Calcium Signaling Is Dispensable for Receptor Regulation of Endothelial Barrier Function*

Endothelial barrier function is tightly regulated by plasma membrane receptors and is crucial for tissue fluid homeostasis; its dysfunction causes disease, including sepsis and inflammation. The ubiquitous activation of Ca2+ signaling upon phospholipase C-coupled receptor ligation leads quite naturally to the assumption that Ca2+ signaling is required for receptor-regulated endothelial barrier function. This widespread hypothesis draws analogy from smooth muscle and proposes the requirement of G protein-coupled receptor (GPCR)-generated Ca2+ signaling in activating the endothelial contractile apparatus and generating interendothelial gaps. Notwithstanding endothelia being non-excitable in nature, the hypothesis of Ca2+-induced endothelial contraction has been invoked to explain actions of GPCR agonists that either disrupt or stabilize endothelial barrier function. Here, we challenge this correlative hypothesis by showing a lack of causal link between GPCR-generated Ca2+ signaling and changes in human microvascular endothelial barrier function. We used three endogenous GPCR agonists: thrombin and histamine, which disrupt endothelial barrier function, and sphingosine-1-phosphate, which stabilizes barrier function. The qualitatively different effects of these three agonists on endothelial barrier function occur independently of Ca2+ entry through the ubiquitous store-operated Ca2+ entry channel Orai1, global Ca2+ entry across the plasma membrane, and Ca2+ release from internal stores. However, disruption of endothelial barrier function by thrombin and histamine requires the Ca2+ sensor stromal interacting molecule-1 (STIM1), whereas sphingosine-1-phosphate-mediated enhancement of endothelial barrier function occurs independently of STIM1. We conclude that although STIM1 is required for GPCR-mediated disruption of barrier function, a causal link between GPCR-induced cytoplasmic Ca2+ increases and acute changes in barrier function is missing. Thus, the cytosolic Ca2+-induced endothelial contraction is a cum hoc fallacy that should be abandoned.

The endothelial layer of blood vessels is a highly regulated barrier between the bloodstream and the interstitial tissue, controlling transvascular passage of fluids, solutes, and cells. A significant contribution to endothelial permeability resides in the paracellular diffusion pathway facilitated via intercellular gaps (1)(2)(3). Paracellular permeability is essentially mediated by cellcell junctional proteins, which are in turn regulated by intracellular signaling pathways that impact on cytoskeletal architecture (2,4). The balance between competing tethering and disassembling mechanisms determines the degree of endothelial barrier function and thus the extent of vascular leakage. Disruption of endothelial barrier function causes increased vascular permeability and is associated with reorganization of the actin cytoskeleton and disassembly of adherens junctions which are contributed by vascular endothelial cadherin (VEcadherin)⅐catenin 2 complexes (2,4). These events are under the control of various signaling pathways that are activated by diverse paracrine and autocrine mediators in blood and interstitial tissue, many of which act on heterotrimeric G proteincoupled receptors (GPCR) or receptor tyrosine kinase and play crucial roles in the control of vascular permeability, tone, angiogenesis and inflammation (2)(3)(4).
Barrier disrupting GPCR agonists including thrombin and histamine activate downstream signaling that regulate actin reorganization and VE-cadherin localization triggered by activation of the small GTPase RhoA (1,2). GPCR-mediated activation of the heterotrimeric G protein G␣ 12/13 family subtype causes the association and activation of p115Rho guanine nucleotide exchange factor (p115RhoAGEF), leading to RhoA activation (1,2). RhoA then activates Rho kinase (ROCK), which phosphorylates the regulatory subunit of the myosin light chain phosphatase and inhibits its activity. This in turn enhances the phosphorylation of myosin light chain (MLC), causing actin-myosin interactions and disruption of endothe-lial barrier function (5). Furthermore, ROCK can directly phosphorylate MLC to cause barrier disruption (6). The signaling pathways activated downstream barrier-stabilizing GPCR agonists such as the sphingosine-1-phosphate (S1P) are less understood. S1P acts on heteromultimeric G␣ i proteins leading to the activation of the small GTPase Rac and focal adhesion kinase, thus promoting formation of focal adhesions, cortical actin, and adherens junction assembly (3,7).
Although it is obvious that endothelial cells develop force in response to barrier-disrupting agonists such as thrombin and histamine, the tension developed by endothelial cells in response to the most powerful agonist thrombin is at least 50-fold weaker than the bona fide tension generated by smooth muscle cells during contraction (8). Nevertheless, during the past three decades Ca 2ϩ -dependent endothelial contraction, a concept extrapolated from studies on muscle cells, has been invoked to explain changes in endothelial barrier function downstream GPCR agonists. Barrier disrupting GPCR agonists such as thrombin and histamine activate G␣ q,11 protein and induce the production of inositol 1,4,5-trisphosphate (IP 3 ) through the action of phospholipase C. This will result in Ca 2ϩ release from the IP 3 -sensitive internal stores of the endoplasmic reticulum (ER) and activation of Ca 2ϩ entry across the plasma membrane through the ubiquitous store-operated Ca 2ϩ entry (SOCE) pathway activated by ER store depletion (9). It is now appreciated that ER store depletion causes the ER-resident Ca 2ϩ sensor stromal-interacting molecule 1 (STIM1) to move toward ER-plasma membrane junctional spaces to trap and directly activate Orai1 Ca 2ϩ entry channels (10 -17). According to the Ca 2ϩ -dependent model, the sustained Ca 2ϩ entry signal thus generated (but not Ca 2ϩ release) activates a key Ca 2ϩ -and calmodulin-dependent kinase, the myosin light chain kinase (MLCK) leading to MLC phosphorylation, formation of actin stress fibers, and endothelial contraction resulting in formation of intercellular gaps (3, 18 -21). For the barrier-stabilizing agonist S1P, Ca 2ϩ release from internal stores, but not Ca 2ϩ entry, was proposed to induce Rac activation, thus promoting assembly of adherens junctions and strengthening of endothelial barrier function (22).
