biotech (Shanghai, China) and are show in Supplemental Table 2. The relative gene expression was determined by the 2−ΔΔCt method and was normalized to GAPDH. ELISA assays Mouse TGF-β, IL-10, and CCL2 levels were measured using commercial ELISA kits according to the manufacturers’ protocols. Specifically, TGF-β was quantified using the Mouse TGF-β ELISA kit (ML057830, Enzyme-linked Biology), IL-10 using the Mouse IL-10 ELISA kit (ML037873, Enzyme-linked Biology), and CCL2 using the Mouse MCP-1/CCL2 ELISA kit (ML037840, Enzyme-linked Biology). Transwell assays To evaluate the effect of NRP2 overexpression or MERTK knockdown in macrophages on the transmigration ability of CFSE-labeled CRC cells, a Transwell co-culture assay was performed. BMDMs with NRP2 overexpression or MERTK knockdown were resuspended at a final concentration of 1 × 10⁵ cells/mL. MC38 cells were labeled with CFSE (ab113853, Abcam), after which the culture medium was removed and the cells were resuspended in serum-free medium to a final concentration of 1 × 10⁵ cells/mL. For the assay, 700 µL of complete MC38 culture medium containing 10% FBS was added to the lower chamber (24-well plate). The upper chamber of the Transwell insert was seeded with a 1:1 mixture of 250 µL BMDMs and 250 µL CFSE-labeled MC38 cells, followed by incubation for 48 h in a CO₂ incubator. After incubation, the inserts were carefully removed with forceps, and the medium in the upper chamber was gently aspirated. The upper surface of the membrane was carefully wiped with a moist cotton swab to remove non-migrated cells. Migrated cells on the lower surface were observed under a fluorescence microscope. To investigate whether macrophages regulate CRC cell migration and invasion through paracrine effects independent of efferocytosis, conditioned media (CM) were collected from macrophages after efferocytosis inhibition and used to treat CRC cells in Transwell assays. Cells were divided into four groups: blank CM (CD5L⁻ macrophages co-cultured with untreated apoptotic MC38 cells), CD5L⁺ macrophages CM (CD5L⁺ macrophages co-cultured with untreated apoptotic MC38 cells), efferocytosis-inhibited CD5L⁺ macrophages CM (CD5L⁺ macrophages co-cultured with Annexin V–pretreated apoptotic MC38 cells), and hypoxic migrasomes + efferocytosis-inhibited CD5L⁺ macrophages CM (CD5L⁺ macrophages co-cultured with Annexin V–pretreated apoptotic MC38 cells and treated with 100 µg/mL hypoxic migrasomes). Macrophages were cultured in serum-free medium throughout. To inhibit efferocytosis, apoptotic MC38 cells were pretreated with 3 µg/mL Annexin V (abbkine, PRP3003), while untreated apoptotic MC38 cells served as controls. After 48 h of co-culture according to the grouping above, supernatants were collected and used as CM. For the Transwell assays, 700 µL of complete MC38 culture medium containing 10% FBS was added to the lower chamber (24-well plate). The upper chamber was seeded with a 1:1 mixture of 250 µL macrophage-CM and 250 µL MC38 cell suspension, followed by incubation for 48 h in a CO₂ incubator. Migrated and invaded cells were visualized and quantified under a fluorescence microscope. Migration assays were performed using 8.0 μm 24-well Transwell inserts (353097, FALCON), and invasion assays using 8.0 μm 24-well Matrigel-coated inserts (354480, FALCON). Statistical analysis Statistical analysis was performed by GraphPad 10.0 using a two-tailed ANOVA following Tukey’s test to assess significant differences between three or more groups. The significant differences between two groups were assessed by a two-tailed t test. The p < 0.05 was considered significant. Results Identification of migrasomes derived from CRC cells As a novel organelle, the study of migrasomes in tumors is still in its early stages, and their role in tumor immunity and CRC has not yet been reported. Therefore, the first step in our investigation is to determine whether migrasomes are present in CRC tumor tissues. To this end, we performed TEM on ultrathin sections of adjacent normal tissues, clinical non-metastatic CRC primary tumors, and liver metastatic tumor tissues from CRC patients. We observed large, oval, pomegranate-like vesicles with diameters ranging from 0.5 to 3.0 μm in all three tissues, and these vesicles were either attached to the contracted fibers of cell extensions or were free in the intercellular space after rupture (Fig. 1A), which is consistent with the reported morphology of migrasomes [5]. Thus, we observed the presence of migrasomes in the tissue microenvironment, with an increased number of migrasomes in the liver metastatic tissues (Fig. 1B). These findings confirm the presence of migrasomes within CRC tumor tissues, which may be associated with liver metastasis. Fig. 1Identification of migrasomes derived from CRC cells. (A) Transmission electron microscopy (TEM) images showing migrasome-like vesicular structures (indicated by red arrows) in peritumoral, primary CRC, and colorectal cancer liver metastasis (CRLM) tissues from CRC patients. (B) Quantification of migrasome numbers per field in TEM images from panel (A). (C) Live-cell imaging of primary CRC cells labeled with WGA-488, demonstrating migrasome formation. (D) Representative images of migrasomes labeled with WGA-488 in CT26, HCT116, and MC38 cells cultured under normoxic or hypoxic conditions, and TSPAN4-GFP-labeled migrasomes in P53R cells. (E) Quantitative analysis of migrasome numbers per cell under normoxia and hypoxia from panel (E). (F) Western blot analysis of migrasome-specific markers (PIGK, CPQ, EOGT) and classical exosome markers (CD9, TSG101, CD63) to validate the identity of isolated migrasomes. * p < 0.05, ** p < 0.01, *** p < 0.001 Identification of migrasomes derived from CRC cells. (A) Transmission electron microscopy (TEM) images showing migrasome-like vesicular structures (indicated by red arrows) in peritumoral, primary CRC, and colorectal cancer liver metastasis (CRLM) tissues from CRC patients. (B) Quantification of migrasome numbers per field in TEM images from panel (A). (C) Live-cell imaging of primary CRC cells labeled with WGA-488, demonstrating migrasome formation. (D) Representative images of migrasomes labeled with WGA-488 in CT26, HCT116, and MC38 cells cultured under normoxic or hypoxic conditions, and TSPAN4-GFP-labeled migrasomes in P53R cells. (E) Quantitative analysis of migrasome numbers per cell under normoxia and hypoxia from panel (E). (F) Western blot analysis of migrasome-specific markers (PIGK, CPQ, EOGT) and classical exosome markers (CD9, TSG101, CD63) to validate the identity of isolated migrasomes. * p < 0.05, ** p < 0.01, *** p < 0.001 To identify which cell types in tumor tissues, contribute to the altered migrasome production, we utilized scRNA-seq data from GSE234804 to calculate the potential for migrasome generation by different cell types. As shown in Supplemental Fig. 1A and 1B, six cell types were identified, including B cells, T cells, myeloid cells, fibroblasts, endothelial cells, and CRC cells. Migrasome score analysis revealed that, although endothelial cells exhibited the highest scores overall, their scores significantly decreased in liver metastases compared with primary CRC tissues, contradicting the TEM findings showing increased migrasome abundance in metastatic lesions. Consequently, endothelial cells were excluded. Notably, both tumor and myeloid cells displayed markedly elevated migrasome scores, suggesting that migrasomes associated with CRC liver metastasis may be mainly derived from these two cell types (Supplemental Fig. 1C). To further demonstrate which specific cell types contribute to the presence of migrasomes in CRC tumor tissues, we focused on three cell populations—cancer cells, endothelial cells, and myeloid-derived macrophages—that exhibited significantly different migrasome scores in the scRNA-seq GSE234804 dataset. IF staining was performed on clinical tumor samples to assess the colocalization of the migrasome marker TSPAN4 with cell-type-specific markers of these three populations. IF results demonstrated that TSPNA4 predominantly co-localized with the CRC cell marker COX-2, and COX-2 expression was significantly higher than other type-cells (Supplemental Fig. 1D). Moreover, we also observed a colocalization of TSPAN4 with the macrophage marker CD163, whereas TSPAN4 expression was scarcely detected in endothelial cells (Supplemental Fig. 1D). These findings suggest that migrasomes in the CRC tumor microenvironment primarily originate from CRC cells. To further confirm that CRC cells can secrete migrasomes, we isolated primary CRC cells from CRC tumor tissues and labeled migrasomes with WGA-488. The results revealed that CRC cells emitted a substantial number of filopodia and small vesicles adorned with WGA-488 (Fig. 1C), thereby validating that primary CRC cells are capable of secreting migrasomes. Next, we investigated and identified