Effective Treatment of Edema and Endothelial Barrier Dysfunction With Imatinib

Background— Tissue edema and endothelial barrier dysfunction as observed in sepsis and acute lung injury carry high morbidity and mortality, but currently lack specific therapy. In a recent case report, we described fast resolution of pulmonary edema on treatment with the tyrosine kinase inhibitor imatinib through an unknown mechanism. Here, we explored the effect of imatinib on endothelial barrier dysfunction and edema formation. Methods and Results— We evaluated the effect of imatinib on endothelial barrier function in vitro and in vivo. In human macro- and microvascular endothelial monolayers, imatinib attenuated endothelial barrier dysfunction induced by thrombin and histamine. Small interfering RNA knock-downs of the imatinib-sensitive kinases revealed that imatinib attenuates endothelial barrier dysfunction via inhibition of Abl-related gene kinase (Arg/Abl2), a previously unknown mediator of endothelial barrier dysfunction. Indeed, Arg was activated by endothelial stimulation with thrombin, histamine, and vascular endothelial growth factor. Imatinib limited Arg-mediated endothelial barrier dysfunction by enhancing Rac1 activity and enforcing adhesion of endothelial cells to the extracellular matrix. Using mouse models of vascular leakage as proof-of-concept, we found that pretreatment with imatinib protected against vascular endothelial growth factor–induced vascular leakage in the skin, and effectively prevented edema formation in the lungs. In a murine model of sepsis, imatinib treatment (6 hours and 18 hours after induction of sepsis) attenuated vascular leakage in the kidneys and the lungs (24 hours after induction of sepsis). Conclusions— Thus, imatinib prevents endothelial barrier dysfunction and edema formation via inhibition of Arg. These findings identify imatinib as a promising approach to permeability edema and indicate Arg as novel target for edema treatment.

T he endothelium tightly controls the exchange of fluid from the circulation to the surrounding tissues. Dysfunction of this barrier leads to uncontrolled fluid extravasation and edema, [1][2][3] and characterizes life-threatening conditions like sepsis 1 and acute lung injury. 4 Despite high mortality rates-up to 50% in sepsis-no treatment is currently available for endothelial barrier dysfunction and edema. 1 However, in a recent case report we described fast resolution of pulmonary edema on treatment with imatinib. 5

Editorial see p 2677 Clinical Perspective on p 2738
Imatinib is a small molecule inhibitor, blocking the ATPase activity of the kinases c-Abl, Abl-related gene (Arg/ Abl2), platelet-derived growth factor receptor (PDGFR), c-KIT, and discoid domain receptor-1. 6 Thus far, imatinib has found its major application in the treatment of Brc-Abl positive chronic myeloid leukemia and gastro-intestinal stromal tumors, 6 whereas nonmalignant proliferative disorders like lung fibrosis 7 and pulmonary hypertension 8 may form future applications of imatinib. Although designed as a smart drug specifically targeting overactive kinases, imatinib is associated with several side effects. Long-term treatment with imatinib may lead to cardiac failure by inducing cardiomyocyte apoptosis, 9 and, of note, long-term treatment with imatinib was associated with subcutaneous edema. 10 In the light of these studies the association of imatinib treatment with resolution of edema is surprising. Yet, increas-ing evidence indicates that imatinib may protect against edema. [11][12][13] A second case report revealed clinical improvement of acute lung injury on initiation of imatinib, 11 whereas two experimental studies demonstrated that imatinib protects against brain edema after stroke. 12,13 The mechanism by which imatinib may protect against edema remains largely unclear. The protective effect of imatinib on brain edema was mainly attributed to PDGFR-␣ inhibition on perivascular astrocytes, 12,13 which is unlikely to explain the protective effects of imatinib observed in pulmonary edema. Otherwise, the descriptive character of mentioned case reports 5,11 limited mechanistic interpretation, although an effect of imatinib on endothelial barrier function was suggested. 5 Little is known about the direct effects of imatinib on the endothelial barrier as main regulator of fluid exchange. In the current study we hypothesized that imatinib reduces edema formation via direct preservation of endothelial barrier integrity. Using in vitro and in vivo models of endothelial barrier dysfunction, we show that imatinib effectively protects against endothelial barrier dysfunction and edema formation.

