Myosin di-phosphorylation and peripheral actin bundle formation as initial events during endothelial barrier disruption

The phosphorylation of the 20-kD myosin light chain (MLC) and actin filament formation play a key role in endothelial barrier disruption. MLC is either mono- or di-phosphorylated (pMLC and ppMLC) at T18 or S19. The present study investigated whether there are any distinct roles of pMLC and ppMLC in barrier disruption induced by thrombin. Thrombin induced a modest bi-phasic increase in pMLC and a robust mono-phasic increase in ppMLC. pMLC localized in the perinuclear cytoplasm during the initial phase, while ppMLC localized in the cell periphery, where actin bundles were formed. Later, the actin bundles were rearranged into stress fibers, where pMLC co-localized. Rho-kinase inhibitors inhibited thrombin-induced barrier disruption and peripheral localization of ppMLC and actin bundles. The double, but not single, mutation of phosphorylation sites abolished the formation of peripheral actin bundles and the barrier disruption, indicating that mono-phosphorylation of MLC at either T18 or S19 is functionally sufficient for barrier disruption. Namely, the peripheral localization, but not the degree of phosphorylation, is suggested to be essential for the functional effect of ppMLC. These results suggest that MLC phosphorylation and actin bundle formation in cell periphery are initial events during barrier disruption.

dothelial electrical resistance (TEER), which showed a maximum decrease at 3-5 min after thrombin stimulation (Fig. 1a). The barrier disrupting effect of thrombin was observed at concentrations higher than 0.01 u ml −1 , and the maximum effect was obtained with 1 u ml −1 thrombin (Fig. 1b). PAR 1 and PAR 4 are the signaling receptors for thrombin 24,25 . PAR 1 -activating peptide (PAR 1 -AP; 30 μ M) induced a decrease in the TEER to a level comparable to that seen following treatment with 1 u ml −1 thrombin, while PAR 4 -AP had no significant effect (Fig. 1c). The thrombin-induced decrease in the TEER was abolished by treatment with 1 μ M of a PAR 1 antagonist, E5555 (Fig. 1c). These observations indicated that thrombin induces barrier disruption mainly through PAR 1 in PAECs.

Thrombin-induced MLC phosphorylation in PAECs. The MLC phosphorylation was analyzed with
Phos-tag SDS-PAGE followed by immunoblotting with an anti-pan MLC antibody. The anti-pan MLC antibody detected three bands in PAECs both before and after thrombin stimulation (Fig. 2a, Phos-tag, IB: MLC). The lower, middle and upper bands ran as the size of 20 kD, 23 kD and 29 kD, respectively, according to the mobility of the molecular weight markers (New England BioLabs, Ipswich, MA, USA). The density of the uppermost band apparently increased, while the density of the lowest band decreased after thrombin stimulation. The middle band was also recognized by the anti-pMLC S19 antibody, while the upper band was recognized by the anti-pMLC T18+S19 antibody (Fig. 2a, Phos-tag). Accordingly, the lower, middle and upper bands represent non-, mono-and di-phosphorylated MLC, respectively. Furthermore, these observations corroborated the mono-specificity of the anti-pMLC S19 and anti-ppMLC T18+S19 antibodies, which was essential for the immunofluorescence staining in the subsequent investigations.