Early studies from our group and others demonstrated that in endothelial cells from various vascular beds (human pulmonary artery, human dermal microvasculature, and human umbilical vein) thrombin, VEGF, and the store-depleting drug thapsigargin activate SOCE encoded by STIM1 and Orai1 (11,(23)(24)(25). In a recent study, we have challenged the hypothesis that SOCE is required for endothelial contraction in response to the powerful barrier-disrupting agonist thrombin (23). We demonstrated using molecular tools that thrombin-mediated endothelial barrier disruption required the ER-resident STIM1 protein but occur independently of SOCE, Orai1, and MLCK (23). We also showed that STIM1 is required for RhoA activation, MLC phosphorylation, actin reorganization, and disruption of intercellular adhesions (23).
In the current study, we set out to determine whether these findings are unique to thrombin or shared by other barrieraltering or barrier-enhancing GPCR agonists and whether Ca 2ϩ release from the ER is required for agonist-mediated effects on endothelial barrier function. We thus used high throughput impedance measurements to determine the role of Ca 2ϩ release and Ca 2ϩ entry mechanisms in regulating endothelial barrier function downstream of three GPCR agonists, namely thrombin, histamine, and S1P. Thrombin and histamine are two typical inflammatory agonists that cause transient barrier disruption, whereas the platelet-derived agonist S1P enhances endothelial barrier function. These three agonists are of major relevance to vascular pathologies such as inflammation, allergy, and atherosclerosis. We compared side by side the effects of these three agonists on endothelial barrier function using electrical measurements and fluorescence microscopy. We also monitored the Ca 2ϩ release from stores and Ca 2ϩ entry across the plasma membrane induced by these agonists. We report that although Ca 2ϩ signaling in response to these agonists coincides with changes in barrier function, neither Ca 2ϩ entry nor Ca 2ϩ release are necessary for GPCR-mediated disruption or enhancement of barrier function. We show that Orai1 plays no significant role in thrombin-, histamine-, and S1P-induced changes in endothelial barrier function. However, STIM1 is required for disruption of endothelial barrier function by thrombin and histamine but is not involved in S1P-mediated enhancement of endothelial barrier function.

Results
Thrombin, Histamine, and S1P Evoke Distinct Impedance Response Profiles in HDMECs-Impedance measurements were performed in monolayers of human dermal microvascular endothelial cells (HDMECs) before and after addition of GPCR agonists using ECIS protocols as described in methods and outlined in details recently (26). Upon stimulation of confluent HDMEC monolayers with GPCR agonists, thrombin, histamine, and S1P, each agonist showed a typical response as shown in Fig. 1. Although thrombin and histamine lead to a transient disruption of endothelial barrier function (Fig. 1, A and B), S1P enhanced (or stabilized) endothelial barrier function (Fig. 1C). Furthermore, the barrier-disrupting response of HDMEC monolayers to thrombin and histamine differed in severity and kinetics; thrombin effects were more profound, and the barrier took over an hour to recover, whereas the histamine effects were modest in comparison, and recovery of the barrier to preagonist levels occurred within 10 min (Fig. 1, A and B). Changes in electrical resistance on ECIS electrodes are caused by changes in cell morphology and changes at the intercellular junctions caused by rearrangements of actin and adherens junction proteins. The opening of cell-cell junctions increases paracellular current flow and thus causes decreases in resistance. Reciprocally, strengthening of cell-cell junctions reduces paths for current flow and leads to resistance increases. These results are consistent with cell morphology changes that were assessed microscopically: immunofluorescence staining for actin (green) and VE-cadherin (red) under control conditions and 2 min after addition of the respective agonists showed that although thrombin caused a disruption of adherens junctions, spanning the full length of cell-cell contact sites, histamine caused localized small oval-shaped openings that seem lacking VE-cadherin staining (Fig. 1D). S1P on the other hand led to increased VE-cadherin staining at the cell-cell junctions (Fig.  1D). To illustrate the changes in VE-cadherin and actin distri-bution during the time course of HDMEC response to thrombin and histamine, cell layers were stained at specific time points after agonist addition as indicated by red arrows in Fig.  1E for thrombin and Fig. 1F for histamine. HDMEC immunofluorescence images representing time points at 2, 5, 10, and 30 min after thrombin stimulation and 1, 5, and 20 min after histamine stimulation are shown in Fig. 1 (G and H), respectively. For comparison, typical control non-stimulated HDMEC monolayers are shown in Fig. 1I. According to the Ca 2ϩ -induced contractility model, tension produced through actin stress fiber formation upon agonist stimulation of endothelial cells is required for barrier disruption (3). The fluorescence micrographs show that within the first 2-5 min after thrombin stimulation, actin redistributes into the cytosol away from cellcell contacts, which also leaves a staining pattern of interrupted VE-cadherin/actin co-localization (yellow) at the cell borders (Fig. 1G). Increased proportion of distinct actin bundles became visible after 10 min of thrombin stimulation, during the monolayer recovery phase (Fig. 1G). Additional quantification of fluorescence intensity from actin and VE-cadherin across cell cross-sections reveals actin accumulation in central regions of the cell at 2-10 min after thrombin stimulation, whereas in untreated and S1P-stimulated cell layers actin is predominantly co-localized with VE-cadherin at the cell borders, as indicated by overlap of intensity peaks (Fig. 1J). After 30 min of thrombin stimulation, at which time HDMEC monolayer resistance was . Typical responses of HDMEC to GPCR agonists thrombin, histamine, and S1P. A-C, electrical resistance of confluent HDMEC cell layers after stimulation with 5 nM thrombin (A), 10 M histamine (B), and 1 M S1P (C) where indicated by arrows. Electrical resistance was normalized to the last values before agonist addition. Absolute baseline resistances were 4202 ⍀ (A), 3841 ⍀ (B), and 2198 ⍀ (C). Note that for thrombin and histamine stimulation, mature HDMEC cell layers were used after 4 days in culture, whereas S1P experiments were intentionally conducted with cell layers after 2 days in culture, which explains the lower initial baseline resistance. S1P experiments were performed in less mature HDMEC monolayers to better resolve S1P-mediated enhancement of barrier function. D, confocal fluorescence images of confluent HDMEC cell layers 2 min after stimulation with thrombin, histamine, or S1P compared with the control. The cells were stained for VE-cadherin (red), actin (green), and Hoechst 33342 (blue). E-I, kinetics of changes in HDMEC cell layer morphology and junctions after 5 nM thrombin and 10 M histamine stimulation were followed at different time points as indicated by red arrows in ECIS traces (E and F). Thrombin and histamine addition, respectively, is indicated by black arrows. Distribution of VE-cadherin and actin is shown for the selected time points after thrombin (G) or histamine (H) stimulation and compared with an untreated control (I). The white arrows point to areas with increased amount of actin bundles and fiber structures. Asterisks indicate cells used for quantification of actin and VE-cadherin distribution in J. J, quantification of actin (green traces) and VE-cadherin (red traces) fluorescence after thrombin, histamine, or S1P stimulation. Intensity histograms (in arbitrary units, a.u.) were generated for the green (actin) and red (VE-cadherin) fluorescence channel images as function of pixel position (a.u.) along cell cross-sections indicated by the white solid lines drawn into the corresponding fluorescence image. Within the intensity plots, cell borders are indicated by dashed lines.