Endothelial Barrier Function Assays
Endothelial barrier function was evaluated with horseradish peroxidase (HRP) passage and electric cell-substrate impedance sensing. For measurement of HRP passage, confluent cells were seeded in 1:1 density on 0.33 cm 2 Costar polycarbonate filters, pore-size 3.0 m (Corning, Lowell, MA), and grown to confluence in 5 days. For pretreatment, pharmacological inhibitors or vector were dissolved in M199 (Biowhittaker/Lonza, Verviers, Belgium) supplemented with 1% human serum albumin (HSA; Sanquin Blood Supply, Amsterdam, The Netherlands), and added to the upper compartment of the filters during 60 minutes. For stimulation, pretreatment medium was changed for 1%HSA/M199 containing designated inhibitors, HRP 5 g/mL (Sigma Aldrich, Zwijndrecht, the Netherlands) and thrombin 1 U/mL (Sigma Aldrich). 1%HSA/M199 was added to the lower compartment. At indicated time points, samples were taken from the lower compartment. The HRP concentration was detected by measuring chemoluminiscence after addition of TMB/E (Upstate/Millipore, Temecula, CA).
For electric cell-substrate impedance sensing measurements, cells were seeded in 1:1 density on gelatin-coated electric cell-substrate impedance sensing arrays, each containing 8 wells with 10 gold electrodes per well (Applied Biophysics, Troy, NY). Culture medium was renewed 24 hours after seeding, and experiments were performed 48 hours after seeding. For pretreatment, pharmacological inhibitors or vector were dissolved in 1%HSA/M199. After 90 minutes of pretreatment, thrombin or histamine were added directly to the wells for final concentrations of 1 U/mL or 10 Ϫ5 mol/L, respectively. During stimulation, resistance was measured at multiple frequencies to allow for calculation of resistance attributable to cell-cell adhesion (Rb) and to cell-matrix interaction (Alpha). 14,15

Evans Blue/Albumin Extravasation in Mouse Skin
Extravasation of albumin was visualized in the Miles assay by extravasation of Evans blue. 3 Male Balb/cByJ mice (Charles River, 25-30 g) were anesthetized with fentanyl, midazolam, and acepromazine. Anesthetized mice were treated with imatinib mesylate (20 mg/kg in PBS, intraperitoneally) or vector. Five minutes later 150 L Evans Blue (Merck, Darmstadt, Germany, 0.5% in PBS) was administered via the tail vein and left circulating for 30 minutes. Subsequently, vascular endothelial growth factor (VEGF) was injected intradermally in the back skin. Mice were euthanized after 30 minutes of VEGF stimulation, and circular skin patches (ø 8 mm) from the injection sites were incubated in formamide for 24 hours. Extracted Evans Blue and hemoglobin was measured spec-trophotometrically at 610 and 450 nm, respectively. The Evans Blue/hemoglobin ratio is given.

Edema Formation in the Isolated Perfused Mouse Lung
Pulmonary vascular permeability for fluid was analyzed as described earlier. 16 In brief, mixed C57/Bl6 mice (Jackson Laboratories, 25-30 g) were anesthetized with ketamine and xylazine. Anesthetized mice were treated with imatinib mesylate (50 mg/kg in PBS, intraperitoneally) or vector. Thirty minutes after imatinib administration, thrombin receptor-1 activating peptide (TRAP, TFLLRN, 5 mg/kg) was administered via the jugular vein and left circulating for 30 minutes. After 30 minutes, mice were intubated by tracheostomy, and heart and lungs were removed en bloc. The pulmonary artery and left atrium were cannulated, and the heart and lungs were positioned on a weighing scale. The lungs were ventilated and perfused with RPMI1640/HEPES buffer (2 mL/min). After 20 minutes of equilibration the lung weight was zeroed, and the outflow pressure was increased with 8 cm H 2 O for 20 minutes while lung weight was monitored. The weight increase of the lungs over time yields the K fc (mL/min/cmH 2 O/g) reflecting pulmonary vascular permeability for fluid.

