Inter-Cellular Exchange of Cellular Components via VE-Cadherin-Dependent Trans-Endocytosis

Cell-cell communications typically involve receptor-mediated signaling initiated by soluble or cell-bound ligands. Here, we report a unique mode of endocytosis: proteins originating from cell-cell junctions and cytosolic cellular components from the neighboring cell are internalized, leading to direct exchange of cellular components between two adjacent endothelial cells. VE-cadherins form transcellular bridges between two endothelial cells that are the basis of adherence junctions. At such adherens junction sites, we observed the movement of the entire VE-cadherin molecule from one endothelial cell into the other with junctional and cytoplasmic components. This phenomenon, here termed trans-endocytosis, requires the establishment of a VE-cadherin homodimer in trans with internalization proceeding in a Rac1-, and actomyosin-dependent manner. Importantly, the trans-endocytosis is not dependent on any known endocytic pathway including clathrin-dependent endocytosis, macropinocytosis or phagocytosis. This novel form of cell-cell communications, leading to a direct exchange of cellular components, was observed in 2D and 3D-cultured endothelial cells as well as in the developing zebrafish vasculature.


Introduction
Intercellular communications are critically important for fundamental functions of multicellular organisms including plants and humans. These signals are commonly mediated by transmembrane channels transferring small molecules or by receptors binding ligands, such as soluble, cell surface proteins, or extracellular matrix components. Endothelial and epithelial cells are unique with regard to their ability to form tight cell-cell adhesion structures which are dynamically remodeled during tissue morphogenesis. These structures also play an important role in regulation of a number of biological processes, including permeability, cell trafficking, and signal transduction from soluble proteins and extracellular matrix components. The main adhesive structure of endothelial junctions is the vascular endothelial cadherin (VEC)-based complex. This complex is capable of controlling a number of endothelial functions via VEC internalization, endocytic trafficking, and recycling [1,2].
Among its many biological functions, VEC plays an important role in the maintenance of vascular integrity due to its involvement in the formation of adherens junction [3]. The VEC extracellular domain includes five cadherin repeats, with the most N-terminal repeat being critically involved in the formation of homophilic adhesions in trans with VEC expressed on neighboring cells [4,5]. The intracellular domain of VEC forms protein complexes with a number of cytoplasmic proteins, including b-catenin and p120 catenin, that can then bind to a-catenin, linking VE-cadherin to the actin cytoskeleton [1,6].
As with most transmembrane proteins, cadherins have been reported to be internalized and recycled to the plasma membrane [7][8][9]. In the process of studying VEC endocytosis using fluorescently tagged VEC examined by spinning disk confocal microscopy in live cells, we observed that VEC tagged proteins were sometimes observed in a different endothelial cell than those they were originally expressed in. Examination of this phenomenon uncovered an alternative mode of VEC trafficking, here termed trans-endocytosis, involving direct internalization of the VEC-VEC dimer bridging the two endothelial cells into one of the cells. The resulting vesicles contained not only the whole VEC molecule, but also VEC-associated proteins and even cytoplasmic components, including EGFP and siRNAs. Importantly, the process of VEC-dependent trans-endocytosis was not affected by the use of inhibitors of clathrin-dependent endocytosis, macropinocytosis or phagocytosis but did require actomyosin force generation and Rac1 activity.