migrasomes derived from CRC cells in vitro. As expected, prominent WGA-488-labeled retraction fibers and associated migrasomes were observed in the CRC cell lines CT26, HCT116, and MC38 (Fig. 1D). Additionally, stable expression of the migrasome marker TSPAN4-GFP via lentiviral transduction in the P53R CRC cell line further confirmed the presence of migrasomes (Fig. 1D). Since hypoxia is a hallmark of solid tumors and a key driver of tumor metastasis [25, 26], we explored the formation of migrasomes under hypoxic conditions. Notably, hypoxic conditions markedly increased the number of migrasomes compared to normoxia (Fig. 1E), indicating that hypoxia promote migrasome formation in CRC cells. Subsequently, we performed WB analysis to examine the expression of migrasome markers in migrasomes derived from P53R cells. The results demonstrated that these vesicles expressed the migrasome-specific marker EOGT, CPQ, and PIGK, but not the exosomal marker CD9, TSG101, and CD63 (Fig. 1F), confirming the successful isolation of migrasomes. Hypoxic migrasomes promote CRC liver metastasis What is the impact of migrasomes produced by CRC under a hypoxic microenvironment on the tumor microenvironment and the tumor itself? To address this question, we treated normal C57BL/6J mice with migrasomes (Fig. 2A). In vivo tracking showed that DIR-labeled migrasomes could accumulate in the liver (Fig. 2B). Next, a CRC liver metastasis model was established via intrasplenic injection of tumor cells, followed by tail vein administration of migrasomes produced by MC38 cells under normoxic and hypoxic conditions (Fig. 2C). Since the MC38 cell line demonstrated reliable capacity for generating splenic liver metastasis models whereas CT26 proved unsuccessful in model establishment, we selected MC38 for all subsequent experimental procedures. In vivo bioluminescence imaging and fluorescence statistics revealed a predominant metastatic burden in the liver in the hypoxic CRC migrasomes group (Fig. 2D and E). Ex vivo examination of liver tissues showed that, compared to normoxic migrasomes, hypoxic CRC migrasomes significantly promoted CRC metastasis to the liver (Fig. 2F). Overall, these results indicate that hypoxic CRC migrasomes promote CRC liver metastasis. Fig. 2Hypoxic migrasomes promote CRC liver metastasis in vivo. (A) Schematic diagram illustrating the experimental workflow of labeling normoxic and hypoxic CRC-derived migrasomes with DiR dye and injecting them into normal C57BL/6J mice to evaluate in vivo biodistribution. (B) Representative in vivo fluorescence imaging of whole-body mice after administration of DiR-labeled normoxic (Nor-Mig) and hypoxic CRC-derived migrasomes (Hyp-Mig). (C) Timeline of CRC liver metastasis model establishment using MC38 cells followed by tail intravenous injection (IV.) with normoxic or hypoxic migrasomes. (D) In vivo bioluminescence imaging showing liver metastases in mice, with corresponding quantification of fluorescence intensity (E). (F) Representative images of ex vivo liver tissues showing metastatic tumor nodules (white arrows) and quantification of visible tumor numbers per liver. n = 5, ** p < 0.01 Hypoxic migrasomes promote CRC liver metastasis in vivo. (A) Schematic diagram illustrating the experimental workflow of labeling normoxic and hypoxic CRC-derived migrasomes with DiR dye and injecting them into normal C57BL/6J mice to evaluate in vivo biodistribution. (B) Representative in vivo fluorescence imaging of whole-body mice after administration of DiR-labeled normoxic (Nor-Mig) and hypoxic CRC-derived migrasomes (Hyp-Mig). (C) Timeline of CRC liver metastasis model establishment using MC38 cells followed by tail intravenous injection (IV.) with normoxic or hypoxic migrasomes. (D) In vivo bioluminescence imaging showing liver metastases in mice, with corresponding quantification of fluorescence intensity (E). (F) Representative images of ex vivo liver tissues showing metastatic tumor nodules (white arrows) and quantification of visible tumor numbers per liver. n = 5, ** p < 0.01 Single-cell transcriptomic profiling of CRC liver metastases following normoxic or hypoxic migrasome treatment To investigate the cellular landscape and transcriptional alterations in CRC liver metastases following treatment with normoxic or hypoxic CRC-derived migrasomes, we performed scRNA-seq on tumor tissues harvested from the established intrasplenic injection-induced CRC liver metastasis model. Quality control metrics of the scRNA-seq data are presented in Supplemental Fig. 2A. A negative correlation was observed between percentage of mitochondrial gene expression and nCount-RNA (r = -0.25), while nFeature-RNA and nCount-RNA showed a strong positive correlation (r = 0.96), indicating high-quality and consistent transcriptomic profiles Supplemental Fig. 2B. Unsupervised clustering analysis using UMAP identified 17 distinct cell clusters across all samples (Supplemental Fig. 2C). Subsequent cell-type annotation classified these clusters into six major cell populations: epithelial cells, endothelial cells, fibroblasts, myeloid cells, T cells, and a small fraction of hepatocytes (Fig. 3A). The top five DEGs for each cell type are illustrated in the bubble plot (Supplemental Fig. 2D), demonstrating clear cell-type-specific transcriptional signatures. Cellular composition analysis revealed that epithelial cells constituted the predominant cell type, followed by myeloid cells and fibroblasts (Fig. 3B). Differential gene expression analysis between the two groups identified 173 upregulated and 142 downregulated DEGs in the hypoxic migrasome-treated group relative to the normoxic group (Fig. 3C). The top five upregulated DEGs in the hypoxic migrasome-treated group, including CSN3, CD55, AQP5, SLIT2, and AR, are highlighted in the bubble plots (Fig. 3D), reflecting distinct gene expression changes induced by hypoxic migrasome exposure. GO enrichment analysis of DEGs revealed significant enrichment in the biological process “regulation of epithelial cell proliferation,” suggesting enhanced proliferative signaling in the metastatic microenvironment (Fig. 3E). Consistently, KEGG pathway analysis identified robust enrichment of oncogenic signaling cascades, including the PI3K-AKT, MAPK, and JAK-STAT signaling pathways, indicating that hypoxic migrasomes may facilitate tumor progression through the activation of multiple pro-survival and pro-proliferative pathways (Fig. 3F). In short, these findings suggest that hypoxia-induced migrasomes alter the cellular composition of CRC liver metastatic tumor tissues and activation of oncogenic signaling pathways. Fig. 3Single-cell transcriptomic profiling of CRC liver metastases following normoxic or hypoxic migrasome treatment. A liver metastasis mouse model was established using MC38 cells, and the mice were treated with normoxic (Nor-Mig) or hypoxic migrasomes (Hyp-Mig). CRC liver metastatic tumor tissues from the two groups were then subjected to scRNA-seq. (A) In tumor tissues, cell-type annotation identified six cell types, including epithelial cells, endothelial cells, fibroblasts, myeloid cells, T cells, and hepatocytes. (B) Pie chart summarizing the overall cell-type composition, with epithelial cells being the most abundant, followed by myeloid cells and fibroblasts. (C) Volcano plot of DEGs between hypoxic and normoxic migrasome treatment groups. Red and blue dots represent upregulated and downregulated genes in the hypoxic group, respectively. (D) Bubble plot illustrating the expression distribution of the top five DEGs between the two groups. (E–F) Gene enrichment analysis of DEGs in liver metastatic tumors treated with hypoxic versus normoxic CRC-derived migrasomes. (E) Gene Ontology (GO) enrichment analysis of DEGs reveals significant enrichment in biological processes including the regulation of epithelial cell proliferation. (F) KEGG pathway enrichment analysis shows significant enrichment in signaling pathways such as PI3K-AKT, MAPK, and JAK-STAT Single-cell transcriptomic profiling of CRC liver metastases following normoxic or hypoxic migrasome treatment. A liver metastasis mouse model was established using MC38 cells, and the mice were treated with normoxic (Nor-Mig) or hypoxic migrasomes (Hyp-Mig). CRC liver metastatic tumor tissues from the two groups were then subjected to scRNA-seq. (A) In tumor tissues, cell-type annotation identified six cell types, including epithelial cells, endothelial cells, fibroblasts, myeloid cells, T cells, and hepatocytes. (B) Pie chart summarizing the overall cell-type composition, with epithelial cells being the most abundant, followed by myeloid cells and fibroblasts. (C) Volcano plot of DEGs between hypoxic and normoxic migrasome treatment groups. Red and blue dots represent upregulated and downregulated genes in the hypoxic group, respectively. (D) Bubble plot illustrating the expression distribution of the top five DEGs