Cecal Ligation and Puncture
Male C57/BL6J mice (Harlan, 25-30 g) were anesthetized with isoflurane (4% vol/vol in air during induction and 1.5%-2% maintenance) and oxygen 0.5 L/min. The abdominal cavity was opened with a 1-cm cut over the medial line, and the cecum was positioned outside the abdominal cavity. For cecal ligation and puncture (CLP), 75% of the cecum was ligated with 5-0 vicryl (Johnson-Johnson Intl, New Brunswick, NJ) and perforated through-and-through with a 21-G needle. 17 After extrusion of a column of 1 mm feces, the cecum was repositioned in the abdominal cavity. For sham surgery, the cecum was only positioned outside the abdominal cavity and repositioned. The abdominal cavity was closed with a continuous ligature through the abdominal muscle wall and single ligatures in the skin. After surgery, mice received fluid resuscitation and analgesia. At tϭ6 hours and tϭ18 hours after surgery, mice were treated with imatinib mesylate (50 mg/kg in PBS) or vector by subcutaneous injection in the neck. At tϭ23 hours, 100 L Evans Blue (1% in PBS) was administered via the tail vein and left circulating for 1 hour. At 24 hours after surgery, mice were anesthetized with fentanyl, midazolam, and acepromazine, and euthanized by withdrawal of 0.5 to 1 mL blood from the heart. Whole blood was collected in heparinized tubes, centrifuged for 10 minutes at 1800g and 4°C. Plasma (75 L) was added to 150 L formamide for determination of the Evans Blue concentration in the plasma. The kidneys, liver, and lungs were collected and thoroughly washed in saline. Evans Blue was extracted from organ tissue by incubating organs in 300 L (kidneys and lungs) or 500 L (liver lobe) formamide at 55°C. After 48 hours the organs were removed; the remaining formamide was centrifuged (13 500 rpm for 5 minutes) and analyzed spectrophotometrically at 610 nm (Evans Blue) and 740 nm (overlap of hemoglobin in the Evans Blue range). The corrected Evans Blue absorbance was calculated by the following formula: OD610Ϫ[1.426ϫOD740ϩ0.03]. 18 After Evans Blue measurement, organs were washed to remove the formamide and air dried at 90°C to determine dry weight. Vascular leakage is presented as the amount of organ Evans Blue absorbance, corrected for organ dry weight and plasma Evans Blue absorbance.
All animal experiments were performed with approval of the Animal Ethical Committees of the VU University Medical Center or the University of Illinois at Chicago.

Statistical Analyses
Data are reported as meanϮstandard error of the mean (SEM). n refers to the number of independent experiments with cells from different donors, unless stated otherwise. With the hypothesis that imatinib decreases endothelial hyperpermeability via an effect on Arg activation, the effect of interventions (imatinib, siRNAs) on endothelial barrier function and activation of specific signaling molecules was tested for statistical significance. For comparison of 2 conditions a Student t test was used, for comparison of Ͼ2 conditions a 1-way ANOVA with Tukey post hoc test or a repeated measures ANOVA with Bonferroni post hoc test was used when appropriate, as indicated in the figure legends. P values Ͻ0.05 were considered statistically significant.
Additional methods and materials used for this study can be found in the online-only Data Supplement.

