Eph-B4 regulates adaptive venous remodeling to improve arteriovenous fistula patency

Low rates of arteriovenous fistula (AVF) maturation prevent optimal fistula use for hemodialysis; however, the mechanism of venous remodeling in the fistula environment is not well understood. We hypothesized that the embryonic venous determinant Eph-B4 mediates AVF maturation. In human AVF and a mouse aortocaval fistula model, Eph-B4 protein expression increased in the fistula vein; expression of the arterial determinant Ephrin-B2 also increased. Stimulation of Eph-B-mediated signaling with Ephrin-B2/Fc showed improved fistula patency with less wall thickness. Mutagenesis studies showed that tyrosine-774 is critical for Eph-B4 signaling and administration of inactive Eph-B4-Y774F increased fistula wall thickness. Akt1 expression also increased in AVF; Akt1 knockout mice showed reduced fistula diameter and wall thickness. In Akt1 knockout mice, stimulation of Eph-B signaling with Ephrin-B2/Fc showed no effect on remodeling. These results show that AVF maturation is associated with acquisition of dual arteriovenous identity; increased Eph-B activity improves AVF patency. Inhibition of Akt1 function abolishes Eph-B-mediated venous remodeling suggesting that Eph-B4 regulates AVF venous adaptation through an Akt1-mediated mechanism.

Understanding venous remodeling remains a significant clinical challenge. For example, arteriovenous fistulae (AVF) remain the gold standard for hemodialysis access in patients with end stage renal disease yet outcomes for AVF are poor, with up to 60% of fistulae failing to attain suitability for dialysis 1 . Failure of fistula maturation requires patients to have repeated interventions and increases catheter-based dialysis 2,3 . A greater understanding of the molecular events guiding successful venous remodeling is needed to develop strategies to improve AVF maturation 4 .
Studies of vein graft remodeling provide some understanding of venous adaptation to the arterial environment. Vein graft remodeling is stimulated by mechanisms that are thought to be similar to those active after blood vessel injury [5][6][7] . Human studies using high resolution imaging have documented venous wall thickening and outward remodeling as consistent components of successful venous remodeling 8 . Venous remodeling is characterized by increased deposition of smooth muscle cells and changes in extracellular matrix component expression resulting in wall thickening and reduced compliance 9,10 . However, it is not currently understood which aspects of remodeling are critical for successful venous adaptation to the arterial environment, whereas excessive remodeling may result in neointimal hyperplasia and clinical failure. The disappointing results of the PREVENT-III and -IV trials suggest that strategies to prevent vein graft failure that focus on inhibition of smooth muscle cell proliferation are not likely to be clinically useful 11,12 .
We next confirmed the regulatory role of Eph-B4 during AVF maturation using a molecular approach. To determine which of the conserved fourteen intracellular tyrosines are critical to Eph-B4 function during AVF maturation, COS cells were transfected with a plasmid expressing WT murine Eph-B4 and then treated with Ephrin-B2/Fc for 3 min. LC-MS/MS identified that tyrosines 581, 590, and 774 were phosphorylated (Supplementary Figure 4A and B). Fourteen separate mutant Eph-B4 plasmids were created in which a single tyrosine residue was replaced with phenylalanine. COS cells were transfected with the individual WT-or mutant-Eph-B4 plasmids and then treated with Ephrin-B2/Fc for 3 min. Three of these mutants, Y774F, Y821F, and Y924F, showed complete absence of Eph-B4 phosphorylation with Ephrin-B2/Fc treatment (Supplementary Figure 4C). Since tyrosine 774 was identified as both an early and critical site of Eph-B4 phosphorylation, we determined whether mutation of tyrosine 774 altered Eph-B4 function in vitro. COS cells were transfected with WT-Eph-B4 or Y774F-Eph-B4 plasmids; the Y774F-Eph-B4 transfected cells showed a sustained lack of tyrosine phosphorylation in response to Ephrin-B2/Fc (Fig. 3C) that was not due to altered receptor localization during biotinylation of surface-expressed proteins (Supplementary Figure 4D). Similarly, cells transfected with the Y774F-Eph-B4 plasmid showed lack of caveolin-1 colocalization (Supplementary Figure 4E), reduced Ephrin-B2/ Fc-stimulated migration (Fig. 3D), as well as reduced Akt and ERK1/2 phosphorylation (Supplementary Figure 4F). Since these data suggest that Eph-B4 function is dependent on tyrosine 774 activity in vitro, we tested the function of Eph-B4 tyrosine activity in vivo. Lentivirus was used to deliver HA-tagged Eph-B4 via a