typically restored by ϳ80% of basal levels (Fig. 1E), accumulation of actin toward the cell borders was evident (Fig. 1, G and J). Similarly, in HDMEC monolayers stimulated with histamine, the appearance of stress fibers occurred 20 min after agonist stimulation (Fig. 1H), but was clearly absent in early stages, where decrease in resistance was maximal (Fig. 1F) and fluorescence intensity of actin and VE-cadherin staining decreased at the cell borders (Fig. 1J). These data suggested to us that the rapid decrease in barrier function in response to thrombin and histamine precedes stress fiber formation. Initial breakdown of endothelial barrier might be rather due to redistribution of actin away from the cell-cell junctional sites, thereby weakening junctional complexes. In turn, actual actin fiber formation might be more important during the barrier recovery phase when actin redistributes back toward the cell junctions for their stabilization. Thus, actin stress fiber tension might not be necessary for GPCR-mediated barrier breakdown.
Thrombin, Histamine, and S1P Activate SOCE in HDMECs-We tested whether Ca 2ϩ signaling is activated in HDMECs by our three model agonists and whether Orai1 and STIM1 mediate SOCE in response to these agonists as was shown for virtually all cell types, including endothelial cells (10,11,13,27,28). We and others have shown that GPCR and receptor tyrosine kinase agonists activated Ca 2ϩ entry into endothelial cells (from various vascular beds including HDMEC) through SOCE bearing typical pharmacological signature and mediated by Orai1 and STIM1 (11,12,23,24). Like thrombin, the other two GPCR agonists considered under this study (histamine and S1P) also couple to phospholipase C activation with generation of IP 3 , activation of Ca 2ϩ release from internal stores, and subsequent activation of Ca 2ϩ entry across the plasma membrane presumably through the SOCE pathway. One of the most common means to activate SOCE without stimulating membrane receptors and without generating second messengers is to passively deplete the internal Ca 2ϩ stores using the sarcoplasmic/ endoplasmic reticulum Ca 2ϩ -ATPase (SERCA) inhibitor thapsigargin (29). Treatment of HDMEC with thapsigargin in the absence of extracellular Ca 2ϩ led to a typical transient response corresponding to Ca 2ϩ release from internal stores ( Fig. 2A, black trace). Restoration of 2 mM Ca 2ϩ to the external buffer revealed a second response corresponding to Ca 2ϩ entry across the plasma membrane ( Fig. 2A, black trace). These typical Ca 2ϩ release and Ca 2ϩ entry profiles can be readily observed when HDMEC are stimulated with thrombin, histamine, or S1P under the same recording protocol (Fig. 2, B-D, black traces). Note that although 5 nM thrombin induces maximal changes in endothelial barrier function (Fig. 1A), concentrations of 100 nM thrombin were required to obtain significant Ca 2ϩ responses to thrombin.
To show involvement of SOCE upon GPCR activation, as a first step, we systematically tested side by side the effect of different commonly used SOCE inhibitors on the HDMEC response to thapsigargin and the three GPCR agonists, thrombin, histamine, and S1P (Fig. 2). We used three different SOCE channel inhibitors; the lanthanide ion Gd 3ϩ at relatively low concentrations of 5 M is a specific inhibitor of SOCE (30,31), BTP2 (32-35), and 2-APB (36, 37). Ca 2ϩ entry (but not Ca 2ϩ release) in response to thapsigargin and all three agonists was inhibited by 5 M Gd 3ϩ (Fig. 2, A-D, red traces), 10 M BTP2 (Fig. 2, E-H, red traces), and 25 M 2-APB (Fig. 2, I-L). As shown in Fig. 2, these commonly used SOCE inhibitors inhibited the Ca 2ϩ entry phase when cells were either treated with drugs 2 min before the addition of agonist (red traces) or when drugs were added after Ca 2ϩ entry has developed (control black traces; arrow indicates addition of drug). Specifically, Ca 2ϩ entry in response to thapsigargin, thrombin, and histamine was potentiated by 2-APB at 5 M and inhibited at 25 M (Fig. 2, I-K, black traces), a defining peculiarity of Orai1-mediated SOCE (36,38). Furthermore, when cells were preincubated with 5 M 2-APB before the addition of thapsigargin, thrombin, or histamine (Fig. 2, I-K, blue traces), the Ca 2ϩ entry phase was enhanced, but when 25 M 2-APB was subsequently added, Ca 2ϩ entry was inhibited. In conclusion, all three model GPCR agonists thrombin, histamine, and S1P can activate Orai1-mediated SOCE in HDMEC in a manner reminiscent of passive store depletion with thapsigargin.
Effect of SOCE Inhibitors on GPCR-mediated Endothelial Barrier Regulation-We previously showed that blockade of SOCE in HDMEC with 5 M Gd 3ϩ did not affect thrombinmediated disruption of barrier function (23). This original finding led to the hypothesis that Ca 2ϩ entry is not necessary for mediating typical acute barrier responses to GPCR agonists. Therefore, we decided here to systematically test the effects of all three SOCE inhibitors (Gd 3ϩ , 2-APB, and BTP2) on the barrier-modulating effects of the three GPCR agonists, thrombin (5 nM), histamine (10 M), and S1P (1 M) by using impedance measurements. HDMEC cell layers grown onto ECIS electrodes were subjected to the three SOCE inhibitors: Gd 3ϩ , BTP2, and 2-APB. Dose-response curves showed that when these inhibitors were added to HDMEC monolayers, without additional GPCRstimulation,theysignificantlyalteredthebaselineresistance of the monolayers at higher concentrations, 25 M for Gd 3ϩ , 5 M for BTP2, and 25 M for 2-APB ( Fig. 3, A-C). BTP2 effects were transient with a recovery to baseline values within 2-3 h (Fig. 3B). Because of these intrinsic inhibitor effects, in all subsequent data related to effects of inhibitors on GPCR-mediated changes in barrier function, for the sake of transparency, we thought it more appropriate to represent raw traces that are not background subtracted (i.e. without subtracting from these data resistance values obtained with inhibitors alone).
