Fig 2: ECs release EVs with dysregulated miRNA levels when cultured in the presence of cancer cells, in particular miR‐142‐5p, miR‐183‐5p and miR‐222‐3p. (a), Overview of the procedure of endothelial EV isolation and small RNA profile analysis. b‐e, Characterization of HUVEC EVs by cryo‐transmission electron microscopy analysis bar represents 100 nm (b), dynamic light scattering analysis (c), and western blotting of the exosomal markers CD9, CD81, CD63 and syntenin in lysates from purified HUVEC EVs and cells from mono‐ and coculture conditions (d). EV lysates: 1: mono‐HUVEC EVs, 2: co‐HUVEC EVs, 3: MDA‐MB‐231‐EVs; cell lysates: 4: mono‐HUVECs, 5: co‐HUVECs, 6: MDA‐MB‐231. (e), Western blotting of endothelial marker CD31 and tumour marker EpCAM in lysates from purified HUVEC EVs and cells from mono‐ and coculture conditions. EV lysates: 1: mono‐HUVEC EVs, 2: co‐HUVEC EVs, 3: MDA‐MB‐231‐EVs; cell lysates: 4: mono‐HUVECs, 5: co‐HUVECs, 6: MDA‐MB‐231. (f), Heatmap of miRNAs with differential abundance between mono‐HUVEC EVs and co‐HUVEC EVs. (g), RT‐qPCR analysis of the miRNAs dysregulated in co‐HUVEC EVs. Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 calculated by One‐sample t‐test (n = 3). (h), Gene targets shared by miRNAs with increased abundance in co‐EVs (nodes). Edge thickness is scaled by the number of genes targeted in common by the connected nodes. Node size is scaled by number of predicted targets. i, j, Top canonical pathways (i) and top diseases and disorders (j) by IPA analysis of genes targeted by at least five miRNAs upregulated in co‐EVs. k, RT‐qPCR analysis of miRNAs dysregulated in EVs from MS1 cells in coculture (with 4T1 mouse breast cancer cell line) versus mono‐culture conditions. Data are presented as mean ± SEM. *P < 0.05 and **P < 0.01 calculated by One‐sample t‐test (n = 3). l, Relative levels of miR‐142‐5p, miR‐183‐5p and miR‐222‐3p in endothelial EVs isolated from plasma of mouse model of 4T1 breast cancer compared to those from control, uninjected BALB/c mice. Data are presented as median ± IQR. **P < 0.01 calculated by unpaired two‐tailed Mann‐Whitney test. ‐4T1 (empty circle) n = 14; +4T1 (black circle) n = 15

Fig 3: Claudin-2 is enriched in extracellular vesicles isolated from patients with replacement-type colorectal cancer liver metastases.a Nanoparticle tracking analysis of EVs detected in plasma derived from either a RHGP liver lesion bearing patient or a DHGP liver lesion bearing patient. b Representative immunoblot analysis of Claudin-2, Claudin-4, CD81, EpCAM, and TSG101 using lysates prepared from EVs isolated from patients with the indicated lesions. c Claudin-2 levels in EVs concentrated from plasma of liver-metastatic CRC patients with the indicated lesions were normalized to TSG101 levels. d Immunoblot analysis of Claudin-2 and TSG101 using lysates prepared from EVs concentrated from patients, pre- and post-resection, with the indicated lesions. e Claudin-2 levels in EVs concentrated from plasma, pre- and post-resection, of liver-metastatic CRC patients with the indicated lesions were normalized to TSG101 levels.

Fig 4: Ectopic expression of EpCAM and Sox9 rescues the self-renewal ability of JMJD2D-knockdown LCSCs.A, Ectopic expression of EpCAM partly rescued the tumorsphere formation ability of JMJD2D-knockdown cells. B, Ectopic expression of Sox9 partly rescued the tumorsphere formation ability of JMJD2D-knockdown cells. C, ectopic expression of both EpCAM and Sox9 significantly rescued the tumorsphere formation ability of JMJD2D-knockdown cells. D, ectopic expression of both EpCAM and Sox9 in JMJD2D-knockdown cells partly rescued subcutaneous tumor growth and tumor weight in vivo. These experiments were performed at least twice with similar results. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

Fig 5: Co‐culture with cancer cells induces an activation of ECs via inflammatory signalling pathways. (a), Schematic of the in vitro coculture system used to study the effect of TME generated by human breast cancer cells (MDA‐MB‐231) on human endothelial cells (HUVECs). HUVECs were cultured with/without MDA‐MB‐231 for 48 h (i.e. coculture vs. monoculture), followed by endothelial cell isolation using magnetic beads coated with anti‐CD31 antibody. Then, RNA‐Seq was performed to assess transcriptomic changes induced by MDA‐MD‐231. (b), Flow cytometry analysis of HUVEC, MDA‐MB‐231 and cocultured HUVEC/MDA‐MB‐231 cells stained for the CD31 endothelial marker and EpCAM. (c), Western blotting of endothelial marker CD31 and breast cancer marker EpCAM in lysates from HUVECs purified from mono or cancer cell coculture and from MDA‐MB‐231 cells. HSP70 is a loading control. (d), Selected gene sets significantly enriched in co‐ versus mono‐HUVECs from GSEA of RNA‐Seq data obtained as in (a). (e), GSEA enrichment of inflammation‐associated pathways in co‐HUVECs. (f), Expression levels (log2TPM) of several adhesion molecules in HUVECs monocultures or in HUVECs cocultured with MDA‐MB‐231 as shown by RNA‐Seq data. TPM = Transcripts per Million. (g), RT‐qPCR analysis of expression of the adhesion molecules ICAM‐1, VCAM‐1 in HUVECs cocultured with MDA‐MB‐231 relative to HUVECs alone. Data are mean ± SEM. **P < 0.01 calculated by One‐Sample T‐test (n = 6)