Inflammation-induced endothelial cell-derived extracellular vesicles modulate the cellular status of pericytes

Emerging lines of evidence have shown that extracellular vesicles (EVs) mediate cell-to-cell communication by exporting encapsulated materials, such as microRNAs (miRNAs), to target cells. Endothelial cell-derived EVs (E-EVs) are upregulated in circulating blood in different pathological conditions; however, the characteristics and the role of these E-EVs are not yet well understood. In vitro studies were conducted to determine the role of inflammation-induced E-EVs in the cell-to-cell communication between vascular endothelial cells and pericytes/vSMCs. Stimulation with inflammatory cytokines and endotoxin immediately induced release of shedding type E-EVs from the vascular endothelial cells, and flow cytometry showed that the induction was dose dependent. MiRNA array analyses revealed that group of miRNAs were specifically increased in the inflammation-induced E-EVs. E-EVs added to the culture media of cerebrovascular pericytes were incorporated into the cells. The E-EV-supplemented cells showed highly induced mRNA and protein expression of VEGF-B, which was assumed to be a downstream target of the miRNA that was increased within the E-EVs after inflammatory stimulation. The results suggest that E-EVs mediate inflammation-induced endothelial cell-pericyte/vSMC communication, and the miRNAs encapsulated within the E-EVs may play a role in regulating target cell function. E-EVs may be new therapeutic targets for the treatment of inflammatory diseases.

These miRNAs can be carried by E-EVs; however, their roles in E-EV-mediated cell-to-cell communication are not yet known.
Vascular endothelial cells and pericytes/vascular smooth muscle cells (vSMCs) are juxtapositioned to each other in blood vessels 22 . The interactions between these two cell types are important for the regulation of vascular integrity, and perturbation of their interaction has been observed in many diseases, including inflammatory diseases that cause vascular dysfunction, such as disturbance of the blood brain barrier (BBB) in cerebral blood vessels [23][24][25][26] . Here, we aimed to determine the involvement of EVs in cerebrovascular endothelial cell-pericyte communication in inflammatory disease. We found that the E-EVs induced by inflammatory stimuli carry numerous specific miRNAs and can induce pericyte responses to endothelial cells. These results suggest that E-EVs are an important mediator of vascular cell communication in inflammatory conditions.

Results
Induction of inflammatory responses in cerebrovascular endothelial cells. To analyze the pathobiology of E-EVs released in inflammatory conditions, we developed a reproducible in vitro system to induce the production of E-EVs from b.End5 cells, a cerebrovascular endothelial cell line. First, we confirmed that b.End5 cells expressed the LPS receptor TLR4/MD-2 complex under unstimulated conditions by immunocytofluorescence (Fig. 1a). The mRNAs of the inflammatory cytokine receptors Tnfr1 (for TNF-a), Il1r (for IL-1b), and Ifngr1 (for IFN-c) were detected in unstimulated b.End5 cells by conventional RT-PCR (Fig. 1b). The gene expression levels were consistent up to 24 hours after stimulation with a high dose of CytoCombo 1 LPS (a mixture of TNF-a, IL-1b, IFN-c, and LPS; Supplementary Table 1).
As inflammatory stimuli have been reported to upregulate IL-6 and ICAM-1 expression levels 27,28 , we determined the inflammatory responses in b.End5 cells to inflammatory cytokine and endotoxin exposure by measuring Il6 and Icam1 gene expression. In real-time PCR analyses, a high dose of CytoCombo 1 LPS (Supplementary Table 1) induced Il6 and Icam1 mRNA expression, which peaked after 3 hours of stimulation ( Fig. 1c and d). LPS alone also induced Il6 and Icam1 mRNA expression, but at lower levels than CytoCombo 1 LPS ( Fig. 1c and d). In accordance with the Icam1 mRNA expression (Fig. 1d), the immunoreactivity of ICAM-1 in b.End5 cells was detectable at 3 hours and peaked at 6 to 12 hours after CytoCombo 1 LPS stimulation (Fig. 1e).
