Gastric cancer-derived exosomal miR-519a-3p promotes liver metastasis by inducing intrahepatic M2-like macrophage-mediated angiogenesis

Tissue samples were collected from GC patients diagnosed by gastroscopy, puncture pathology and CT at First Affiliated Hospital of Nanjing Medical University (including 30 patients with stage III and 30 GC patients with liver metastases). The tissue samples mainly included gastric cancer tissue, normal paracancer tissue, liver metastasis tissue and normal liver tissue. The fresh tissue was frozen and stored in liquid nitrogen until use. Blood samples were collected preoperatively and on the seventh postoperative day from patients who underwent surgical treatment. In addition, healthy blood samples were collected from 15 volunteers without malignancy or acute/chronic disease. All blood samples were centrifuged at 2,500 g for 10 min to extract serum and then stored at -80 °C until use. Prior approval was obtained from the Nanjing Medical University Ethics Committee, and informed consent was obtained from each patient.

The human gastric epithelial cell line GES -1 and GC cell lines (validated by Short Tandem Repeat DNA fingerprinting) were purchased from the Cell Bank of the Type Culture Collection of the Chinese Academy. MKN45, MKN45-HL, HGC-27, and NCI-N87 were cultured in RPMI-1640 (w/o Hepes). KATOIII and AGS were cultured in Dulbecco's Modified Eagle's medium (DME H-21 4.5 g/liter glucose) and Nutrient Mixture F-12 K, respectively. To the nutrient medium, 10% fetal bovine serum (Gibco, USA), 1% antibiotics (HyClone, USA), and 0.1% mycoplasma antibiotics (Invitrogen, USA) were added. All cells were proliferated in an environmental incubator (humidified atmosphere with 5% CO₂, 37 °C).

Gastric cancer cells with high liver metastatic potential (MKN45-HL) were successfully established in our previous study [21]. In brief, 2 million MKN45 cells stably transfected with a double-labeled empty GFP-Flag lentivirus (Shanghai Gene Pharma) were injected into the hepatic portal vein of 6-week-old female BALB/c nude mice. 3 weeks later, the mice with liver metastases were sacrificed, and the livers were harvested and then processed into single cell suspensions. The labeled GC cells were sorted out from the single cell suspensions by flow cytometry (FC). The extracted GFP-positive MKN45 cells were then re-injected into the hepatic portal vein to obtain cells with a greater capacity for liver metastasis. After repeating the above steps three times, the flow cytometrically sorted GFP-positive cells were classified as GC cells with high liver metastatic potential.

MKN45 and MKN45-HL cells were cultured in normal medium until they were 80% confluent. Then the medium was replaced with exosome-depleted medium (RPMI1640 containing 10% exosome-depleted FBS). Two days later, the conditioned medium (approximately 15 ml) was collected from each dish and subjected to polymerization precipitation, ultrafiltration, and ultracentrifugation to harvest exosomes. Briefly, the conditioned medium was successively centrifuged at 1,00 × g for 10 min, 2,000 × g for 10 min, and 18,000 × g for 30 min to remove cell fragments, cell debris, and large extracellular vesicles (lEVs). The supernatant was then concentrated using an ultrafiltration tube (Merck Millipore) to < 3 mL (4,000 × g, 4 °C) to separate the exosome fraction from the soluble exosome fraction. Exosomes were then purified according to the experimental protocol of Total Exosome Isolation Reagent (Invitrogen; 4,478,359). The pellet was resuspended in particle-free PBS or stored at -80 °C in the refrigerator for subsequent use. For exosome purification from plasma, blood samples must be successively centrifuged at 150 g for 15 min and 1,200 g for 15 min to obtain PPP (platelet-poor plasma). The separation and concentration of exosomes in PPP were then similar to that of the cell-conditioned medium.

The NanoSight LM10 system (Malvern Instruments, UK) was used to analyze the amount and concentration of exosomes. Exosomes were loaded into the sample chamber of the LM10 unit, and the fast video recording and particle tracking software measured the speed of Brownian motion with the same detection threshold. The exosome concentration was calculated as an average value. In addition, the morphology of exosomes was determined by transmission electron microscopy (TEM, JEOL, Japan). The expression of the exosome markers CD81, TSG101, and the lEVs marker calnexin was detected by Western blot after measuring the protein concentration of exosomes using the BCA protein assay kit (ThermoFisher Scientific, USA). Information on antibodies detecting the above markers is provided in Table S1 in the supplemental material.