In conventional SDS-PAGE, MLC ran as a single band (Fig. 2a, conventional). The amounts of total MLC were similar before and after thrombin stimulation, while the levels of pMLC and ppMLC apparently increased after thrombin stimulation (Fig. 2a, conventional, IB: pMLC S19 , pMLC T18+S19 ). When evaluated 3 min after thrombin stimulation, there was a concentration-dependent increase in ppMLC, and the maximum effect was obtained around 0.3 u ml −1 thrombin (Fig. 2b). ppMLC was scarcely detected before thrombin stimulation (Fig. 2c). However, upon stimulation with 1 u ml −1 thrombin, the level of ppMLC substantially increased to a peak around 3 min after stimulation, and gradually decreased to the prestimulation level thereafter (Fig. 2c). The PAECs contained 25% pMLC before stimulation. The level of pMLC slightly increased after thrombin stimulation, and exhibited a bi-phasic response, with a dip at 10 min (Fig. 2c). However, the changes in pMLC were modest. Similar changes in pMLC and ppMLC were observed with conventional SDS-PAGE followed by three separate detections with the anti-pan MLC, pMLC S19 , and pMLC T18+S19 antibodies (Fig. 2d). PAR 1 -AP also induced changes in pMLC and ppMLC similar to those seen with thrombin, while PAR 4 -AP had no significant effect (Fig. 2e,f). phosphorylation were examined (Fig. 3). Rho-kinase inhibitors, Y27632 and H1152, were applied 30 min prior to and during stimulation with 1 u ml −1 thrombin. The resting level of ppMLC slightly decreased during the pretreatment with Y27632 (Fig. 3b, time 0); otherwise, no significant effects were observed for the resting levels of pMLC or ppMLC. However, both Y27632 and H1152 concentration-dependently inhibited the increase in ppMLC observed 3 min after thrombin stimulation (Fig. 3a). Almost complete inhibition of ppMLC was obtained with 10 μ M Y27632 and 1 μ M H1152. However, these inhibitors had no significant effect on the levels of pMLC. During the time course after thrombin stimulation, Y27632 (10 μ M) and H1152 (1 μ M) almost completely abolished the increase in ppMLC (Fig. 3b, ppMLC). The increase in pMLC seen at 3 min appeared to be inhibited by H1152 and Y27632; however, their inhibitory effects were not statistically significant. In contrast, Y27632 and H1152 consistently and significantly decreased the level of pMLC seen in the late phase, i.e., 15-30 min after thrombin stimulation (Fig. 3b, pMLC). The pretreatment with Y27632 and H1152 had no significant effect on the resting TEER; however, they almost completely abolished the thrombin-induced decrease in the TEER (Fig. 3b, TEER).

Effects of Rho-kinase inhibitors and the removal of extracellular Ca
The removal of extracellular Ca 2+ had no significant effect on the resting levels of both pMLC and ppMLC (Fig. 3c, time 0). The early phase (3 min) of pMLC was resistant to the Ca 2+ removal, while the late phase (>10 min) significantly decreased in the absence of extracellular Ca 2+ (Fig. 3c, pMLC). However, the thrombin-induced increase in ppMLC was resistant to Ca 2+ removal (Fig. 3c, ppMLC). When the extracellular Ca 2+ was removed, the resting TEER gradually decreased during the 30-min incubation (Fig. 3c, TEER). The subsequent thrombin stimulation induced a decrease in the TEER similar to that observed in the presence of extracellular Ca 2+ (Fig. 3c, TEER). As a result, the level of ppMLC, but not pMLC, was correlated with the degree of barrier disruption induced by thrombin. pMLC and ppMLC exhibited different sensitivities toward Rho-kinase inhibition and the removal of extracellular Ca 2+ .
Effects of a myosin ATPase inhibitor, blebbistatin, on thrombin-induced and barrier disruption in PAECs. The effect of a myosin ATPase inhibitor, blebbistatin, on the thrombin-induced decrease in the TEER was examined to address the involvement of myosin-based contraction in the barrier disruption 28 . Pretreatment with 100 μ M blebbistatin 30 min prior to and during thrombin stimulation substantially inhibited The concentration-dependent effects of thrombin on the TEER at 3 min after stimulation (n = 3-7). (c) A summary of the temporal changes in the TEER induced by 1 u ml −1 thrombin without (Thrombin; n = 12) and with 1 μ M E5555 (Thrombin + E5555; n = 4), 30 μ M PAR 1 -AP (n = 4) and 30 μ M PAR 4 -AP (n = 6). The data are the means ± SEM. *P < 0.05 vs. thrombin, according to an ANOVA followed by Dunnett's post-hoc test.
the thrombin-induced decrease in the TEER (see supplementary Fig. S1). However, blebbistatin had no effect on the levels of pMLC and ppMLC either before or 3 min after thrombin stimulation (see supplementary Fig. S1).