Imatinib Attenuates Disruption of the Endothelial Barrier by Thrombin and Histamine
The direct effect of imatinib on endothelial barrier function was evaluated in isolated human endothelial cell monolayers under basal and stimulated conditions. Short-term treatment of human lung microvascular endothelial cells and human umbilical vein endothelial cells (HUVECs) with imatinib did not affect endothelial barrier function under basal conditions ( Figure 1A and 1D). However, imatinib dose-dependently attenuated endothelial barrier disruption by thrombin with an optimal dose at 10 mol/L in HUVECs ( Figure I in the online-only Data Supplement). Imatinib 10 mol/L effectively protected against endothelial barrier dysfunction, shown by a 46% and 44% reduction in thrombin-induced macromolecule passage ( Figure 1A and 1D) and a 9% and 28% attenuation of the thrombin-induced decrease in endothelial electric resistance ( Figure 1B and 1E). Immunostaining of the cell-cell junctional proteins ␤-catenin and VEcadherin revealed that imatinib prevents the formation of intercellular gaps after thrombin stimulation. ( Figure 1C and Figure II in the online-only Data Supplement). Imatinib also attenuated endothelial barrier dysfunction in microvascular endothelial cells isolated from human skin ( Figure III in the online-only Data Supplement) and endothelial barrier dysfunction induced by histamine ( Figure 1F). Together these data show that imatinib effectively protects against endothe- lial barrier dysfunction, independent of endothelial cell type or barrier-disruptive agent.

Imatinib Exerts its Protective Effect via Inhibition of the Tyrosine Kinase Abl-Related Gene (Arg)
To elucidate the kinase through which imatinib exerts its protective effect on endothelial barrier function, we performed siRNA knock-downs of the known imatinib-sensitive tyrosine kinases (c-Abl, Arg, PDGFR, c-KIT, discoid domain receptor-1) and evaluated the effects on thrombin-induced endothelial barrier dysfunction. Knock-down of PDGFR-␣/␤, c-Abl, c-KIT, or discoid domain receptor-1 did not affect the thrombin response (Figure IVA-IVH in the online-only Data Supplement). In contrast, knock-down of Arg attenuated the thrombin-induced increase in macromolecule passage ( Figure  2A) and decrease in endothelial resistance ( Figure 2B). Knockdown of Arg and treatment with imatinib similarly attenuated the thrombin response, whereas imatinib had no additive protective effect in Arg-depleted cells ( Figure 2C), indicating that imatinib exerts its protective effects predominantly via inhibition of Arg. To establish whether Arg is activated during endothelial barrier dysfunction, we measured CrkL phosphorylation at Tyr207 (an exclusive target for c-Abl and Arg 19 ) in c-Abl-depleted cells. Thrombin induced robust CrkL phosphorylation in c-Abl-depleted cells, which could be prevented by imatinib or combined c-Abl/Arg knock-down ( Figure 2D and 2E). Arg was also activated on stimulation with the barrier-disruptive agents VEGF and histamine ( Figure 2F). These findings identify Arg as a novel mediator of endothelial barrier dysfunction. Arg-mediated endothelial barrier dysfunction can be effectively inhibited with imatinib.

Arg Inhibition Prevents Loss of Cell-Matrix Interaction During Endothelial Stimulation
Our next step was to analyze the effect of imatinib on processes regulating endothelial barrier function. A functional endothelial barrier is characterized by low actomyosin tension and stable cell-cell junctions. During endothelial barrier dysfunction increased actomyosin contraction and disruption of cell-cell junctions result in gap formation. 1,2 Tight adhesion of endothelial cells to the subcellular matrix counteracts cell retraction, and as such limits junction disruption and gap formation. 2,20,21 Imatinib did not affect RhoA/ Rho kinase activity or calcium-dependent signaling ( Figure  3), as main determinants of actomyosin contraction. Moreover, imatinib did not visibly change the morphology of actin fibers (data not shown). Resolving the endothelial resistance measurements into separate components reflecting cell-cell contact and cell-matrix interaction 14,15 displayed that Arg inhibition with siRNA or imatinib predominantly attenuated the loss of cell-matrix interaction during thrombin stimulation ( Figure 4A and 4B and Figure V in the online-only Data Supplement). Using immunofluorescence and live-cell imaging of the focal adhesion marker paxillin we indeed observed that imatinib enhanced the formation of focal adhesions, in particular at the cell periphery ( Figure 4C and 4D and Figure  VI and Movie I in the online-only Data Supplement). Furthermore, the activity of Rac1-a GTPase known to reinforce both cell-matrix interaction 21 and cell-cell junctions 1,2 -was enhanced by imatinib during thrombin stimulation ( Figure  4E). Together, these data indicate that imatinib limits Argmediated endothelial barrier dysfunction by enhancing Rac1 activity and by enforcing adhesion of endothelial margin areas to the extracellular matrix.