Exosome tracking assays were performed as previously described [21]. A total of 10 μg exosomes were first mixed with Diluent C buffer containing 4 μl PKH26/67 dye (Sigma, USA) for 5 min and then incubated with 1% BSA solution at room temperature. For exosome tracking assays in vivo, 10 μg PKH26-labelled exosomes resuspended in 100 μl PBS were injected into BALB/c nude mice (6 weeks old) via the tail vein. Twenty-four hours later, the mice were sacrificed, and the brain, lung, liver, spleen, and bone tissues were harvested for in vivo imaging. To determine whether exosomes were taken up by macrophages in vitro, 10 µg of PKH-26/67-labelled exosomes were resuspended in PBS solution and incubated with PMA-treated THP-1 cells at 37℃ for 4 h. THP-1 cells were then stained with DAPI and observed by confocal laser microscopy.

Exosomes were extracted from the conditioned medium of MKN45 and MKN45- HL cells. Total RNAs in exosomes were extracted after exosome characterization. Small RNA sequencing was performed in an Illumina HiSeq 2500 system from Huayin Health Co (Guangzhou, China). Briefly, 35 ng of exosome RNA was converted to cDNA after ligation with RNA 3' and 5' adapters and amplified with Illumina primers. Raw data were generated and then subjected to initial quality control to weed out sequences less than 16 nt in length. The miRBase database was used as a reference genome for human miRNA sequences. Raw data normalized by TPM were analyzed using the edgeR package (Bioconductor Software).

TRIzol reagent (Technologies, USA) was used to extract total RNA from GC cells and tissues. The miRNA and mRNA were transcribed into cDNA using the New Poly(A) Tailing Kit (ThermoFisher, USA) and PrimeScript RT Master Mix Kit (TaKaRa, Japan), respectively, according to the manufacturer's protocol. The above cDNAs were then amplified using the FastStart Universal SYBR Green Master Kit (Roche, Mannheim, Germany) and detected using the ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Waltham, MA, USA). The relative expression of miRNA and mRNA was normalized to U6 and β-actin, respectively. Primer information is shown in Table S2 in the supplemental material.

RIPA Lysis buffer (Beyotime, Shanghai, China) supplemented with protease and phosphatase inhibitors (NCM Biotech, Suzhou, China) was used to extract proteins from cells and tissues. SDS-PADE Protein buffer (Beyotime, Shanghai, China) was added to the lysate, and proteins were then denatured in a water bath at 100 °C for 5 min. Protein samples were transferred to polyvinylidene difluoride membranes (Thermo Fisher, MA, USA) after separation by SDS–polyacrylamide gel electrophoresis. The above membranes were blocked with 5% fast blocking buffer (Beyotime, Shanghai, China) for 45 min and then incubated with primary antibodies overnight at 4 °C in a refrigerator and then immunoblotted with secondary antibodies for 2 h. Finally, images of the blots on the membranes were acquired using a Tanon-4600 Imaging System (Biotanon, China). Information on the primary and secondary antibodies used in the Western blot is given in Table S1 in the supplemental material.

To detect CD206 + macrophages, PMA-treated THP-1 cells were first harvested and fixed overnight in 1% paraformaldehyde (PFA) at 4 °C. They were then resuspended in flow cytometry buffer (1 × PBS buffer containing 1% FSA) and stained with anti-CD206 (BD Biosciences, USA) for 30 min at room temperature.

To investigate the effect of exosomes on the expression of TGF-β, VEGFA, and VEGFD in THP-1 cells, enzyme-linked immunosorbent assay (ELISA) was performed using the TGF beta-1 Human ELISA kit, the vascular endothelial growth factor A Human ELISA kit, and the VEGF-D (FIGF) Human ELISA kit (Thermo Fisher Scientific, USA). Absorbance was measured at 490 nm using a spectrophotometer (Thermo Fisher Scientific, USA) within 20 min after completion of the reaction. Cytokine concentrations were scaled according to the standard curve.

