Cancer-derived small extracellular vesicles promote angiogenesis by heparin-bound, bevacizumab-insensitive VEGF, independent of vesicle uptake

Abstract

Cancer-derived small extracellular vesicles (sEVs) induce stromal cells to become permissive for tumor growth. However, it is unclear whether this induction solely occurs through transfer of vesicular cargo into recipient cells. Here we show that cancer-derived sEVs can stimulate endothelial cell migration and tube formation independently of uptake. These responses were mediated by the 189 amino acid isoform of vascular endothelial growth factor (VEGF) on the surface of sEVs. Unlike other common VEGF isoforms, VEGF₁₈₉ preferentially localized to sEVs through its high affinity for heparin. Interaction of VEGF₁₈₉ with the surface of sEVs profoundly increased ligand half-life and reduced its recognition by the therapeutic VEGF antibody bevacizumab. sEV-associated VEGF (sEV-VEGF) stimulated tumor xenograft growth but was not neutralized by bevacizumab. Furthermore, high levels of sEV-VEGF were associated with disease progression in bevacizumab-treated cancer patients, raising the possibility that resistance to bevacizumab might stem in part from elevated levels of sEV-VEGF.

Keywords: Cancer microenvironment; Extracellular signalling molecules; Tumour angiogenesis.

Fig. 1

Cancer cell-derived sEVs can stimulate endothelial cell migration and tube formation independently of uptake. a Immunoblot of TSG101 and flotillin-1 in fractions of the indicated buoyant densities that were isolated from media conditioned by ovarian (ES2), colorectal (HCT116), and renal (786-0) cancer cell lines, and from ovarian cancer patient ascites. b Immunogold labeling of CD63 on vesicles in sEV fractions (i.e., density of 1.09–1.13 g/mL). Scale bar = 100 nm. c–f Human umbilical vein endothelial cells (HUVEC) were pretreated with endocytosis inhibitors (chlorpromazine, CPZ; dynasore, DYN) or with dimethyl sulfoxide (DMSO) solvent, and then stimulated with sEVs of ES2, HCT116, and 786-0 cells. Shown are numbers (c) and representative images (d) of migrating HUVEC at 5 h after stimulation, and numbers (e) and representative images (f) of tubes formed at 4 h after stimulation. Mean ± SD of n = 4 independent experiments are shown. Scale bar = 100 μm. g, h HUVEC were treated as in c and evaluated for uptake of PKH26 dye-labeled sEVs by flow cytometry at 5 h thereafter. Shown are mean fluorescence intensity (MFI) values of PKH26 fluorescence detected in HUVEC in n = 3 independent experiments (g) and representative histogram plots (h). Gating strategy and contour plots are shown in Supplementary Fig. 3b and 4b, respectively. **P < 0.01, ***P < 0.001, ****P < 0.0001, by ANOVA with Bonferroni’s corrections; one-way in g, two-way in c and e. ns: not significant. Source data used for graphs in c, e, and g can be found in Supplementary Data 1

Fig. 2

VEGF is present on the surface of cancer cell-derived sEVs. a Detection of angiogenesis-related proteins in sEVs of ES2 cells by Ab array. b Levels of angiogenic factors detected by enzyme-linked immunosorbent assays (ELISA) in lysates of sEVs (gray bars) and on the surface of equivalent amounts of intact sEVs (magenta bars) of ES2, HCT116, and 786-0 cells. CD63 and TSG101 were assayed as positive and negative controls for sEV surface protein, respectively. Shown are mean ± SD of n = 3 independent experiments. *P < 0.05, **P < 0.01, ****P < 0.0001, by two-sided unpaired t-test. Source data can be found in Supplementary Data 2. c To detect sEV surface protein by flow cytometry, microbeads were coupled to the indicated Ab, incubated with sEVs of parental (VEGF^+/+) and VEGF-deficient (VEGF^−/−) HCT116 cells, and then stained with exo-FITC dye to label sEV membrane. Binding of Ab to protein on the surface of sEVs was evaluated by analyzing exo-FITC fluorescence in the gated population of Ab-coupled microbeads. Shown are representative histogram plots of fluorescence. Gating strategy, contour plots, and MFI values of n = 3 independent experiments are shown in Supplementary Fig. 6. d Immunogold labeling of VEGF on sEVs isolated from parental cancer cell lines and from ovarian cancer patient ascites. Scale bar = 100 nm