Effect of Imatinib-Sensitive Tyrosine Kinases on Basal Barrier Function
Although imatinib had no effect on basal barrier function ( Figure 1A and 1D), we observed that inhibition of individual imatinib-sensitive kinases did change basal endothelial barrier function. Under nonstimulated conditions PDGFR-and c-KIT-depleted endothelial monolayers displayed improved barrier function, whereas c-Abl depletion reduced barrier function ( Figure 5A). The finding that simultaneous inhibition of all these kinases (cq, by imatinib) has no effect on basal barrier function, suggests that the barrier-impairing effect of c-Abl inhibition is balanced by the beneficial effect of PDGFR and c-KIT inhibition, rendering a net zero effect on basal endothelial barrier function. To test this, we treated endothelial cell monolayers with Tyrphostin AG1296, a selective inhibitor of PDGFR and c-KIT. Tyrphostin AG1296 enhanced endothelial resistance up to 35% ( Figure 5B). These data indicate that under basal conditions c-Abl inhibition opposes the barrier enforcing effects of Arg/PDGFR/c-KIT inhibition, whereas thrombin-induced endothelial barrier dysfunction was only influenced by Arg.

Imatinib Protects Against Vascular Leakage and Pulmonary Edema Formation In Vivo
To establish the protective effect of imatinib on endothelial barrier function in vivo, we tested imatinib in mouse models of vascular leakage and pulmonary edema. VEGF-induced vascular leakage of albumin was measured by intravenous injection of Evans Blue, followed by injection of VEGF in the skin. 3 Vascular leakage was compared between mice pretreated with imatinib and mice pretreated with vector. Imatinib treatment (20 mg/kg) attenuated VEGF-induced extravasation of Evans Blue in the skin by 39% to 55% ( Figure 6A and 6B). Next, we measured the effect of imatinib on pulmonary edema formation. Acute pulmonary edema was induced in vivo by intravenous injection of thrombin-receptor activating peptide in mice pretreated with imatinib or vector. Ex vivo, the weight gain of isolated perfused lungs was measured, reflecting the pulmonary vascular permeability for fluid (K fc ). 16 Imatinib treatment (50 mg/kg) reduced pulmonary edema formation, shown by 66% reduction of K fc ( Figure 6C). Thus, imatinib effectively prevents vascular leakage and edema formation in vivo.
To exclude the possibility that the attenuation of vascular leakage resulted from a smaller hydrostatic pressure difference or a decrease in microvascular perfusion, we measured the effect of imatinib on these parameters in an experimental set-up similar to the Miles assay and the K fc measurements. First, systemic blood pressure was measured using radio telemetry. Blood pressure was monitored before and after administration of imatinib ( Figure VIIA in the online-only Data Supplement). Comparing mean arterial pressure 5 minutes before and 30 minutes after administration of imatinib, no effect of imatinib on mean arterial pressure was observed ( Figure 6D; Figure VIIA in the online-only Data Supplement). Subsequently, the effect of imatinib on microvascular perfusion was evaluated in skin and muscle by contrastenhanced ultrasonography. Comparing microvascular blood volume (as measure of microvascular perfusion) before and after administration of imatinib, we found that imatinib did not decrease microvascular perfusion in skin ( Figure 6E and 6F; Figure  Together, these experiments indicate that the protective effect of imatinib on vascular leakage cannot be explained by smaller hydrostatic pressure differences or decreased microvascular perfusion. As proof-of-concept these data therefore support the hypothesis that imatinib prevents vascular leakage through a direct protective effect on the endothelial barrier.

