A class of extracellular vesicles from breast cancer cells activates VEGF receptors and tumour angiogenesis

Abstract

Non-classical secretory vesicles, collectively referred to as extracellular vesicles (EVs), have been implicated in different aspects of cancer cell survival and metastasis. Here, we describe how a specific class of EVs, called microvesicles (MVs), activates VEGF receptors and tumour angiogenesis through a unique 90 kDa form of VEGF (VEGF_90K). We show that VEGF_90K is generated by the crosslinking of VEGF₁₆₅, catalysed by the enzyme tissue transglutaminase, and associates with MVs through its interaction with the chaperone Hsp90. We further demonstrate that MV-associated VEGF_90K has a weakened affinity for Bevacizumab, causing Bevacizumab to be ineffective in blocking MV-dependent VEGF receptor activation. However, treatment with an Hsp90 inhibitor releases VEGF_90K from MVs, restoring the sensitivity of VEGF_90K to Bevacizumab. These findings reveal a novel mechanism by which cancer cell-derived MVs influence the tumour microenvironment and highlight the importance of recognizing their unique properties when considering drug treatment strategies.

Figure 1. Combining an Hsp90 inhibitor and Bevacizumab synergistically inhibits tumour growth.

(a) MDAMB231 cells (3 × 10⁶) were injected subcutaneously into each flank of nude mice. On detecting tumours of 1–2 mm in diameter, IP injections of vehicle only (control), Bevacizumab (5 mg kg⁻¹), 17AAG (20 mg kg⁻¹) or Bevacizumab plus 17AAG, were initiated and the drugs were re-administered every third day for 15 days. Left: Plots showing mean tumour volumes (mm³) as a function of time for mice treated with the various drug combinations (n=4 for each condition). The tumour volumes for the Bevacizumab plus 17AAG treatment group, versus treatment with Bevacizumab or 17AAG alone, were statistically significant (P<0.05). Right: Histograms showing mean weights of the resulting MDAMB231 xenografts 24 days after being treated with different drug combinations. (b) Human breast cancer HCI-001 tumour grafts were implanted into the fat pads of NOD/SCID mice. When the tumours reached 3 mm in diameter (∼4 weeks after implantation), IP injections with vehicle only (control), Bevacizumab (2.5 mg kg⁻¹), 17AAG (10 mg kg⁻¹) or Bevacizumab plus 17AAG, were initiated every other day for 4 weeks. Left: Plots showing mean tumour volumes (mm³) as a function of time for mice treated with the various drug combinations (n=4 for each condition). The tumour volumes for the Bevacizumab plus 17AAG treatment group, versus treatment with Bevacizumab or 17AAG alone, were statistically significant (P<0.05). Middle and Right: Images and histograms showing the resulting HCI-001 tumours, and mean weights of these tumours, 8 weeks after being implanted and treated with the different drug combinations. (c) Left: Plots showing the mean tumour volumes (mm³) of tumour graft HCI-002 grown in mice treated with vehicle only (control), Bevacizumab, 17AAG or Bevacizumab plus 17AAG, at the indicated intervals (n=8 for each condition). The tumour volumes for the Bevacizumab plus 17AAG treatment group, versus the Bevacizumab treatment group or 17AAG treatment group, were statistically significant (P<0.05). Middle and Right: Images and histograms showing the HCI-002 tumours, and mean weights of these tumours, 6 weeks after being implanted and treated with the different drug combinations.

Figure 2. Reconstitution of Bevacizumab and 17AAG sensitivity.