To determine the optimal stimulatory conditions for inducing inflammatory responses in b.End5 cells, the dose effects of the inflammatory stimuli were determined by real-time PCR analyses of Il6 and Icam1 mRNA at 3 hours after stimulation. The expression of Il6 and Icam1 mRNA was significantly induced by CytoCombo 1 LPS, CytoCombo (a mixture of TNF-a, IL-1b, and IFN-c), and LPS ( Fig. 1f and g). However, the mRNA levels induced by CytoCombo 1 LPS were significantly higher than those induced by the other stimuli at all doses examined. The mRNA levels induced by the low and middle doses of LPS were significantly higher than those induced by the low and middle doses of CytoCombo, whereas the high doses of both stimuli induced equivalent mRNA expression levels ( Fig. 1f and g). The induction of Il6 and Icam1 mRNA was not significant when b.End5 cells were stimulated with either cytokine alone. As an apoptotic cell death triggered by the inflammatory cytokines and LPS 29,30 , has been reported to induce the release of apoptotic bodies 4 , we examined apoptosis in b.End5 cells that were exposed to inflammatory stimuli. As compared with non-stimulated control, 3-hour exposure to any kinds of stimulants did not significantly increase the cleaved caspase 3-positive apoptotic cell. Twenty four-hour exposure to the stimulants except for that of low and middle dose of CytoCombo significantly increased cleaved caspase 3-positive apoptotic cell ( Supplementary Figure 1a and b). Based on these data, we exposed b.End5 cells to CytoCombo 1 LPS, LPS, and CytoCombo for 3 hours when induction of apoptosis is still not evident, and then examined the inflammatory simulation-induced E-EV release from these cells.
Inflammation induced E-EV release from cerebrovascular endothelial cells. We next examined E-EV release from b.End5 cells that were stimulated by CytoCombo 1 LPS. To identify the E-EVs released from b.End5 cells, the plasma membranes of these cells were prelabeled with PKH26 and these cells were then exposed to the inflammatory stimuli. Three hours after exposure to a high dose of CytoCombo 1 LPS, PKH26 fluorescence was detected on the plasma membrane of b.End5 cells (Fig. 2a, upper row). Then, we identified PKH26-positive objects of submicron to micron size within the E-EVs fractionated from the culture medium of PKH26-positive b.End5 cells after a 3-hour stimulation with the high dose of CytoCombo 1 LPS, and confirmed that these fractionated materials were b.End5-derived globules. We further characterized them using the extracellular vesicle markers, such as vascular endothelial growth factor receptor 2 (VEGFR-2) and CD62E (E-selectin) 31 . These globular objects in E-EV fraction were immunoreactive for VEGFR-2 and CD62E under confocal microscopy (Supplementary Figure 2). In addition, immunoreactive bands of VEGFR-2 and CD62E were enriched in the E-EV fraction than in b.End5 cell lysates by western blot analyses (Supplementary Figure 3).