Paraffin-embedded tissue blocks were cut into 2.5 μm sections and spread on glass slides. FISH was performed using miR-519a-3p detection probes (RiboBio, Guangzhou, China) and a fluorescence in situ hybridization kit (FISH) (Bosterbio, USA) according to the manufacturer's protocol. For immunohistochemistry (IHC), endogenous peroxidase activity was first blocked with 3% hydrogen peroxide, and then the sections were incubated with the primary antibody overnight at 4 °C. The next day, the sections were incubated with secondary antibody for 1 h at room temperature and then stained with 3,3-diaminobenzidine solution and hematoxylin. Antibody information is provided in Table S1 in the supplemental material.

For histological analysis, liver tissues embedded in paraffin were first baked at 60 °C for 1 h and then deparaffinized with xylene and anhydrous ethanol. The tissues were sequentially treated with Triton X-100, enhanced citrate antigen retrieval solution, and Enhanced Endogenous Peroxidase Blocking Buffer (Beyotime Biotechnology, China) to break the membrane, repair the antigen, and eliminate endogenous peroxidation. After blocking with QuickBlock™ for one hour, the tissues were incubated overnight at 4 °C with the primary antibody. The next day, tissues were incubated with fluorescently labeled secondary antibody for 2.5 h before staining nuclei with DAPI. Fluorescence images were acquired and analyzed using a Thunder Imager (LEICA, Germany) fast, high-resolution, inverted fluorescence imaging system.

The miR-519a-3p inhibitor/mimics or negative controls were purchased from GenePharma Co. Lentivirus vectors expressing miR-519a-3p and suppressing miR-519a-3p were designed and manufactured by ViGene Biosciences Co. In rescue experiments, PMA-treated THP-1 cells were transfected with miR-519a-3p mimics together with DUSP2 overexpression plasmid (ViGene Biosciences Co). Sequence information is shown in Table S2 in the supplemental material.

PMA-treated THP-1 cells were transfected with miR-519a-3p inhibitor and inhibitor-NC before total RNA was extracted with TRIzol (Life Technologies, USA). After quality control with Nanodrop 2000, total RNA was temporarily stored at -80 °C, and high-throughput mRNA sequencing was performed on the BGIseq 500 platform (BGI-Shenzhen, China).

The wild-type or mutant 3'UTR of DUSP2 was constructed into dual luciferase reporters using pmirGLO vector (Kinkairui, China), and 60 nM miR-519a-3p mimics or negative control (GenePharma, China) was cotransfected with 1 μg of the wild-type or mutant luciferase reporter plasmids in HEK293T cells. Relative luciferase activity was measured using a dual luciferase reporter assay kit according to the manufacturer's protocol (Promega, USA). Fluorescence signal was detected using the GloMax ® 20/20 instrument (Promega, USA).

Before performing cell function assays, an in vitro co-culture system was established. In brief, HUVECs (human umbilical vein endothelial cells) suspended in 200 ml medium were seeded in the upper chamber (diameter = 0.4 μm; Corning, USA), and exosome-incubated or plasmid-transfected THP-1 cells in 800 μl medium containing 10% FBS were added to the lower chamber. After 48 h of coculture, HUVECs were separated from the upper chamber and prepared for subsequent experiments. Transwell chambers (diameter = 8 μm; Corning, USA) were used to study the migration ability of HUVECs. After 24 h of incubation, cells were stained with 0.1% crystal violet for 30 min and 6 random 100 × fields of view were selected to count the number of migrating cells.

Matrigel (Corning, USA) was first thawed at 4 °C and spread into 96-well plates with 50 μl in each well and rested at 37 °C for 30 min to form a gel. The supernatant of PMA-treated THP-1 cells incubated with different exosomes was collected, and 1 × 10⁴ HUVECs were suspended in 100 μl of the above supernatant. After 6-h incubation, tube formation was photographed with a digital camera system (Olympus, Japan), and the total length and number of branches of the tubes were analyzed with Image J software.