Results
We first employed the COS7 cell system commonly used for analyzing the function of ectopically expressed VEC. We transduced cells with fluorescently tagged VEC by lentiviral expression system. COS7 cells normally use N-cadherin to form cell-cell junctions; however, forced expression of VEC excludes N-cadherin from junctions resulting in the formation of VEC-based cell-cell adhesion ( Figure S1A). VEC-EGFP and VEC-TagRFPT were first transduced into separate cell populations. As the VECbased junction is subject to dynamic remodeling with constitutive internalization and recycling, we observed abundant intracellular vesicles containing tagged VEC. In control cells expressing only VEC-EGFP or VEC-TagRFPT, there was almost no ''opposite color'' fluorescence detected, save for very slight auto-fluorescence in red or green channel, respectively (data not shown). However, mixing of the two populations of COS7 cells expressing either VEC-EGFP [10] or VEC-TagRFPT created a cell-cell interface consisting of a mixture of VEC-EGFP and VEC-TagRFPT (data not shown). When the two populations of COS7 cells expressing either VEC-EGFP or VEC-TagRFPT were co-cultured, we observed internalized VEC fused proteins of both flourophores: i.e. VEC-TagRFP molecules in VEC-EGFP expressing cells (data not shown). Because VEC can be cleaved at the membranecytoplasm interface [11,12], EGFP fluorescent protein was fused to the C-terminus of VEC to permit tracking of the full length VEC molecule. VEC-EGFP was found in TagRFPT positive cells, and vice versa, indicating that the full length VEC was transferred from one cell to the adjacent cell.
The same VEC trans-endocytosis was observed in isolated primary microvascular endothelial cells from mice ( Figure 1A and 1B). To prove that these trans-endocytosed VEC-EGFP molecules were internalized by the cells, we performed immunostaining of fixed cells by anti-GFP antibody with/without permeabilization. Primary microvascular endothelial cells were isolated separately from VEC-EGFP knock-in mice and Rosa26-mTmG mice expressing mTomato fluorescent protein in all cells and fixed after 24 hours of the co-culture. Proteins inside of non-permeabilized cells would not be stained by antibodies [13]. The fusion protein, VEC-EGFP, can be detected by 488 nm laser excitation without permeabilization. If VEC-EGFP molecules are transendocytosed by Rosa26-mTmG endothelial cells and then internalized, VEC-EGFP in Rosa26-mTmG endothelial cells will be detected by anti-GFP antibody only in permeabilized cells, but not in cells without permeabilization. The results of the immunostaining showed that the VEC-EGFP molecules were detectable in the EGFP channel and could not be stained by anti-GFP antibody when cells had not been permeabilized before immunostaining (Arrows in Figure 1B, upper panel). However, the VEC-EGFP molecules were clearly stained by anti-GFP antibody when cells had been permeabilized (Arrow heads in Figure 1B, lower panel). These results suggest these trans-endocytosed VEC-EGFP molecules are inside of cells.
Furthermore, when VEC-EGFP and VEC-TagRFPT were transduced by lentivirus in human umbilical venous endothelial cells (HUVECs), the trans-endocytosis of VEC was also observed ( Figure 1C and 1D). And when cells were transduced with VEC with small tag, VEC-FLAG and VEC-HA, the trans-endocytosis of VEC was also observed (unpublished data).
To study the kinetics of this process, two sets of cells, COS7 cells expressing either VEC-EGFP or VEC-TagRFPT were mixed together. Then we tracked the same cell over time and measured the number of trans-endocytosed VEC-TagRFPT molecules in the EGFP positive cell. There was a gradual increase of transendocytosed molecules over time, starting at 1 hr after formation of cell-cell contacts that reached plateau by 3 hours (Figure S1B and S1C). When repeated in HUVECs cells, the kinetics were similar (unpublished data).
To exclude the possibility that a fluorescent signal is arising from cellular auto-fluorescence or bleed-through originating from fluorescent cross-talk, we made a VEC fused protein using mKikGR fluorescent protein. Fluorescence of mKikGR can be irreversibly converted from green to red by photo-activation using UV or violet light (350-410 nm) [14,15]. When the two populations of HUVECs expressing VEC-EGFP and VEC-mKikGR were co-cultured, there was almost no red fluorescence detected before photo-activation ( Figure S1D). After photo-activation, however, we observed red VEC-mKikGR molecules in VEC-EGFP expressing cells ( Figure S1E). The red fluorescence observed is not from auto-fluorescence, because they were not detected before photo-activation. Regardless of the extent of activation of VEC-mKikGR, enhanced red fluorescence in VEC-EGFP expressing cells meant the presence of VEC-mKikGR molecule in the VEC-EGFP expressing cells. This result clearly shows the red fluorescence was not due to auto-fluorescence, but due to the presence of trans-endocytosed whole VEC-mKikGR molecules from neighboring cells. When HUVECs expressing VEC-EGFP and VEC-mKikGR in the normal growth medium were mixed together, there was a gradual increase in the number cells demonstrating the presence of trans-endocytosis molecules over time reaching over 80%, by 8 hours (Figure S1F and S1G). Virtually all cells that had cell-cell contact with neighboring cells demonstrated trans-endocytosis of VEC with a frequency of over 80%, 10 h after plating.
