Gastric Cancer-Derived Extracellular Vesicles (EVs) Promote Angiogenesis via Angiopoietin-2

Simple Summary Angiogenesis is the formation of new blood vessels, which is essential for gastric cancer growth and metastasis. Angiopoietin-2 is a key driver of tumor angiogenesis and has recently emerged as a promising target for antiangiogenic therapy. Extracellular vesicles play an important role in tumor progression including angiogenesis. We explored the crosstalk between gastric cancer and endothelial cells mediated by vesicles, with a specific focus on angiopoietin-2. We show that primary gastric cancer and omental metastasis tissues express angiopoietin-2. We isolated gastric cancer vesicles and demonstrated that they induce the proliferation, migration, invasion, and tube formation of endothelial cells. Characterization of the angiogenic profile of these vesicles revealed high levels of proangiogenic proteins including angiopoietin-2. Using angiopoietin-2 knockdown, we demonstrate that angiopoietin-2 mediates the proangiogenic effects of the gastric cancer vesicles. Our findings suggest a new mechanism via which gastric cancer cells induce angiogenesis. Such a mechanism may be used as a target for cancer therapy. Abstract Angiogenesis is an important control point of gastric cancer (GC) progression and metastasis. Angiopoietin-2 (ANG2) is a key driver of tumor angiogenesis and metastasis, and it has been identified in primary GC tissues. Extracellular vesicles (EVs) play an important role in mediating intercellular communication through the transfer of proteins between cells. However, the expression of ANG2 in GC-EVs has never been reported. Here, we characterized the EV-mediated crosstalk between GC and endothelial cells (ECs), with particular focus on the role of ANG2. We first demonstrate that ANG2 is expressed in GC primary and metastatic tissues. We then isolated EVs from two different GC cell lines and showed that these EVs enhance EC proliferation, migration, invasion, and tube formation in vitro and in vivo. Using an angiogenesis protein array, we showed that GC-EVs contain high levels of proangiogenic proteins, including ANG2. Lastly, using Lenti viral ANG2-shRNA, we demonstrated that the proangiogenic effects of the GC-EVs were mediated by ANG2 through the activation of the PI3K/Akt signal transduction pathway. Our data suggest a new mechanism via which GC cells induce angiogenesis. This knowledge may be utilized to develop new therapies in gastric cancer.


Introduction
Gastric cancer (GC) is the third leading cause of cancer mortality worldwide [1]. Despite considerable improvements in surgical quality and multidisciplinary treatment,

EV Isolation from AGS and SNU-16 Cell Lines
Gastric cancer small EVs (GC-EVs), enriched in exosomes, were obtained by differential centrifugation from culture medium conditioned by AGS and SNU-16 cells grown in 200 mL of serum free medium for 24 h, as we previously described [32]. Briefly, conditioned medium was centrifuged at 3000× g for 10 min to remove cell debris and most "large" EVs. The resulting supernatant was further centrifuged for 30 min at 10,000× g to remove most "medium" EVs; finally, small EVs were enriched by ultracentrifugation at 100,000× g for 70 min. The small EV pellet was suspended in PBS and recentrifuged at 100,000× g for 70 min [33]. All steps were carried out at 4 • C. Using this protocol, EV characterization revealed the presence of particles with morphological and biochemical characteristics of small EVs, in accordance with the guidelines of the Minimal Information for Studies of Extracellular Vesicles (MISEV 2018) [34]. The pelleted exosomes were resuspended in PBS and quantified using the Bradford assay (Bio-Rad, Hercules, CA, USA).

Nanoparticle Tracking Analysis (NTA)
The analysis of size distribution of EVs based on Brownian motion was assayed by NanoSight NS300 (NanoSight, Amesbury, UK) as previously described [32]. Briefly, the EV fraction was diluted 1:1000 with 0.1 µm filtered PBS, and size dispersion was measured at 25 • C. At least three videos of 30 s in duration were taken with a frame rate of 30 frames/second, and particle movement was analyzed by NTA software (version 3.1, NanoSight, Salisbury, UK).

EVs Labeling
EVs labeling was performed as described by Hazan-Halevy et al. and by our group [32,38].