The thoracic aorta was obtained from 6- to 8-week-old BALB/c nude mice in a sterile environment. The aortic rings (cut into 5-mm-long lengths) were placed in wells coated with Matrigel and incubated at 37 °C for 30 min to allow the Matrigel to polymerize. DMEM/F12-K medium (Gibco, USA) containing 10% fetal bovine serum and 1% antibiotics (HyClone, USA) supplemented with 10 ng/ml VEGF and 25 μg/ml heparin were then added to the wells. Twenty-four hours later, aortic rings were cultured with conditioned medium from PMA-treated THP-1 cells incubated with GC cells-derived exosomes or transfected with miR-519a-3p mimics, miR-519a-3p inhibitors, or DUSP2 overexpression plasmids. After one week of culture, vessel growth was quantified by microscopic counting of all buds on each aortic ring.

The animal experimental protocols of this study were reviewed and approved by the Animal Ethics Committee of Nanjing Medical University. All mice were purchased from the Laboratory Animal Centre of Nanjing Medical University and raised under pathogen-free conditions. The exosome-educated mouse models were established as described previously. In brief, mice were first randomly divided into several groups according to the requirements of the experimental design. 10 μg of exosomes from differentially treated MKN45 and MKN45-HL cells were resuspended in 150 μl PBS and then injected via the tail vein of the mice. This injection of exosomes via the tail vein was repeated every 2 days for 2 weeks, which we termed exosome education. At the end of exosome education, 2 million MKN45 cells transfected with luciferase lentivirus were injected into the mice via the spleen. Three weeks after spleen injection, the mice underwent in vivo imaging (Calliper Life Sciences, Hopkinton, MA, USA) and were subsequently sacrificed to harvest their livers for photographs and H&E staining.

For the subcutaneous tumor mouse model, 5-week-old female mice were randomly divided into 4 groups, and then 1 × 10⁷ GC cells stably knockdown- or overexpressed with miR-519a-3p were suspended in 150 μl PBS and then subcutaneously injected into the axilla of the forelimb of each mouse. The volume (V = length × width² × 0.5) of xenograft tumors was measured each week. The mice were sacrificed four weeks later, and the subcutaneous tumors were harvested, weighed, and photographed.

To extract intrahepatic macrophages from mice, mice educated with exosomes were sacrificed and hepatocyte-monocyte suspensions were prepared. Then, 2 mL liver monocyte suspension, 50% Percoll (Sigma, USA), 25% Percoll, and 1 mL PBS were slowly added to a 10 mL centrifuge tube and centrifuged at 1600 rpm for 25 min at 4 °C. After centrifugation, the white cell mass was collected between 50% Percoll and 25% Percoll in a new tube, and 10 mL of PBS was added. The tubes were mixed and centrifuged at 200 rpm for 7 min, and then the supernatant was discarded. Cells were inoculated into 10 cm cell cultures or total RNA, or protein was extracted directly from the cells for subsequent assays.

All experiments were repeated more than 3 times. Quantitative data were expressed as mean ± standard deviation. Student's t test was performed to analyze the statistical difference between two groups, and analysis of variance (ANOVA) was applied to evaluate the differences between multiple groups. Kaplan–Meier method and log-rank test were used to analyze overall survival. The p-value < 0.05 was defined as statistically significant. For exosome miRNA-seq and mRNA-seq, the corrected p-value < 0.05 was defined as statistically significant.

To investigate the mechanisms underlying GC-LM, a previously established GC cell line (MKN45-HL) with increased migration and invasion ability and high liver metastatic potential was used for our study [21]. Exosomes were first isolated from the supernatant of MKN45 and MKN45-HL cells by ultracentrifugation and then characterized and quantified by electron microscopy and nanoparticle tracking analysis (NTA). As shown in Fig. 1A, the typical cup-shaped and double-membrane-structured particles with diameters ranging from 30 to 150 nm were observed by electron microscopy, and NTA also revealed that exosomes from both cell lines were slightly larger than 100 nm (Fig. 1B, C). In addition, the purity of exosomes derived from MKN45 and MKN45-HL was assessed by analysis of exosome markers (TSG101 and CD81) and confirmation of the absence of endoplasmic reticulum (calnexin) (Fig. 1D). To investigate whether GC cell-derived exosomes could promote GC-LM, an exosome-educated GC-LM model were constructed (Fig. 1E). We intravenously injected mice with PBS, MKN45 exosomes, and MKN45-HL exosomes in the tail vein, followed by luciferase-labeled MKN45 cells in the spleen. After three weeks, in vivo imaging was performed, and the results showed that GC-derived exosomes significantly increased the fluorescence intensity in mouse liver compared with the PBS group, and the strongest fluorescence signal was observed in MKN45-HL-derived exosome-educated mice (Fig. 1F). Mice were then sacrificed, livers weighed and subjected to H&E staining. The results showed that GC-derived exosomes significantly increased the size and number of liver metastases, especially MKN45-HL-derived exosomes, compared with the PBS group (Fig. 1G-I). Furthermore, immunohistochemical staining of CD31 revealed that education with GC-derived exosomes resulted in significantly more disorganized blood vessels in and around liver metastases compared with PBS (Fig. 1J). Thus, these results demonstrated that GC-derived exosomes play an important regulatory role in GC-LM.