(a) Schematic of the in vivo angiogenesis assay. (b,c) Relative amounts of endothelial cells that entered implanted angioreactors that lacked any activators (vehicle control; histogram 1 in both graphs), or were loaded with MDAMB231 cells (5 × 10⁴ cells per angioreactor) without (b, histogram 2) or with a pan inactivating VEGF antibody (200 ng per angioreactor; b histogram 3) or were loaded with rVEGF₁₆₅ (4 ng per angioreactor) without (c histogram 2) or with VEGF antibody (200 ng per angioreactor; C, histogram 3). (d) Scanning electron microscopy (s.e.m.) of a MDAMB231 cell. Arrows indicate large EVs. (e) MVs from MDAMB231 cells were examined by fluorescence staining using the membrane dye FM 1-43FX. Arrows indicate MVs. Scale bar, 5 μm. (f) MVs isolated from MDAMB231 cells were examined using rhodamine-conjugated phalloidin. Arrows indicate MVs. Scale bar, 5 μm. (g) The relative amounts of endothelial cells that entered angioreactors that lacked activators (control; histogram 1), contained MDAMB231 cell MVs (2 μg total protein per angioreactor) without (histogram 2) or with Bevacizumab (1 μg per angioreactor; histogram 3) or contained rVEGF₁₆₅ (4 ng per angioreactor) without (histogram 4) or with Bevacizumab (1 μg per angioreactor; histogram 5). (h) Relative amounts of endothelial cells that entered angioreactors that lacked activators (vehicle control; histogram 1), or angioreactors that contained the indicated combinations of MDAMB231 cell MVs (2 μg per angioreactor), 10 μM 17AAG and 1 μg Bevacizumab (histograms 2–4). (i) Tubulogenesis assays of HUVECs left untreated (control; histogram 1), treated with MVs (10 μg ml⁻¹ total protein) from MDAMB231 cells without (histogram 2) or with 0.5 μg ml⁻¹ Bevacizumab (histogram 3) or treated with rVEGF (15 ng ml⁻¹) without (histogram 4) or with 0.5 μg ml⁻¹ Bevacizumab (histogram 5). Left: The relative differences in tube lengths were plotted. Right: Images of the tubulogenesis assays performed under the different conditions tested. (j) Tubulogenesis assays were performed on HUVECs that were untreated (control; histogram 1) or treated with MVs (10 μg ml⁻¹ of MV protein) from MDAMB231 cells pre-treated with various combinations of 10 μM 17AAG and 0.5 μg ml⁻¹ Bevacizumab (histograms 2–5).

Figure 3. MVs shed from breast cancer cells contain an oligomeric VEGF species.

(a) Whole cell lysates (WCL) from MDAMB231 cells (lane 1) and human recombinant VEGF₁₆₅ (rVEGF₁₆₅; lane 2) were immunoblotted with antibodies against VEGF₁₆₅. (b) Concentrated conditioned medium (20 μg total protein) from serum-starved MDAMB231 (lane 1), HeLa (lane 2) or SKBR3 cells (lane 3) were immunoblotted. (c) MDAMB231 cells were analysed by immunofluorescent (IF) microscopy using Rhodamine-conjugated phalloidin (top panels) and either a pan VEGF or anti-VEGF₁₆₅ antibody (middle panels). Arrows indicate VEGF localized on MVs. Scale bar, 2 μm. IF images of MDAMB231 cells stained by secondary antibody (control; bottom panels). Scale bar, 10 μm (d) Exosomes (lane 1) or MVs (lane 2) from MDAMB231 cells were isolated, lysed and immunoblotted (5 μg per samples) with antibodies against VEGF, CD-63, actin and flotillin-2. (e) MVs from SKBR3 (lane 2), HeLa (lane 3) or MDAMB231 (lane 4) cells were isolated and lysed. WCL from MDAMB231 cells (lane 1), as well as MV lysates (10 μg per sample), were immunoblotted with antibodies against VEGF₁₆₅, flotillin-2 and the cytosolic-specific marker IκBα. (f) MVs from MDAMB231 cells transfected with control siRNA (lane 1) or siRNAs targeting VEGF (lanes 2–4) were immunoblotted with a pan VEGF antibody or anti-flotillin-2. WCL were immunoblotted with a pan VEGF antibody and an anti-actin antibody. (g) Tubulogenesis assays on HUVECs that were untreated (control) or treated with MVs from MDAMB231 cells (10 μg ml⁻¹ MV protein) transfected with control siRNA or siRNAs targeting VEGF. (h) MVs from MDAMB231 cells transfected with control siRNA (lane 1) or siRNAs targeting tTG (lanes 2 and 3) or treated with MDC (50 μM) (lane 4) were immunoblotted with antibodies against VEGF or flotillin-2, while WCLs were immunoblotted with antibodies against VEGF, tTG or actin. (i) Tubulogenesis assays of HUVECs treated with MVs (10 μg ml⁻¹ MV protein) from MDAMB231 cells expressing control siRNA, siRNAs targeting tTG or with MVs from MDAMB231 cells treated with 50 μM MDC.

Figure 4. MV-associated VEGF_90K stimulates a sustained activation of VEGFRs that is insensitive to Bevacizumab.