EVs Internalization Assay
Internalization was measured by confocal microscopy and flow cytometry analysis as previously described [32] using EA.hy926 cells (3 × 10 5 and PKH-67-labeled-EVs (2 µg). Data were acquired by FACS Canto II with Diva software (Becton Dickinson, Franklin Lakes, NJ, USA) and analyzed using FlowJo software (Tree Star, Inc., Ashland, OR, USA). For confocal microscopy analysis, the slides were viewed under a Zeiss LSM700 confocal microscope.

Cell Growth Assay
Cell proliferation was measured using the XTT cell proliferation kit (Biological Industries) according to the manufacturer's instructions and as described previously [32] using 5000 EA.hy926 cells/well and 10 µg/mL EVs for 24 h.

Migration and Invasion Assays
Transwell migration and invasion assays were conducted as described previously [32] using 5 × 10 4 EA.hy926 cells suspended in serum-free medium (0.5 mL/chamber) in the upper chamber of 24-well Transwell inserts with 8 mm pore size (BD Biosciences). The lower chamber was filled with 0.75 mL/well DMEM with 2% EV-depleted FBS and 10 µg/mL EVs. In all cases, the cells were incubated for 16 h at 37 • C. The migratory and invasive activities were determined by counting the number of cells in three fields per well (magnification, 100×) in triplicate with the ImageJ 1.48v.Java image processing program.

Matrigel Plug Assay
All animal procedures and care were approved by the Institutional Animal Care and Usage Committee of the Tel Aviv Sourasky Medical Center (protocol # 24-8-19). Six week old male athymic nude mice Foxn1 nu/+ were injected subcutaneously with 500 µL of Matrigel (BD Biosciences) containing 100 µg AGS-EVs or PBS as a control. The mice were sacrificed after 14 days, and the Matrigel plugs were surgically removed and analyzed for hemoglobin content with Drabkin's reagent kit (Sigma-Aldrich) and for the expression of CD31 using immunohistochemistry (IHC) [39].

Human Angiogenesis Antibody Array
The content of EV angiogenesis-related proteins was screened by the Human Angiogenesis Antibody Array G1000 (RayBiotech, Peachtree, GA, USA) according to the manufactures' instructions using 100 µg of total protein obtained from AGS-EVs. Fluorescence intensity was analyzed using a microarray scanner (Innopsys InnoScan ® , Chicago, IL, USA) and its software, which calculated the mean intensity of each spot minus background and the negative control. The results were normalized to positive control spots.

Statistical Analysis
Statistical analysis was performed using GraphPad Prism TM software. Continuous variables were compared using the t-test for significance. All significance tests were twotailed, and a p-value of ≤0.05 was considered statistically significant. Results are presented as the mean ± standard deviation (SD).

ANG2 Is Expressed in Human Gastric Cancer and Omental Metastasis Samples
Initially, the slides were stained for pan-cytokeratin (pan CK) to confirm the presence of tumor and metastatic cells in the evaluated tissues ( Figure 1A,B) [41]. Primary and metastatic cells demonstrated a high expression level of pan-cytokeratin. To evaluate the clinical relevance of angiopoietins in gastric cancer (GC), we assessed the expression levels of ANG2 in tissues of primary gastric tumors and in matched normal gastric tissues (n = 12 each). As depicted in Figure 1A, IHC staining for ANG2 showed higher expression levels in tumor cells of primary gastric tumors compared to normal gastric tissues and revealed a trend toward a positive correlation between poor tumor differentiation and the level of expression. We then stained matched specimens of GC omental metastases, a common site for GC spread (n = 12). As shown in Figure 1B, we observed moderate expression of ANG2 in the metastatic tumoral tissues. These data support the premise that angiopoietins have a potential role in GC progression and metastasis, most probably via induction of angiogenesis. We next sought to investigate whether GC-induced angiogenesis is mediated by angiopoietins secreted and delivered by GC-EVs.

Statistical Analysis
Statistical analysis was performed using GraphPad Prism TM software. Continuous variables were compared using the t-test for significance. All significance tests were twotailed, and a p-value of ≤0.05 was considered statistically significant. Results are presented as the mean ± standard deviation (SD).