To investigate the organs susceptible to GC-derived exosomes, fluorescently labeled exosomes from MKN45 and MKN45-HL cells were injected into mice via the tail vein. Analysis of fluorescence intensity in each organ revealed that GC-derived exosomes were significantly more enriched in the liver than in other organs in all cases. In addition, exosomes derived from MKN45-HL cells were more prone to internalization by the liver than exosomes from MKN45 cells (Fig. 2A and Supplementary Fig. S1A). We next examined the predominant cells most likely to take up tumor exosomes in the liver. As shown in Fig. 2B, immunofluorescence staining of liver tissue revealed that both MKN45-HL- and MKN45-derived exosomes (PKH26-labeled) were colocalized with intrahepatic macrophages (F4/80 +). Subsequently, human THP-1 monocytes treated with 12-myristate-13-acetate (PMA) were used to mimic macrophages. The stimulated cells exhibited markedly adherent morphology and highly expressed the recognized macrophage marker CD68 (Supplementary Fig. S1B). Tumor-derived exosomes were then co-cultured with PMA-treated THP-1 cells, and the results showed that both MKN45-HL- and MKN45-derived exosomes were internalized by macrophages (Fig. 2C, D). Macrophages have been found to transform into classically activated phenotypes (M1) and alternatively activated phenotypes (M2-like) as a consequence of changes in the microenvironment, and M2-like macrophages play an important role in tumor development and metastasis [22‐24]. Our qRT-PCR results revealed that MKN45-HL exosomes significantly increased the expression of CD163, IL-10, TGF-β, and VEGFA (M2 macrophage markers) and significantly decreased the expression of iNOS and TNFA (M1 macrophage markers) in PMA-treated THP-1 cells compared with PBS and MKN45 exosomes (Fig. 2E, F). Subsequently, the expression of CD206 in PMA-treated THP-1 cells was detected by flow cytometry, and the results indicated that MKN45-HL-exo significantly promoted the expression of CD206 (Fig. 2G and Supplementary Fig. S1C). Similar results were observed in vivo. Immunofluorescence staining of livers from exosome-educated mice for CD206 revealed that exosomes derived from GC cells, particularly cells with high hepatic metastatic potential (MKN45-HL), significantly promoted CD206 expression compared with PBS (Fig. 2H and Supplementary Fig. S1D). The above results showed that GC-derived exosomes were mainly taken up by intrahepatic macrophages and could promote their transformation into the M2-like phenotype.