The surface of the plasma membrane of the inflamed endothelial cells was examined by scanning electron microscopy (SEM). In most of unstimulated b.End5 cells (Initial), we observed few cellular processes like microvilli and limited number of sheddings on the surface of the plasma membrane (Fig. 2b, upper), whereas small fraction of unstimulated cells (2.38 6 5.83%) showed similar changes as seen in those with stimulation. At 10 seconds after the addition of high dose CytoCombo 1 LPS to the culture medium, many sheddings were emerged on the surface of plasma membrane, and the microvilli turned to be obscure in all stimulated cells (100%; Fig. 2b, upper row). Higher magnification images further revealed the tiny and dome-shaped swellings (less than 0.1 mm in diameter) on the surface of plasma membrane, which structure may correspond to the early phase of sheddings (Supplementary Figure 4a). The sheddings were further increased at 10 minutes, and most of them ranged between 0.1 mm and 0.4 mm in diameter (Fig. 2b, upper row). At higher magnification, sheddings of larger-size ($0.3 mm) and those of aggregated form (up to 3 mm) were increased as compared with those at 10-second of stimulation (Supplementary Figure 4b). In addition, similar shedding-like structures were observed on the bottom surface of the cells that were exposed to high-dose CytoCombo 1 LPS by using the interference contrast microscopy (DIC) mode of a confocal microscope (Fig. 2b, bottom row). These data strongly suggest that E-EV production is mainly membrane-shedding type 15,32 and may be partly from multivesicular bodies (MVBs) 33 , and the plasma membrane shedding of E-EVs induced by inflammatory stimulants is very acute. In addition, our findings may support the possibility that E-EVs could be produced bilaterally, on both the luminal and abluminal surfaces, from inflammation-stimulated endothelial cells in cerebral blood vessels in vivo.
Next, E-EV release from b.End5 cells prelabeled with PKH26 was quantitatively analyzed by flow cytometry. Based on the above mentioned SEM study, we focused on particles between 0.3 mm and 3.0 mm in diameter, which corresponds to the diameter of the E-EVs released in membrane-shedding mode 32,34,35 . The preculture medium containing CytoCombo 1 LPS included many particles between 0.3 mm and 3.0 mm in diameter that were negative for PKH26, such as FBS-derived EVs, and EVs from MVBs (Fig. 3a, upper graphs). In contrast, after 3 hours of culture with high dose CytoCombo 1 LPS, the culture medium contained a significant number of PKH26-positive particles between 0.3 mm and 3.0 mm in diameter (Fig. 3a, bottom graphs). These findings indicated that the plasma membrane-intercalating nature of PKH26, with which  the b.End5 cells were prelabeled, allowed us to discriminate foreign substances in the medium from E-EVs and to precisely measure the release of E-EVs. In a comparative assay, the three types of stimulants significantly induced the release of PKH26-positive E-EVs compared to that released from the non-stimulated control (stimulant, P-value vs. the control; CytoCombo 1 LPS, P , 0.0001; CytoCombo, P , 0.01; LPS, P , 0.05; Fig. 3b). CytoCombo 1 LPS induced the release of E-EVs in a dose-dependent manner up to the high dose, and the release induced by CytoCombo 1 LPS was significantly higher than that induced by the other stimuli at the high dose ( Fig. 3b). At the high doses, CytoCombo and LPS individually induced the release of E-EVs to similar extents. These changes of E-EV production were well correlated with those of Il6 and Icam1 mRNA induction, both of which were induced by different inflammatory stimuli. Accordingly, these data suggested that E-EV production was correlated with the level of cellular inflammation. At the low and middle doses, CytoCombo induced E-EV release (P , 0.01 and P , 0.001, respectively, vs. the control) but not Il6 and Icam1 mRNA (both n.s. vs. the control; Fig. 1f and g). Thus, E-EV production in cerebrovascular endothelial cells appears to be a more sensitive cellular response than cytokine production based on relative induction by inflammatory stimulants.
Inflammation-induced E-EVs may target cerebrovascular pericytes. Endothelial cells and pericytes are juxtapositioned and functionally correlated in the vasculature 22,36 . Next, we determined if the miRNAs in E-EVs could mediate the inflammatory signaling between cere-  CytoCombo 1 LPS. The E-EV fractions obtained were washed 3 times with PBS, and then added to the culture medium of human brain pericytes (HBPCs). After a 24-hour incubation, immunocytofluorescence showed that many M-EGFP-positive E-EVs were incorporated into the cytoplasmic region of HBPCs (Fig. 4). Some of the M-EGFP-positive E-EVs were localized near the nucleus, as confirmed by the Y-Z cross-sectional image and single-plane sequential Z images (Fig. 4, upper right and bottom row). These findings suggest that the bioactive molecules encapsulated in the E-EVs were transferred into HBPCs.