ANG2 Is Expressed in Human Gastric Cancer and Omental Metastasis Samples
Initially, the slides were stained for pan-cytokeratin (pan CK) to confirm the presence of tumor and metastatic cells in the evaluated tissues ( Figure 1A,B) [41]. Primary and metastatic cells demonstrated a high expression level of pan-cytokeratin. To evaluate the clinical relevance of angiopoietins in gastric cancer (GC), we assessed the expression levels of ANG2 in tissues of primary gastric tumors and in matched normal gastric tissues (n = 12 each). As depicted in Figure 1A, IHC staining for ANG2 showed higher expression levels in tumor cells of primary gastric tumors compared to normal gastric tissues and revealed a trend toward a positive correlation between poor tumor differentiation and the level of expression. We then stained matched specimens of GC omental metastases, a common site for GC spread (n = 12). As shown in Figure 1B, we observed moderate expression of ANG2 in the metastatic tumoral tissues. These data support the premise that angiopoietins have a potential role in GC progression and metastasis, most probably via induction of angiogenesis. We next sought to investigate whether GC-induced angiogenesis is mediated by angiopoietins secreted and delivered by GC-EVs.

Isolation and Characterization of Gastric Cancer-Derived EVs
To investigate the effect of GC-EVs on angiogenesis, we first purified and characterized small EVs from the conditioned medium (CM) of the human gastric cancer cell line AGS. Isolation was performed using differential ultracentrifugation according to the current gold standard for EV isolation and as we have previously described [32,33]. EV count, size, markers, and morphology were characterized by nanoparticle tracking, Western blot, and cryo-TEM in accordance with the Minimal Information for Studies of Extracellular Vesicles (MISEV 2014 and 2018) guidelines [32,34,42]. As can be seen in Figure 2A, the cryo-TEM images revealed the presence of round particles with dimensions of 65-200 nm. Nanoparticle tracking analysis (NTA) demonstrated a homogeneous population of EVs with a modal diameter of 68.9 ± 1.0 nm, which is within the range of small EVs (sEVs, <100 nm) [34], and a concentration of 1.68 × 10 11 ± 2.32 × 10 10 particles/mL ( Figure 2B). The EV preparations also contained 10% large EVs with a diameter >134.3 nm as analyzed by NTA (D90 = 134.3 ± 3.0 nm). As depicted in Figure 2C, CD63, CD81, and CD9 were highly expressed in both AGS cells and AGS-derived EVs. These results confirmed the secretion of EVs from GC cells.

Isolation and Characterization of Gastric Cancer-Derived EVs
To investigate the effect of GC-EVs on angiogenesis, we first purified and characterized small EVs from the conditioned medium (CM) of the human gastric cancer cell line AGS. Isolation was performed using differential ultracentrifugation according to the current gold standard for EV isolation and as we have previously described [32,33]. EV count, size, markers, and morphology were characterized by nanoparticle tracking, Western blot, and cryo-TEM in accordance with the Minimal Information for Studies of Extracellular Vesicles (MISEV 2014 and 2018) guidelines [32,34,42]. As can be seen in Figure 2A, the cryo-TEM images revealed the presence of round particles with dimensions of 65-200 nm. Nanoparticle tracking analysis (NTA) demonstrated a homogeneous population of EVs with a modal diameter of 68.9 ± 1.0 nm, which is within the range of small EVs (sEVs, <100 nm) [34], and a concentration of 1.68 × 10 11 ± 2.32 × 10 10 particles/mL ( Figure 2B). The EV preparations also contained 10% large EVs with a diameter >134.3 nm as analyzed by NTA (D90 = 134.3 ± 3.0 nm). As depicted in Figure 2C, CD63, CD81, and CD9 were highly expressed in both AGS cells and AGS-derived EVs. These results confirmed the secretion of EVs from GC cells.

Uptake of Gastric Cancer Derived EVs by Endothelial Cells
EVs must be taken up by their target cells in order to assess protein delivery via EVs. For that purpose, we incubated fluorescent dye-labeled AGS-EVs with EAhy.926 transformed human umbilical vein endothelial cells (HUVECs) at different timepoints and followed their uptake by quantitative flow cytometry ( Figure 3A) and confocal microscopy analysis ( Figure 3B). We observed EV uptake by endothelial cells as early as 1 h following incubation (2.76%). Longer incubation times resulted in greater accumulation of EVs inside the cells. As depicted in Figure 3A, 10.55% of EVs were taken up by the endothelial cells after 2 h of incubation, rising to 83.42% after 8 h of incubation. These results confirmed the rapid uptake of gastric cancer-derived EVs by endothelial cells (ECs).