(a) Tubulogenesis assays were performed on HUVECs that were untreated (control; histogram 1) or treated with recombinant VEGF₁₆₅ (rVEGF; 15 ng ml⁻¹) (histogram 2) or with MVs (10 μg ml⁻¹ total protein) from MDAMB231 (histogram 3), HeLa (histogram 4) or SKBR3 (histogram 5) cells. (b) Lysates of serum-deprived HUVECs exposed to rVEGF₁₆₅ (5 ng ml⁻¹; lanes 1–3) or MVs from MDAMB231 cells (5 μg ml⁻¹ of MV protein; lanes 4–6) for the indicated lengths of time were immunoblotted with antibodies that recognize phosphorylated VEGFR2 (P-VEGFR2), total VEGFR2, phosphorylated ERK (P-ERK), total ERK or actin. (c) Lysates of serum-deprived HUVECs treated for the indicated time with MVs (5 μg ml⁻¹ of MV protein) from MDAMB231 cells expressing VEGF siRNA were immunoblotted with antibodies that recognize phosphorylated VEGFR2, total VEGFR2 or actin. (d) Lysates of serum-deprived cultures of HUVECs that were untreated (lane 1), treated with MDAMB231 cell MVs (5 μg ml⁻¹ of MV protein) without (lane 2) or with (lane 3) either 200 ng ml⁻¹ anti-pan VEGF antibody, 0.5 μg ml⁻¹ Bevacizumab (lane 4) or rVEGF₁₆₅ (5 ng ml⁻¹) without (lane 5) or with (lane 6) 0.5 μg ml⁻¹ Bevacizumab, for 15 min were immunoblotted with antibodies that recognize phosphorylated VEGFR2, total VEGFR2 or actin. (e) Lysates of serum-deprived HUVECs that were untreated (lane 1), or exposed to SKBR3 cell MVs (5 μg ml⁻¹ total protein) treated without (lane 2) or with (lane 3) 0.5 μg ml⁻¹ Bevacizumab for 15 min were immunoblotted with antibodies that recognize phosphorylated VEGFR2, total VEGFR2 or actin. (f) MDAMB231 cell MVs were evenly divided into two samples. In one sample, immunoprecipitations (IPs) using a pan VEGF antibody were performed on the intact MVs (∼25 μg of MV protein in RPMI medium; lane 1), while in the other sample, MVs were first lysed before immunoprecipitations were performed (lane 2). The immunocomplexes were blotted with a pan VEGF antibody and the MV protein inputs were blotted with an antibody against the MV marker flotillin-2.

Figure 5. Hsp90 localizes to MVs and binds VEGF_90K.

(a) Immunofluorescent (IF) images of primary tumour cells cultured from breast cancer patient-derived xenograft HCI-002 stained with antibodies against Hsp90 and VEGF. Arrows indicate Hsp90 and VEGF localized on MVs. Scale bar, 10 μm. (b) MVs from serum-starved MDAMB231 (lanes 1 and 3) or HeLa (lane 2) cells were isolated and lysed. Immunoprecipitations (IPs) using a pan VEGF antibody were performed on the MV lysates (25 μg MV protein, each; lanes 2 and 3). As a control, MV lysates were incubated without the VEGF antibody (lane 1). The immunocomplexes were blotted with VEGF and Hsp90 antibodies. MV lysates were probed with antibodies against Hsp90, VEGF, flotillin-2 and actin (to confirm equivalent amounts of sample were used in each IP) (bottom panel). (c) rVEGF₁₆₅ (30 ng) was incubated in RPMI medium with recombinant Hsp90 (rHsp90; 30 ng, lanes 2 and 3), and without (lane 2) or with recombinant tTG (100 ng, lane 3), for 1 h on ice to generate VEGF_90K. Immunoprecipitations (IPs) using an Hsp90 antibody were performed on the protein incubations and the immunocomplexes were blotted with antibodies against VEGF and Hsp90. As a control, rVEGF₁₆₅ was incubated without VEGF antibody (lane 1). (d) Lysates of MVs from MDAMB231 cells expressing control siRNA (lanes 1) or VEGF siRNA (lane 2) were immunoblotted with antibodies against VEGF, Hsp90 and the MV marker flotillin-2.

Figure 6. VEGF released from MVs regains its sensitivity to Bevacizumab.