Distinct subcellular localization of pMLC and ppMLC after thrombin stimulation in confluent
PAECs. The mono-and di-phosphorylation of MLC were originally considered to be sequential events in smooth muscle cells 6,12 . However, the observations of the present study suggest that there are distinct regulatory pathway and functional roles of pMLC and ppMLC during thrombin-induced barrier disruption in PAECs. To confirm this finding, any distinct subcellular localization of pMLC and ppMLC and its correlation with actin filament formation was examined (Fig. 4). Before thrombin stimulation, pMLC was observed homogeneously The temporal changes in pMLC and ppMLC (n = 4) and the TEER (n = 6 for control; n = 5 for Y27632; n = 4 for H1152) after stimulation with 1 u ml −1 thrombin in the absence and presence of 10 μ M Y27632 or 1 μ M H1152. (c) The temporal changes in pMLC and ppMLC (n = 4) and the TEER (n = 5) after stimulation with 1 u ml −1 thrombin in the presence and absence of extracellular Ca 2+ . PAECs were exposed to Y27632, H1152 or Ca 2+ -free HBS 30 min prior to and during thrombin stimulation. The data are the means ± SEM. *P < 0.05 vs. control, according to an ANOVA followed by Dunnett's post-hoc test.
in the perinuclear cytoplasmic region, while ppMLC was scarcely detected (Fig. 4a, left column). Three minutes after thrombin stimulation, intense staining of ppMLC was characteristically detected in the peripheral region of the cells (Fig. 4a, middle column). The intensity of pMLC appeared to increase; however, its localization remained unchanged (Fig. 4a, middle column). As a result, the localization of pMLC and ppMLC after thrombin stimulation was distinct, represented by central pMLC staining, surrounded by peripheral ppMLC staining (Fig. 4a, the middle column, top panel). The distinct localization of pMLC and ppMLC was also supported by double staining with VE-cadherin (Fig. 4b, left and middle columns). ppMLC was closely associated with VE-cadherin staining, while some separation was observed between pMLC and VE-cadherin. The double staining of ppMLC and actin revealed their colocalization in a peripheral bundle pattern (Fig. 4a, right column). Notably, actin did not show up in a stress fiber pattern 3 min after thrombin stimulation. The pretreatment with 10 μ M Y27632 abolished the peripheral localization of ppMLC 3 min after thrombin stimulation, while the central cytoplasmic localization of pMLC was not affected by Y27632 treatment (Fig. 4b, right column).
Thrombin induces stress fiber formation in PAECs with loose cell-cell adhesion. The observations in Fig. 4 were apparently inconsistent with those in previous reports, which showed stress fiber formation after thrombin stimulation [1][2][3][4] . The difference in the state of cell-cell adhesion could be one possible explanation for this discrepancy. Therefore, the localization of pMLC, ppMLC and actin filaments was examined in PAECs with loose cell-cell adhesion (Fig. 5). Such cells were prepared by removing the extracellular Ca 2+ in the confluent culture or on an early culture day (day 3) after plating.
pMLC and ppMLC were scarcely detected after a 30-min incubation in the absence of extracellular Ca 2+ and before thrombin stimulation (Fig. 5a, left column). Upon thrombin stimulation, ppMLC showed a stress fiber pattern in 3 min, which was co-localized with pMLC ( Fig. 5a, middle column) as well as actin filaments (Fig. 5a, right column). On culture day 3, pMLC and ppMLC exhibited sparsely filamentous co-localization in a stress fiber pattern before thrombin stimulation in the area of a low cell density (Fig. 5b, left column), where the lining of VE-cadherin was less developed (Fig. 5b, right column). Upon thrombin stimulation, intense co-staining of pMLC and ppMLC in a stress fiber pattern was observed 3 min after stimulation (Fig. 5b, middle column). The actin filaments also showed a stress fiber pattern (Fig. 5b, right column). A similar stress fiber pattern of pMLC, ppMLC and actin was observed in the area of a relatively high cell density, where the lining of VE-cadherin was well developed (Fig. 5c). Collectively, the degree and maturity of cell-cell adhesion appeared to determine which pattern of localization of ppMLC and actin was induced (stress fiber or peripheral bundle).

Effects of mutations of the phosphorylation sites in MLC on thrombin-induced barrier disruption.
In order to obtain further support for the distinct roles for pMLC and ppMLC in the thrombin-induced barrier disruption, the functional effects of mutations of the specific phosphorylation sites in MLC were examined. The human MYL12B isoform of MLC and its mutants were expressed in PAECs using adenoviral expression vectors. The exogenously expressed MLC mobilized faster than the endogenous MLC in conventional SDS-PAGE ( Fig. 6a,b). Approximately 80% of the MLC in PAECs was estimated to be derived from the exogenous MLC for all constructs (Fig. 6a). The expression of endogenous MLC was suppressed in the cells expressing exogenous MLC (Fig. 6a,b).