Imatinib Treatment Attenuates Vascular Leakage During Sepsis
To evaluate the effect of imatinib in a clinically relevant disease model, sepsis was induced by CLP, 17 and mice were treated with imatinib or vector 6 hours and 18 hours after induction of sepsis ( Figure 7A). For evaluation of vascular leakage, Evans Blue was administered intravenously 23 hours after induction of sepsis, and organs were harvested 1 hour after Evans Blue administration. Septic mice treated with vector showed a 2-to 3-fold increase in Evans Blue in the kidneys, which was attenuated by 50% in septic mice treated with imatinib 50 mg/kg ( Figure 7B). A similar trend was observed for the liver, although post hoc analyses did not show a statistical difference between vector-and imatinib-treated mice ( Figure 7C). In the lungs, a slight increase of Evans Blue was observed in septic mice treated with vector. This increase was not significant, mainly because not all mice developed vascular leakage in the lungs ( Figure 7D). Yet, the number of mice that developed pulmonary vascular leakage on sepsis was significantly higher in vector-treated mice than in imatinib-treated mice (0/5 in the sham group versus 3/6 in the CLPϩvector group versus 0/5 in the CLPϩimatinib group; PϽ0.05 in a 2 test). This animal study demonstrates that imatinib attenuates vascular leakage in a clinically relevant disease model and indicates that imatinib is also effective when imatinib treatment is initiated after induction of disease.

Discussion
Here we show that treatment with imatinib is an effective therapeutic approach to endothelial barrier dysfunction and vascular leakage. Imatinib attenuated endothelial barrier dysfunction in human endothelial cells isolated from multiple origins and stimulated with a variety of barrier-disruptive agents. Specifically, we found that imatinib exerts its protective effects via inhibition of Arg, a thus far unknown mediator of endothelial barrier dysfunction. Imatinib limited Argmediated endothelial barrier dysfunction by enhancing Rac1 activity and enforcing adhesion of endothelial cells to the extracellular matrix. The barrier-protective effect of imatinib was established in in vivo models of vascular leakage and pulmonary edema.

Effect of Imatinib on Endothelial Barrier Function
Our finding that short-term treatment with imatinib protects against endothelial barrier dysfunction and edema formation provides first mechanistic insight regarding previous case reports on patients in whom initiation of imatinib treatment was followed by fast resolution of pulmonary edema. 5,11 Combining in vitro and in vivo measurements of endothelial barrier dysfunction and vascular leakage, we found that imatinib protects against edema formation by enforcing the endothelial barrier. Although edema formation and vascular leakage may also be affected by changes in blood pressure, microvascular perfusion, or vascular remodeling, these factors are less likely to underlie the protective effect of imatinib. Alteration of blood pressure and microvascular perfusion as explanation for edema resolution was excluded in this study, because (1) imatinib did not affect systemic blood pressure in an experimental set-up similar to the Miles assay or K fc measurements, (2) the pressure and the flow in the pulmonary circulation was kept constant in the K fc measurements, and (3) no effects of imatinib on microvascular perfusion were observed. The acute character of the in vivo experiments further excludes chronic vascular remodeling as explanation for the protective effects of imatinib on edema formation and vascular leakage. Therefore, we conclude that imatinib protects against edema formation by preservation of endothelial barrier integrity.
Whereas the Miles assay and the K fc measurements serve as proof-of-concept experiments in which imatinib was given as pretreatment and possible confounders were excluded, the clinical relevance of the protective effect of imatinib on endothelial barrier function was evaluated in a murine model of sepsis (CLP). This experiment mimics the clinical setting, because CLP is considered the most reliable disease model available for sepsis, 17 and because the treatment sequence in this experiment mimicked the clinical sequence of development of disease and subsequent initiation of treatment. In septic mice we found that imatinib reduced vascular leakage of Evans Blue in the kidneys by 50%, resembling the attenuating effect found in the Miles assay. In addition, the number of septic mice developing vascular leakage in the lungs was significantly lower in the imatinib-treated group than in the vector-treated group. As reported previously, 22 a high interindividual variation was observed for vascular leakage in liver and the lungs, which may account for the lack of significance in post hoc analyses.
The optimal protective effect of imatinib on endothelial barrier function was already achieved at concentrations be- Figure 7. Imatinib attenuates vascular leakage in a murine model of sepsis. A, Study protocol to assess the effect of imatinib on sepsis-induced vascular leakage of Evans Blue (EB). Sepsis was induced in male C57/BL6J mice by cecal ligation and puncture (CLP). Vascular leakage of EB was compared between vector-and imatinib-treated mice. Imatinib (50 mg/kg subcutaneous) or vector were administered 6 hours and 18 hours after induction of sepsis. EB was administered after 23 hours (100 L of 1% EB solution, intravenous), followed by euthanization of the mice at tϭ24 hours. Vascular leakage of EB was measured in the kidneys, the liver, and the lungs. B, Vascular leakage of EB in the kidneys, corrected for organ dry weight (DW) and the EB plasma concentration. *PϽ0.05, ***PϽ0.001 in Tukey post hoc test of 1-way ANOVA (nϭ5-6 mice per group). C, Vascular leakage of EB in the liver, corrected for organ dry weight (DW) and the EB plasma concentration. NS indicates nonsignificant, *PϽ0.05 in Tukey post hoc test of 1-way ANOVA (nϭ5-6 mice per group). D, Vascular leakage of EB in the lungs, corrected for organ dry weight (DW) and the EB plasma concentration. NS indicates nonsignificant in Tukey post hoc test of 1-way ANOVA (nϭ5-6 mice per group). tween 5 and 10 mol/L in vitro. These concentrations correlate with plasma levels in patients treated with imatinib for chronic myeloid leukemia 23 or gastro-intestinal stromal tumors. 24 Also, the dosage used in our in vivo experiments resembles imatinib dosages used in the clinical setting. The slight difference in treatment concentration between our in vivo experiments (20 -50 mg/kg) and clinical treatment dosage (5-10 mg/kg) is compensated by the higher metabolism and the lower half-life of imatinib in mice (T 1/2 ϭ2-4 hours in mice) 25 compared with human (T 1/2 ϭ18 hours). Therefore, this study not only explains how imatinib may protect against edema, but also proposes imatinib administration as promising approach to edema resulting from endothelial barrier dysfunction.