Because fluorescent proteins were fused to the carboxyl-terminus of VEC, which is located in the cytosol, these results suggest that the entire VEC molecule containing the fluorescent protein is internalized, presumably via homophilic extracellular VEC-VEC interaction. To demonstrate that this indeed is the case, we expressed only one set of VEC constructs (either VEC-EGFP or VEC-TagRFPT) in COS7 cells that do not normally express VEC, and co-cultured these COS7 cells with HUVECs that express endogenous VEC. When we mixed COS7 cells expressing VEC-EGFP with HUVECs, we could observe trans-endocytosis between an endogenous VEC in HU-VECs and an over-expressed VEC-EGFP in COS7 cells (Figure 2A and 2B and Movie S1). The trans-endocytosed VEC-EGFP molecules from a COS7 cell appeared to bud off from cell-cell junctions and be pulled into a HUVEC ( Figure 2B and Movie S1). No trans-endocytosis was detectable when COS7 cells expressing VEC-EGFP were mixed with COS7 cells that did not express endogenous VEC ( Figure 2C). These results demonstrate that VEC-VEC homophilic interaction is required for trans-endocytosis.
Although trans-endocytosis appears to occur at cell-cell contacts, it is possible that the VEC complex is released as an exosome and is then incorporated into other cells by vesicular uptake [16,17]. To evaluate this possibility we first cultured VEC-EGFP-or VEC-TagRFPT-transduced HUVECs on different plates and exchanged the conditioned media. However, no transendocytosis was observed under these conditions (unpublished data). Next, we utilized the transwell plates where VEC-EGFPtransduced HUVECs were seeded on the upper chamber and VEC-TagRFPT-transduced HUVECs were seeded on the lower chamber. Serial observations after plating showed no evidence of trans-endocytosis ( Figure S2A), whereas the control plate in which these cells are mixed and plated clearly showed trans-endocytosed molecules. Additionally, media was fractionated for the exosomal fraction. This faction, while positive for the exosomal marker Syntenin, did not contain detectable VEC ( Figure S2B). Finally, we disrupted cell-cell junction formation by introducing a point mutation in the VEC extracellular domain required for homophilic VEC-VEC binding [18]. Transduction of HUVECs with mutated VEC (VEC-W49A-TagRFPT) completely eliminated trans-endocytosis ( Figure S2C). Taken together, these experiments demonstrate that trans-endocytosis occurs at VEC-formed cell-cell junction and is not driven by an exosome-mediated mechanism, but instead requires hemophilic VEC interactions.
We next set to examine the extent of the protein complex that is trans-endocytosed in this VEC-VEC interaction-driven process. To this end, we used a FLAG-tagged VEC fused to a-catenin (VEC/a-catenin-FLAG). Expression of this construct in HUVECs leads to formation of highly stabilized adherens junctions [20]. Nevertheless, the entire VEC/a-catenin-FLAG construct was observed to be trans-endocytosed by adjacent cells ( Figure 3D). Furthermore, cytosolic components such as EGFP were also transendocytosed by adjacent cells ( Figure 3E). When HUVECs expressing EGFP were co-cultured with HUVECs expressing VEC-TagRFPT, EGFP was detected in VEC-TagRFPT expressing cells ( Figure 3E). The trans-endocytosis of cytosolic components, like EGFP molecule, suggests that genomic components like mRNA and miRNA also trans-endocytosed by adjacent cells which would potentially convert the property of the adjacent cells. We confirmed that transfected siRNA molecules were also transendocytosed by adjacent cells ( Figure 3F). When HUVECs expressing iRFP were co-cultured with VEC-EGFP expressing HUVECs transfected with scramble siRNA labeled with Cy3, siRNA molecules were trans-endocytosed by iRFP expressing cells, and those siRNA molecules co-localized with VEC-EGFP ( Figure 3F). These data suggest that not only the entire VEC molecules, but also a large part of the protein complex associated with VEC including cytosolic components was trans-endocytosed by adjacent cells.
We next examined whether the VEC complex trans-endocytosis involved one of the standard endocytic pathways. The transendocytosed VEC from the neighboring cell did not co-localize with internalized transferrin (Figure S4A), while a small subset of Rab5-positive endosomes in the recipient cells co-localized with trans-endocytosed VEC ( Figure 4A and 4B). This suggests that the VEC trans-endocytosis is occurring via an unconventional clathrin-independent pathway, and that some of the internalized VEC-containing structures may be merging with the Rab5 pathway [21] after clathrin-independent internalization. We carried out co-culture of VEC-EGFP-expressing HUVECs with mRFP-Rab5 or mRFP-Rab5-DN expressing cells and found that trans-endocytosis was not inhibited by Rab5-DN expression ( Figure S4B). These results suggest Rab5 is not involved in the process of VEC trans-endocytosis.