Uptake of Gastric Cancer Derived EVs by Endothelial Cells
EVs must be taken up by their target cells in order to assess protein delivery via EVs. For that purpose, we incubated fluorescent dye-labeled AGS-EVs with EAhy.926 transformed human umbilical vein endothelial cells (HUVECs) at different timepoints and followed their uptake by quantitative flow cytometry ( Figure 3A) and confocal microscopy analysis ( Figure 3B). We observed EV uptake by endothelial cells as early as 1 h following incubation (2.76%). Longer incubation times resulted in greater accumulation of EVs inside the cells. As depicted in Figure 3A, 10.55% of EVs were taken up by the endothelial cells after 2 h of incubation, rising to 83.42% after 8 h of incubation. These results confirmed the rapid uptake of gastric cancer-derived EVs by endothelial cells (ECs).

Gastric Cancer-Derived EVs Promote Proliferation, Migration, and Invasion of Endothelial Cells
Next, we examined whether uptake of GC-EVs induces proangiogenic properties in endothelial cells. For that purpose, we used EVs isolated from two different gastric cancer cell lines: AGS and SNU-16. AGS-EVs and SNU-16-EVs were cocultured with EAhy.926 endothelial cells (ECs), and their effect on proliferation was detected using an XTT cell proliferation kit. Our results demonstrated that both AGS-EVs and SNU-16-EVs increased endothelial cellular proliferation by 30% (p < 0.001) and 50% (p = 0.0288), respectively ( Figure 4A). Similarly, AGS-EVs and SNU-16-EVs increased migration and invasion of endothelial cells in a Boyden chamber assay, migration by 1.5-fold (p = 0.0352) and 1.3-fold (p = 0.0328), respectively, and invasion by 2.6-fold (p = 0.0027) and 1.9-fold (p = 0.0132), respectively ( Figure 4B,C). Taken together, these results demonstrated that GC-EVs induce phenotypic changes in ECs favoring angiogenesis.
x FOR PEER REVIEW 7 of 16

Gastric Cancer-Derived EVs Promote Proliferation, Migration, and Invasion of Endothelial Cells
Next, we examined whether uptake of GC-EVs induces proangiogenic properties in endothelial cells. For that purpose, we used EVs isolated from two different gastric cancer cell lines: AGS and SNU-16. AGS-EVs and SNU-16-EVs were cocultured with EAhy.926 endothelial cells (ECs), and their effect on proliferation was detected using an XTT cell proliferation kit. Our results demonstrated that both AGS-EVs and SNU-16-EVs increased endothelial cellular proliferation by 30% (p < 0.001) and 50% (p = 0.0288), respectively (Figure 4A). Similarly, AGS-EVs and SNU-16-EVs increased migration and invasion of endothelial cells in a Boyden chamber assay, migration by 1.5-fold (p = 0.0352) and 1.3-fold (p = 0.0328), respectively, and invasion by 2.6-fold (p = 0.0027) and 1.9-fold (p = 0.0132), respectively ( Figure 4B,C). Taken together, these results demonstrated that GC-EVs induce phenotypic changes in ECs favoring angiogenesis.

Gastric Cancer-Derived EVs Promote Tube Formation In Vitro and Blood Vessels In Vivo
We next performed an in vitro endothelial tube formation assay. Both AGS-EVs and SNU-16-EVs increased the formation of endothelial tubes by 3.7-fold (p < 0.001) and 3.2-fold (p < 0.001), respectively ( Figure 5A). We then performed an in vivo Matrigel plug assay in which Matrigel mixed with PBS (control) or AGS-EVs was injected subcutaneously into nude mice. As shown in Figure 5B, AGS-EVs significantly increased vascularization within the Matrigel plugs. Figure 5C depicts a significant increase in the hemoglobin content within the Matrigel plugs injected with AGS-EVs: 17.5 mg/dL in the AGS-EVs vs. 3.8 mg/dL in the control-treated mice (p = 0.046). In addition, CD31 immunohistochemical (IHC) staining of sections from the Matrigel plugs demonstrated a significant increment in the number of blood vessels: nine CD31-positive vessels/field in the AGS-EVs vs. four in the control-treated mice (p = 0.0036) ( Figure 5D). Taken together, these data suggest that GC-EVs promote angiogenesis both in vitro and in vivo.