The addition of 5 M Gd 3ϩ to HDMEC monolayers, which fully blocked SOCE without affecting the basal monolayer resistance, did not affect the typical changes in cell layer resistance upon stimulation with thrombin, histamine, or S1P (Fig. 4, A-C). All traces from Fig. 4 represent averages from 3-4 wells/ condition from a single HDMEC isolation. Even higher concentrations of Gd 3ϩ (10 -50 M) were without effect on changes in barrier function induced by all three GPCR agonists (Fig. 4, A-C). Statistical analysis on maximum change in normalized resistance in the absence or presence of Gd 3ϩ at concentrations of 5, 10, or 50 M from several independent experiments, 4 -12, involving independent HDMEC isolations showed no statistical significance for all agonists (Fig. 4, D-F).
The decrease in monolayer resistance in response to thrombin and histamine was enhanced in the presence of BTP2 (Fig.   OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44   A and B). The transient decrease of resistance observed when BTP2 is added alone (Fig. 3B) is likely responsible for this enhanced decrease in resistance seen when HDMEC monolayers are stimulated with either thrombin or histamine in the presence of BTP2. Indeed, when this effect of BTP2 on basal monolayer resistance is taken into consideration, BTP2 even at higher concentrations that fully inhibit SOCE has virtually no effect on thrombin-and histamine-mediated disruption of HDMEC barrier function (Figs. 5, A and B, and 3B). The lowest concentration of 1 M BTP2 used in experiments reduced SOCE, without causing full block (data not shown). This was consistent with reported BTP2 effect on Jurkat cells where 1 M BTP2 resulted in ϳ25% inhibition of Orai1-mediated Ca 2ϩ release-activated Ca 2ϩ current (39). With all concentrations of BTP2 used, the S1P response profile of HDMEC monolayers was not affected (Fig. 5C). The effect of BTP2 at different concentrations of 1, 5, and 10 M in 6 -12 experiments involving HDMEC from several independent isolations is summarized in Fig. 5 (D-F).

5,
As discussed in Fig. 3C, 2-APB at 25 M led to a slow decrease of cell layer resistance over the time course of the measurements and thus seemed to be a side effect of 2-APB, independently of the GPCR agonist used (Fig. 6). Even when taking into consideration this slow effect of 2-APB on basal monolayer resistance (Fig. 3C), treatment of HDMEC on ECIS electrodes with 25 M 2-APB significantly inhibited the acute and rather fast decrease in resistance after thrombin and histamine stim-ulation (Fig. 6, A and B) but had no significant effect on the S1P-induced increase in monolayer resistance (Fig. 6C). The effect of 2-APB at different concentrations of 5, 10, and 25 M in 6 -11 experiments involving HDMEC from several independent isolations is summarized in Fig. 6 (D-F).
In summary, of the three well known SOCE inhibitors, only 2-APB reliably inhibited the initial fast decrease in HDMEC monolayer resistance induced by thrombin and histamine with marginal effect on S1P-induced enhancement of barrier function. Therefore, the effects of 2-APB are clearly not related to inhibition of SOCE and could be due either to nonspecific effects of 2-APB on endothelial barrier function or to the ability of 2-APB to block STIM1 oligomerization and reorganization (40), consistent with the reported STIM1-mediated control of barrier dysfunction in response to thrombin that occur independently of Orai1 and Ca 2ϩ entry (23). To confirm results with SOCE inhibitors, we decided to complement these pharmacological data with molecular knockdown in HDMEC monolayers of SOCE components Orai1 and STIM1, as discussed in the following.
Effect of Knockdown of Orai1 and STIM1 on GPCR-mediated Endothelial Ca 2ϩ Signaling and Barrier Regulation-For molecular knockdown of Orai1 and STIM1, we used siRNA sequences previously characterized in our laboratory, which achieved significant knockdown in 80 -90% of HDMEC monolayers without off-target effects, specifically on other STIM/ Orai isoforms (11,23,(41)(42)(43). Non-targeting siRNA (siNT) were used as control, and knockdown efficiency was quantified by Western blot and shown for both STIM1 and Orai1 (Fig. 7, A and C); densitometry results from three (for STIM1) and four (for Orai1) independent transfections were quantified statistically (Fig. 7, B and D). Functional evaluation of SOCE activated by passive store depletion using thapsigargin (Fig. 7, E-H) or agonist stimulation with histamine (Fig. 7, I-L) using Fura-2 imaging showed significant inhibition of Ca 2ϩ entry after knockdown of either STIM1 or Orai1. The effects of knockdown of STIM1 and Orai1 were quantified from a large number of cells originating from 3-4 independent transfections with 6 -8 independent recordings (Fig. 7, F, H, J, and L); the two numbers on the bar graphs separated by comma "x, y" represent the number of independent runs (x) and the total number of cells averaged (y). STIM1 and Orai1 knockdown was performed using a second set of sequences (siSTIM1 # 2 and siOrai1 # 2), and effective inhibition of thrombin and histamine-mediated Ca 2ϩ entry was confirmed with these additional sequences (Fig.  7, M-T).
Knockdown of either STIM1 or Orai1 had a slight effect on the basal HDMEC monolayer resistance, but this effect was somewhat significant only for Orai1 knockdown (Fig. 8, A-D), suggesting a role for Orai1 in proper maturation of HDMEC monolayers. An important observation is that initial baseline values vary significantly depending on the primary cell isolation. The importance of paying attention to initial baseline values and its effect on change in resistance after GPCR activation has been recently discussed in detail (26). A lower initial resistance leads to a smaller resistance change after agonist stimulation, when compared with a cell layer with high initial resis- tance when all other conditions are equal (26). We quantified the response intensities (⌬R) for each agonist and found a clear dependence of the signal intensity on initial resistance values of the HDMEC cell layers for thrombin and histamine, which was also evident for all types of siRNA used (siNT, siOrai1, and siSTIM1) (Pearson's correlation coefficients 0.58 -0.87; data not shown). Because siRNA treatment itself had no impact on basal HDMEC barrier development after cell seeding compared with siRNA-untreated control HDMEC (Fig. 8, E and F), we ascribed differences in baseline unstimulated resistance to variation in cell isolations between human donors. We intentionally chose to conduct knockdown experiments with HDMEC monolayers from different isolations generating different basal resistance values to show the general applicability of our findings. Therefore, and as discussed in the next paragraph, ⌬R values showed wide variation among different individual experiments.