The internalized complexes containing the VEC molecule from the neighboring cell co-localized with a subset of Rab7-positive endosomes and a small subset of Rab5-and Rab11-positive endosomes, in the receiving cells ( Figure S5). This suggests that trans-endocytosed complexes behave in a way similar to endosomes and may cycle back to the membrane or be degraded. To determine if trans-endocytosis is driven by contractile force generated by the actomyosin cytoskeleton, we used LifeAct polypeptide [27] to visualize F-actin bundles in living cells. Coculture of HUVECs expressing VEC-EGFP and LifeAct-TagRFPT showed a clear association between VEC positive trans-endocytosed structures with F-actin fibers ( Figure 5A and Movie S2). Treatment with (2)-blebbistatin, a selective inhibitor of non-muscle myosin II [28,29], inhibited trans-endocytosis of VEC ( Figure 5B, 5C and Movie S3). There were many transendocytosed molecules moving in both directions, i.e., VEC-EGFP molecules were trans-endocytosed by the RFP-positive cell and VEC-TagRFPT molecules were trans-endocytosed by the EGFPpositive cells, before adding (2)-blebbistatin at 20 minutes of timepoint in the Movie S3. However, after adding (2)-blebbistatin, the movement decreased dramatically ( Figure 5C and Movie S3). These data suggest that trans-endocytosis is driven by actomyosin contractility.
Vinculin is thought to be involved in the transfer of actomyosin contractile force to cadherin-complex [30,31]. To evaluate vinculin involvement in trans-endocytosis, we over-expressed acatenin constructs with mutations in the vinculin binding site (VBS) (a-catenin-DVBS-EGFP) [30]. Overexpression of a-catenin-DVBS-EGFP precludes binding of endogenous vinculin to the VEC-complex [30]. Trans-endocytosis of VEC occurred when native a-catenin or a-catenin-EGFP was expressed in HUVECs, but not when a-catenin-DVBS-EGFP was over-expressed in HUVECs ( Figure 5D, 5E and 5F). These results suggest that actin-mediated pulling force on the VEC complex is involved in the trans-endocytosis process.
Further studies revealed that treatment with NSC23766, a specific inhibitor of Rac1 that targets its activation by guanine nucleotide exchange factors, (GEFs), Tiam1 and TrinoN [32], inhibited the trans-endocytosis in a dose-dependent manner, whereas ML141, an inhibitor of cdc42 Rho family GTPase, had no effect ( Figure 6A and 6B). Of note, adherens junctions were not disturbed by NSC23766 treatment ( Figure 6A). Another Rac1 inhibitor, W56, also inhibited the VEC trans-endocytosis in a time-dependent manner ( Figure S6). These results suggest that VEC-driven trans-endocytosis is a Rac1-dependent process.
In order to further define the role of Rac1 in VEC transendocytosis, we used photo-activatable (PA) Rac1 constructs, which were fused with the photoreactive domain from phototropin. These constructs block Rac1 interactions until photoactivation [33]. Co-culture of HUVECs expressing PA-constitutively active Rac1 (Rac1-CA) and VEC-EGFP revealed that PA-Rac1-CA accumulated at cell-cell junction and co-localized with VEC-EGFP after its activation, while PA-dominant negative Rac1 (Rac1-DN) did not ( Figure 6C and 6D, Movie S4). This further demonstrates that Rac1 activation is directly involved in VEC trans-endocytosis at the cell-cell junction.
Finally, we confirmed that the existence of trans-endocytosis occurs in a three-dimensional culture using a HUVEC sprouting assay, and in vivo, using live zebrafish embryos (Figure 7). Coculture of HUVECs expressing VEC-EGFP and VEC-TagRFPT, in a three-dimensional fibrin gel culture showed VEC-EGFP molecules were trans-endocytosed by VEC-TagRFPT expressing HUVECs, forming tube-like structures in the fibrin gel ( Figure 7A and 7B, Movie S5).
Zebrafish VEC-EGFP (zVEC-EGFP) plasmids were injected into Tg(flkl:myr-mCherry) zebrafish using the Tol2 system to induce transient mosaic expression of zVEC-EGFP. At the one cell stage, VEC-EGFP plasmids were injected into Tg(flkl:myr-mCherry) zebrafish, which express ras-mCherry in all endothelial cells. For time lapse analysis, embryos were placed in 3.5 cm glass bottom dishes in egg water with 0.016% tricaine to prevent movement. Zebrafish VEC-EGFP endothelial expression was analyzed at 30-80 hr post fertilization using spinning disk microscopy focusing on the connection between the dorsal longitudinal anastomotic vessel (DLAV) and an intersegmental vessel (ISV), where was the best region to observe vessels to detect signals from VEC-EGFP. Time-lapse images demonstrated a VEC-EGFP positive structure budding to adjacent endothelial cells ( Figure 7C and 7D, Movie S6). This is similar to the manner in which VEC-EGFP molecules were trans-endocytosed by adjacent cells in cultured cells ( Figure 2B).