Gastric Cancer-Derived EVs Promote Tube Formation in Vitro and Blood Vessels in Vivo
We next performed an in vitro endothelial tube formation assay. Both AGS-EVs and SNU-16-EVs increased the formation of endothelial tubes by 3.7-fold (p < 0.001) and 3.2fold (p < 0.001), respectively ( Figure 5A). We then performed an in vivo Matrigel plug assay in which Matrigel mixed with PBS (control) or AGS-EVs was injected subcutaneously into nude mice. As shown in Figure 5B, AGS-EVs significantly increased vascularization within the Matrigel plugs. Figure 5C depicts a significant increase in the hemoglobin content within the Matrigel plugs injected with AGS-EVs: 17.5 mg/dL in the AGS-EVs vs. 3.8 mg/dL in the control-treated mice (p = 0.046). In addition, CD31 immunohistochemical (IHC) staining of sections from the Matrigel plugs demonstrated a significant increment in the number of blood vessels: nine CD31-positive vessels/field in the AGS-EVs vs. four in the control-treated mice (p = 0.0036) ( Figure 5D). Taken together, these data suggest that GC-EVs promote angiogenesis both in vitro and in vivo. Representative images are shown on the left (magnification, 100×; scale bar represents 100 µm). IHC quantification is shown on the right. * p < 0.05; ** p < 0.01; *** p < 0.001.
PEER REVIEW 9 of 16 [28]. The expression levels of most proangiogenic proteins (approximately 64%) were higher than those of the antiangiogenic factors. Moreover, according to the exosome and extracellular vesicles database ExoCarta and EVpedia [43,44], 14 proteins (33%) were previously found in GC-EVs ( Figure 6B). In contrast, ANG1 and ANG2, as well as several other proangiogenic proteins, have not been previously reported in relation to GC-EVs. In support of our hypothesis and previous findings, these data demonstrate that GC-EVs contain numerous proangiogenic proteins, including ANG1 and ANG2.

Gastric Cancer-Derived EVs Increase the Expression of ANG2 in Endothelial Cells
We first examined the effect of AGS-EVs on the expression of ANG2 and two of its downstream proteins: PI3K and p-Akt, in naïve endothelial cells. As depicted in Figure  7A, there was an increase in the expression of ANG2, PI3K, and p-Akt in these AGS-EVtreated endothelial cells. Next, in order to investigate the specific role of GC-EV-derived ANG2, we created an in vitro model of ANG2 knockdown GC and endothelial cells (ECs). Figure 7B demonstrates a successful ANG2 knockdown in both cell lines. Consequently, lower expression levels of ANG2 were also observed in EVs from knockdown GC cells (shANG2-EVs) compared to the control EVs (shCtrl-EVs; Figure 7C). Next, we treated ANG2 knockdown ECs with shANG2-EVs vs. shCtrl-EVs and evaluated the expression levels of ANG2 and two of its downstream proteins, PI3K and p-Akt, in these endothelial cells. As shown in Figure 7D, there was a significantly higher expression of ANG2, PI3K, and p-Akt in the endothelial cells that were treated with the ANG2-positive EVs (shCtrl-EVs). Lastly, an inhibitor of TIE2 was used in order to specifically assess the involvement of the PI3K/Akt pathway. As can be seen in Figure 7E, treatment of the shANG2-ECs with the TIE2 inhibitor (TIE2-I) reduced the expression of both PI3K and p-Akt. Taken together, these results suggest that GC-EVs transfer ANG2 to endothelial cells, thus activating the PI3K/Akt signal transduction pathway.