(a) Non-permeabilized MDAMB231 cells were analysed by immunofluorescent confocal microscopy using anti-VEGF and anti-Hsp90 antibodies. Top images: VEGF_90K and Hsp90 are detected on MVs (arrows). Bottom images: MDAMB231 cells treated with 10 μM 17AAG overnight were fixed and stained. Scale bar, 10 μm. Far-right: Blow-ups of the MVs. (b) MDAMB231 cell MVs treated without (lane 2) or with (lane 3) 17AAG at 37 °C for 2 h were lysed and immunoprecipitations were performed using a Hsp90 antibody (25 μg MV protein, each). (c) VEGF_90K was generated by tTG-catalysed crosslinking of rVEGF₁₆₅, incubated with Hsp90 (30 ng), either without (lane 1) or with 17AAG (10 μM) (lane 2), at 37 °C for 1 h and immunoprecipitated using an anti-Hsp90 antibody. (d) MDAMB231 cell MVs treated without (lane 1) or with 17AAG (lane 2) at 37 °C for 2 h were collected on a 0.22 μm filter. The filtered MVs and the flow-through were immunoblotted. (e) Serum-deprived HUVECs were untreated (lane 1), or exposed to VEGF_90K (∼10 ng ml⁻¹) released from 17AAG-treated MVs and present in the flow-through from the experiment shown in Fig. 5d, in the presence (lane 2) or absence (lane 3) of Bevacizumab, for 15 min, lysed and immunoblotted. (f) Serum-starved HUVECs, untreated (lanes 1, 2, 5 and 6) or pre-treated with 10 μM GA for 1 hour (lanes 3 and 4), were incubated without (lane 1) or with MVs (5 μg ml⁻¹ protein) from MDAMB231 cells (lane 2), together with 200 ng ml⁻¹ pan inactivating VEGF antibody (lane 4), 10 μM 17AAG (lane 5) or both the VEGF antibody and 17AAG (lane 6), for 15 min. Cell extracts were lysed and immunoblotted. (g) Serum-starved HUVECs were incubated with the VEGF_90K–Hsp90 complex (∼10 ng ml⁻¹) without (lane 1) or with Bevacizumab (0.5 μg ml⁻¹) (lane 2) for 15 min, lysed and immunoblotted. (h) Left: Relative amounts of tubulogenesis for HUVECs untreated (control; histogram 1) or treated with the VEGF_90K–Hsp90 complex (∼20 ng ml⁻¹ total protein) without (histogram 3) or with 0.5 μg ml⁻¹ Bevacizumab (histogram 4). Right: Images of the tubulogenesis assays.

Figure 7. The role of MVs in Bevacizumab insensitivity extends to PDXs.

(a) Lysates of serum-deprived HUVECs treated with conditioned medium (CM; 20 μg ml⁻¹ total protein) from cell cultures of PDX samples HCI-001-003 and HCI-005 (Left) or HCI-006-011 and HCI-013 (Right), described in Table 1, were supplemented without or with 0.5 μg ml⁻¹ Bevacizumab, as indicated, for 15 min. HUVEC lysates were immunoblotted with antibodies that recognize phosphorylated VEGFR2, total VEGFR2 or actin. (b) MVs from cultures of cells established from PDX samples HCI-001 and HCI-002 were isolated, lysed and immunoblotted with antibodies against pan VEGF or the MV marker flotillin-2. (c) Serum-deprived HUVECs were incubated with serum-free medium alone (control; lane 1), or with serum-free medium containing the indicated combinations of MVs from PDX samples HCI-001 or HCI-002 (5 μg ml⁻¹ of MV protein), 0.5 μg ml⁻¹ Bevacizumab (lanes 3 and 6) and 10 μM 17AAG (lanes 4 and 7), for 15 min and then lysed. Cell extracts were immunoblotted with antibodies that recognize phosphorylated VEGFR2, total VEGFR2 or actin. (d) Plots showing relative mean tumour volumes (mm³) of tumour graft HCI-003 in NOD/SCID mice that were untreated (vehicle control only), or treated with either Bevacizumab or Bevacizumab plus 17AAG. The differences between the tumour volumes for the Bevacizumab treatment group versus the vehicle only (control) group were statistically significant (P<0.05).

Figure 8. Diagrams depicting cancer cells generating VEGF_90K and shedding MVs with associated VEGF_90K.

(a) VEGF₁₆₅ is crosslinked by tTG to generate VEGF_90K (top). MVs with associated VEGF_90K are budding and shed from cancer cell plasma membrane (bottom). (b) Diagram depicting cancer cells shedding MVs with associated VEGF_90K that engage recipient endothelial cells and activate VEGFRs, thereby promoting angiogenesis. MV-associated Hsp90 binds to VEGF_90K, enabling the vesicles to activate VEGFRs on endothelial cells and stimulate the formation of new blood vessels. This stimulation is insensitive to Bevacizumab (top). 17AAG causes the release of VEGF_90K from MVs. The free VEGF_90K activates VEGFRs and stimulates angiogenesis but is sensitive to the inhibitory actions of Bevacizumab (bottom).