between the two groups. (E–F) Gene enrichment analysis of DEGs in liver metastatic tumors treated with hypoxic versus normoxic CRC-derived migrasomes. (E) Gene Ontology (GO) enrichment analysis of DEGs reveals significant enrichment in biological processes including the regulation of epithelial cell proliferation. (F) KEGG pathway enrichment analysis shows significant enrichment in signaling pathways such as PI3K-AKT, MAPK, and JAK-STAT Hypoxic CRC-derived migrasomes promote the expansion of Tmem45a⁺ fibroblasts with myofibroblast-like properties Fibroblasts are critical mediators of tumor metastasis, as they contribute to extracellular matrix remodeling, immune modulation, and the establishment of a pro-metastatic microenvironment [27]. To further investigate the heterogeneity and functional relevance of fibroblasts in CRC liver metastases, we performed subclustering analysis of fibroblast populations. As shown in the UMAP plot (Fig. 4A), five distinct fibroblast subtypes were identified. The cluster-specific heatmap revealed that the Kif20a⁺ subtype was characterized by high expression of Kif20a, Cenpa, Cenpf, Cdc20, and Ccnb2, whereas the Tmem45a⁺ subtype exhibited enriched expression of Tmem45a, Sulf1, Cxcl12, Cdh11, and Bmp5 (Fig. 4B). Quantification of cell abundance showed that the proportion of Tmem45a⁺ fibroblasts was markedly increased in the hypoxic migrasome-treated group, whereas the other four fibroblast subtypes decreased in abundance (Fig. 4C, Supplemental Fig. 3A), suggesting that hypoxia-induced migrasomes preferentially promote the expansion of the Tmem45a⁺ fibroblasts subtype. Cell–cell communication analysis using a chordal network mapping based on ligand-receptor flux quantification indicated that fibroblasts communicated with epithelial tumor cells and macrophages primarily through the PDGF signaling pathway (Fig. 4D). Furthermore, GSVA revealed significant activation of the “adipogenesis”, and “complement signaling pathway” in the Tmem45a⁺ fibroblast (Fig. 4E). To further characterize these fibroblast subsets, we assessed the expression of the myofibroblast marker α-SMA (Acta2). Notably, α-SMA expression was significantly elevated in the Tmem45a⁺ subtype compared to other fibroblast subsets (Fig. 4F), indicating that this population may exhibit an activated, pro-tumorigenic myofibroblast-like phenotype. Consistently, pseudotime trajectory analysis further supported the notion that Tmem45a⁺ fibroblasts represent a terminally differentiated state within the fibroblast lineage. As illustrated in trajectory topology mapping, these cells were predominantly located at the endpoint of the differentiation trajectory (Fig. 4G and H), suggesting a mature and potentially functionally distinct phenotype. Notably, Tmem45a⁺ fibroblasts selectively expressed high levels of genes associated with myofibroblast transdifferentiation, including COL15A1 and Grem1 (Fig. 4I). These findings reinforce the hypothesis that Tmem45a⁺ fibroblasts may play a critical role in remodeling the tumor microenvironment and promoting CRC liver metastasis through a myofibroblast-like phenotype. Fig. 4Hypoxic CRC-derived migrasomes promote the expansion of Tmem45a⁺ fibroblasts with myofibroblast-like properties. (A) UMAP plot showing five fibroblast subpopulations identified by subclustering analysis. (B) Heatmap displaying the expression profiles of representative marker genes across the five fibroblast clusters. The Kif20a⁺ cluster showed high expression of cell cycle-related genes (Kif20a, Cenpa, Cenpf, Cdc20, Ccnb2), while the Tmem45a⁺ cluster was enriched for Tmem45a, Sulf1, Cxcl12, Cdh11, and Bmp5. (C) Bar plot showing the relative abundance of each fibroblast subset in the normoxic and hypoxic migrasome groups. The Tmem45a⁺ fibroblasts were significantly enriched in the hypoxic group. CY: normoxic migrasomes treatment group (Nor-Mig); DY: hypoxic migrasomes treatment group (Hyp-Mig). (D) Chord plot illustrating predicted intercellular communication networks, revealing PDGF signaling between fibroblasts and both tumor epithelial cells and macrophages. (E) GSVA analysis indicating enrichment of the complement signaling pathway in the Tmem45a⁺ fibroblast subset. (F) Violin plot showing elevated expression of α-SMA (Acta2) in Tmem45a⁺ fibroblasts, suggesting their activated myofibroblast-like phenotype. (G–H) Pseudotime trajectory analysis demonstrating that Tmem45a⁺ fibroblasts are primarily localized at the terminal differentiation state. (G) The trajectory from pseudotemporal ordering, with the dark-to-light color scheme representing the inferred progression of cell states. (H) The composition and distribution of fibroblast subtypes across the pseudotime continuum. (I) Feature plots showing selective high expression of COL15A1 and Grem1 in Tmem45a⁺ fibroblasts, supporting their involvement in myofibroblast transdifferentiation Hypoxic CRC-derived migrasomes promote the expansion of Tmem45a⁺ fibroblasts with myofibroblast-like properties. (A) UMAP plot showing five fibroblast subpopulations identified by subclustering analysis. (B) Heatmap displaying the expression profiles of representative marker genes across the five fibroblast clusters. The Kif20a⁺ cluster showed high expression of cell cycle-related genes (Kif20a, Cenpa, Cenpf, Cdc20, Ccnb2), while the Tmem45a⁺ cluster was enriched for Tmem45a, Sulf1, Cxcl12, Cdh11, and Bmp5. (C) Bar plot showing the relative abundance of each fibroblast subset in the normoxic and hypoxic migrasome groups. The Tmem45a⁺ fibroblasts were significantly enriched in the hypoxic group. CY: normoxic migrasomes treatment group (Nor-Mig); DY: hypoxic migrasomes treatment group (Hyp-Mig). (D) Chord plot illustrating predicted intercellular communication networks, revealing PDGF signaling between fibroblasts and both tumor epithelial cells and macrophages. (E) GSVA analysis indicating enrichment of the complement signaling pathway in the Tmem45a⁺ fibroblast subset. (F) Violin plot showing elevated expression of α-SMA (Acta2) in Tmem45a⁺ fibroblasts, suggesting their activated myofibroblast-like phenotype. (G–H) Pseudotime trajectory analysis demonstrating that Tmem45a⁺ fibroblasts are primarily localized at the terminal differentiation state. (G) The trajectory from pseudotemporal ordering, with the dark-to-light color scheme representing the inferred progression of cell states. (H) The composition and distribution of fibroblast subtypes across the pseudotime continuum. (I) Feature plots showing selective high expression of COL15A1 and Grem1 in Tmem45a⁺ fibroblasts, supporting their involvement in myofibroblast transdifferentiation ScRNA-seq reveals a novel subtype of CD5L+ macrophage whose efferocytotic function response to hypoxic CRC migrasomes induction Given that myeloid cells represent the second most abundant cell population in the tumor microenvironment after epithelial tumor cells (Fig. 3B), we further focused on characterizing the composition of the myeloid compartment. As a result of reclustering analysis of the myeloid compartment, a total of 10 distinct cell subtypes were identified (Fig. 5A). The expression heatmap of representative marker genes for each myeloid subset is shown in Fig. 5B. We found that the CD5L+ macrophage subset exhibited the most markedly increase following hypoxic migrasome treatment, with an approximately eightfold enrichment compared to the normoxic group (Fig. 5C, Supplemental Fig. 3B). Therefore, we focused on the CD5L+ macrophage subtype. GSVA revealed significant activation of the “coagulation” and “complement signaling pathway” within the CD5L⁺ macrophages (Fig. 5D). Cell-cell communication analysis using a chord diagram demonstrated that CD5L+ macrophage subset potentially interacts with fibroblasts, T cells, and other immune cell populations through activation of the GDF signaling network (Fig. 5E). Furthermore, CD5L⁺ macrophages specifically exhibited high expression of IGF1, suggesting activation of the IGF signaling pathway, which mediate intercellular communication with tumor cells (Fig. 5F and G). CD5L specifically highly expressed in CD5L⁺ macrophages also observed (Fig. 5H). Notably, the migrasome marker TSPAN4 was found to be expressed across distinct macrophage subpopulations (Fig. 6A and Supplemental Fig. 3C-3D). Other migrasome marker such as NDST1 and CD81 were also found to be expressed in CD5L+ macrophage subpopulations (Supplemental Fig. 3E). Subsequently, functional analysis revealed that the CD5L+ macrophage subtype exhibited the strongest efferocytotsis score among all the 10 myeloid cell subtypes (Fig. 6B). Indeed, CD5L+ macrophage specifically highly expressed efferocytosis marker such as CD300LB, MERTK, and CD300LD (Fig. 6C), indicating the potential involvement in efferocytic activity. Fig. 5Characterization of myeloid cell subsets and identification of CD5L⁺ macrophages as potential migrasome-responsive cells. (A) UMAP