To examine the bioactive molecules encapsulated in the E-EVs, we examined the molecules in the E-EV fractions from b.End5 cells with an miRNA array. Many miRNAs were upregulated in the E-EVs from cells stimulated by a high dose of CytoCombo 1 LPS compared to those in the E-EVs from non-stimulated control cells (Fig. 5a). Some of them were also upregulated in the E-EVs from cells stimulated with a high dose of CytoCombo, which induced an inflammatory response in b.End5 cells, although to a lesser extent than CytoCombo 1 LPS (Fig. 5a). These miRNAs sorted into the E-EVs were classified by Ingenuity Pathway Analysis (IPA). A significant proportion of the miRNAs were approved to be related to Inflammatory Disease and Inflammatory Response (bright red) in Disease and Disorders of the IPA (Supplementary Figure 5). Moreover, the miRNAs were abundant in those related to Organismal Injury and Abnormalities, which may associate with inflammation, over-represented (dark red). These data suggest that inflammatory stimuli are able to sort the inflammation-related miRNAs into E-EVs.
Among the various inflammation-related miRNAs detected (Fig. 5a), we focused on two, miR-328-3p and let-7d-3p, because these miRNAs were extensively induced after stimulation with CytoCombo 1 LPS, and were significantly induced, but to a lesser extent, after CytoCombo stimulation. Real-time PCR showed that the cultured HBPCs expressed Yin and yang 1 (YY1), which is a target mRNA of let-7d-3p and is an orthologue of mouse Yy1 that was identified through in silico analysis (Figure 5b), but not the target mRNAs of miR-328-3p including Proviral integration site for Moloney murine leukemia virus 1 (PIM1), Testis-specific kinase 2 (TESK2), and Ataxin-2 binding protein 1 (A2BP1) (data not shown). The well-conserved, mature let-7d-3p sequences in mice and humans are identical (Fig. 5c) and were shown to target YY1 mRNA through in silico analyses (Fig. 5d). VEGF family proteins are synthesized in pericytes/vSMCs 37,38 and are believed to be the downstream targets of YY1, since YY1 knockdown upregulates their expression levels, in which the upregulation of VEGF-B was most striking among VEGF family proteins 39 . Therefore, we investigated whether inflammation-induced E-EVs modulated the expression of VEGF family mRNAs and proteins. We observed that VEGF-B mRNA and protein levels were low in the controls, but were extensively induced in the E-EV-supplemented group (Fig. 6b, e, f). In contrast, E-EV supplementation did not affect VEGFA, VEGFC, or PLGF mRNA levels (Fig. 6a, c, d). The miRNAs carried by E-EVs may mediate the inflammatory responses of endothelial cells by regulating gene expression profiles in pericytes/vSMCs, especially in cerebrovascular pericytes.

Discussion
Emerging evidence has suggested the importance of EVs in cell-tocell communication as carriers of small molecules such as peptides, proteins, lipids, mRNAs, and miRNAs 40,41 . The bioactive molecules encapsulated in EVs can be transferred to and function in target recipient cells 6,[42][43][44] . In addition, miRNAs act as regulatory molecules of diverse biological phenomena, such as immune responses, inflammation, and tumor growth 18,19 . However, our knowledge of EV-and encapsulated miRNA-mediated cell-to-cell communication is mainly limited to tumor cell biology. Although it is known that the proinflammatory agent TNF-a induces the release of E-EVs 45 , the significance of these E-EVs remains to be elucidated. To the best of our knowledge, we demonstrated here, for the first time, that the inflammation-related miRNAs carried by the E-EVs from endothelial cells mediated inflammatory responses in pericytes. The induction of E-EV shedding was very acute and sensitive compared to previously established inflammation markers like IL-6 and ICAM-1 27,28 , and it may be useful as a very sensitive marker for inflammatory diseases.