Fig. 3

sEV-VEGF is signaling competent. a Immunoblot of phosphorylated VEGFR2 (p-VEGFR2) and total VEGFR2 in HUVEC at 5 min following stimulation with sEVs of ES2, HCT116, and 786-0 cells or with recombinant VEGF₁₆₅ (rVEGF₁₆₅). b–e HUVEC were pretreated with inhibitors of VEGFR tyrosine kinase activity (b, c) and with neutralizing Ab to VEGFR2 (d, e), stimulated with sEVs or rVEGF₁₆₅, and then assayed for tube formation at 4 h thereafter. In b and d, mean ± SD of n = 4 independent experiments. In c and e, representative images of tube formation. Scale bar = 100 μm. f VEGF levels in conditioned media (CM) of ES2 and HCT116 cells that were depleted of sEVs (sEV-dep) or left non-depleted (whole). Shown are mean ± SD of n = 3 independent experiments. g–i HUVEC were stimulated with whole and sEV-depleted conditioned media, and then assayed for VEGFR2 phosphorylation (g) and tube formation (h, i). In h, mean ± SD of n = 4 independent experiments. In i, representative images of tube formation. Scale bar = 100 μm. **P < 0.01, ***P < 0.001, ****P < 0.0001, by two-way ANOVA with Bonferroni’s corrections in b, d, and h, and by a two-sided unpaired t-test in f. Source data used for graphs in b, d, f, and h can be found in Supplementary Data 3

Fig. 4

Stimulatory effects of cancer cell-derived sEVs on endothelial cells and tumor growth depend on VEGF. a–c HUVEC were stimulated with equivalent amounts of sEVs of VEGF^+/+ cancer cells, sEVs of VEGF^−/− cancer cells or rVEGF₁₆₅, and then assayed for VEGFR2 phosphorylation (a) and tube formation (b, c). In b, mean ± SD of n = 4 independent experiments. In c, representative images of tube formation. Scale bar = 100 μm. d–f Nude mice were inoculated i.p. with ES2 VEGF^−/− cells that stably expressed GFP. At 7 days thereafter when tumors were palpable, mice were randomized into groups (n = 6 mice per group) and then administered equivalent amounts of sEVs of ES2 VEGF^+/+ cells or sEVs of ES2 VEGF^−/− cells, three times a week for 2 weeks. Negative and positive control groups of tumor-bearing mice were administered saline and rVEGF₁₆₅, respectively. In d, representative images of GFP-expressing tumors in the abdominal cavity viewed under a fluorescence stereomicroscope. Arrows indicate tumors on the omentum. Scale bar = 10 mm. In e, immunofluorescence staining of CD31 (red) in sections of omental tumors (Ome T) adjacent to the pancreas (Panc). Scale bar = 100 μm. In f, amount of i.p. tumor burden, numbers of intratumoral CD31⁺ cells, and volume of ascites in each mouse in each of the groups. I.p. tumor burden is expressed as % of area of GFP fluorescence in the abdominal cavity. Numbers of CD31⁺ cells were scored in five random 100× fields per omental tumor section and an average score was determined for each mouse. Error bars in f represent SD. ***P < 0.001, ****P < 0.0001, by one-way ANOVA with Bonferroni’s corrections in b and f. Source data used for graphs in b and f can be found in Supplementary Data 4

Fig. 5

sEV-VEGF predominantly comprises dimeric VEGF₁₈₉. a Immunoblot of cellular VEGF in lysates of cells of parental cancer cell lines that were treated with brefeldin A to block protein secretion. Recombinant VEGF proteins were included as controls. Overexposure shows VEGF₁₂₁ and VEGF₁₆₅ dimers. b Immunoblot of VEGF in sEVs isolated from the same cell lines as in a but without brefeldin A treatment. c Immunoblot of VEGF in tumor tissues (T) of three patients with ovarian cancer and in sEVs isolated from ascites of the same patients. d Immunoblot of VEGF in sEVs isolated from serum or plasma of five patients with colorectal carcinoma (CRC) and two patients with renal cell carcinoma (RCC). TSG101 was assayed as a control in c, d

Fig. 6

Selective localization of VEGF₁₈₉ in sEVs is mediated by heparin-binding and increases ligand stability. a VEGF levels in conditioned media and sEVs of ES2 VEGF^−/− and HCT116 VEGF^−/− cells transfected with VEGF₁₂₁, VEGF₁₆₅, or VEGF₁₈₉. TSG101 was assayed in sEVs as a control. Mean ± SD of n = 3 independent experiments are shown. b, c VEGF levels in whole and sEV-depleted conditioned media of VEGF^−/− cells transfected with VEGF₁₈₉ (b) and VEGF₁₂₁ (c). Mean ± SD of n = 3 independent experiments are shown. d Conditioned media of non-transfected VEGF^−/− cells was incubated with addition of recombinant VEGF₁₈₉ (rVEGF₁₈₉) at a concentration equivalent to the VEGF concentration in conditioned media of VEGF₁₈₉-transfected VEGF^−/− cells (2500 pg/mL for ES2; 1500 pg/mL for HCT116; see data in a). Thereafter, sEVs were isolated. Amounts of VEGF₁₈₉ detected in these sEVs were compared with VEGF content in sEVs secreted by VEGF₁₈₉-transfected VEGF^−/− cells. Mean ± SD of n = 3 independent experiments are shown. e Levels of human VEGF₁₈₉ in conditioned media and sEVs of CHO-K1 and pgsD-677 cells transfected with human VEGF₁₈₉. Mean ± SD of n = 3 independent experiments are shown. f, g PKH67-labeled sEVs of parental ES2 cells were pretreated with heparinase, chondroitinase, or no enzyme, and then incubated with VEGF Ab coupled to microbeads. VEGF on sEVs was detected by flow cytometric analysis of PKH67 fluorescence in Ab-coupled microbeads (solid histograms). Dotted histograms show background fluorescence when sEVs were incubated with control Ig-coupled beads. As a negative control for enzymatic digestion, the same approach was used to detect the transmembrane protein CD63. In f, representative histogram plots. In g, MFI values of n = 3 independent experiments (mean ± SD). Contour plots are shown in Supplementary Fig. 12. h rVEGF₁₈₉ and sEVs with an equivalent content of VEGF were added to healthy donor plasma. Following incubation at 37 °C for the indicated times, VEGF levels in plasma were assayed. Shown are mean of n = 2 independent experiments. **P < 0.01, ***P < 0.001, ****P < 0.0001, by one-way ANOVA with Bonferroni’s corrections in a and g, by two-sided unpaired t-test in b–e. Source data used for graphs in a–e, g, and h can be found in Supplementary Data 5