Discussion
To our knowledge, our results show a new paradigm of cellular communication that achieves a direct transfer of cellular materials between adjacent cells via cell-cell adhesion-dependent transendocytosis. VEC expression and homophilic interactions are necessary for this process, which depends on formation of a VEC bridge. Two cells making cell-cell contacts are pulling each other by actomyosin force via the VEC bridges. When the force from one of the cell exceeds that of the other cell, the cell internalizes the VEC complex with its partner in an actin/myosin-and Rac1- dependent manner. The model of VEC-driven trans-endocytosis is summarized in Figure 8.
The best way to provide further evidence of the VEC transendocytosis is to carry out a thorough EM analysis. This will provide strong evidence for this model and is the scope of future investigation. To further clarify the precise mechanism of transendocytosis, it is of particular interest to understand whether transendocytosis requires tension across VEC. It has been reported that fluid shear stress reduces the tension across VEC adhesion [34]. It would be interesting to demonstrate the relationship between tension across VE-cadherin adhesion and the rate of transendocytosis under physiological condition. Also, the distribution of pulling and pushing forces at the site of trans-endocytosis needs to be clarified. Using a validated FRET-based VEC tension sensor [34] will address this question.
We first observed ''trans-endocytosis'' in confluent endothelial monolayers and then demonstrated that it also occurred when endothelial cells are subconfluent using time-lapse imaging. It is very difficult to distinguish trans-endocytosis occurring while cells are subconfluent vs. the same process taking place in the confluent monolayer since fluorescent molecule trans-endocytosed during subconfluent period remain in the cells after the confluent monolayer forms. Distinguishing these two possibilities would require development of a more sophisticated approach. Most of data in the manuscript has been done in the subconfluent conditions. Further experiments to clarify whether ''trans-endocytosis'' occurs in contiguous monolayers would require the use of SNAP- [35] and/or CLIP-tags [36] to label membrane VEcadherin molecule only after endothelial cells become confluent.
We often observed that the trans-endocytosed proteins in the recipient cell exclusively contained the fluorescence arising from the VEC-fusion protein of the neighboring cell and not the one expressed by the recipient cell, and vice versa. We speculate that this is attributed to the chemical property of fluorescent proteins known to be affected by pH. After molecules enter the endocytic process, they are trafficked to early endosomes and subsequently late endosomes followed by lysosomes. The pH in the vesicles decreases as they progress to the later stage, from about pH 6. 8-5.9 in early endosomes to about pH 6.0-4.9 in late endosomes and lysosomes [37,38]. GFP is acid-sensitive, pKa = 5.8 [39], and TagRFP is highly acid-insensitive, pKa ,4.0 [40]. Since most of our observations were done in live cell settings, the intensity of a fluorescent tagged protein might have been variably affected, with the highest susceptibility in EGFP, depending on the pH environment in their vesicles. Another explanation for this unbalanced fluorescence is the presence of endogenous VEC that occurred when we used the HUVEC system. The formation of adherens junctions between fluorescent tagged VEC and endogenous native VEC resulted in the fluorescence of only one color.
To study the role of Rac1, we first tried over-expression of Rac1-CA and Rac1-DN constructs. The expression of either resulted in a major change in cell morphology, rendering transendocytosis studies nearly impossible. To avoid that problem, we used the photo-activatable form of Rac1 constructs, namely PA-Rac1-CA and PA-Rac1-DN ( Figure 6C and 6D). The benefit of PA-Rac1 constructs are that they act as constitutively or dominant negative form of Rac1 only when they are photo-activated by 488 nm laser. These results using PA-Rac1 constructs in Figure 6C and 6D clearly showed the role of Rac1 in trans-endocytosis. Because the use of Rac1 siRNA would result in the same morphology change problems that were observed with Rac1DN, we did not use Rac1 siRNA for the Rac1 signal depletion experiment.
The VEC trans-endocytosed structures may have a similar structure to the annular gap junction, with double bilayer lipid membranes containing some cytosolic components [41,42]. The exchange of cellular components may play a role as a primitive form of cell-cell communication similar to the process by which Escherichia coli exchange their sex factor with transmission of Ffactor plasmid by cell-cell contact [43]. Various kinds of processes similar to ''trans-endocytosis'' have been reported. Most are mediated by ligand/receptor interaction, such as EphrinB-Eph receptor [44], Notch-DLL ligand [45,46], CD47-SHPS-1 ligand [47], and CD86-CTLA-4 [48]. All of these events are due to binding between different proteins. In distinction, VEC-driven trans-endocytosis is due to homophilic binding between two molecules of the same protein in trans. This suggests that cadherins are important for the cell-cell communication between the same types of cells expressing the same type of cadherins, which are conventionally considered to be important for the segregation of different cell populations. Similar to our finding, fragments of membranes called argosomes are transported over large distances intercellularly in the Drosophila wing disk to form morphogen gradients [49]. Also, in epithelial cells, claudin, a component of tight junctions, is co-endocytosed into adjacent cells while other components of tight junctions, such as JAM and ZO-1, are not co-endocytosed with claudin [50]. Recently, b-integrindependent transfer of neutrophil myeloperoxidase from neutrophils to ECs has been reported [51].
In trans movement of cellular components in this type of communication has also been reported in the transfer of doublemembrane structures via very small spines, called spinules, in the adult rat hippocampus when studied by electron microscopy [52]. Tunneling of nano-tube like structures in immune cells [53,54] and epithelial cells [55], the cellular bridge in neural crest cells [56], and direct cell-cell contact dependent intercellular transfer of cellular components are other reported examples of similar in trans cellular communication [57]. However, none of these have identified the mechanism responsible for the observed effects. The mechanism described here underlying VEC-dependent transendocytosis may be the driving force behind these observations perhaps utilizing VEC-related molecules such as N-or Ecadherins.
Using this mechanism, neighboring cells can exchange cellular material, including genetic material such as miRNAs, a function previously attributed only to gap junctions [58][59][60] and exosomes [61,62]. The vesicles that are internalized from adjacent cells and have double bilayer lipid membranes may merge into an endosome, allowing the release of cytoplasmic components from neighboring cells to the recipient cell. While the full biological importance of this novel mode of cell-cell communication remains to be defined, the ability to exchange cellular components, exemplified by the transfer of cytosolic proteins (EGFP) and siRNA, argues for its potential significance.