In recent years, a variety of pathogens and disease states have been shown to affect the composition and function of these EVs. In vitro studies demonstrated that environmental stresses such as heat, hypoxia, irradiation, and changing of conditioned medium, differentially modified the composition of EVs [46][47][48][49] . Specific coding and non-coding RNAs, retrotransposon RNAs, Alu transposable elements, and chromosomal and mitochondrial DNA have been reported to be enriched in tumor cell-derived EVs [50][51][52][53][54] . The results of our present study showed that when exposed to inflammatory stimuli, cerebrovascular endothelial cells could sort many inflammation-related miRNAs into E-EVs ad initium. Moreover, the type and amount of miRNAs that were sorted into the E-EVs was strongly correlated with the severity of endothelial inflammation.
EVs from tumor cells could educate the bone marrow progenitor cells to mobilize out from their niche 55 . Vascular endothelial cells in tumors could be ''educated'' by EVs derived from tumor cells to exhibit a highly proliferative phenotype and karyotype abnormalities 7,56 . Along these same lines, our results show that the molecules in the inflammation-induced E-EVs function in target pericytes to increase VEGF-B mRNA and protein expression. Furthermore, this ''education'' was suggested to be mediated by the miRNA encapsulated in the E-EVs since VEGFB is thought to be a downstream target of the highly-accumulated miRNA in the inflammation-induced E-EVs. Similarly, various organ-derived EVs have been shown to be incorporated into bone marrow cells in vitro 57 . As a result, the mRNAs were delivered and the expression of organ-specific mRNAs was induced in the bone marrow cells. E-EVs and encapsulated miRNA were shown to be involved in the cascade of cellular events in the inflammatory vascular response.
VEGF-B is a specific ligand of VEGFR-1 that is expressed in endothelial cells and pericytes/vSMCs 58,59 . VEGFR-1 mediates pathological angiogenesis by targeting both endothelial cells and pericytes in the pathological neovascularization of choroidal and retinal tissues 60 . Similarly, VEGFR-1 stimulates the endothelial differentiation and the formation of blood vessels in infantile hemangioma 61 and vascular formation in implanted tumors 62 . Furthermore, VEGFR-1 ablated vascular pericytes and aggravated vascular leakage to induce cancer-associated retinopathy 63 . Accordingly, inflammation-induced E-EVs enhanced VEGF-B expression in pericytes, which may mediate pathological neovascularization and/or vascular leakage in response to inflammation of endothelial cells. Furthermore, since E-EVs are carried in the blood stream in many pathological conditions, E-EVs may spread the local inflammatory responses of endothelial cells to the systemic vasculature. Inhibition of E-EVs may limit the extent of inflammation-induced cellular responses from endothelial cells to the surrounding cells, and E-EVs may be a new therapeutic target for inflammatory diseases.
In conclusion, we developed a reproducible in vitro E-EV production system and a quantitative E-EV measurement method. Using these tools, we demonstrated that inflammation-induced E-EVs could ''educate'' target cells to modulate growth factor production. We also showed evidence to suggest that part of this cell education signal might be an miRNA encapsulated in the E-EVs. Our findings offer a new clinical approach, targeting E-EVs, for the prevention and treatment of acute and chronic inflammatory diseases. Further experiments are needed to elucidate the molecular mechanisms by which E-EVs transfer their ''educating'' messages. Methods Cell lines and culture conditions. The mouse brain endothelial cell line, b.End5, and human brain pericytes (HBPCs) were purchased from DS Pharma Biomedical (Osaka, Japan), and from Applied Cell Biology Research Institute (Kirkland, WA), respectively. The cells were cultured using the EBM-2 bullet kit (Takara, Kyoto, Japan) at 37uC in a humidified atmosphere containing 5% CO 2 . Membrane-localized EGFP (M-EGFP)-expressing b.End5 cells were generated by transfection of the GAP43 membrane targeting sequence-Egfp fusion gene, which was constructed as follows. Briefly, the human GAP43 membrane targeting sequence was cloned from the plasmid (Clone ID No. IRAL020F10, RIKEN BRC, Tsukuba, Japan) and inserted into the multi-cloning site of the pEGFP-N1 vector (Takara). The vector containing the resultant fusion gene, consisting of the GAP43 membrane targeting sequence and Egfp, was transfected into b.End5 cells using X-tremeGENE transfection reagent (Roche, Basel, Switzerland) according to the manufacturer's protocol.