pluronic gel to the IVC. En face staining validated delivery of the lentivirus to the venous wall, including the endothelium (Supplementary Figure 4G). The adventitia of both the arterial and venous limbs of AVF were treated with pluronic gel containing lentivirus with WT-Eph-B4 or Y774F-Eph-B4 at the time of AVF creation. AVF treated with WT-Eph-B4 lentivirus had thin walls similar to control mice, whereas AVF treated with Y774F-Eph-B4 lentivirus had markedly thicker walls (Fig. 3E), with increased proliferating cells and no diminution in apoptotic cells (Supplementary Figure 3D and E). Interestingly, fistulae treated with WT-Eph-B4 showed less outward remodeling of both the venous limb (Fig. 3F) and the arterial limb (Supplementary Figure 3F) compared to WT mice whereas Y774F-Eph-B4 treatment did not show reduced outward remodeling. These results suggest that Eph-B4 tyrosine 774 is critical for Eph-B4 function in vivo during AVF maturation, and that diminished Eph-B4 function during AVF maturation promotes venous wall thickening and outward remodeling.

Eph-B4 signaling is mediated by Akt1 in vitro. Since Eph-B4 stimulates Akt phosphorylation in vitro
and Akt is a critical factor promoting endothelial cell survival and function 17 , we examined whether Akt is a critical factor that mediates Eph-B4 function in an in vitro static environment (Fig. 4A). Endothelial cells (EC) derived from WT and Akt1 KO mice were stimulated with Ephrin-B2/Fc; as expected, WT EC showed increased Akt1, eNOS, and ERK1/2 phosphorylation; Akt1 KO EC did not show Akt1 or eNOS phosphorylation but had increased ERK1/2 phosphorylation ( Fig. 4B and C). In WT EC, Akt1 phosphorylation colocalized with Eph-B4 in the cytoplasm after Ephrin-B2/Fc stimulation (Fig. 4D, white arrowheads). Similarly, Ephrin-B2/Fc stimulated NO release in WT cells that was inhibited in Akt1 KO EC (Fig. 4E), suggesting that Akt1 regulates some Eph-B4 functions in EC in vitro.
We next determined if Eph-B4 similarly stimulates Akt1 phosphorylation in an in vitro flow environment. Compared to the basal amount of Akt1 phosphorylation in EC under static conditions (0 dynes/cm 2 ), there was increased Akt1 phosphorylation in response to arterial magnitudes of laminar shear stress (20 dynes/cm 2 ); however, pretreatment of EC with the Eph-B4 inhibitor NVP-BHG712 resulted in further increased Akt1 phosphorylation ( Fig. 4F and G). Conversely, pretreatment of EC with Ephrin-B2/Fc to activate Eph-B4 resulted in decreased shear stress inducted Akt1 phosphorylation (Fig. 4H). These results suggest that Eph-B4 inhibits laminar shear stress-induced Akt1 phosphorylation, e.g. Akt1 mediates some functions of Eph-B4 in EC in vitro, under both static and flow conditions, but the effect of Eph-B4 on Akt1 activity in vitro may differ under static and flow conditions.  5A). Similarly, there was increased Akt1 protein expression with no significant change in Akt2 expression at days 7 and 21 (Fig. 5B). Immunofluorescence showed increased Akt1 immunoreactivity predominantly in the AVF endothelium at days 7 and 21; at day 42 there was less endothelial Akt1 immunoreactivity, although there was sustained Akt1 expression in the AVF wall, below the neointima (P = 0.0314; Fig. 5C). These data show that Akt1 expression is regulated during AVF maturation, in a time course similar to Eph-B4 expression; therefore, we determined if altered Akt1 activity directly affects AVF maturation. AVF in WT mice were treated with an adenovirus containing either constitutively active Akt1, dominant negative Akt1, or control virus. AVF treated with constitutively active Akt1 showed increased venous wall thickening compared to AVF treated with control or dominant negative Akt1 virus (Fig. 5D). AVF treated with constitutively active Akt1 showed similar outward remodeling compared to AVF treated with control virus, but AVF treated with dominant negative Akt1 showed reduced outward remodeling (Fig. 5E); there was no difference in arterial remodeling in any of these groups (Supplementary Figure 5A-B). Similarly, there was increased eNOS activity in CA-Akt treated mice, similar to the increased activity in AVF in WT mice, which was not present in mice treated with DN-Akt (Supplementary Figure 5C,D, and E). These results suggest that Akt1 signaling has a functional effect during AVF maturation, and that Akt1 activity may mediate venous remodeling. To determine if Akt1 mediates venous remodeling, AVF were performed in Akt1 KO and WT mice. There was a trend toward reduced venous wall thickening ( Fig. 5F) in Akt1 KO mice compared to WT mice. In addition, there was reduced venous outward remodeling in Akt1 KO mice (Fig. 5G). These results support a mechanistic role for Akt1 during adaptive venous remodeling.