Ethics Statement
All animal experiments are performed accordingly to the protocol approved by the institutional animal care and use committee of Yale University.

Cell culture and cells
Pooled human umbilical vein endothelial cells (HUVECs) from different donors (LONZA, Walkersville, MD, USA) were cultured in EBM-2 culture medium supplemented with EGM-2 bullet kit (LONZA) on gelatin-coated tissue dishes. HUVECs up to passage 6 were used for experiments. COS-7 cells, transformed African green monkey kidney fibroblast cells (ATCC, Manassas, VA, USA), and HEK293T cells (a gift from Dr. Yuanfei Wu, Yale University, New Haven, CT, USA) were cultured in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, Grand Island, NY, USA) supplemented with 10% fetal bovine serum and penicillin/streptomycin (Invitrogen). Primary mouse lung microvascular endothelial cells were isolated from mice lung by the method previously described [63,64] and cultured in DMEM supplemented with 20% fetal bovine serum, penicillin/streptomycin, non-essential amino acids (Invitrogen) and 20 mg/ml endothelial mitogen (Biomedical Technologies, Inc., Stoughton, MA, USA).

Immunofluorescence
Cells were washed with PBS, fixed with 2% paraformaldehyde (PFA) (18814, Polysciences, Inc, Warrington, PA, USA) in PBS for 10 minutes, permeabilized with 0.1% triton X-100 in PBS containing 2% PFA at room temperature for 5 minutes, and blocked with 3% bovine serum albumin (BSA) (IgG-Free, Protease-Free BSA (001-000-162), Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA) at room temperature for 30 minutes. Cells were washed with PBS and incubated with diluted primary antibody at 4uC overnight, washed three times with PBS and incubated with diluted Alexa Fluor-conjugated secondary antibodies (1:500 dilution) (Invitrogen) for 1 hour at room temperature. Then the slides were washed three times with PBS and mounted using Fluor-Gel (Electron Microscopy Sciences, Hatfield, PA, USA). Images were taken by Zeiss 510 laser scanning confocal microscopy, Zeiss 780 laser scanning confocal microscopy and PerkinElmer spinning disk microscopy confocal systems. Except images for Figure 7, images were taken at only one focal plane.