STIM1 knockdown inhibited the typical decrease in HDMEC monolayer resistance but only upon stimulation with thrombin and histamine with no effect on S1P-induced enhancement of HDMEC monolayer resistance (Fig. 9, A-C). However, Orai1 knockdown failed to affect the changes in endothelial barrier function in response to all three GPCR agonists (Fig. 9, D-F). Traces in Fig. 9 (A-F) represent averages from n ϭ 4 -8/condition from a single HDMEC isolation. To establish solid statistical coverage over different measurements from different cell isolations, statistical analyses were performed on 4 -10 independent HDMEC isolations for each experimental condition (with 23-86 independent wells/condition). The collected data are shown as density plots (Fig. 9, G-I).
High densities indicate frequent occurrence of respective response intensities (⌬R), and vertical lines represent the mean response intensity. As shown in Fig. 9 (G and H), STIM1 knockdown caused a left shift in response intensity upon thrombin and histamine stimulation toward lower ⌬R values. The respective means of ⌬R values from siSTIM1-treated cell layers were significantly different from those of siNT-treated cells, as revealed by ANOVA analysis. In contrast, Orai1 knockdown does not affect the response intensity to thrombin or histamine (Fig. 9, G and H). Neither STIM1 nor Orai1 knockdown significantly influenced ⌬R values in response to S1P stimulation (Fig. 9I). Qualitatively the same results were obtained using a second set of siRNAs (siSTIM1 # 2 and siOrai1 # 2) (Fig. 9, J-O).
In summary, our statistical analysis confirmed that significant effects on GPCR-mediated changes in endothelial barrier function were only detectable upon STIM1 knockdown and only for stimulation with thrombin and histamine. STIM1 obviously plays an important role in mediating the opening of endothelial barrier upon GPCR stimulation, but its effect is independent from Ca 2ϩ entry. In the following, we set out to further support our finding that HDMEC barrier breakdown is independent from extracellular Ca 2ϩ rise.
Role of Extracellular Ca 2ϩ in GPCR-mediated Endothelial Barrier Regulation-We have previously refrained from performing long term experiments that omit Ca 2ϩ from the external solution as means to test the involvement of Ca 2ϩ entry in endothelial barrier function (23). Ca 2ϩ ions are required co-factors for adhesion molecules, and long term removal of external Ca 2ϩ will inevitably lead to gradual decrease in HDMEC monolayer resistance. Therefore, here we used a protocol involving short term removal of external Ca 2ϩ (for ϳ8 min) while monitoring changes in cytosolic Ca 2ϩ in HDMEC, the monolayer resistance by ECIS, and visual inspection of monolayers by fluorescence microscopy. As seen in Fig. 10, the addition of thrombin in the absence of external Ca 2ϩ caused an even bigger decrease in HDMEC monolayer resistance than obtained when thrombin was added in presence of external Ca 2ϩ ( Fig. 10A; 11 min zoom in is shown in Fig. 10B). Restoration of 2 mM Ca 2ϩ to the external solution did not cause additional decrease in monolayer resistance (Fig. 10B). Fluorescence images showed that the disruption of cell-cell junctions after thrombin is added in the absence of external Ca 2ϩ was much more substantial than that observed in the presence of 2 mM Ca 2ϩ (Fig. 10D). This is likely due to monolayers being less stable in the absence of external Ca 2ϩ as a result of weak cell-cell and cell-substrate adhesion. The Ca 2ϩ response of HDMEC in the absence and presence of external Ca 2ϩ showed the expected Ca 2ϩ release and Ca 2ϩ entry phases (Fig. 10C). Our conclusion from this experiment is that although extracellular Ca 2ϩ is required for proper cell adhesion and monolayer stability and integrity, GPCR-activated Ca 2ϩ entry across the plasma membrane is not required to open the endothelial barrier.
Role of Ca 2ϩ Release from Internal Stores in GPCR-mediated Endothelial Barrier Regulation-The use of Ca 2ϩ -free extracellular solutions, SOCE inhibitors, and STIM1/Orai1 molecular knockdown used thus far only prevent Ca 2ϩ entry across the plasma membrane with no effect on Ca 2ϩ release from IP 3sensitive stores. Therefore, we set out next to determine whether Ca 2ϩ release from internal stores is required for changes in endothelial barrier function in acute response to GPCR agonists. We therefore incubated HDMEC monolayers with the membrane permeant form of the fast pH-insensitive Ca 2ϩ chelator BAPTA (BAPTA-AM). Incubation of HDMEC monolayers with 10 M of BAPTA-AM for 10 min caused a substantial inhibition of both Ca 2ϩ release and Ca 2ϩ entry in response to thapsigargin as measured with Fura-2 cytosolic Ca 2ϩ imaging (Fig. 11A, red trace); 1 M of BAPTA-AM had partial effects on inhibiting these thapsigargin-induced Ca 2ϩ signals in HDMEC (Fig. 11A, blue trace) and was therefore not used any further. The incubation of HDMEC with 10 M of BAPTA-AM for 10 min also substantially inhibited both Ca 2ϩ release and Ca 2ϩ entry in response to thrombin (Fig. 11B), histamine (Fig. 11C), and S1P (Fig. 11D). However, 10 min of preincubation of monolayers with 10 M BAPTA-AM failed to inhibit the changes in endothelial barrier function in response to thrombin (Fig. 11E), histamine (Fig. 11F), and S1P (Fig. 11G). Thus, buffering of Ca 2ϩ release from internal stores does not prevent the acute agonist-mediated changes in endothelial barrier function.