Eph-B4-mediated venous remodeling depends on Akt1. Since Eph-B4 regulates venous wall thickening and outward remodeling, and Akt1 also mediates venous wall thickening and outward remodeling, we determined if Akt1 is a mechanism by which Eph-B4 regulates adaptive venous remodeling during AVF maturation (Fig. 6A). AVF in WT mice treated with Ephrin-B2/Fc showed reduced phosphorylated Akt1 immunoreactivity as well as less total Akt1 immunoreactivity compared to control AVF (Fig. 6B), suggesting that Eph-B activation   inhibits Akt1 expression during venous remodeling in vivo. To confirm that Eph-B4 inhibits Akt expression during venous remodeling, AVF were treated with WT-Eph-B4 or mutant Y774F-Eph-B4 lentivirus. AVF in WT mice treated with mutant Y774F-Eph-B4 lentivirus showed increased Akt1 phosphorylation and total Akt immunoreactivity compared to mice treated with WT-Eph-B4 lentivirus (Fig. 6C). These results are consistent with Eph-B4 inhibiting Akt1 function during venous remodeling in vivo.
To directly test whether Akt1 mediates Eph-B4 regulated venous remodeling, Eph-B4 function was stimulated with Ephrin-B2/Fc after AVF creation in WT and Akt1 KO mice. In WT mice, Ephrin-B2/Fc treatment was associated with thin venous walls (Fig. 2) with less α-actin compared to control veins that were thick and had more α-actin; however, veins of Akt1 KO mice treated with Ephrin-B2/Fc did not have thinner walls nor less actin staining compared to control Akt1 KO veins (Fig. 6D-F). Similarly, WT mice treated with Ephrin-B2/Fc had less venous outward remodeling compared to WT mice (Fig. 2), whereas Akt1 KO mice treated with Ephrin-B2/ Fc did not show less venous outward remodeling (Fig. 6G). In toto, these data show that Eph-B4 inhibits Akt1 phosphorylation and expression during AVF maturation, and that reduced Eph-B4 function during normal AVF maturation is associated with increased Akt1 function, suggesting that Eph-B4 inhibits Akt1-mediated venous remodeling.
We have previously shown, using selective knockdown of Akt1 from endothelial cells, that Akt1 is critical for angiogenesis 23 . AVF created in mice that underwent tamoxifen-inducible Akt1 deletion in endothelial cells showed similar wall thickness and dilation (21 days) compared to control mice that received vehicle alone (Fig. 7). However, AVF created in mice that underwent tamoxifen-inducible Akt1 deletion in smooth muscle cells showed reduced wall thickness and dilation (21 days) compared to control mice (Fig. 7). These results are consistent with smooth muscle cell Akt1 playing a critical role in venous remodeling in the fistula environment.