Spinning disk live cell microscopy
For live cell imaging microscopy, cells were seeded on 3.5 cm glass bottom culture dishes (MatTek Corporation, Ashland, MA, USA) coated with 0.1% gelatin (G1393, Sigma). Time-lapse live cell imaging was performed with a PerkinElmer UltraView VoX spinning disk confocal system with a Nikon eclipse Ti inverted microscope and a Prior NanoScanZ 250 Micron motorized stage, enclosed by temperature controlled chamber with a CO 2 environment system (IVS-3001) (In Vivo Scientific, LLC., St. Louis, MO, USA). The time-lapse images were acquired with a 60x oil NA 1.4 Plan Apochromat VC objective lens by using a chamber heater at 37uC. All images were enhanced for display with contrast adjustments in Volocity 6 software (PerkinElmer).

Photo-activation of photo-activatable constructs
For photo-activation of mKikGR tagged protein, cells were photo-activated by 100% of 405 nm laser for 5 sec using PerkinElmer UltraView VoX spinning disk confocal system. For photo-activation of PA-Rac1 constructs, cells were photo-activated by 488 nm laser at the same laser power (10%) for other acquiring condition using PerkinElmer UltraView VoX spinning disk confocal system. Only photo-activated molecules in VEC-EGFP positive cells were counted in order to avoid counting autofluorescence in cells.

Exosomal fractionation
To collect the exosomal fraction, we carried out fractionation using an ultracentrifuge (Sorvall MTX150, Thermo Scientific, Waltham, MA, USA) according to general methods for the exosomal fractionation [66]. Briefly, the culture medium, supplemented with exosomal free FBS, 24 hours after medium change was collected and centrifuged at 3,0006g for 10 min at 4uC. The supernatant was carefully transferred into a new tube and centrifuged at 10,0006g for 20 min at 4uC. The resulting supernatant was the carefully transferred into a new tube for ultracentrifuge and centrifuged at 100,0006g for 90 min at 4uC. The supernatant was carefully removed and pellet was resuspended with sample buffer for western blotting and used as the exosomal fraction.

siRNA transfer
HUVECs transduced with a lentiviral vector containing VEC-EGFP (HUVECs-VEC-EGFP) and without any lentiviral treatment (HUVECs-NT, non-treated) were maintained separately in 3.5 dishes. HUVECs-NT were transfected with Cy3 Labeled GAPDH siRNA according to Lipofectamine RNAiMax procedure. 24 hours after siRNA transfection, HUVECs-VEC-EGFP and HUVECs-NT were washed with PBS, trypsinized and mixed into 3.5 cm glass bottom dish. 4 hours after mix, the digital fluorescent images were captured by a PerkinElmer UltraView VoX spinning disk confocal system with a Nikon eclipse Ti inverted microscope.

Inhibitor experiments
Two portions of HUVECs, expressing VEC-EGFP and VEC-TagRFPT cultured in 10 cm tissue culture dishes, were pretreated with inhibitors (ML141, 1.25-10 mM; NSC23766, 12.5-200 mM) for an hour, then trypsinized and mixed into gelatincoated glass bottom dishes. Cells were further incubated in a CO 2 incubator at 37uC for 4 hour with inhibitors and analyzed using spinning disk microscopy.
The three-dimensional fibrin gel culture of HUVECs The three-dimensional culture experiments of HUVECs were performed using dextran-coated beads and a fibrin gel [67]. Two portions of HUVECs, expressing VEC-EGFP and VEC-TagRFPT cultured in 10 cm tissue culture dishes, were trypsinized and mixed with dextran-coated Cytodex3 microcarrier beads (Sigma) in 5 ml of EGM-2 MV media in 50 ml tube. Beads with cells were mixed by turning the tube upside down every 20 minutes for 3 hours at 37uC in 5% CO 2 incubator. After incubation, beads with cells were transferred to a petri dish and incubated overnight at 37uC in 5% CO 2 incubator. The following day, beads with cells were washed three times with EBM-2 supplemented with 2% FBS and re-suspended in 3 ml of EBM-2 supplemented with 2% FBS, 5 mg/ml fibrinogen (Sigma) and 50 mg/ml aprotinin (Sigma). Five hundred ml of bead solution was seeded on a 3.5 cm glass bottom dish and added to 0.5 units of thrombin (Sigma). After 15 minutes incubation at 37uC in 5% CO 2 incubator, fibroblasts (WI-38 cells, 400,000 cells/ml) in 1 ml of EGM-2 MV media were harvested onto fibrin gels. After 3-7 days incubation, sprouting HUVECs were observed using spinning disk microscopy.