Discussion
Previous work from our laboratory showed that STIM1 and Orai1 are the molecular components of SOCE in several human endothelial cell types, including endothelial cells from the umbilical vein, the pulmonary artery, and the dermal microvas-culature (HDMECs) (11,23). We also showed that thrombin and VEGF trigger Ca 2ϩ entry into endothelial cells through SOCE and its biophysical correlate, the Ca 2ϩ release-activated Ca 2ϩ current (11,23). We performed ECIS studies using both human endothelial cells from the umbilical vein and HDMEC monolayers and showed that STIM1 controls thrombin-mediated disruption of endothelial barrier function independently of Orai1 and Ca 2ϩ entry, suggesting that SOCE and Ca 2ϩ entry in general were not required for thrombin-mediated disruption of endothelial barrier function.  . For this specific experiment, specialized ECIS arrays were used (2W4 ϫ 1E PC). These arrays had wells with the same dimensions as the chambers used for Ca 2ϩ imaging, and ECIS measurements were performed at room temperature. C, corresponding Ca 2ϩ imaging trace upon addition of 100 nM thrombin using the same ϪCa 2ϩ / ϩCa 2ϩ protocol. D, corresponding immunofluorescence staining of VE-cadherin (red), actin (green) and nuclei stain Hoechst (blue) on cells stimulated for 2 min with 100 nM thrombin in the Ca 2ϩ -free (upper panels) and Ca 2ϩ -containing (lower panels) HBSS.
In the current study, we aimed to determine whether the requirement of STIM1 and the dispensable aspect of Orai1 and Ca 2ϩ entry in thrombin-mediated barrier disruption is generally applicable to other GPCR agonists, including those that enhance barrier function and whether Ca 2ϩ release from internal stores is required for GPCR-mediated changes in barrier function. Thus, in addition to thrombin, we chose to focus on two additional GPCR agonists, histamine which disrupts endothelial barrier function but does so with less intensity and quicker recovery time than thrombin, and S1P, which enhances endothelial barrier function. We found that all three GPCR agonists, thrombin, histamine, and S1P activate Ca 2ϩ entry, which displays the typical pharmacological profiles of SOCE. Inhibition of SOCE and Ca 2ϩ entry in general with lanthanides or BTP2 had no effects on changes in barrier function caused by all three GPCR agonists. However, the uses of 25 M 2-APB (a concentration that inhibits Ca 2ϩ entry with no effect on Ca 2ϩ release) inhibited thrombin-and histamine-mediated disruption of barrier function with no effect on S1P-mediated enhancement of barrier function. In light of the results obtained with Gd 3ϩ and BTP2, the effect of 2-APB is likely unrelated to its ability of inhibiting Ca 2ϩ entry. 2-APB was shown to inhibit STIM1 function by preventing STIM1 aggregation and movements in response to store depletion (40), which is consistent with our STIM1 molecular knockdown inhibiting barrier disruption in response to thrombin and histamine. STIM1 knockdown had no effect on the barrier-stabilizing function of S1P, suggesting that STIM1 plays a role in processes involving dis-ruption of barrier function but not in the molecular pathways that control stabilization of endothelial barrier function. Knockdown of Orai1 failed to affect the disruption of barrier function in response to thrombin and histamine and had no bearing on the enhancement of barrier function in response to S1P.
We also showed that the use of a fast cytosolic Ca 2ϩ buffer, BAPTA-AM on endothelial monolayers led to a strong inhibition of the Ca 2ϩ release and Ca 2ϩ entry signals in response to thapsigargin and GPCR agonists. However, this Ca 2ϩ buffering failed to affect the typical response of endothelial monolayers to thrombin, histamine, and S1P, strongly arguing that Ca 2ϩ release and Ca 2ϩ entry are not necessary for GPCR-mediated changes in endothelial barrier function. These results are consistent with previous studies showing that thrombin-mediated disruption of endothelial barrier function occurs independently of G q -mediated Ca 2ϩ signaling (44) and a recent study showing that histamine-mediated disruption of endothelial barrier function requires RhoA and ROCK, whereas phospholipase C pharmacological inhibition had a marginal effect (45). Our previous studies showed that knockdown of STIM1 caused inhibition of RhoA activity and MLC phosphorylation in response to thrombin. We also showed that STIM1 knockdown led to increased basal phosphorylation of focal adhesion kinase and paxillin, consistent with a role for STIM1 in mediating GPCRinduced endothelial barrier disruption through RhoA independently of receptor-activated Ca 2ϩ signaling (23). Likewise, the stabilizing effects of S1P on barrier function occur via Rac-me- diated pathways and are independent of STIM1, Orai1, and Ca 2ϩ signaling. Collectively, the data in the literature support that regulation of endothelial barrier disruption and stabilization is likely mediated by the balance between the activities of Rho and Rac/Cdc42.
The "increased cytosolic Ca 2ϩ -induced endothelial contraction" hypothesis stipulates that Ca 2ϩ entry upon GPCR ligation activates MLCK, which in turn phosphorylates MLC to cause endothelial contraction and disruption of barrier function (3,18,21). Earlier studies that proposed a role for MLCK in controlling endothelial barrier function downstream GPCR agonists relied on nonspecific approaches such as the use of MLCK inhibitory peptides or MLCK pharmacological inhibitors, namely ML7 and ML9 (46 -48). Subsequently, ML9 was shown to inhibit STIM1 aggregation and movement upon store depletion in an MLCK-independent manner (40). Work in our laboratory showed that effective knockdown of endothelial MLCK failed to alter the disruption of endothelial barrier function in response to thrombin, ruling out the requirement of MLCK in GPCR-mediated acute control of barrier function (23).