Discussion
We report that expression of the Eph-B4 receptor as well as its ligand Ephrin-B2 increase in the vein wall during AVF maturation (Fig. 1), consistent with acquisition of dual arterial-venous identity in the fistula environment; increased Eph-B activity regulates venous remodeling, leading to improved long-term patency (Fig. 2). Directed point mutations of the Eph-B4 receptor identified cytoplasmic tyrosine 774 as a critical phosphorylation site for Eph-B4 function and delivery of mutant Eph-B4 receptors with a nonfunctional tyrosine 774 resulted in altered venous remodeling, similar to that seen in heterozygous Eph-B4 mice (Fig. 3). In vitro, Eph-B4 inhibits shear stress-induced Akt1 phosphorylation, a different effect on Akt1 phosphorylation observed under static conditions (Fig. 4). In vivo, Akt1 is a critical mediator of venous remodeling (Fig. 5), particularly smooth muscle cell Akt1 (Fig. 7), and Akt1 is a mechanism of Eph-B-mediated venous remodeling (Fig. 6). In toto, these data show that veins adapt to the fistula environment by acquisition of dual arterial-venous identity, whereas veins adapt to the arterial environment by loss of vessel identity 7,20 ; in addition these data show that Eph-B4 regulates venous remodeling by an Akt1-mediated mechanism.
Our data showing that Eph-B4 regulates venous remodeling during AVF maturation are consistent with previous reports showing that Eph-B4 inhibits venous remodeling during vein graft adaptation 19,20 . Interestingly, the expression pattern of Eph-B4 and Ephrin-B2 is different between the AVF and vein graft models. We previously showed that vein graft adaptation is characterized by loss of venous identity, with reduced Eph-B4 expression, but without a gain of arterial identity 7,20 . Here we report that during AVF maturation the vein acquires dual vessel identity, with increased venous and arterial markers (Fig. 1, Table 1). We verified this finding using an alternative model by placing a patch into a rat IVC, without or with a fistula, to confirm that the presence of a fistula increases both Eph-B4 and Ephrin-B2 expression (Fig. 1). Acquisition of dual arterial and venous identity has not been previously described during physiological venous adaptation such as occurs during AVF maturation; increased expression of both Eph-B4 and Ephrin-B2 has been reported in some cancers [24][25][26][27] . The different gain of dual arterial-venous identity during AVF maturation that is distinct from the loss of vessel identity during vein graft adaptation demonstrates the distinct biology of AVF maturation in the fistula environment from that of vein graft adaptation to the arterial environment; these molecular differences correlate with the distinct clinical outcomes of these two procedures, with AVF associated with distinctly worse outcomes compared to vein grafts 12,28 . However, the differential mechanisms by which vessel identity is regulated in these two environments is not known. Notably, our data regarding vessel identity (Table 1 and Fig. 1) is derived from the whole AVF wall; it is not clear whether different cell components or layers within the vessel have different identities or are differentially altered during vein graft adaptation or AVF maturation, and thus could be mechanistic or simply markers of these processes.
There are numerous differences between the arterial and fistula environments, including differences in flow, pressure, resistance of the runoff bed, oxygen tension, and regular environmental injury 10 . Hemodynamics can regulate Ephrin-Eph signaling and vessel function 29 ; for example, laminar shear stress can reduce Ephrin-B2 and Eph-B4 expression [30][31][32] , and Ephrin-B2-Eph-B4 interactions promote monocyte adhesion to the endothelium under flow conditions 33 . Ephrin-Eph signaling is also involved in blood pressure regulation 34 , and stretch-induced Ephrin-B2 expression limits smooth muscle cell migration and monocyte extravasation, promoting arteriogenesis 35 . Other mechanisms may also be functional, as endothelial Ephrin-B2 also regulates vasodilation and NO signaling 36 . Since both Ephrin-B2 and Eph-B4 expression increase during AVF maturation (Fig. 1), it is likely that numerous factors regulate EphrinB-EphB signaling in vivo, and that EphrinB-EphB signaling is only a single component of this complex adaptive process.