The intended mutations were validated by examining their phosphorylation in response to thrombin stimulation (Fig. 6b). The anti-pan MLC antibody detected doublets in PAECs expressing exogenous MLC (Fig. 6b, uppermost immunoblot). The immunoreactivity of the lower band of wild type MLC to both the anti-pMLC S19 and anti-pMLC T18+S19 antibodies increased 3 min after thrombin stimulation (Fig. 6b, conventional). The double mutation of T18 and S19 abolished the immunoreactivity of the lower band to the phospho-specific antibodies. For the S19A mutant, the immunoreactivity of the lower band for the anti-pMLC S19 antibody was abolished, while immunoreactivity toward the anti-pMLC T18+A19 antibody was observed. However, this band exhibited a mobility similar to that for pMLC in Phos-tag SDS-PAGE (Fig. 6b, Phos-tag), indicating that the S19A mutant detected by the anti-pMLC T18+A19 antibody was mono-phosphorylated at T18. For the T18A mutant, immunoreactivity of the lower band with the anti-pMLC S19 antibody was observed, while the immunoreactivity with the anti-pMLC T18+S19 antibody was abolished.
In PAECs expressing wild type MLC, thrombin induced a decrease in the TEER similar to that seen with LacZ-transfected PAECs (Fig. 6c). When both phosphorylation sites (T18 and S19) were mutated to alanine, the thrombin-induced decrease in the TEER was substantially inhibited (Fig. 6c). The T18A mutant had no significant effect on the thrombin-induced decrease in the TEER (Fig. 6c). The S19A mutant exhibited modest inhibitory effects on thrombin-induced barrier disruption, but its effects were not statistically significant (Fig. 6c).

Localization of MLC mutants in PAECs.
The subcellular localization of pMLC (i.e., detection with the anti-pMLC S19 antibody), ppMLC (i.e., detection with the anti-pMLC T18+S19 antibody) and actin in PAECs expressing wild type and mutant MLC was examined before and 3 min after thrombin stimulation (Fig. 7). PAECs expressing wild type MLC exhibited pMLC, ppMLC and actin localizations similar to those seen with control non-infected cells (shown in Fig. 4), both before and 3 min after thrombin stimulation. Namely, ppMLC and actin filaments exhibited a peripheral bundle pattern after thrombin stimulation, while pMLC exhibited a perinuclear cytoplasmic localization (Fig. 7). The T18A+ S19A mutant abolished the peripheral localization of ppMLC and actin filaments after thrombin stimulation (Fig. 7a,c). The residual staining with the anti-pMLC T18+S19 antibody was conceivably derived from residual endogenous MLC (Fig. 7a). The immunoreactivity with the anti-pMLC S19 antibody was also substantially decreased compared to that seen with wild type MLC (Fig. 7b).
In PAECs expressing the S19A or T18A mutants, the immunoreactivity with the anti-pMLC T18+S19 antibody was scarcely detected before and after thrombin stimulation (Fig. 7a). Regarding the S19A mutant, this observation was apparently inconsistent with that seen with immunoblots using this antibody, which demonstrated the  mono-phosphorylation at T18 in the S19A mutant (Fig. 6b). The reason for this apparent discrepancy remains unclear. The reactivity of the anti-pMLC T18+S19 antibody might differ between immunoblotting and immunofluorescence staining. In contrast, the anti-pMLC S19 antibody homogeneously bound to the cytoplasm of PAECs expressing the T18A mutant after thrombin stimulation, while this antibody exhibited only residual staining in the cells expressing the S19A mutant (Fig. 7b). These observations seen with the anti-pMLC S19 antibody are consistent with those seen in the immunoblot analysis (Fig. 6b). Most important for the functional relevance, the thrombin-induced peripheral actin bundle formation was abolished only by the double mutation. PAECs expressing wild type MLC and the T18A and S19A mutants all exhibited peripheral actin bundles after thrombin stimulation (Fig. 7c). The effects of the MLC mutations on the peripheral actin bundle formation were therefore well correlated with those on the barrier disruption (Fig. 6c).  or phosphorylation defective mutants (T18A + S19A, S19A, T18A) of MLC, before (Control) and 3 min after stimulation with 1 u ml −1 thrombin (Thrombin). pMLC and ppMLC indicate the fluorescence images obtained with the anti-pMLC S19 antibody and anti-pMLC T18+S19 antibody, respectively. n = 5 (a,b), n = 2 (c).