visualization showing the reclustering of myeloid cells into 10 distinct subpopulations. (B) Heatmap displaying the expression profiles of representative marker genes for each myeloid subset. (C) Bar plot showing the relative abundance of each myeloid subpopulation in normoxic versus hypoxic migrasome-treated groups. A notable 8-fold increase in the proportion of CD5L⁺ macrophages was observed in the hypoxic group. (D) Gene Set Variation Analysis (GSVA) indicating significant activation of the complement signaling pathway in CD5L⁺ macrophages. (E) Chord diagram depicting inferred cell–cell communication networks, highlighting potential interactions between CD5L⁺ macrophages and fibroblasts, T cells, and other immune cells via GDF signaling. (F) Activation IGF signaling pathway network mediated crosstalk between CD5L⁺ macrophages and epithelial tumor cells. (G) In the IGF signaling pathway network, CD5L⁺ macrophages exhibit specific overexpression of IGF1. (H) Among the six macrophage cell-types, CD5L⁺ macrophages exhibit specific overexpression of CD5L Characterization of myeloid cell subsets and identification of CD5L⁺ macrophages as potential migrasome-responsive cells. (A) UMAP visualization showing the reclustering of myeloid cells into 10 distinct subpopulations. (B) Heatmap displaying the expression profiles of representative marker genes for each myeloid subset. (C) Bar plot showing the relative abundance of each myeloid subpopulation in normoxic versus hypoxic migrasome-treated groups. A notable 8-fold increase in the proportion of CD5L⁺ macrophages was observed in the hypoxic group. (D) Gene Set Variation Analysis (GSVA) indicating significant activation of the complement signaling pathway in CD5L⁺ macrophages. (E) Chord diagram depicting inferred cell–cell communication networks, highlighting potential interactions between CD5L⁺ macrophages and fibroblasts, T cells, and other immune cells via GDF signaling. (F) Activation IGF signaling pathway network mediated crosstalk between CD5L⁺ macrophages and epithelial tumor cells. (G) In the IGF signaling pathway network, CD5L⁺ macrophages exhibit specific overexpression of IGF1. (H) Among the six macrophage cell-types, CD5L⁺ macrophages exhibit specific overexpression of CD5L Fig. 6Phagocytic activity and immunofluorescence validation of CD5L⁺ macrophages in CRC liver metastases following migrasome treatment. (A) UMAP blot showing the expression of migrasome marker TSPAN4 in myeloid subsets. (B) Boxplot showing efferocytosis scores across the 10 identified myeloid cell subtypes. (C) Violin plots depicting the expression of efferocytosis markers CD300B, MERTK, and CD300D across 10 distinct myeloid cell subtypes. (D) UMAP plots displaying the expression patterns of three efferocytosis-associated marker genes specifically enriched in CD5L⁺ macrophages. (E) Immunofluorescence staining of tumor tissues from MC38-tumor bearing mice showing colocalization of CD163, CD5L, and the migrasome marker MERTK in both treatment groups. Increased MERTK expression is observed in the hypoxic group, indicating enhanced migrasome targeting of CD5L⁺ macrophages Phagocytic activity and immunofluorescence validation of CD5L⁺ macrophages in CRC liver metastases following migrasome treatment. (A) UMAP blot showing the expression of migrasome marker TSPAN4 in myeloid subsets. (B) Boxplot showing efferocytosis scores across the 10 identified myeloid cell subtypes. (C) Violin plots depicting the expression of efferocytosis markers CD300B, MERTK, and CD300D across 10 distinct myeloid cell subtypes. (D) UMAP plots displaying the expression patterns of three efferocytosis-associated marker genes specifically enriched in CD5L⁺ macrophages. (E) Immunofluorescence staining of tumor tissues from MC38-tumor bearing mice showing colocalization of CD163, CD5L, and the migrasome marker MERTK in both treatment groups. Increased MERTK expression is observed in the hypoxic group, indicating enhanced migrasome targeting of CD5L⁺ macrophages To further elucidate the transcriptional differences between CD5L⁺ macrophages and other macrophage subsets (other group), we performed a DEG analysis (Supplemental Fig. 4A). Compared with the other group, CD5L⁺ macrophages exhibited 690 significantly upregulated and 492 significantly downregulated genes (Supplemental Fig. 4B). The top five most upregulated genes in CD5L⁺ macrophages were CLEC4F, PAQR9, C6, CD5L, and VSIG4 (Supplemental Fig. 4C). Notably, a large number of efferocytosis-related genes were highly expressed in CD5L⁺ macrophages, including MERTK, NR1H, C4B, CD36, ARHGAP25, and PLA2G15 (Fig. 6D and Supplemental Fig. 4D). GO enrichment analysis of biological processes revealed a significant enrichment of the term “phagocytosis,” (Supplemental Fig. 4E) and KEGG pathway analysis identified “efferocytosis” as a significantly enriched pathway (Supplemental Fig. 4F). These findings indicate that CD5L⁺ macrophages may possess a distinct transcriptional profile that endows them with enhanced efferocytic potential. Collectively, these findings led us to hypothesize that CD5L⁺ macrophages may serve as recipient cells for hypoxic CRC cell-derived migrasomes and that stimulation by hypoxic migrasomes may significantly enhance the efferocytic capacity of CD5L⁺ macrophages. To validate this hypothesis, we performed IF staining on tumor tissues from murine CRC liver metastases both hypoxic and normoxic migrasomes groups using antibodies against CD163 and CD5L to label CD5L⁺ macrophages, along with an antibody against the efferocytosis marker MERTK. The results showed clear colocalization of CD163, CD5L, and the efferocytosis marker MERTK in tumor tissues from both normoxic and hypoxic migrasome-treated groups (Fig. 6E). Notably, CD5L and MERTK expression was markedly elevated in the hypoxic group compared to the normoxic group (Fig. 6E), suggesting that hypoxia-induced migrasomes provoked efferocytotic function of CD5L⁺ macrophages. Taken together, scRNA-seq has identified a previously unrecognized subtype of CD5L-positive macrophages, which exhibit an efferocytotic function in response to the induction of hypoxic migrasomes in CRC. Hypoxic CRC-derived migrasomes promote the differentiation and efferocytosis of CD5L⁺ macrophages Given the findings in Fig. 6 suggesting that CD5L⁺ macrophages with enhanced efferocytosis score may serve as recipient cells for CRC-derived migrasomes, we further investigated whether hypoxic CRC-derived migrasomes could induce the differentiation and efferocytic function of this macrophage subset. To this end, mouse bone marrow cells were isolated and differentiated into macrophages in vitro. These macrophages were then treated with MC38-derived migrasomes for 24 h. Flow cytometry analysis revealed that treatment with hypoxic CRC migrasomes significantly increased the proportion of CD5L⁺ macrophages compared with normoxic controls (Fig. 7A), indicating a potent role in promoting CD5L⁺ subset differentiation. To determine whether these induced CD5L⁺ macrophages also exhibited enhanced efferocytic capacity, we sorted the CD5L⁺ macrophage population by flow cytometry (Fig. 7B) and co-cultured them with apoptotic PI-labeled MC38 tumor cells. IF staining for F4/80 demonstrated that CD5L⁺ macrophages exposed to hypoxic CRC migrasomes showed a markedly increased uptake of apoptotic tumor cells, evidenced by internalized red fluorescence (Fig. 7C-D). Collectively, these findings suggest that hypoxic CRC-derived migrasomes promote both the differentiation of CD5L⁺ macrophages and their efferocytic activity. Fig. 7Hypoxic CRC-derived migrasomes promote the differentiation and efferocytosis of CD5L⁺ macrophages. (A) Flow cytometry analysis of the proportion of CD5L⁺ macrophages after treatment of normoxic or hypoxic CRC-derived migrasomes for 24 h. (B) Flow cytometric gating strategy for the isolation of CD5L⁺ macrophages. (C) Immunofluorescence analysis of efferocytosis by CD5L⁺ macrophages. Sorted CD5L⁺ macrophages were co-cultured with PI-labeled apoptotic MC38 tumor cells after migrasome treatment. F4/80 (green) marks macrophages; PI (red) indicates apoptotic tumor cells. Representative images show enhanced uptake of apoptotic cells by CD5L⁺ macrophages in the hypoxic migrasome-treated group. Scare bar: 50 μm. (D) The efferocytosis ability of CD5L⁺ macrophage was quantified by phagocytic index. ns p > 0.05, ** p < 0.01 Hypoxic CRC-derived migrasomes promote the differentiation and efferocytosis of CD5L⁺ macrophages. (A) Flow cytometry analysis of the proportion of CD5L⁺ macrophages after treatment of normoxic or hypoxic CRC-derived migrasomes for 24 h. (B) Flow cytometric gating strategy for the isolation of CD5L⁺ macrophages. (C) Immunofluorescence analysis of efferocytosis by CD5L⁺ macrophages. Sorted CD5L⁺ macrophages were co-cultured with PI-labeled apoptotic MC38 tumor cells after migrasome treatment. F4/80 (green) marks