In metastatic target organs, angiogenesis forms the developmental basis for the transformation of dormant micrometastases into rapidly growing, clinically significant lesions, and M2-like macrophages have been reported to directly influence tumor cell survival, growth, and metastasis by promoting tumor angiogenesis [25]. In addition, TGF-β and VEGF have been shown to be involved in the activation, migration, tube formation, and maturation of the intercellular junctions and surrounding basement membrane of endothelial cells, thus playing an important regulatory role in tumor angiogenesis [26‐30]. As shown in Fig. 3A, MKN45-HL-exo significantly increased the secretion of vascular-associated cytokines in PMA-treated THP-1 cells compared with PBS and MKN45-exo. Subsequently, the indirect in vitro co-culture systems were further applied to investigate the role of GC exosome-treated macrophages in angiogenesis (Supplementary Fig. S1E, F). Transwell assays indicated that the supernatants from PMA-treated THP-1 cells pre-incubated with MKN45-HL-exo significantly increased the migration of HUVECs (Fig. 3B, C). Tube formation assays and aortic ring assays were then performed to detect whether macrophages polarized by GC-derived exosomes affect angiogenesis. Compared with PBS and MKN45-exo, conditioned medium from PMA-treated THP-1 cells pre-incubated with MKN45-HL-exo significantly increased the number of tubules formed in HUVECs and sprouts from mouse aortic rings (Fig. 3D and Supplementary Fig. S1G, H). In addition, it was found that direct incubation of aortic rings with GC-derived exosomes or addition of Panobinostat [31] (an HDAC inhibitor that has been shown to strongly block M2-type polarization of macrophages in vivo and in vitro) to PMA-treated THP-1 cells did not promote sprouting of aortic rings (Fig. 3E, F). Subsequently, mice educated with exosomes were simultaneously administered Panobinostat or DMSO intraperitoneally. CD31 staining of the livers of these mice showed that MKN45-HL-derived exosomes significantly promoted the formation of disorganized neovascularizations in the livers of mice, but this pro-angiogenic effect was significantly diminished when the mice were treated with Panobinostat (Fig. 3G). This suggested that M2-like polarization of macrophages mediated by GC cell-derived exosomes is involved in the regulation of angiogenesis. To further investigate the effect of angiogenesis mediated by M2-like macrophages on GC-LM, KRN-633, a potent VEGF receptor inhibitor, was used in our exosome-educated GC-LM model. As shown in Fig. 3H, in vivo imaging of mice revealed that KRN-633 treatment significantly reduced the fluorescence intensity of mouse livers. Subsequently, the weight and H&E staining of the livers of the sacrificed mice exhibited consistent results (Fig. 3 I and Supplementary Fig. S1I). In addition, immunohistochemical staining for CD31 showed that treatment with KRN-633 markedly inhibited angiogenesis in and around liver metastases in mice (Fig. 3J). Collectively, these results indicated that GC-derived exosomes promote angiogenesis via M2-like macrophages to facilitate GC-LM.

The contents of exosomes, especially miRNA, play an important role in intercellular communication [32]. Therefore, exosomes isolated from MKN45 and MKN45-HL cells were subjected to miRNA sequencing to determine the targets involved in GC-LM (Fig. 4A). A total of 176 differentially expressed miRNAs were identified (Foldchange > 2 or < 0.5; FDR p-value < 0.05), including 71 down-regulated and 105 up-regulated miRNAs in MKN45-HL exosomes (Fig. 4B). The top 6 up-regulated miRNAs were selected for preliminary validation in serum exosomes from 15 GC-LM patients and 15 GC patients without LM, indicating that miR-519a-3p was the most upregulated miRNA (Fig. 4C and Supplementary Table S3). Subsequently, the expression of miR-519a-3p was detected, and the results showed that it was significantly upregulated in GC cells and tissues compared with normal gastric mucosal epithelial cells and tissues (Supplementary Fig. S2A-C). In addition, higher expression of miR-519a-3p was observed in GC-LM tissue compared with GC tissue without LM (Supplementary Fig. S2D-F). Subsequently, the expression of miR-519a-3p in exosomes (exo-miR-519a-3p) was further validated in GC cells and serum from expanded samples. The results showed that it was significantly highly expressed in cells with high LM potential and in serum from GC-LM (Fig. 4D, E). Moreover, the receiver operating characteristic curve (ROC) indicated that exo-miR-519a-3p expression in serum could serve as a potential diagnostic marker for GC-LM (Fig. 4F, AUC = 0.7461, p = 0.0011). Kaplan–Meier survival analysis showed that exo-miR-519a-3p expression in serum was negatively correlated with the endpoint of GC-LM patients (Fig. 4G, cut-off = 0.1048, p = 0.0426). An important function of exosomes is to influence cellular functions by delivering bioactive molecules to other cells [33]. In our previous study [21], GC-derived exosomes were found to be predominantly taken up by intrahepatic macrophages. Therefore, we next investigated whether miR-519a-3p in exosomes could also be delivered into macrophages via exosomes. As shown in Fig. 4H, PMA-treated THP-1 cells incubated with exosomes from MKN45-HL and MKN45 cells expressed more miR-519a-3p than with PBS. However, when the exosomes in the supernatant were removed by ultracentrifugation, the concentration of miR-519a-3p in PMA-treated THP-1 cells was not further increased (Fig. 4I). Moreover, it was observed by immunofluorescence that exo-miR-519a-3p labeled with Cy3 was internalized by PMA-treated THP-1 cells after coculture with them (Fig. 4J). These results indicate that miR-519a-3p is highly expressed in GC-derived exosomes, and its expression level in serum is closely associated with the prognosis of GC-LM patients and can be delivered to macrophages via exosomes.