Using the w804a point mutation in the Eph-B4 caveolin-binding domain, we previously showed that Eph-B4 phosphorylation is critical to its function; however, in that study we did not identify the critical tyrosine that is responsible for the activity of the Eph-B4 kinase domain 20 . Here we used mass spectrometry to determine that three Eph-B4 intracellular tyrosines, Y581, Y590, and Y774, are phosphorylated within 3 min of stimulation with the Eph-B4 ligand Ephrin-B2/Fc; this pattern is consistent with canonical Eph receptor activation 13   mutation of all Eph-B4 intracellular tyrosines showed that only Y774 was both phosphorylated within 3 minutes of activation and was also critical for Eph-B4 receptor activity (Supplementary Figure 4). Eph-B4 tyrosine 774 is critical for Eph-B4 receptor tyrosine phosphorylation, Akt and ERK1/2 phosphorylation (Supplementary Figure 4), and cell migration in vitro (Fig. 3), as well as venous remodeling in vivo (Fig. 3), e.g. tyrosine 774 is a critical mediator of Eph-B4 function both in vitro and in vivo. It is possible that the effects of Eph-B4-Y774F are pleotropic among all the cell types of the vessel wall, and that overexpressed Eph-B4 may act as a sink for Ephrin-B2 ligands, altering multiple forward and reversed signaling pathways. However, in toto, these results are consistent with previous reports showing the importance of tyrosine phosphorylation for activation and intracellular signaling of Eph-B receptors 13 , and complement our data using Ephrin-B2/Fc as an activator of Eph-B signaling (Figs 2, 4, 6) as well as our data using heterozygous Eph-B4 mice with reduced Eph-B4 signaling (Fig. 3). Nonetheless, it is possible that Eph-B receptors on the many components of the vessel wall, including endothelial, smooth muscle, mesenchymal and immune cells, alters AVF remodeling, and this is consistent with our finding that smooth muscle Akt1 is critical for venous remodeling (Fig. 7). Acquisition of dual arterial-venous identity during AVF maturation, e.g. colocalization of cells expressing EphrinB ligands with cells expressing EphB receptors, may be an important mechanism of venous remodeling; however, the relative contributions of EphrinB-and EphB-expressing cells that are native cell components of the vessel wall, or infiltrating cells such as inflammatory cells or M2 macrophages 37 , is not clear.
Our data also show that Akt1 function is critical for venous remodeling that occurs during AVF maturation and is a mechanism of Eph-B4 regulation of venous remodeling. Akt promotes cell survival, protein synthesis, and growth that are necessary for adaptive remodeling of the vein to the fistula environment. VEGF activates the Akt-eNOS pathway, while inhibiting ERK1/2, to promote maintenance of the vascular system 38 ; similarly Eph-B4 activates the Akt-eNOS pathway both in vitro and in vivo 17,21 . Interestingly, Akt expression is increased during the venous remodeling that occurs during vein graft adaptation, but under these different hemodynamic conditions neither Akt nor eNOS is phosphorylated 19 . However, it is not clear why Akt is activated during venous remodeling in the fistula environment whereas Akt is not activated during venous remodeling in the arterial environment. We also observed that Eph-B4 stimulates Akt1 phosphorylation under static conditions but inhibits and Akt KO mice without and with Ephrin-B2/Fc stimulation (day 21); P = 0.0318 (ANOVA). *P = 0.0234 (WT control vs Ephrin-B2/Fc; post hoc). n = 3-4. (F) Bar graph showing quantification of α-actin density in the wall of AVF in WT and Akt1 KO mice without or with Ephrin-B2/Fc (day 21); P < 0.0001 (ANOVA). *P < 0.0001 (WT control vs Ephrin-B2/Fc; post hoc). n = 3. (G) Line graph showing infrarenal IVC diameter in WT and Akt1 KO mice without or with Ephrin-B2/Fc; *P < 0.0001 (ANOVA, WT); NS, P = 0.137 (ANOVA, Akt1 KO). n = 3-5. Data represent mean ± SEM.
Akt1 phosphorylation under laminar shear stress conditions in vitro (Fig. 4), suggesting that Eph-B4 may have mechanosensory functions. Nevertheless, our data shows that loss of Akt1 function abolishes Eph-B4 regulation of venous remodeling (Fig. 6), suggesting that Akt1 is a mechanism of Eph-B4-mediated venous remodeling such as occurs during AVF maturation in vivo. Interestingly, selective deletion of Akt1 from smooth muscle cells, but not endothelial cells, abolished venous remodeling (Fig. 7); this data is consistent with the recent description of differentiated smooth muscle cells contributing to medial wall thickening during AVF maturation 39 . This data also confirms the critical importance of Akt1 to venous remodeling during AVF maturation.