Role of Arg in Endothelial Barrier Dysfunction
The protective effects of imatinib on endothelial barrier function resulted predominantly from inhibition of the nonreceptor tyrosine kinase Arg. Knock-down of Arg mimicked the effect of imatinib on endothelial barrier function, and imatinib did not have an additive effect in Arg-depleted cells.
To the best of our knowledge, this is the first report showing that Arg is involved in endothelial barrier dysfunction. The importance of Arg as mediator of endothelial barrier dysfunction was illustrated by the fact that Arg inhibition with imatinib reduced the thrombin response up to 44%, whereas the finding that Arg is activated on endothelial stimulation with the barrier-disruptive agents thrombin, VEGF, and histamine stresses its relevance.
In search for signaling pathways underlying the barrierdisruptive actions of Arg, we found that inhibition of Arg by genetic knock-down or imatinib treatment prevented the loss of cell-matrix interaction during endothelial stimulation. This was accompanied by enhanced formation of focal adhesions (FAs), particularly at the periphery of the cell. As proposed by Ingber, 20 adhesion of cells to the subcellular matrix is one of the ways for a cell to remain cell shape and counteract contractile forces during cell retraction. Cell-matrix interaction is mainly achieved through FAs, multifaceted protein complexes that connect extracellular matrix proteins to the intracellular cytoskeleton. 2 The spatial distribution of FAs is an important determinant of endothelial barrier function, as redistribution of FAs to the cell periphery has previously been associated with improved endothelial barrier integrity. 21,26 Fibroblast studies have demonstrated that Arg inhibits this redistribution by reducing formation and increasing turnover of peripheral FAs. 27 Compared with wild-type, Arg-deficient fibroblasts show larger and denser FAs, mainly located at the cell periphery. 28 These studies support our finding that Arg inhibition with imatinib increases the number of FAs at the cell periphery.
In addition, we found that imatinib enhanced the activity of Rac1, a GTPase known to enforce both cell-cell interaction 1,2 and cell-matrix interaction. 2,21 Rac1 activity may enforce endothelial cell-cell junctions via mediators like angiopoietin-1. 29 Of note, Rac1 was also described to mediate peripheral accumulation of FAs, thereby enhancing endothelial barrier function. 21 A direct interrelation between Arg, Rac1, and integrin-mediated adhesion was recently suggested in a fibroblast study that demonstrated that Arg inhibits Rac1 activity and integrin function. 30 Figure 8 shows an overview of the protective effect of imatinib during endothelial barrier dysfunction as proposed in this study. Arg is activated upon binding of barrier-disruptive agents to their receptor. A likely mediator of Arg activation is Src, a tyrosine kinase that is activated by the thrombin receptor and the VEGF receptor 2 and that is able to bind and activate Arg. 31 Arg activation leads to disassembly of peripheral FAs, 27,28 thereby reducing cell-matrix interaction. Imatinib inhibits Arg, which directly leads to preservation of peripheral FAs and improved cell-matrix interaction. In Figure 8. Overview of the protective effect of imatinib during endothelial barrier dysfunction. Imatinib inhibits Abl-related gene (Arg), which is activated upon binding of barrier-disruptive agents to their receptor. Arg activation leads to disassembly of peripheral focal adhesions (FAs) and can be inhibited with imatinib. Peripheral FAs improve cell-matrix interaction and contribute to endothelial barrier integrity by counteracting contractile forces and supporting cell-cell contacts. In addition, Arg inhibition with imatinib enhances Rac1 activity, which supports both cell-cell contacts and cell-matrix interaction.
addition, imatinib enhances Rac1 activation, which in turn improves cell-cell contact and cell-matrix interaction. The enhanced cell-matrix interaction, by supporting cell-cell contacts and counteracting contractile forces, limits cell retraction and gap formation. 20