Gastric Cancer-Derived EVs Increase the Expression of ANG2 in Endothelial Cells
We first examined the effect of AGS-EVs on the expression of ANG2 and two of its downstream proteins: PI3K and p-Akt, in naïve endothelial cells. As depicted in Figure 7A, there was an increase in the expression of ANG2, PI3K, and p-Akt in these AGS-EVtreated endothelial cells. Next, in order to investigate the specific role of GC-EV-derived ANG2, we created an in vitro model of ANG2 knockdown GC and endothelial cells (ECs). Figure 7B demonstrates a successful ANG2 knockdown in both cell lines. Consequently, lower expression levels of ANG2 were also observed in EVs from knockdown GC cells (shANG2-EVs) compared to the control EVs (shCtrl-EVs; Figure 7C). Next, we treated ANG2 knockdown ECs with shANG2-EVs vs. shCtrl-EVs and evaluated the expression levels of ANG2 and two of its downstream proteins, PI3K and p-Akt, in these endothelial cells. As shown in Figure 7D, there was a significantly higher expression of ANG2, PI3K, and p-Akt in the endothelial cells that were treated with the ANG2-positive EVs (shCtrl-EVs). Lastly, an inhibitor of TIE2 was used in order to specifically assess the involvement of the PI3K/Akt pathway. As can be seen in Figure 7E, treatment of the shANG2-ECs with the TIE2 inhibitor (TIE2-I) reduced the expression of both PI3K and p-Akt. Taken together, these results suggest that GC-EVs transfer ANG2 to endothelial cells, thus activating the PI3K/Akt signal transduction pathway.

ANG2 Derived from Gastric Cancer EVs Mediates Gastric Cancer Induced Angiogenesis
We used our in vitro model of ANG2 knockdown GC cells and ECs to specifically assess whether ANG2 is essential for AGS-EV-induced angiogenesis. First, we compared the ability of the ANG2 knockdown ECs (shANG2-ECs) to form capillary-like structures compared to ANG2-expressing ECs (shCtrl-ECs). As depicted in Figure 8A (upper panel), knockdown of ANG2 in ECs reduced their ability to form tubes by fourfold compared to ANG2-expressing ECs (p = 0.0001). Next, we treated ANG2 knockdown ECs with shANG2-EVs vs. shCtrl-EVs and evaluated their ability to restore the formation of capillary-like structures. As shown in Figure 8A (lower panel), ANG2-positive EVs (shCtrl-EVs) increased the ability of the ANG2 knockdown endothelial cells to form capillary-like structures by 3.4-fold (p = 0.0001). Similarly, human recombinant ANG2 increased their tube formation by threefold (p = 0.0003), whereas treatment with TIE2 inhibitor prior to the addition of shCtrl-EVs completely diminished this effect. These results suggest that GC-EV-derived ANG2 may mediate GC-induced angiogenesis.

Discussion
In the present study, we demonstrated the proangiogenic effect of gastric cancer-derived EVs (GC-EVs). We describe, for the first time, the angiogenic protein profile of these

ANG2 Derived from Gastric Cancer EVs Mediates Gastric Cancer Induced Angiogenesis
We used our in vitro model of ANG2 knockdown GC cells and ECs to specifically assess whether ANG2 is essential for AGS-EV-induced angiogenesis. First, we compared the ability of the ANG2 knockdown ECs (shANG2-ECs) to form capillary-like structures compared to ANG2-expressing ECs (shCtrl-ECs). As depicted in Figure 8A (upper panel), knockdown of ANG2 in ECs reduced their ability to form tubes by fourfold compared to ANG2-expressing ECs (p = 0.0001). Next, we treated ANG2 knockdown ECs with shANG2-EVs vs. shCtrl-EVs and evaluated their ability to restore the formation of capillary-like structures. As shown in Figure 8A (lower panel), ANG2-positive EVs (shCtrl-EVs) increased the ability of the ANG2 knockdown endothelial cells to form capillary-like structures by 3.4-fold (p = 0.0001). Similarly, human recombinant ANG2 increased their tube formation by threefold (p = 0.0003), whereas treatment with TIE2 inhibitor prior to the addition of shCtrl-EVs completely diminished this effect. These results suggest that GC-EV-derived ANG2 may mediate GC-induced angiogenesis.

ANG2 Derived from Gastric Cancer EVs Mediates Gastric Cancer Induced Angiogenesis
We used our in vitro model of ANG2 knockdown GC cells and ECs to specifically assess whether ANG2 is essential for AGS-EV-induced angiogenesis. First, we compared the ability of the ANG2 knockdown ECs (shANG2-ECs) to form capillary-like structures compared to ANG2-expressing ECs (shCtrl-ECs). As depicted in Figure 8A (upper panel), knockdown of ANG2 in ECs reduced their ability to form tubes by fourfold compared to ANG2-expressing ECs (p = 0.0001). Next, we treated ANG2 knockdown ECs with shANG2-EVs vs. shCtrl-EVs and evaluated their ability to restore the formation of capillary-like structures. As shown in Figure 8A (lower panel), ANG2-positive EVs (shCtrl-EVs) increased the ability of the ANG2 knockdown endothelial cells to form capillary-like structures by 3.4-fold (p = 0.0001). Similarly, human recombinant ANG2 increased their tube formation by threefold (p = 0.0003), whereas treatment with TIE2 inhibitor prior to the addition of shCtrl-EVs completely diminished this effect. These results suggest that GC-EV-derived ANG2 may mediate GC-induced angiogenesis.