In summary, we show that Eph-B4 expression increases during AVF maturation and that Eph-B4 activity, mediated by tyrosine 774, regulates adaptive venous remodeling through an Akt1-mediated mechanism. Stimulation of Eph-B4 attenuated Akt1 expression in vivo, leading to improved long-term patency in this animal model. Eph-B4 is normally present in adult veins and may be a potential therapeutic target to reduce pathologic venous remodeling that is associated with failure of fistula maturation. These results suggest that strategies to manipulate vessel identity by altering Eph-B4 activity may improve AVF maturation.

Methods
Human specimens. The principles outlined in the Declaration of Helsinki were followed, and approval of the Veterans Affairs Human Investigation Committee was obtained. Deidentified discarded specimens of vein and AVF were obtained during revision operations of functional AVF; informed consent to use the samples was obtained.
Infrarenal aorto-caval fistula. All animal experiments were performed in strict compliance with federal guidelines and with approval from the Yale University IACUC. Mice used for this study included wild type C57BL6/J (WT), Eph-B4 heterozygous (Eph-B4 het) 20 , or Akt1 knockout (Akt1 KO) mice 40 . Smooth muscle-specific knockouts were generated by breeding Akt1 flox/flox with the Myh11-CreERT2 strain 23 . Deletion of Akt1 was induced by injecting 4-5 week old mice with tamoxifen (100 ug/g total body weight) for 5 consecutive days. Akt1 flox/flox control mice were similarly injected with vehicle. The Akt1 flox/flox were bred to the Cdh5-Cre-ERT2 mice to obtain inducible, endothelial-targeted Akt1 mice, as previously described 23 . Young adult, male mice (~4-5 wks of age) were injected with tamoxifen (100 ug/g total body weight) via intraperitoneal delivery for 7 consecutive days. Phenotyping assessments were performed 6 weeks post-tamoxifen administration (~10-12 wks of age).
All mice were male and 9-12 weeks of age when the infrarenal aorto-caval fistulae were created as previously described 22,41 . Briefly, AVF were created by needle puncture from the aorta into the IVC. Visualization of pulsatile arterial blood flow in the IVC was assessed as a technically successful AVF.

Measurement of fistula dilation in vivo. Doppler ultrasound (40 MHz; Vevo770 High Resolution
Imaging System; VisualSonics Inc., Toronto, Ontario, Canada) was used to confirm the presence of the AVF and to measure the diameter of the vessels as previously described 22,41 . Doppler ultrasound was performed the day prior to operation (pre-op values) and serially post-operatively. Increased diastolic flow through the aorta and a high velocity pulsatile flow within the IVC were used to confirm the presence of an AVF during post-operative interrogations, and AVF that were not patent on Doppler study were explanted; patency was confirmed at AVF explantation by direct visualization of pulsatile arterial blood flow in the IVC, and in all cases correlated with the ultrasound findings.
Histology. After euthanasia, the circulatory system was flushed under pressure with PBS followed by 10% formalin and the AVF was extracted en bloc. The tissue was then embedded in paraffin and cut in 5 μm cross sections. Hematoxylin and eosin (H&E) staining was performed for all samples. For wall thickness measurements, cross sections were obtained 125 μm cranial to the AVF and stained with elastin van Gieson (EVG) stain. Eight equidistant points per cross section were averaged to obtain the mean outer wall diameter encompassing the intima and media, as previously described 7,20 . Additional unstained cross sections in this same region (100-150 μm cranial to the AVF) were used for immunohistochemical or immunofluorescence microscopy.

RNA extraction and quantitative PCR.
Total RNA from the venous limb of the AVF was isolated using the RNeasy Mini kit. Reverse transcription was performed using the SuperScript III First-Strand Synthesis Supermix (Invitrogen, Carlsbad, CA). Real-time quantitative PCR was performed using SYBR Green Supermix (Bio-Rad Laboratories, Hercules, CA) and amplified for 40 cycles using the iQ5 Real-Time PCR Detection System (Bio-Rad Laboratories). Correct target amplification and exclusion of nonspecific amplification was confirmed by 1.5% agarose gel electrophoresis, and primer efficiencies were determined by melt curve analysis. All samples were normalized by GAPDH RNA amplification. Rat venous patch model. All animal experiments were performed in strict compliance with federal guidelines and with approval from the Yale University IACUC. 6-8 week old male Wistar rats were used for patch implantation as previously described 21 . Rats were divided into three groups and underwent bovine pericardial patch (LeMaitre Vascular, Burlington, MA) implantation into the IVC, the aorta, or the IVC in the presence of an aorto-caval fistula (IVC + AVF). For the IVC + AVF group, after the patch was implanted into the IVC, the distal micro-clamp of the IVC was removed first to flush out the air, then the proximal micro-clamp was removed, and IVC flow was restored for 5 minutes. Next, the aorta was exposed and the infrarenal aorta was clamped. A 20-gauge needle was used to puncture through the aorta into the adjacent IVC just above the aorta bifurcation. 10-0 suture was used to close the puncture site on the aorta and the aortic clamp was then removed. The abdomen was closed using 5-0 Dacron sutures. Animals were sacrificed on postoperative day 14 for histology with IHC and IF as described above. No immunosuppressive agents, antibiotics or heparin were given at any time.