Fig. 7

Heparin-bound sEV-VEGF is not neutralized by bevacizumab in vitro. a, b rVEGF₁₈₉ was captured by HMW heparin or by VEGF capture Ab (positive control), and then incubated with bevacizumab or VEGFR1/R2-Fc. Bevacizumab bound to VEGF₁₈₉ and VEGFR1/R2-Fc bound to VEGF₁₈₉ were detected by anti-human IgG. In a, experimental scheme. In b, relative levels of bevacizumab bound to VEGF₁₈₉ and VEGFR1/R2 bound to VEGF₁₈₉. Shown are mean ± SD of n = 6 independent experiments. c, d Microbeads were coupled to bevacizumab, incubated with sEVs of VEGF^+/+ cells or sEVs of VEGF^−/− cells, and then stained with exo-FITC dye to label sEV membrane. The same procedure was performed using microbeads coupled to VEGFR1/R2-Fc (positive control). Binding of bevacizumab and VEGFR1/R2-Fc to VEGF on the surface of sEVs was evaluated by flow cytometric analysis of exo-FITC fluorescence in the gated population of microbeads. In c, representative histogram plots. In d, MFI values of n = 3 independent experiments (mean ± SD). Gating strategy is shown in Supplementary Fig. 6a. Contour plots are shown in Supplementary Fig. 14a. e Bevacizumab and VEGFR1/R2-Fc were incubated with recombinant VEGF and with sEVs that have a VEGF content equivalent to the range of amounts of recombinant VEGF. Following incubation, levels of unbound bevacizumab and unbound VEGFR1/R2-Fc were assayed. Shown are mean ± SD of n = 3 independent experiments. f–h HUVEC were stimulated with sEVs or rVEGF₁₈₉ that were pre-incubated with control Ig, bevacizumab, or VEGFR1/R2-Fc, and then assayed for phosphorylated and total VEGFR2 (f) and tube formation (g, h). In g, mean ± SD of n = 3 independent experiments. In h, representative images of tube formation. Scale bar = 100 μm. *P < 0.05, ***P < 0.001, ****P < 0.0001 by ANOVA with Bonferroni’s corrections; one-way in d; two-way in b and g. Source data used for graphs in b, d, e, and g can be found in Supplementary Data 6

Fig. 8

sEV-VEGF is not neutralized by bevacizumab in vivo and is associated with disease progression in bevacizumab-treated cancer patients. a–e Nude mice were inoculated i.p. with GFP-expressing ES2 VEGF^−/− cells. At 7 days thereafter when tumors were palpable, mice were randomized into groups (n = 6 mice per group) and then administered sEVs of ES2 VEGF^+/+ cells in combination with either normal human IgG (negative control) or bevacizumab, or rVEGF₁₈₉ in combination with either normal human IgG or bevacizumab, three times a week for 2 weeks. In a, representative images of GFP-expressing tumors in the abdominal cavity. Arrows indicate tumors on the omentum. Scale bar = 10 mm. In b, immunofluorescence staining of CD31 (red) in sections of omental tumors (Ome T) adjacent to the pancreas (Panc). Scale bar = 100 μm. Amount of i.p. tumor burden (c), numbers of intratumoral CD31⁺ cells (d) and volume of ascites (e) in each mouse in control (Cont) and bevacizumab (Bev) treatment groups. ***P < 0.001, ****P < 0.0001, by two-sided unpaired t-test. f, g Baseline plasma levels of total VEGF (f) and sEV-VEGF (g) in 17 patients with newly diagnosed metastatic renal cell carcinoma, who were treated presurgically with single-agent bevacizumab for 8 weeks and thereafter restaged. P-values were determined by Mann–Whitney U-test. Source data used for graphs in c–g can be found in Supplementary Data 7