Discussion
In the present study, we demonstrated the proangiogenic effect of gastric cancer-derived EVs (GC-EVs). We describe, for the first time, the angiogenic protein profile of these

Discussion
In the present study, we demonstrated the proangiogenic effect of gastric cancerderived EVs (GC-EVs). We describe, for the first time, the angiogenic protein profile of these EVs and show that their proangiogenic effect is specifically mediated through the transference of ANG2 and PI3K/Akt activation.
Growing evidence points toward the role of cancer-derived EVs in intercellular communication and cancer progression. These studies described the role of EVs as specialized messengers, covering the different stages of cancer progression, including growth and metastasis, tumor metabolism, chemoresistance, and angiogenesis [45]. While several recent studies have shown that cancer-derived EVs modulate the tumor microenvironment by enhancing angiogenesis [46], data on GC-EVs and angiogenesis are scarce. We now showed that GC-EVs are transferred into endothelial cells, thus inducing endothelial cell proliferation, migration, invasion, and tube formation both in vitro and in vivo. These data support recent reports that showed that angiogenesis was promoted by the treatment of HUVECs with GC cell-derived exosomes. Those authors demonstrated that transmission of specific miRNAs and other noncoding RNAs by GC-EVs prompts angiogenesis and gastric cancer tumor progression [47][48][49][50][51]. While these few reports focused on the role of transported RNAs, there is only a single report that described the effect of GC-EV-derived proteins on angiogenesis. In that study, Xue et al. showed that Y-box-binding protein 1 (YB-1) transferred by GC exosomes into HUVECs promotes angiogenesis by enhancing the expression of VEGF, ANG1, MMP-9, and IL-8 in endothelial cells [52].
We used a comprehensive approach to characterize angiogenic proteins in GC-EVs. Our proteomic analysis revealed the expression of various pro-and antiangiogenic factors within GC-EVs. Interestingly, we observed a high expression of proangiogenic proteins (i.e., ANG1, ANG2, VEGF, MMP1, uPAR, CXCL11, IL-6, IGF-I, and TIMP-1) compared with a relatively low expression of antiangiogenic proteins (i.e., angiostatin, endostatin, IL-4, TIMP-2, and INF-gamma). Our findings are supported by a few other studies that demonstrated the angiogenic profile of EVs from different tumor cells. Analysis of exosomes from glioblastoma cells showed that they are enriched with the proangiogenic factors VEGF, angiogenin, TGFβ, IL-8, IL-6, MMP2, MMP9, TIMP-1, TIMP-2, and CXCR4 [53,54]. Another study showed that melanoma-derived exosomes contain VEGF, IL-6, and MMP2 [55]. As for the ratio between pro-and antiangiogenic properties of tumor-derived EVs, a recent study demonstrated that exosomes from nasopharyngeal carcinoma cells contain high levels of proangiogenic proteins including CD44 isoform 5, ICAM-1, and MMP13, in contrast to low levels of the antiangiogenic protein, thrombospondin-1 [56]. Our results support existing data by showing that cancer cells reprogram their surroundings into a tumor-promoting microenvironment via the secretion of EVs [57]. In order to create their own vascular system, tumors change their microenvironment from an antiangiogenic to a proangiogenic one via a process termed "angiogenic switch" [29,58]. Our report supports the premise that EVs may be specifically involved in such a cancer-related angiogenic switch; moreover, this is the first study to demonstrate the role of GC-EVs in this process.
ANG2 was one of the most abundant angiogenic proteins in the present proteomic analysis. Significant upregulation of tissue and blood ANG2 levels has been reported in many types of cancer, including melanoma, glioblastoma, breast cancer, renal cell carcinoma, colorectal cancer, and GC [14,17,59,60], but the transference of ANG2 via tumor EVs into the microenvironment has only been reported in hepatocellular carcinoma (HCC). In their study, Xie et al. demonstrated that ANG2 existed in HCC-derived exosomes and was delivered into HUVECs via exosome endocytosis to stimulate angiogenesis [61].