Eph-B4 stimulation in vivo.
Eph-B4 was stimulated with Ephrin-B2/Fc (R&D, Minneapolis, MN). 24 hr prior to AVF creation, mice underwent ultrasound for measurement of preoperative vessel diameter as described above. While still under general isoflurane anesthesia, 20 µg of Ephrin-B2/Fc diluted in 200 µL PBS was delivered by intraperitoneal injection (IP). Control mice received an equal volume injection of vehicle (PBS) as control. The next day, an infrarenal aorto-caval AVF was created as described above. Additional IP injections of Ephrin-B2/Fc were delivered every 48 hr beginning on postoperative day 1 and continued throughout the study period.
Plasmid transfection. Cells were plated on 6-well plates at a concentration of 300,000 cells per well. The following day, cells were treated with plasmid and Lipofectamine 2000 (Life Technologies) diluted in OptiMEM (Life Technologies). Plates were incubated for 4-5 hr at 37 °C, 5% CO 2 . The media was then replaced with complete media and incubated overnight at 37 °C, 5% CO 2 . The following day the media was aspirated and the cells starved for 24 hr by the addition of 2 ml of OptiMEM per well.
Eph-B4 stimulation in vitro. Following starvation, cells were stimulated with either mouse Ephrin-B2/Fc (2 µg/ml) or mouse IgG/Fc (2 µg/ml; R&D Systems) in PBS, for the designated amount of time, after which either the cell lysates or the conditioned media were collected.
In vitro shear stress. EC were cultured to confluence on collagen I-coated glass plates (StreamerTM Culture Slips, Flexcell Corporation). Confluent EC were serum-starved for 24 hr and then pre-treated with NVP-BHG712 (1 µM; Sigma Aldrich, St. Louis, MO) for 1 hr or stimulated with 2 µg/ml mouse Ephrin-B2/Fc (R&D Systems) for 1 hr and compared to control. After pre-treatment (37 °C, 5% CO 2 ), cells were exposed to steady laminar flow in a parallel-plate flow chamber with circulating EBM-2 medium (0% FBS, 37 ± 0.5 °C, 1 hr). Wall shear stress was calculated by the formula τ = 6 μQ/bh 2 , where μ is the viscosity of the fluid, Q is the flow volume (ml/s), b is the width of the flow channel in cm, and h is the height of the flow channel in cm; shear stress was set at 0 dynes/ cm 2 or 20 dynes/cm 2 . After shear stress treatment, the EC on the slides were removed with RIPA lysis buffer and analyzed.
In vitro whole cell lysate isolation. Cells were treated with RIPA lysis buffer containing protease and phosphatase inhibitor cocktail. Cells were then scraped, sonicated and centrifuged.
Mutagenesis of Eph-B4 tyrosine residues. The cDNA sequence of mouse Eph-B4 was purchased from Open Biosystems (GE Dharmacon, Lafayette, CO) and corresponds to transcript variant 2 (NCBI Reference Sequence: NM_010144.6). The cDNA was amplified using appropriate primers. The PCR product was prepared for insertion into the pShuttle-IRES-hrGFP-2 plasmid vector (Agilent Technologies, Santa Clara, CA) by digestion with SpeI and PvuI. The QuikChange Site-Directed Mutagenesis Kit (Agilent) was utilized for generating phenylalanine substitutions for cytoplasmic tyrosine residues. Sense and antisense mutant primers were designed