To investigate the role of exo-miR-519a-3p in GC-LM, MKN45 and MKN45-HL cell lines with stable overexpression and knockdown of miR-519a-3p, respectively, were constructed (Supplementary Fig. S3A, B). We then investigated whether aberrant intracellular miR-519a-3p expression could affect the biological behavior of MKN45 and MKN45-HL cells by CCK-8 assay, transwell assay, and subcutaneous implantation of xenograft tumors. The results showed that knockdown or overexpression of miR-519a-3p in MKN45 and MKN45-HL cells had little effect on the proliferation, migration, invasion, and subcutaneous tumor proliferation of GC cells (Supplementary Fig. S3C-H). In our study, GC-derived exosomes were shown to promote LM by inducing M2-like polarization of intrahepatic macrophages, so here we investigated whether this effect is mediated by miR-519a-3p in exosomes. The qRT-PCR results of PMA-treated THP-1 cells incubated with exosomes derived from miR-519a-3p-overexpressed MKN45 cells showed that the expression of M2 markers was dramatically upregulated and M1 markers were downregulated (Fig. 5A). The opposite trend of M2 and M1 marker expression was observed when PMA-treated THP-1 cells were incubated with miR-519a-3p-inhibited exosomes (Fig. 5B). In addition, detection of CD206 in exosome-incubated THP-1 cells by flow cytometry and in exosome-educated mouse liver by immunofluorescence showed that miR-519a-3p-knockdown exosomes suppressed CD206 expression, whereas miR-519a-3p-overexpressed exosomes promoted the expression of CD206 (Fig. 5C and Supplementary Fig. S4A-C). Subsequently, TGF-β, VEGFA, and VEGFD were detected by ELISA in the supernatant of PMA-treated THP-1 cells incubated with conditioned exosomes. This showed that miR-519a-3p-overexpressing and miR-519a-3p-suppressing exosomes increased and suppressed their expression, respectively, compared with the negative control (Fig. 5D). To determine the effect of exo-miR-519a-3p on angiogenesis, HUVECs were first co-cultured with PMA-treated THP-1 cells pre-incubated with exosomes. Transwell assays showed that THP-1 cells pre-incubated with miR-519a-3p-overexpressed and miR-519a-3p-inhibited exosomes increased and decreased the migratory capacity of HUVECs, respectively (Fig. 5E and Supplementary Fig. S4D). Consistent with the above observations, the conditioned medium of PMA-treated THP-1 cells pre-incubated with miR-519a-3p-overexpressed exosomes significantly increased the number of tubules formed by HUVECs and sprouts from aortic rings (Fig. 5F, G). Meanwhile, an exosome-educated GC-LM model was used to investigate the role of exo-miR-519a-3p in GC-LM. The results of IVIS showed that education with miR-519a-3p-overexpressed exosomes significantly increased the fluorescence signal in the liver compared with controls, whereas miR-519a-3p-suppressed exosomes attenuated the fluorescence signal (Fig. 5H and Supplementary Fig. S4E). Consistently, the weight and H&E staining of the livers of the sacrificed mice showed that education with miR-519a-3p-overexpressed exosomes promoted liver metastasis in mice compared with control exosomes, whereas LM was inhibited after education with miR-519a-3p-inhibited exosomes (Fig. 5I, J and Supplementary Fig. S4F). Furthermore, by immunohistochemical staining for CD31, we found that miR-519a-3p-overexpressing exosomes significantly promoted angiogenesis in liver metastases and surrounding tissues, whereas miR-519a-3p-suppressing exosomes inhibited angiogenesis compared with negative controls (Supplementary Fig. S4G). These results indicate that exo-miR-519a-3p induces polarization of M2 macrophages to promote angiogenesis and liver metastasis, whereas intracellular miR-519a-3p has a limited effect on the phenotype of primary GC cells.