The expression of ANG2 in primary and metastatic tumors has been described in breast and prostate cancers. There was a significant correlation between ANG2 expression, lymph node metastasis, tumor grade, and lymph-vascular invasion in breast cancer specimens [62], whereas ANG2 was correlated to the Gleason score and metastases in prostate cancer [63]. The expression of ANG2 has been reported in primary GC tissues [14,15], and our analysis of primary GC tissues corroborates those findings. In addition, we now show that ANG2 is also expressed in GC omental metastasis specimens. Taken together, these results imply that ANG2 may promote gastric cancer metastasis by stimulating angiogenesis in the omental metastatic niche; however, further studies are necessary to validate this preliminary finding.
The angiogenic activity of ANG2 was demonstrated through binding to TIE2 and activation of the PI3K/Akt signaling pathway [11]. In the current study, GC-EVs induced the expression of ANG2 and the downstream proteins PI3K and p-Akt in ANG2 knockdown endothelial cells. We hypothesize that ANG2 delivered by GC-EVs may be recycled by endothelial cells after internalization and excreted to bind TIE2, thus activating the PI3K/Akt signal transduction pathway. An additional support for our hypothesis was demonstrated in the tube formation assay of the ANG2 knockdown endothelial cell model. Recombinant human ANG2, as well as EVs containing ANG2, could restore tube formation of ANG2 knockdown endothelial cells. This effect was completely eliminated when the cells were treated with TIE2 inhibitor before treatment.
Antiangiogenic therapies play a role in the treatment of metastatic cancer patients. Most of these therapies target the VEGF signaling pathway and have shown efficacy in many types of cancer. However, their long-term efficacy is limited due to drug resistance. Since angiogenesis is crucial for cancer progression and metastasis, there is a clear need for novel antiangiogenic therapeutic targets. Over the past decade, the ANG-TIE signaling pathway has been the subject of much research in search of a complement to the current VEGF-based anti-angiogenic therapies. To date, several drugs targeting the ANG2/TIE2 signaling pathway are under clinical trials as adjunct cancer therapies [16]. We focused on GC microenvironment, specifically endothelial cells, as a potential target for therapy. Our data suggest that ANG2 and the ANG/TIE pathway comprise a potential target for therapy for gastric cancer patients. Moreover, we present novel findings which support the activation of this signaling pathway as being mediated by gastric cancer EVs. Thus, these EVs should be further investigated as a potential angiogenic target for therapy. To date, there are no existing anti-EV therapies in clinical practice or clinical trials which evaluate their efficacy. However, over the past years, various inhibitors of EVs have been studied, including pharmacological inhibitors of EV release, particularly inhibitors of EV trafficking and those that affect lipid metabolism [64]. Hopefully, understanding the biogenesis of EVs in the near future will unravel specific mechanisms of EV-induced angiogenesis, a key factor in the development of new gastric cancer therapy strategies. We believe that the transportation mechanism should be further investigated in the next years as a potential therapeutic approach for gastric cancer patients.
The present study harbors several limitations. Firstly, we performed immunohistochemistry for ANG2 in primary and metastatic GC human tissues and used it as a proof of principle. Thus, conclusions should be carefully made regarding the correlation between cancer differentiation and ANG2 expression. To do so, a large cohort of gastric cancer patients should be investigated. Secondly, the ANG-TIE signaling pathway is complex and affected by numerous molecules such as ANG1, VEGF, and HIF1α. These interact as agonists and antagonists to the TIE receptor. Therefore, further investigation is needed prior to the potential development of efficient anti-TIE therapies. Lastly, we used EVs which were produced from two GC cell lines. It is possible that EVs from fresh human gastric cancer tumors are affected by their native tumor microenvironment; thus, their content differs. We plan to adopt this approach to improve our biological model.

Conclusions
We demonstrated that GC-EVs promote angiogenesis by enhancing endothelial cells migration, invasion, and tube formation. Proteomic analysis of these EVs revealed high expression of a proangiogenic protein, ANG2, supporting the premise that these EVs are important mediators of tumor angiogenesis. This study provides a new mechanism via which GC cells induce angiogenesis. Such a mechanism may be used as a target for antiangiogenic therapy.