Time-lapse imaging of zebrafish embryos
All experiments using zebrafish were approved by the IACUC of Yale University. Zebrafish (Danio rerio) embryos were raised as previously described [68]. The Tg(kdrl:ras-mCherry) [69] transgenic line was used for injection. The Tol2 system was used for the injection of the plasmid for transient expression in embryos [70]. Briefly, the zebrafish VEC gene was cloned from zebrafish cDNA and cloned into pDestTol2pA2, the transposon-donor plasmid with a fli1 promoter sequence and SV40 polyA signal. About 1 nl of a DNA/RNA solution, containing 25 ng/ul transposon-donor plasmid and 25 ng/ul transposase mRNA in 10 mM HEPES buffer, were co-injected into one-cell stage embryos using microinjection. The injected embryos were raised at 28uC in egg water with 0.002% 1-phenyl-2-thiourea to prevent pigment development. Embryos were placed for time lapse analysis in 3.5 cm glass bottom dishes in egg water containing 1% low melting point agarose with 0.016% tricaine to inhibit movement of the embryos. Zebrafish VEC-EGFP expression in endothelial cells of embryos was analyzed at 30-80 h post fertilization (hpf) using spinning disk microscopy. The embryos maintained heartbeat and circulation throughout the imaging period.

Statistical analysis
All experiments were performed independently at least three times. Data were expressed as mean 6 standard deviation (SD). Comparison between groups was performed with a two-tailed Student's t test. Results were considered significant at P,0.01. Movie S1 Time-lapse imaging of co-culture of COS7 cells expressing VEC-EGFP and HUVECs expressing VEC-TagRFP657. The interaction between endogenous VEC in HUVECs and over-expressed VEC-EGFP in COS7 cells can induce trans-endocytosis. The trans-endocytosed VEC-EGFP molecules from a COS7 cell appeared to bud off from cell-cell junctions and be pulled into a HUVEC. The time-lapse images were acquired with a 60x objective lens every 13 seconds for 25 minutes by spinning disk confocal microscopy and were played back as movie at 10 frames per second.

(MOV)
Movie S2 Time-lapse imaging of co-culture of HUVECs expressing VEC-EGFP and HUVECs expressing LifeAct-TagRFPT. The internalized VEC-EGFP vesicle clearly associated with F-actin visualized by LifeAct-TagRFPT. The time-lapse images were acquired with a 60x objective lens every 1 minute for 42 minutes by spinning disk confocal microscopy and were played back as movie at 10 frames per second.

(MOV)
Movie S3 Time-lapse imaging of co-culture of HUVECs expressing VEC-EGFP and HUVECs expressing VEC-TagRFPT with (2)-blebbistatin. 100 mM of (2)-blebbistatin was added at 20 minutes during the time-lapse acquisition and inhibited transendocytosis of VEC. The time-lapse images were acquired with a 60x objective lens every 80 seconds for 60 minutes by spinning disk confocal microscopy and were played back as movie at 4 frames per second. (MOV) Movie S4 Time-lapse imaging of co-culture of HUVECs expressing VEC-EGFP and HUVECs expressing PA-Rac1-CA or PA-Rac1-DN. Upper images are time-lapse imaging of coculture of HUVECs expressing VEC-EGFP and HUVECs expressing PA-Rac1-CA. PA-Rac1-CA accumulated at cell-cell junction and co-localized with VEC-EGFP after its activation. Lower images are time-lapse imaging of co-culture of HUVECs expressing VEC-EGFP and HUVECs expressing PA-Rac1-DN. PA-Rac1-DN did not accumulate at cell-cell junction, nor did it co-localize with VEC-EGFP after its activation. The time-lapse images were acquired with a 60x objective lens every 36 seconds for 29 minutes by spinning disk confocal microscopy and were played back as movie at 7 frames per second.

(MOV)
Movie S5 Time-lapse imaging of sprouting HUVECs in the three-dimensional fibrin gel culture. Some vesicles were transendocytosed from VEC-EGFP positive HUVECs into VEC-TagRFPT positive HUVECs in the three-dimensional fibrin gel culture. The time-lapse images were acquired with a 60x objective lens every minute for 32 minutes by spinning disk confocal microscopy and were played back as movie at 10 frames per second.

(MOV)
Movie S6 Time-lapse imaging of live zebrafish embryo expressing VEC-EGFP. Endothelial cells expressing VEC-EGFP at intersegmental vessel of a zebrafish embryo were observed for 17 minutes. Arrows showed a zVEC-EGFP positive vesicle budding into inside of the endothelial cell. The time-lapse images were acquired with a 60x objective lens every 1 minute for 17 minutes by spinning disk confocal microscopy and were played back as movie at 5 frames per second. (MOV)