Other Ca 2ϩ -sensitive enzymes, including Ca 2ϩ /calmodulindependent kinase II (CaMKII) and PKC, were proposed to regulate the cell-cell junctional VE-cadherin⅐catenin complex disassembly in endothelial cells in response to thrombin (49 -55). For instance, the endothelial CaMKII␦6 isoform was shown to play a role in thrombin-mediated endothelial barrier disruption. However, CaMKII␦6 contribution was through RhoA/ ROCK-dependent mechanism and was apparent only at low concentrations of thrombin (2.5 nM) where the Ca 2ϩ signal activated by thrombin was negligible, but not at higher concentrations of thrombin (56). Therefore, increased cytosolic Ca 2ϩ and MLCK are not absolutely required for triggering the acute and rapid disruption in barrier function in response to GPCR agonists; RhoA activation appears necessary and sufficient. However, we cannot obviously rule out subtle regulatory roles for cytosolic Ca 2ϩ and MLCK (or other Ca 2ϩ -activated enzymes) during the signaling processes impacting on cytoskeletal rearrangements such as endothelial permeability or migration. In most in vivo studies, Ca 2ϩ signals and endothelial permeability to agonists, measured for instance on mesenteric venules of anesthetized rats, are merely correlative and are based on the use of pharmacological compounds with questioned specificity (57)(58)(59). However, a regulatory role for cytosolic Ca 2ϩ rise and subcellular activation of Ca 2ϩ -activated enzymes in endothelial barrier function might be more significant under in vivo conditions that are more complex and where subtle rearrangements of the barrier at specific sites are sufficient to allow passage of cells and solutes. We are arguing herein that the idea of acute and rapid changes in endothelial barrier function being triggered by a Ca 2ϩ signal in a manner analogous to contraction of smooth muscle is not supported by the evidence and therefore the model of "Ca 2ϩ -induced endothelial contraction" should be abandoned.
For impedance measurements, the cells were grown onto ECIS cultureware arrays from Applied BioPhysics Inc. (Troy, NY). If not specified otherwise electrode type 8W10Eϩ PET were used. The other array types used were 96w20idf PET and 2W4 ϫ 10E PC. Before use, arrays were treated with sterile filtered 10 mM cysteine in water for 10 min to generate stable electrode impedance. Afterward electrodes were rinsed three times with sterile water and coated with 300 l of 0.2% gelatin in PBS (8.0 g/liter NaCl, 0.2 g/liter KCl, 1.42 g/liter Na 2 HPO 4 , and 0.24 g/liter KH 2 PO 4 pH 7.4) for 10 min (gelatin was purchased from Acros Organics, gelatin for analysis, granular; catalog no. AC410875000, cas 9000-70-8). HDMEC were seeded onto the ECIS cultureware at a density of 100,000 cells/cm 2 in 400 l of complete medium. To allow uniform settling of cells at the bottom of the wells, the arrays were left for ϳ10 min on the bench. The cells were maintained in culture for 2-4 days (depending on the agonist studied; see below) in a humidified incubator with 5% CO 2 at 37°C. Every 24 h, half of the medium was exchanged by removing 200 l of old medium and adding back 200 l of fresh medium. Exchanging only 50% of medium instead of a full exchange reduces mechanical stress on the cell layer. Without medium changes, the cell layers established lower baseline resistances and higher variability in baseline values, as well as in kinetics of response to agonists. 6 -24 h before the experiment complete medium was exchanged with low serum (0.3%) medium.
siRNA Transfections-For specific knockdown of STIM1 and Orai1, we used siRNA sequences (see "Reagents") that have been established and tested for knockdown efficiency as previously described (11,23). Transfections with siRNA sequences were performed with the Amaxa 4D-Nucleofector (Lonza, Amaxa Biosystems) using transfection solution P5 and the program EH100, following the manufacturer's instructions. 0.5 g of GFP plasmid construct (Amaxa) was co-transfected with siRNA to control transfection efficiency, which typically exceeded 80%. Protein knockdown efficiency was quantified by Western blot.
For impedance measurements, transfected cells were directly seeded onto the gelatin-coated electrodes for thrombin or histamine experiments. For S1P experiments where 2-dayold cell layers were used, the cells were first grown for 2 days on regular culture flasks, before seeding for impedance measurements.
Impedance Measurements-Impedance measurements were performed using the electric cell-substrate impedance sensing (ECIS) Z machine (Applied BioPhysics Inc., Troy, NY) with an array holder station for two 8-well arrays of type 8W10Eϩ (or 2W4 ϫ 1E PC) and an alternative 96-well plate holder for 96W20idf type electrode arrays (26). The measurements were performed in a humidified cell culture incubator unless stated otherwise. Barrier function of HDMEC cell layers was measured at 4000 Hz when using 8W10Eϩ electrodes or at 2000 Hz when using 96W20idf electrode arrays, which were the respective frequencies with maximum dynamic range between the cell-free and cell-covered electrodes (26). Single frequency time collect mode allowed detecting HDMEC response to GPCR agonists with a time resolution of 8 s for 16 wells or 48 s for 96 wells, respectively. For data presentation the resistance, the real part of the impedance data, which is most indicative of paracellular currents via cell-cell junctions was normalized to the last value before agonist addition (normalized resistance) and plotted versus time. For measurements of the complete attachment, growth and cell layer maturation process electrodes were pretreated with 10 mM cysteine solution 1 day before cell seeding. After 10 min of incubation with 150 l of cysteine solution, arrays were rinsed three times with water and incubated in HDMEC medium overnight to support establishment of stable electrode-electrolyte interface. HDMEC attachment, spreading, and cell layer maturation was measured with the multiple frequency time collect mode. A baseline of the cell-free array with 400 l of medium in each well was measured over at least 1 h. Data acquisition was paused for inoculation with single-cell suspension (100,000 cells/cm 2 ) and for medium changes.
Monitoring of HDMEC Response to GPCR Agonists-Thromin (catalog no. T464), histamine (catalog no. 53300) and S1P (catalog no. A8806) were purchased from Sigma-Aldrich. Thrombin stock solutions in MilliQ water (100 M) were stored at Ϫ80°C. Immediately before use in an experiment, 100 nmol/ liter working solution (for a final concentration of 5 nmol/liter, 20 l/well) was prepared in prewarmed 0.3% serum medium (50 ml of EBM-2 150 l of FBS, 500 l of 100 ϫ antibioticsantimycotics, and 500 l of 100 ϫ L-glutamine solution). 1 mol/ liter histamine stock solutions in MilliQ water were stored at Ϫ20°C. Before use, a 200 mol/liter working 20ϫ solution was prepared in serum-free medium (for a final concentration of 10 mol/liter and addition of 20 l/well). The solution was prewarmed to 37°C for ϳ5 min before addition to the cells. S1P was solubilized in methanol, aliquoted, dried under nitrogen, and stored at Ϫ20°C. For use in stimulation experiments, 20 M S1P (20 ϫ) working solution was made in HBSS (Fisher Scientific; catalog no. 14025-092) with 0.03 mg/ml fatty acidfree BSA (from bovine plasma, Sigma-Aldrich; catalog no. A8806). Solubilization was supported by brief sonication for 5 min at 37°C.