Potential Further Improvement of Endothelial Barrier Function by Imatinib Derivatives
Considering previous reports on subcutaneous edema as side effect of imatinib, it is also important to note that in the concentrations used in this study, imatinib did not affect basal endothelial barrier integrity. This difference might first of all be explained by treatment duration. Subcutaneous edema as side effect may result from chronic PDGFR inhibition (several months to years) in pericytes and consequent disturbed vascular support. 32 Second, because c-Abl inhibition impairs endothelial barrier function 33 and imatinib inhibits both Arg and c-Abl, the protective effect of imatinib may depend on the balance of Arg and c-Abl expression in a specific vascular bed. In none of the various macro-and microvascular endothelial cell types that we tested, c-Abl inhibition with imatinib impaired barrier function. This suggests that c-Abl inhibition by imatinib has a limited effect on barrier function. However, the opposing effects of Arg/PDGFR/c-KIT versus c-Abl on endothelial barrier function suggest that imatinibderivatives lacking c-Abl as target may further improve treatment of endothelial barrier dysfunction.

Clinical Implications
For several reasons, this study may have direct clinical value. Imatinib had an optimal protective effect at 10 mol/L, which correlates with plasma concentrations in patients on imatinib treatment. 23,24 The barrier protective effect observed in our study was independent of anatomic location, species, endothelial phenotype, and barrier-disruptive agent, indicating a broad applicability of imatinib. Endothelial barrier protection was already achieved after short-term treatment (30 minutes pretreatment in vivo), whereas initiation of imatinib treatment after induction of sepsis was also shown to be effective. This may facilitate edema treatment in acute conditions like sepsis, but also limit side-effects. As noted before, this study elucidates previous clinical observations favoring imatinib treatment in edema. 5,11 Combining our study with these clinical observations provides bench-to-bedside evidence for a protective effect of imatinib on endothelial barrier dysfunction and supports clinical development of imatinib as therapeutic approach to edema.

Conclusion
Thus, imatinib prevents endothelial barrier dysfunction and edema formation via inhibition of Arg. These findings identify imatinib treatment as a promising approach to permeability edema and indicate Arg as novel target for edema treatment.