To determine the downstream targets of miR-519a-3p, PMA-treated THP-1 cells transfected with miR-519a-3p inhibitor or negative controls were subjected to mRNA sequencing (Fig. 6A and Supplementary Table S4). And 5 genes potentially targeted by miR-519a-3p were identified by combining the genes upregulated in mRNA-seq (|log₂Foldchange|> 1.5) with those predicted by the TargetScan and PITA databases (Fig. 6B). These five genes were further validated by qRT-PCR in PMA-treated THP-1 cells, and the results showed that only the expression of DUSP2 was significantly altered after knockdown and overexpression of miR-519a-3p (Fig. 6C, D). Based on the potential binding site of miR-519a-3p in the 3'UTR of DUSP2 predicted by bioinformatic analysis (Fig. 6E), luciferase reporter assays were performed. Luciferase activity in HEK293T cells was significantly attenuated when co-transfected with Luc-DUSP2-3'UTR-WT and miR-519a-3p mimics, whereas little change was observed when co-transfected with Luc-DUSP2-3'UTR-MUT and miR-519a-3p mimics, suggesting that miR-519a-3p directly targets DUSP2 (Fig. 6F). Then the results of Western blot revealed that mimics and inhibitors of miR-519a-3p transfected into PMA-treated THP-1 cells promoted and inhibited the expression of DUSP2, respectively (Fig. 6G). Similarly, DUSP2 was observed to be upregulated or downregulated in intrahepatic macrophages when mice were educated with miR-519a-3p-deficient or enriched exosomes, respectively (Fig. 6H, I). Moreover, the contribution of CD206 expression by miR-519a-3p mimics was found to be diminished by flow cytometry after co-transfection with DUSP overexpression plasmid (Fig. 6J). And it was observed that overexpression of DUSP2 in PMA-treated THP-1 cells impaired the pro-secretory effects of miR-519a-3p mimics on TGF-β, VEGF-A, and VEGF-D (Fig. 6K). Thus, miR-519a-3p induces M2-like polarization of macrophages and increases their secretion of angiogenesis-related factors by targeting DUSP2. DUSP2 is a nuclear phosphatase that specifically inactivates ERK by direct dephosphorylation of phosphothreonine and phosphotyrosine residues [34]. To investigate whether miR-519a-3p is involved in the activation of the MAPK/ERK pathway, key pathway-related proteins, including p-ERK1/2, p–c-FOS, and p–c- JUN, were measured by Western blot. The results showed that overexpression of miR-519a-3p in THP-1 cells significantly activated the MAPK/ERK pathway, but this activation was attenuated by overexpression of DUSP2 (Fig. 6L).

The MAPK cascade is an important oncogenic factor in human cancers and is involved in M2-like activation of macrophages [35‐38]. Therefore, targeting inhibition of this pathway is an important strategy for the treatment of GC-LM. PD0325901, a well-established non-ATP competitive MEK inhibitor with efficient inhibition of ERK1/2 phosphorylation, was used for our in vitro and in vivo experiments. The results of Western blot showed that PD0325901 treatment significantly inhibited the activation of MAPK/ERK pathway in THP-1 cells, and the addition of this inhibitor significantly blocked the activation of this pathway by exosome miR-519a-3p (Fig. 7A). Subsequently, the supernatants of PMA-treated THP-1 cells incubated with exosomes and treated with PD0325901 were used for angiogenesis assays and aortic ring assays. Supernatants of THP-1 cells preincubated with miR-519a-3p-rich exosomes significantly increased the number of tubules formed in HUVECs and the budding of aortic rings. However, when PMA-treated THP-1 cells were treated with PD0325901, exo-miR-519a-3p was barely able to exert a pro-angiogenic effect (Fig. 7B, C). Similar in vivo results were observed in the GC-LM model. We attempted to educate mice with miR-519a-3p-riched exosomes in the presence of inhibitors to partially promote LM of GC cells. However, the results of IVSI, H&E and CD31 staining of mouse liver showed that exo-miR-519a-3p could hardly exert its role in promoting angiogenesis and LM after oral administration of PD0325901 (Fig. 7D-H). Collectively, these results suggested that exo-miR-519a-3p creates a vascular-rich environment for the premetastatic niche by activating the DUSP2-MAPK/ERK axis in intrahepatic macrophages.