Reagents. The inflammatory cytokines TNF-a, IL-1b, and IFN-c were purchased from PeproTech (Rocky Hill, NJ). Lipopolysaccharide (LPS) was obtained from List Biological Laboratories (Campbell, CA). TNF-a, IL-1b, and IFN-c were used separately, as a mixture (referred to as CytoCombo), or as a mixture of CytoCombo and LPS (CytoCombo 1 LPS) (Supplementary Table 1). The plasma membrane intercalating fluorescent dye PKH26 was purchased from Sigma-Aldrich (St. Louis, MO) and was used according to the manufacturer's protocol. The cell tracer fluorescence dye 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CSFE) was purchased from eBioscience (San Diego, CA) and was used according to the manufacturer's protocol.
Fractionation of endothelial cell-derived extracellular vesicles (E-EVs). Tencentimeter dishes coated with type I collagen (BD Biosciences, San Jose, CA) of confluent b.End5 cells were rinsed with PBS, and stimulated with medium containing CytoCombo 1 LPS, CytoCombo, or LPS alone to induce inflammation. Culture medium without any stimulant was used as a control. After a 3-hour incubation, the conditioned culture medium was centrifuged at 1,000 rpm for 10 minutes at 4uC to eliminate the cellular debris. The supernatant was then ultracentrifuged at 100,000 3 g for 90 minutes at 4uC, and the pellets were used as the E-EV fraction in the following analyses.
FACS analysis. E-EVs released into culture media were analyzed by flow cytometry. E-EVs in the media were gated as events of 0.3 to 3.0 mm in diameter, corresponding to the size of the membrane shed EVs [32][33][34] . E-EVs were further gated as PKH26positive events, by using b.End5 cells pre-labeled with PKH26, a membrane intercalating dye. This gating allowed us to discriminate E-EVs from the other microparticles of similar diameters in the medium. We used 3.0-to 3.4-mm rainbow calibration particles (BD Biosciences) as benchmarks. Flow cytometry was performed with a FACSCanto II (BD Biosciences) and FlowJo software (Tree Star, Ashland, OR) was used for data analysis. Scanning electron microscopy. Endothelial b.End5 cells were seeded on cover slips and cultured until confluent. Confluent cells were then exposed to inflammatory stimuli (CytoCombo 1 LPS) for 10 seconds or 10 minutes. After stimulation, the cells were fixed with 2% formaldehyde (Wako Pure Chemical Industries, Osaka, Japan) and 2.5% glutaraldehyde (Wako Pure Chemical Industries) prepared in 0.1 M sodium phosphate buffer (pH 7.4) for 2 hours at room temperature, and then postfixed in 1% osmium tetroxide in the same buffer for 2 hours on ice. Cells were then dehydrated in a graded series of ethanol, substituted with t-butyl alcohol and freeze-dried. Cells were sputter-coated with osmium and then observed in a S4200 scanning electron microscope (Hitachi).