For experiments with thrombin, HDMEC cell layers were used after 4 days in culture. The cells were changed to serumfree medium 12-24 h before start of the experiment. Cell layer quality was checked by phase contrast microscopy and by either recording a single full impedance spectrum or collecting data for 10 -30 min at multiple frequency time mode. The resistance at 4000 Hz should be in the range of 2500 -4500 ⍀ for HDMEC grown on 8W10Eϩ electrodes for 4 d. On 96W20idf the resistance at 2000 Hz should be above 1000 ⍀; if these basal resistances were not attained, the experiment was aborted. A baseline was recorded for at least 10 min. Thrombin was added as a 20ϫ stock solution. Although the measurement was running, 20 l of old medium were removed from the initial 400 l of each well. 20 l of 20ϫ thrombin solution (100 nmol/liter) were then added to the remaining 380 l to give a final concentration of 5 nmol/liter. Cell responses was monitored for at least 1 h after addition of thrombin. For experiments with histamine, HDMEC cell layers were used after 3-4 days in culture. The cells were changed to serum-free medium 6 -12 h before start of the experiment. Histamine was added as a 20ϫ stock solution of 200 mol/liter histamine in low serum medium, following the procedure as described for thrombin. For S1P experiments, HDMEC were used after 2 days in culture, where cell layers were confluent, but cell-cell junctions were not fully mature (26). The cells were changed to serum-free medium 6 h before stimulation with 1 M S1P. The experiment was conducted as described above for thrombin and histamine, using a 20ϫ stock solution. For ECIS experiments using preincubation with Gd 3ϩ and corresponding controls, a HEPES-based medium was used (130 mM NaCl, 3 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 5 mM glucose, and 20 mM Na-HEPES, pH 7.4) as previously described (23).
Calcium Imaging-For measuring changes in intracellular calcium, HDMECs were grown on gelatin-coated 30-mm round coverslips as previously described (61)(62)(63). The coverslips with cells were mounted in a Teflon chamber and incubated with 4 M Fura-2/AM in culture medium for 35 min at 37°C protected from light. Afterward, the cells were gently washed three times with HEPES-buffered saline (140 mM NaCl, 1.13 mM MgCl 2 , 4.7 mM KCl, 2 mM CaCl 2 , 10 mM D-glucose, and 10 mM HEPES, pH 7.4) and let sit for 10 min at room temperature and protected from light. Data acquisition was performed using InCyt Im2 Fluorescence Imaging System and IntCyt Im2 software (Intracellular Imaging Inc., Cincinnati, OH). Fura-2loaded cells were alternately illuminated with 340 and 380 nm. Fluorescence emission was collected at 510 nm. 340/380 ratio images were obtained on pixel by pixel basis. Figures showing Ca 2ϩ traces show averages from several cells per coverslip and are representative of several independent recordings of several cells as indicated in the respective figure legends. Thrombin and histamine solutions were made in HEPES-buffered saline, and S1P solutions additionally contained 0.03 mg/ml fatty acidfree BSA.
Immunofluorescence Staining-HDMECs were grown on gelatin-coated cover glasses for 4 days (initial seeding density 100,000 cells/cm 2 ). The cell layers were stimulated with different agonists under the same conditions as used for ECIS experiments. At the indicated time after stimulation, cells were fixed with 4% (w/v) paraformaldehyde in PBS for 20 min. Fixed cells were rinsed with PBS three times and permeabilized with 0.02% Triton X-100 in PBS. Permeabilized cells were first treated with Image-iT FX signal enhancer (Fisher Scientific; catalog no. I36933) for 30 min and afterward blocked with a mixture of 3% BSA and 5% serum in TBS for 1 h. The cell layers were rinsed two times before incubated with goat anti-VE-cadherin primary antibody (1:200) in TBS for 2 h at room temperature. Secondary antibody (Alexa Fluor 594-labeled rabbit anti-goat antibody from Molecular Probes) incubation was for 1 h at room temperature together with Alexa Fluor 488 phalloidin and Hoechst 33342 incubation. The cells were rinsed three times and embedded with anti-fade mounting medium (Sigma-Aldrich). Images were taken on a Zeiss confocal microscope.
Western Blotting-The cells were transfected with respective siRNAs as described above. After 4 days the cells were lysed in radioimmune precipitation assay lysis buffer (Sigma-Aldrich) with 10% protease inhibitor mixture and 10% phosphatase inhibitor mixture (both from Roche Diagnostics). Suspensions were subjected to a 30 s sonication on ice. Cell debris was removed by 1.5 min centrifugation at 14,000 rpm and 4°C. Protein concentration in the supernatant was determined using BCA assay (Pierce) and SpectraMax Paradigm multimode detection platform (Molecular Devices), and equal amounts of proteins were loaded in each well.
Lysates (50 g/well) were subjected to SDS-PAGE using the Precision Plus TM protein dual color standard (Bio-Rad) as reference (with 10,15,20,25,37,50,75,100,150, and 250 kDa) and transferred to immunoblot PVDF membranes (Bio-Rad). The membranes were blocked (5% nonfat milk/TBS/Tween-20) at 4°C overnight, incubated with respective antibodies, washed, and subsequently incubated with HRP-linked secondary antibody in 2% nonfat milk/TBS/Tween-20 for 2 h. Bound antibody was detected by enhanced chemiluminescence with Super Signal West Pico or Femto reagents (Pierce). Signal intensity was measured using a Bio-Rad ChemiDoc MP imaging system (Bio-Rad). The membranes were then stripped and reprobed with anti-␤-actin, anti-GAPDH, or anti-HSP 70/HSC 70 antibody to verify equal protein loading. Quantitative densitometry analysis was performed using ImageJ software.
Statistical Analysis-Experimental data are presented as means Ϯ S.D. Only for data sets with very high N numbers were means Ϯ S.E. used, and the values are indicated as such in the figure legends. To test for statistical significance, data sets were subjected to ANOVA analysis. Levels of significance were indicated with asterisks as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.005. A p value of Ͻ0.05 was considered statistically significant. ANOVA analysis and Pearson's correlation calculations were performed with OriginLab Pro software. Density plots were generated with R software using Kernel density smoothing for histograms. All data were plotted using Origin Pro software.