macrophages; PI (red) indicates apoptotic tumor cells. Representative images show enhanced uptake of apoptotic cells by CD5L⁺ macrophages in the hypoxic migrasome-treated group. Scare bar: 50 μm. (D) The efferocytosis ability of CD5L⁺ macrophage was quantified by phagocytic index. ns p > 0.05, ** p < 0.01 CRC-derived migrasomes transfer NRP2 to macrophages To elucidate the molecular mechanisms by which hypoxic CRC-derived migrasomes promote CD5L⁺ macrophage differentiation and efferocytosis, we analyzed proteomic data of CRC-derived migrasomes from the ProteomeXchange database. A total of 601 proteins were identified to be highly enriched in migrasomes (Fig. 8A). By intersecting these proteins with efferocytosis-associated proteins, we identified two candidates—Neuropilin-2 (NRP2) and Calreticulin (CALR)—that were enriched in migrasomes (Fig. 8A). WB analysis revealed that both CALR and NRP2 were significantly upregulated in hypoxic CRC-derived migrasomes compared to normoxic controls under equal cell loading (Fig. 8B). Importantly, this trend remained consistent regardless of whether GAPDH or the migrasome marker CPQ was used as an internal reference, supporting the robustness of the normalization. Furthermore, coefficient of variation analysis demonstrated comparable variability between GAPDH- and CPQ-normalized data, indicating that subcellular localization differences of GAPDH did not introduce normalization artifacts. However, only NRP2, but not CALR, was significantly upregulated in macrophages after co-incubation with hypoxic CRC-derived migrasomes (Fig. 8C), suggesting a potential functional role for NRP2 in hypoxic CRC-derived migrasome-mediated macrophage reprogramming. Moreover, clinical CRC tumor specimens confirmed that NRP2 expression was significantly upregulated in tumor tissues compared to adjacent non-tumorous tissues (Fig. 8D). As expected, analysis of the TCGA-COAD dataset revealed that patients with high NRP2 expression exhibited shorter overall survival (Supplemental Fig. 5A). Fig. 8CRC-derived migrasomes transfer NRP2 to macrophages. (A) Venn diagram showing the intersection between migrasome-enriched proteins and efferocytosis-associated proteins, identifying NRP2 and CALR. (B) Western blot analysis of NRP2 and CALR expression in migrasomes derived from normoxic and hypoxic CRC cells under equal cell conditions. GAPDH and the migrasome marker CPQ were used as internal references to evaluate normalization consistency. Quantitative analysis revealed comparable expression trends and similar coefficients of variation (CV) between GAPDH- and CPQ-normalized data. (C) Western blot analysis of NRP2 and CALR expression in macrophages following incubation with normoxic or hypoxic CRC-derived migrasomes. (D) Western blot analysis of NRP2 expression in clinical CRC tumor specimens. (E) Schematic diagram illustrating the experimental design for tracking NRP2-GFP–labeled migrasomes internalized by macrophages. (F) Representative immunofluorescence (IF) images showing colocalization of NRP2-GFP (green) and DiI-labeled migrasomes (red) in macrophages. Hoechst 33,342 stains nuclei (blue). Scare bar: 50 μm. (G) Flow cytometry (FCM) analysis of GFP-positive macrophages after co-incubation with NRP2-GFP–labeled migrasomes, indicating successful uptake. ns p > 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001 CRC-derived migrasomes transfer NRP2 to macrophages. (A) Venn diagram showing the intersection between migrasome-enriched proteins and efferocytosis-associated proteins, identifying NRP2 and CALR. (B) Western blot analysis of NRP2 and CALR expression in migrasomes derived from normoxic and hypoxic CRC cells under equal cell conditions. GAPDH and the migrasome marker CPQ were used as internal references to evaluate normalization consistency. Quantitative analysis revealed comparable expression trends and similar coefficients of variation (CV) between GAPDH- and CPQ-normalized data. (C) Western blot analysis of NRP2 and CALR expression in macrophages following incubation with normoxic or hypoxic CRC-derived migrasomes. (D) Western blot analysis of NRP2 expression in clinical CRC tumor specimens. (E) Schematic diagram illustrating the experimental design for tracking NRP2-GFP–labeled migrasomes internalized by macrophages. (F) Representative immunofluorescence (IF) images showing colocalization of NRP2-GFP (green) and DiI-labeled migrasomes (red) in macrophages. Hoechst 33,342 stains nuclei (blue). Scare bar: 50 μm. (G) Flow cytometry (FCM) analysis of GFP-positive macrophages after co-incubation with NRP2-GFP–labeled migrasomes, indicating successful uptake. ns p > 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001 To further determine whether migrasomes could mediate intercellular transfer of NRP2 into macrophages, CRC MC38 cells cultured under hypoxic conditions were transfected with NRP2-GFP, and the resulting migrasomes were labeled with DiI and co-cultured with BMDMs (Fig. 8E). IF revealed strong colocalization of NRP2-GFP (green) and DiI-labeled migrasomes (red) within BMDM, indicating internalization of NRP2-loaded migrasomes (Fig. 8F). Moreover, flow cytometry analysis showed a significantly increased proportion of GFP-positive macrophages in the NRP2-migrasome treatment group compared to the PBS control (Fig. 8G). These results suggest that hypoxic CRC-derived migrasomes can transfer NRP2 into macrophages, potentially mediating downstream effects. CRC-derived migrasomes transfer NRP2 to macrophages and promote efferocytosis-associated reprogramming To further validate whether CRC-derived migrasomes regulate CD5L⁺ macrophage differentiation and efferocytosis via NRP2 transfer, we performed NRP2 knockdown in MC38 cells under hypoxic conditions (Fig. 9A and B) and isolated migrasomes lacking NRP2. When these NRP2-deficient migrasomes were co-incubated with BMDMs, the proportion of CD5L⁺ macrophages were significantly reduced compared to the control group treated with wild-type migrasomes (Fig. 9C). Consistently, efferocytosis of apoptotic tumor cells by CD5L⁺ macrophages was also impaired in the absence of migrasomal NRP2 (Fig. 9D). Furthermore, mRNA and protein levels of key efferocytosis receptors—including MERTK, TYRO3, OLR1, CD36, AXL, and TIM3—were downregulated in macrophages treated with NRP2-deficient migrasomes, confirming the crucial role of NRP2 in mediating efferocytosis phenotype (Fig. 9E and F). Collectively, these results demonstrate that hypoxic CRC-derived migrasomes promote the differentiation and efferocytotic function of CD5L⁺ macrophages through the delivery of NRP2. Fig. 9Migrasomal NRP2 is required for CRC-induced CD5L⁺ macrophage differentiation and efferocytosis. (A) RT-qPCR analysis confirming efficient knockdown of NRP2 in MC38 cells under hypoxia. (B) Western blot analysis confirming efficient knockdown of NRP2 in MC38 cells under hypoxia. (C) Flow cytometry analysis of CD5L⁺ macrophage proportion after treatment with control or NRP2-deficient hypoxic migrasomes from MC38 cells. (D) Immunofluorescence assay of efferocytosis by CD5L⁺ macrophages following treatment with control or NRP2-deficient hypoxic migrasomes. F4/80 (green) labels macrophages; PI (red) labels apoptotic tumor cells. Scare bar: 50 μm. (E) Quantification of mRNA expression of efferocytosis receptors (MERTK, TYRO3, OLR1, CD36, AXL, and TIM3) in macrophages treated with control or NRP2-deficient migrasomes by RT-qPCR. (F) Quantification of protein expression of efferocytosis receptors (MERTK, TYRO3, OLR1, CD36, AXL, and TIM3) in macrophages treated with control or NRP2-deficient migrasomes by Western blot. * p < 0.05, ** p < 0.01 Migrasomal NRP2 is required for CRC-induced CD5L⁺ macrophage differentiation and efferocytosis. (A) RT-qPCR analysis confirming efficient knockdown of NRP2 in MC38 cells under hypoxia. (B) Western blot analysis confirming efficient knockdown of NRP2 in MC38 cells under hypoxia. (C) Flow cytometry analysis of CD5L⁺ macrophage proportion after treatment with control or NRP2-deficient hypoxic migrasomes from MC38 cells. (D) Immunofluorescence assay of efferocytosis by CD5L⁺ macrophages following treatment with control or NRP2-deficient hypoxic migrasomes. F4/80 (green) labels macrophages; PI (red) labels apoptotic tumor cells. Scare bar: 50 μm. (E) Quantification of mRNA expression of efferocytosis receptors (MERTK, TYRO3, OLR1, CD36, AXL, and TIM3) in macrophages treated with control or NRP2-deficient migrasomes by RT-qPCR. (F) Quantification of protein expression of efferocytosis receptors (MERTK, TYRO3, OLR1, CD36, AXL, and TIM3) in macrophages treated with control or NRP2-deficient migrasomes by Western blot. * p < 0.05, ** p < 0.01 NRP2 interacts with PROX1 to regulate CD5L⁺ macrophage differentiation and efferocytosis To further elucidate the molecular mechanism by which NRP2 regulates the differentiation and efferocytosis of CD5L⁺ macrophages, we predicted