RNA extraction, cDNA synthesis, conventional PCR, and real-time PCR. Total RNA was isolated from b.End5 cells or HBPCs using the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Total RNA (100 ng) was reverse-transcribed as described in the PrimeScript RT reagent Kit protocol (Takara). Conventional PCR was performed with a Bio-Rad C100 Thermal Cycler (Bio-Rad Laboratories, Hercules, CA). The cDNAs were diluted 1520 with the reaction mixture from the Ex Taq PCR kit (Takara), and the PCR was performed by standard 3 step amplification protocol consisting of hot start enzyme activation at 95uC for 10 minutes followed by 35 cycles of amplification at 95uC for 30 seconds, 60uC for 60 seconds, and 72uC for 60 seconds. For real-time PCR, which was performed with a Takara Thermal Cycler Dice Real Time System TP800 (Takara), the cDNAs were diluted 1525 in the reaction mixture containing SYBR Premix EX Taq II (Takara). The real-time PCR program consisted of hot start enzyme activation at 95uC for 10 seconds followed by 45 cycles of amplification at 95uC for 10 seconds and 60uC for 40 seconds. Finally, to obtain a dissociation curve, a final cycle was performed at 95uC for 1 minute, 60uC for 30 seconds, and 95uC for 10 seconds. For data analysis, the mouse or human glyceraldehyde-3-phosphate dehydrogenase (Gapdh or GAPDH) housekeeping gene was used as an internal control. Induction values were calculated using analysis software (Takara). Primer sequences are available upon request from the TAKARA BIO INC. website (http://www.takara-bio. co.jp/).
MiRNA array. The miRNA expression profile was obtained by using a Mouse miRNA microarray 8 3 15 K (Agilent Technologies, Santa Clara, CA) according to the manufacturer's protocol. Total RNA was prepared from fractionated E-EVs using the miRNA easy Mini Kit (Qiagen). RNA quality was measured with an Agilent 2100 Bioanalyzer (Agilent Technologies). Synthesis of probes, and hybridization to the Mouse miRNA microarray 8 3 15 K were performed by using the miRNA Complete Labeling and Hyb Kit (Agilent Technologies). Hybridized samples were analyzed using a Microarray Scanner (G2505B, Agilent Technologies) and GeneSpring GX software (Agilent Technologies) according to standard protocols. The miRNA array data were analyzed using Ingenuity Pathway Analysis (Ingenuity Systems, http:// www.ingenuity.com/).
Western blotting. For E-EV marker analyses, E-EVs that were fractionated from the culture media of b.End5 cells after CytoCombo 1 LPS stimulation were used as sample. For the control, b.End5 cells were used. For VEGF-B analyses, E-EVs that were fractionated from the culture media of b.End5 cells after CytoCombo stimulation were added (100 mg) to HBPCs cultured in a collagen type I-coated 24well plate (BD Bioscience), and the plate was incubated at 37uC for 48 hours. Sample preparations and all other procedures for western blotting have been described elsewhere 64 . Briefly, cells were lysed using RIPA buffer (Thermo Fisher, Waltham, MA) on ice. Whole cell lysates were separated by SDS-PAGE and then electrophoretically transferred to polyvinylidene difluoride membranes. The membranes were incubated for 1 hour at room temperature in a 50 mM Tris buffer containing 150 mM NaCl, 0.05% Tween 20, and 5% BSA for blocking. The membranes were then probed with rabbit anti-VEGFR-2 (15500; Cell Signaling Technology), rabbit anti-CD62E (15100; Abcam), rabbit anti-VEGF-B (15100; Abcam), and mouse anti-GAPDH (155,000; Merck Millipore, Billerica, MA) antibodies. The membranes were washed in a 50 mM Tris buffer containing 150 mM NaCl, and 0.05% Tween 20, and were incubated with the appropriate horseradish peroxidase-conjugated secondary antibodies. Immunoreactive bands were detected using enhanced chemiluminescence reagents (GE Healthcare Bio-Sciences AB, Uppsala, Sweden) according to the manufacturer's instructions.
Statistical analysis. Statistical significance was determined using Student's t-test, or one-or two-way analysis of variance (ANOVA) with Tukey's or Newman-Keuls multiple comparisons as a post hoc analysis for ANOVA. P values less than 0.05 were considered statistically significant. Graphs were drawn using GraphPad Prism 6 (GraphPad Software, Inc., La Jolla, CA). Quantified data are presented as mean 6 SEM.