potential NRP2-interacting proteins using the STRING database, identifying 73 candidate partners (Supplemental Fig. 5B). Given that cell fate determination is largely governed by transcription factors (TFs), which may also exert transcriptional regulation over efferocytosis-related genes, we intersected the 73 NRP2-interacting genes with the TF list obtained from the AnimalTFDB database. The resulting Venn diagram revealed three overlapping proteins—NR2F2, PROX1, and SOX18 (Supplemental Fig. 5C). Expression analysis using GEPIA2 demonstrated that among these three candidates, only PROX1 was significantly upregulated in COAD tissues compared with controls (Supplemental Fig. 5D). Therefore, PROX1 was selected as a downstream interacting protein of NRP2 for subsequent analyses. To verify the physical interaction between NRP2 and PROX1, we performed Co-IP assays using an anti-NRP2 antibody in macrophages with or without hypoxic migrasome treatment (Fig. 10A). Western blot analysis showed that NRP2 efficiently enriched PROX1 in both conditions, confirming their binding interaction. Additionally, IF co-localization analysis provided direct spatial evidence supporting the physical association between NRP2 and PROX1 (Fig. 10B). To determine whether NRP2 regulates macrophage differentiation and efferocytosis through PROX1, we simultaneously overexpressed NRP2 and silenced PROX1 in macrophages. The efficiency of NRP2 overexpression and PROX1 knockdown was confirmed by RT-qPCR (Supplemental Fig. 5E–F). As expected, NRP2 overexpression markedly increased CD5L expression (Supplemental Fig. 5G) and elevated the proportion of CD5L⁺ macrophages (Fig. 10C), whereas PROX1 knockdown exerted the opposite effect. Moreover, PROX1 silencing partially abrogated the promotive effect of NRP2 overexpression. Similarly, RT-qPCR analysis showed that NRP2 overexpression enhanced the mRNA levels of efferocytosis-related genes AXL, MERTK, and TYRO3, whereas PROX1 knockdown reduced their expression; co-treatment mitigated these individual effects (Fig. 10D). IF staining for F4/80-positive macrophages and PI-MC38 apoptotic debris further supported that PROX1 knockdown not only significantly impaired the efferocytic capacity of CD5L⁺ macrophages but also partially attenuated the enhancing effect of NRP2 overexpression (Fig. 10E). Collectively, these findings reveal that NRP2 promotes CD5L⁺ macrophage differentiation and efferocytosis via a PROX1-dependent transcriptional mechanism. Fig. 10NRP2–PROX1 interaction promotes CD5L⁺ macrophage differentiation and enhances efferocytosis. (A) Co-immunoprecipitation (Co-IP) assays showing that NRP2 interacts with PROX1 in macrophages under normoxic and hypoxic migrasome-treated conditions. (B) Immunofluorescence co-localization images confirming the spatial association between NRP2 (green) and PROX1 (red) in macrophages. Nuclei were counterstained with DAPI (blue). Scale bar, 50 μm. (C) Flow cytometry analysis showing the proportion of CD5L⁺ macrophages following NRP2 overexpression and/or PROX1 knockdown. (D) RT-qPCR analysis of efferocytosis-related genes (AXL, MERTK, and TYRO3) in macrophages with indicated treatments. (E) Immunofluorescence staining of F4/80⁺ macrophages (green) engulfing PI-labeled apoptotic MC38 debris (red). Knockdown of PROX1 suppressed efferocytic activity and attenuated the NRP2-induced enhancement. Scale bar, 50 μm. (F) Representative fluorescence images and quantification of transwell assay. Fluorescently labeled CRC cells were co-cultured with macrophages overexpressing NRP2, MERTK-knockdown macrophages, or macrophages with combined NRP2 overexpression and MERTK knockdown, and CRC cell transmigration was assessed using a transwell assay. * p < 0.05, ** p < 0.01, *** p < 0.001 NRP2–PROX1 interaction promotes CD5L⁺ macrophage differentiation and enhances efferocytosis. (A) Co-immunoprecipitation (Co-IP) assays showing that NRP2 interacts with PROX1 in macrophages under normoxic and hypoxic migrasome-treated conditions. (B) Immunofluorescence co-localization images confirming the spatial association between NRP2 (green) and PROX1 (red) in macrophages. Nuclei were counterstained with DAPI (blue). Scale bar, 50 μm. (C) Flow cytometry analysis showing the proportion of CD5L⁺ macrophages following NRP2 overexpression and/or PROX1 knockdown. (D) RT-qPCR analysis of efferocytosis-related genes (AXL, MERTK, and TYRO3) in macrophages with indicated treatments. (E) Immunofluorescence staining of F4/80⁺ macrophages (green) engulfing PI-labeled apoptotic MC38 debris (red). Knockdown of PROX1 suppressed efferocytic activity and attenuated the NRP2-induced enhancement. Scale bar, 50 μm. (F) Representative fluorescence images and quantification of transwell assay. Fluorescently labeled CRC cells were co-cultured with macrophages overexpressing NRP2, MERTK-knockdown macrophages, or macrophages with combined NRP2 overexpression and MERTK knockdown, and CRC cell transmigration was assessed using a transwell assay. * p < 0.05, ** p < 0.01, *** p < 0.001 Next, we sought to elucidate how macrophage efferocytosis contributes to metastatic progression. Previous studies have shown that MERTK-mediated efferocytosis not only enhances the immunosuppressive phenotype of macrophages but also promotes tumor metastasis by shaping a suppressive tumor microenvironment [21]. In CRC, efferocytic activity of M2-like macrophages has been reported to suppress cytotoxic T cell function through inhibition of the STING/IFN-I pathway [22], while pharmacological blockade of the efferocytic receptor MERTK markedly suppresses tumor growth and liver metastasis [23]. These findings collectively highlight efferocytosis as a key driver of metastasis. To verify whether NRP2 promotes CRC metastasis through macrophage-mediated efferocytosis, we co-cultured fluorescently CFSE-labeled CRC cells with macrophages overexpressing NRP2 or with MERTK-knockdown macrophages, and assessed CRC cell transmigration using a transwell assay. Fluorescence imaging showed that NRP2 overexpression in macrophages markedly enhanced CRC cell migration, whereas MERTK knockdown significantly inhibited it. Moreover, co-manipulation of NRP2 overexpression and MERTK silencing in macrophages partially reversed the NRP2-induced enhancement of CRC migration (Fig. 10F, Supplemental Fig. 5H), indicating that NRP2-overexpressed macrophages promote CRC cell migration at least in part through a MERTK-dependent efferocytic mechanism. Given that the paracrine activity of macrophages is a major mechanism driving tumor metastasis [28], we next investigated whether NRP2-induced CD5L⁺ macrophages exerted additional paracrine functions beyond efferocytosis. To this end, Annexin V/PS liposomes were used to competitively block the “eat-me” signal recognition, thereby inhibiting efferocytosis of CD5L⁺ macrophages, and the conditioned media (CM) from different groups were collected (Supplemental Fig. 6A). Treatment of CRC cells with these CMs revealed that CM from CD5L⁺ macrophages significantly increased the invasion and migration of CRC cells compared with control CM (Supplemental Fig. 6B). Efferocytosis inhibition in CD5L⁺ macrophages reduced this promotive effect. Notably, CM from hypoxic migrasomes-treated combined with efferocytosis-inhibited CD5L⁺ macrophages partially rescued the invasion and migration of CRC cells.After the blockade of efferocytosis, hypoxia-induced migrasomes were still able to promote cell migration, indicating the presence of a pro-paracrine mechanism that enhances CRC migration independent of the efferocytosis pathway. Consistently, ELISA analysis showed that secretion of immunosuppressive cytokines IL-10, TGF-β, and CCL2 was markedly increased in CD5L⁺ macrophage CM compared with control CM. Inhibition of efferocytosis decreased the levels of these cytokines, whereas treatment with hypoxic migrasomes in the context of efferocytosis inhibition partially restored cytokine secretion (Supplemental Fig. 6C). To further confirm that hypoxic migrasomes regulate cytokine secretion in CD5L⁺ macrophages through the NRP2/PROX1 signaling axis, we examined IL-10, TGF-β, and CCL2 mRNA expression in macrophages transfected with NRP2 overexpression or PROX1 knockdown constructs. Compared with the control group, NRP2 overexpression significantly increased IL-10, TGF-β, and CCL2 expression, while PROX1 knockdown reduced their expression (Supplemental Fig. 6D). Simultaneous PROX1 knockdown partially reversed the cytokine upregulation induced by NRP2 overexpression, suggesting that NRP2/PROX1 axis also enhances the paracrine immunosuppressive function of CD5L⁺ macrophages. Discussion The liver is the predominant site of distant metastasis in CRC and represents the leading cause of CRC-related mortality. Although considerable progress has been made in understanding the molecular mechanisms driving CRC liver metastasis—including intestinal bacteria [29], cancer-associated fibroblasts [14], and immune modulation [30]—how the tumor microenvironment, particularly under hypoxic stress, shapes the metastatic niche remains incompletely understood. In particular, the dynamic crosstalk between tumor-derived migrasomes and immune cells during metastatic progression has not been fully elucidated. In this study, we demonstrate that hypoxia-induced CRC-derived migrasomes can transfer NRP2 to macrophages, thereby promoting the differentiation of a CD5L⁺ macrophage subpopulation with enhanced its efferocytic capacity through interaction with PROX1 (Fig. 11). These findings reveal a novel intercellular communication pathway mediated by migrasomes and suggest that targeting tumor–macrophage signaling via migrasome cargos such as NRP2 may provide new therapeutic opportunities for CRC liver metastasis. Fig. 11Graphical summary of the proposed mechanism. Schematic illustration summarizing the key findings of this study. Hypoxic CRC cells release migrasomes enriched with NRP2, which are preferentially taken up by macrophages in the liver metastatic microenvironment. Migrasomal NRP2 promotes the differentiation of CD5L⁺ macrophages and enhances their efferocytic activity. These CD5L⁺ macrophagescontribute to a pro-metastatic niche that facilitates CRC liver metastasis. Created with Figdraw (http://www.figdraw.com) for technical support (WIRYY15135) Graphical summary of the proposed mechanism. Schematic illustration summarizing the key findings of this study. Hypoxic CRC cells release migrasomes enriched with NRP2, which are preferentially taken up by macrophages in the liver metastatic microenvironment. Migrasomal NRP2 promotes the differentiation of CD5L⁺ macrophages and enhances their efferocytic activity. These CD5L⁺ macrophagescontribute to a pro-metastatic niche that facilitates CRC liver metastasis. Created with Figdraw (http://www.figdraw.com) for technical support (WIRYY15135) Here, we show that CRC migrasome–NRP2 axis orchestrates macrophage plasticity and contributes to the immunological remodeling of the liver metastatic microenvironment. Given the relatively recent discovery of migrasomes, their biological roles remain poorly understood, and to our knowledge, no prior studies have examined their involvement in CRC liver metastasis. Nevertheless, evidence from other malignancies has highlighted the critical role of migrasomes in promoting cancer progression and metastasis. For instance, elevated expression of CD151 has been shown to enhance migrasome biogenesis, and CD151-enriched migrasomes derived from hepatocellular carcinoma cells have been reported to promote tumor invasion and angiogenesis [16]. Notably, this study is the first to demonstrate that CRC-derived migrasomes can reprogram macrophage subpopulations. To date, only two studies have explored the crosstalk mediated by migrasomes in tumor microenvironment. Zhang et al. revealed that migrasomes derived from pancreatic cancer cells are enriched with various chemokines, cytokines, and signaling proteins, which drive macrophage polarization toward an immunosuppressive M2-like phenotype and promote tumor progression [31]. Another study, utilizing scRNA-seq and spatial transcriptomics, identified a subset of TSPAN4+ fibroblasts with high migrasome output that interact with multiple immune cell populations, including macrophages, thereby contributing to pancreatic cancer metastasis and immune modulation [32]. Collectively, these findings support our results that CRC-derived migrasomes promote reshaping the liver metastatic niche by directly reprogramming macrophage phenotypes. Our study identifies NRP2 as a critical migrasome cargo responsible for enhancing the differentiation and efferocytic activity of CD5L⁺ macrophages. NRP2 belongs to the neuropilin (NRP) family, a class of single-pass transmembrane receptors with short intracellular domains [33]. These NRPs lack intrinsic kinase activity and require co-receptors, such as VEGF receptors or plexins, to initiate downstream signaling. NRP2 has been extensively documented for its pivotal roles in angiogenesis, lymphangiogenesis, and axonal guidance [33]. Beyond its roles in vascular and neural systems, NRP2 has also been implicated in immune regulation. As reviewed by Schellenburg and colleagues, NRP2 contributes to the maturation of macrophages, activation of dendritic cells, and the development and activation of T lymphocytes [33]. Increasing evidence further supports the involvement of NRP2 in macrophage polarization; for instance, in the context of diabetic foot ulcers, FOXM1-mediated induction of M1 macrophage polarization has been shown to depend on the SEMA3C/NRP2/Hedgehog signaling axis [34]. Moreover, NRP2 isoforms are selectively upregulated in macrophage subsets and has been associated with phagocytic activity and breast cancer metastasis [35]. To our knowledge, however, this study is the first to describe a CD5L⁺ macrophage subpopulation with enhanced efferocytic capacity in the CRC metastatic liver microenvironment. Consequently, the role of NRP2 in regulating the differentiation and functional programming of CD5L⁺ macrophages has not been previously reported. Nevertheless, insights from existing literature on NRP2’s involvement in classical M1/M2 polarization and macrophage function provide indirect support for its broader immunomodulatory capacity. Our findings thus extend the current understanding of NRP2 beyond conventional macrophage subsets, highlighting its novel role in modulating the phenotypic and functional plasticity of a newly defined efferocytic macrophage population. It is noteworthy that NRP2 functions as a co-receptor for vascular endothelial growth factors C and D (VEGF-C/D), and previous studies have demonstrated that the NRP2/VEGF axis can contribute to tumor progression by regulating angiogenesis [36, 37]. In the context of our study, hypoxia-migrasomes of CRC promote metastasis primarily through inducing CD5L⁺ macrophage efferocytosis. While we did not investigate potential angiogenic effects of migrasomes in this work, the established role of NRP2 in endothelial signaling provides a foundation for future studies to explore whether migrasomes may also influence tumor vascularization. Moreover, our study identifies NRP2 as a key mediator of hypoxic CRC-derived migrasome-induced macrophage reprogramming and liver metastasis. These findings raise the possibility that therapeutic targeting of NRP2 or modulation of macrophage efferocytosis could inhibit metastatic progression. However, as NRP2 is also expressed in normal endothelial and lymphatic tissues and efferocytosis is essential for tissue homeostasis, systemic inhibition may result in off-target effects. Therefore, developing strategies for selective targeting, such as tumor-specific delivery systems, will be critical. Future preclinical studies are warranted to evaluate the efficacy and safety of such NRP2 interventions in CRC. In summary, our study uncovers a previously unrecognized mechanism by which hypoxia-induced CRC–derived migrasomes promote liver metastasis through the NRP2/PROX1 axis-mediated reprogramming of CD5L⁺ macrophages. We demonstrate that migrasomes function as intercellular communication vectors, transferring NRP2 to recipient CD5L⁺ macrophages, thereby enhancing their differentiation and efferocytic activity through interaction with PROX1. This study not only deepens the understanding of TAM heterogeneity in CRC but also highlights the migrasome–NRP2 axis as a novel driver of tumor–immune communication in the metastatic setting. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1: Supplemental Fig. 1. Single-cell transcriptomic analysis of GSE234804 dataset reveals migrasome signature distribution among cell clusters. (A) UMAP plot showing cell-type annotation of the GSE234804 scRNA-seq dataset, identifying six distinct cellular clusters. (B) Bubble plot displaying representative marker genes for each of the six annotated cell populations. (C) Migrasome signature scores across the six cell populations in both primary CRC and liver metastasis samples, revealing differential enrichment patterns. (D) Immunofluorescence was performed to examine the colocalization of the migrasome marker TSPAN4 with the CRC cell marker COX-2, the macrophage marker CD163, and the endothelial cell marker CD31 in CRC tumor tissues. Supplementary Material 1: Supplemental Fig. 1. Single-cell transcriptomic analysis of GSE234804 dataset reveals migrasome signature distribution among cell clusters. (A) UMAP plot showing cell-type annotation of the GSE234804 scRNA-seq dataset, identifying six distinct cellular clusters. (B) Bubble plot displaying representative marker genes for each of the six annotated cell populations. (C) Migrasome signature scores across the six cell populations in both primary CRC and liver metastasis samples, revealing differential enrichment patterns. (D) Immunofluorescence was performed to examine the colocalization of the migrasome marker TSPAN4 with the CRC cell marker COX-2, the macrophage marker CD163, and the endothelial cell marker CD31 in CRC tumor tissues. Supplementary Material 2: Supplemental Fig. 2. Quality control and correlation analysis of scRNA-seq data. (A) Distribution of key quality control metrics after quality control, including nFeature-RNA, nCount-RNA, and mitochondrial gene percentage. (B) Correlation plots showing a negative correlation between mitochondrial percentage and nCount-RNA (r = -0.25), and a strong positive correlation between nFeature-RNA and nCount-RNA (r = 0.96). (C) UMAP visualization of 17 distinct cell clusters identified from scRNA-seq analysis of liver metastatic tumor tissues. (D) Bubble plot showing the top 5 DEGs for each of the six major cell types. CY: normoxic migrasomes treatment group (Nor-Mig); DY: hypoxic migrasomes treatment group (Hyp-Mig). Supplementary Material 2: Supplemental Fig. 2. Quality control and correlation analysis of scRNA-seq data. (A) Distribution of key quality control metrics after quality control, including nFeature-RNA, nCount-RNA, and mitochondrial gene percentage. (B) Correlation plots showing a negative correlation between mitochondrial percentage and nCount-RNA (r = -0.25), and a strong positive correlation between nFeature-RNA and nCount-RNA (r = 0.96). (C) UMAP visualization of 17 distinct cell clusters identified from scRNA-seq analysis of liver metastatic tumor tissues. (D) Bubble plot showing the top 5 DEGs for each of the six major cell types. CY: normoxic migrasomes treatment group (Nor-Mig); DY: hypoxic migrasomes treatment group (Hyp-Mig). Supplementary Material 3: Supplementary Fig. 3. Expression profile of migrasome markers in macrophage subpopulations in scRNA-seq data. (A) Bar plot showing the proportions of different fibroblast subpopulations between the two groups. (B) Bar plot showing the proportions of different macrophage subpopulations between the two groups. (C) Bubble plot of migrasome marker TSPAN4 expression in all myeloid subpopulations. (D) Violin plot of migrasome marker TSPAN4 expression in all myeloid subpopulations. (E) Violin plot of expression of other migrasome markers in all myeloid subpopulations. Supplementary Material 3: Supplementary Fig. 3. Expression profile of migrasome markers in macrophage subpopulations in scRNA-seq data. (A) Bar plot showing the proportions of different fibroblast subpopulations between the two groups. (B) Bar plot showing the proportions of different macrophage subpopulations between the two groups. (C) Bubble plot of migrasome marker TSPAN4 expression in all myeloid subpopulations. (D) Violin plot of migrasome marker TSPAN4 expression in all myeloid subpopulations. (E) Violin plot of expression of other migrasome markers in all myeloid subpopulations. Supplementary Material 4: Supplemental Fig. 4. Transcriptomic profiling of CD5L⁺ macrophages reveals enhanced efferocytic signature compared to other macrophage subpopulations. (A) UMAP plot showing the distribution of CD5L⁺ macrophages and other macrophage subpopulations. (B) Volcano plot depicting differentially expressed genes (DEGs) between CD5L⁺ macrophages and other macrophage subsets. (C) Violin plots showing the expression levels of the top five most significantly upregulated and downregulated genes in CD5L⁺ macrophages. (D) Violin plots displaying the expression patterns of three efferocytosis-associated marker genes specifically enriched in CD5L⁺ macrophages. (E) GO enrichment analysis of DEGs in CD5L⁺ macrophages, highlighting significant enrichment in the biological process term “phagocytosis.” (F) KEGG pathway enrichment analysis of DEGs in CD5L⁺ macrophages, with the “efferocytosis” pathway prominently enriched. Supplementary Material 4: Supplemental Fig. 4. Transcriptomic profiling of CD5L⁺ macrophages reveals enhanced efferocytic signature compared to other macrophage subpopulations. (A) UMAP plot showing the distribution of CD5L⁺ macrophages and other macrophage subpopulations. (B) Volcano plot depicting differentially expressed genes (DEGs) between CD5L⁺ macrophages and other macrophage subsets. (C) Violin plots showing the expression levels of the top five most significantly upregulated and downregulated genes in CD5L⁺ macrophages. (D) Violin plots displaying the expression patterns of three efferocytosis-associated marker genes specifically enriched in CD5L⁺ macrophages. (E) GO enrichment analysis of DEGs in CD5L⁺ macrophages, highlighting significant enrichment in the biological process term “phagocytosis.” (F) KEGG pathway enrichment analysis of DEGs in CD5L⁺ macrophages, with the “efferocytosis” pathway prominently enriched. Supplementary Material 5: Supplementary Fig. 5. Identification of PROX1 as the downstream transcriptional partner of NRP2. (A) Kaplan–Meier analysis of overall survival (OS) in TCGA-COAD patients stratified by NRP2 expression levels. Patients with high NRP2 expression showed significantly shorter OS. (B) Protein–protein interaction (PPI) network of NRP2-interacting proteins predicted using the STRING database, revealing 73 candidate partners. (C) Venn diagram showing the overlap between NRP2-interacting proteins and transcription factors (TFs) obtained from the AnimalTFDB database, identifying NR2F2, PROX1, and SOX18 as potential common factors. (D) Expression levels of NR2F2, PROX1, and SOX18 in COAD tissues and normal controls analyzed using GEPIA2. Only PROX1 showed significant upregulation in COAD tissues. (E–F) Verification of NRP2 overexpression and PROX1 knockdown efficiency in macrophages by RT-qPCR. (G) CD5L mRNA expression levels in macrophages with NRP2 overexpression or PROX1 knockdown. (H) Verification of MERTK knockdown efficiency in macrophages by RT-qPCR. ** p < 0.01, *** p < 0.001. Supplementary Material 5: Supplementary Fig. 5. Identification of PROX1 as the downstream transcriptional partner of NRP2. (A) Kaplan–Meier analysis of overall survival (OS) in TCGA-COAD patients stratified by NRP2 expression levels. Patients with high NRP2 expression showed significantly shorter OS. (B) Protein–protein interaction (PPI) network of NRP2-interacting proteins predicted using the STRING database, revealing 73 candidate partners. (C) Venn diagram showing the overlap between NRP2-interacting proteins and transcription factors (TFs) obtained from the AnimalTFDB database, identifying NR2F2, PROX1, and SOX18 as potential common factors. (D) Expression levels of NR2F2, PROX1, and SOX18 in COAD tissues and normal controls analyzed using GEPIA2. Only PROX1 showed significant upregulation in COAD tissues. (E–F) Verification of NRP2 overexpression and PROX1 knockdown efficiency in macrophages by RT-qPCR. (G) CD5L mRNA expression levels in macrophages with NRP2 overexpression or PROX1 knockdown. (H) Verification of MERTK knockdown efficiency in macrophages by RT-qPCR. ** p < 0.01, *** p < 0.001. Supplementary Material 6: Supplemental Fig. 6. NRP2-mediated paracrine activity of CD5L⁺ macrophages promotes CRC cell migration and invasion. (A) Schematic representation of the experimental design. CD5L⁺ macrophages were treated with Annexin V/PS liposomes to inhibit efferocytosis, and conditioned media (CM) were collected from control macrophages, CD5L⁺ macrophages, efferocytosis-inhibited CD5L⁺ macrophages, and hypoxic migrasomes co-cultured with efferocytosis-inhibited CD5L⁺ macrophages. These CMs were used to stimulate CRCs in transwell assays. (B) ELISA analysis of IL-10, TGF-β, and CCL2 levels in the CM. (C) Transwell assays showing the migration and invasion of CRC cells treated with different CMs. (D) RT-qPCR analysis of IL-10, TGF-β, and CCL2 mRNA expression in macrophages with NRP2 overexpression, PROX1 knockdown, or combined NRP2 overexpression and PROX1 knockdown. *p < 0.05, **p < 0.01, ***p < 0.001. Supplementary Material 6: Supplemental Fig. 6. NRP2-mediated paracrine activity of CD5L⁺ macrophages promotes CRC cell migration and invasion. (A) Schematic representation of the experimental design. CD5L⁺ macrophages were treated with Annexin V/PS liposomes to inhibit efferocytosis, and conditioned media (CM) were collected from control macrophages, CD5L⁺ macrophages, efferocytosis-inhibited CD5L⁺ macrophages, and hypoxic migrasomes co-cultured with efferocytosis-inhibited CD5L⁺ macrophages. These CMs were used to stimulate CRCs in transwell assays. (B) ELISA analysis of IL-10, TGF-β, and CCL2 levels in the CM. (C) Transwell assays showing the migration and invasion of CRC cells treated with different CMs. (D) RT-qPCR analysis of IL-10, TGF-β, and CCL2 mRNA expression in macrophages with NRP2 overexpression, PROX1 knockdown, or combined NRP2 overexpression and PROX1 knockdown. *p < 0.05, **p < 0.01, ***p < 0.001. Supplementary Material 7 Supplementary Material 7 Supplementary Material 8 Supplementary Material 8