Scientific RepoRts | 6:20989 | DOI: 10.1038/srep20989 Actin stress fiber formation in the late phase after thrombin stimulation. Loose or immature cell-cell adhesion predisposes actin filaments to form stress fibers in response to thrombin (Fig. 5). Once peripheral actin bundles are formed in the early phase, the tight cell-cell adhesion might become loosened due to concentric contraction 29 . In this case, actin filaments might be rearranged from peripheral bundles to stress fibers in the later phase. In fact, the time course analysis of actin filament organization revealed that such reorganization was found in the late phase after thrombin stimulation (Fig. 8a). The peripheral actin bundles were observed as early as 1 min after stimulation and up to 5 min after. Thereafter, the actin filaments were rearranged as stress fibers (Fig. 8a). At 20 min after thrombin stimulation, pMLC was localized in a stress-fiber pattern (Fig. 8a). When Rho-kinase inhibitors were applied 10 min after thrombin stimulation, the level of pMLC at 20 min after thrombin stimulation was significantly suppressed Fig. 8b).

Discussion
The main finding of the present study is that MLC di-phosphorylation and actin bundle formation at the cell periphery play a critical role as initial events during the thrombin-induced barrier disruption. This finding is novel in two respects. First, the present study demonstrated a specific role of ppMLC in barrier disruption, which is distinct from that of pMLC. The temporal profile, subcellular localization and sensitivity to Rho-kinase inhibitors and removal of extracellular Ca 2+ were different between pMLC and ppMLC. The decrease in the TEER during the early phase of the thrombin-induced barrier disruption (<10 min) exhibited sensitivity to Rho-kinase inhibitors and resistance to extracellular Ca 2+ removal similar to those seen with ppMLC, but not pMLC. These observations suggest the specific involvement of ppMLC in the early phase of thrombin-induced barrier disruption.
Second, the present study proposes that the peripheral actin bundle formation is an initial event during barrier disruption. Stress fiber formation was observed in the later phase (>10 min). This is a new finding in terms of the previous understanding of the role of stress fiber formation in endothelial barrier disruption 1,2,5 . Notably, the peripheral actin bundles were temporally and spatially associated with ppMLC. These results suggest that the thrombin-induced barrier disruption starts from ppMLC at the cell periphery, which in turn causes actin bundle formation. This event has been overlooked in the earlier views of the mechanism underlying endothelial barrier disruption.
Distinct functional roles of pMLC and ppMLC have previously been shown for the regulation of myosin ATPase activity 11,12,21 , filament formation of myosin or actin 18,21-23 , cytokinesis 20 , cellular stiffness 17,19 , and cell migration 16 , mainly by using either unphosphorylatable (A) or phosphomimetic (D) mutants of MLC at T18 and S19. These differences were sometimes quantitative, but other times were qualitative. However, many reports have shown a quantitative difference in such aspects as the regulation of myosin ATPase activity 11,12,21,23 , cellular stiffness 17,19 , filament formation of actin 22 and stabilization of myosin filaments 16,18 . In these cases, ppMLC exerts a similar but greater effect than pMLC. The present study demonstrated the qualitative differences between pMLC and ppMLC. The most striking difference is their subcellular localization. This is in contrast to the findings of many other reports showing the co-localization of pMLC and ppMLC in stress fibers or the contractile ring of cytokinesis 20,22,30 .
Only a few reports have shown a differential localization of pMLC and ppMLC, and these were in situations unlike those in the present study 16,18 . For example, in CHO cells, the adhesion to fibronectin-coated substratum increased the levels of pMLC and ppMLC 16 . Under such situations, pMLC localized in the area of the lamellipodia, while ppMLC localized in the rear or side area. The preferential localization of ppMLC at the region of the cell-cell junction was shown in Madin Darby canine kidney (MDCK) II epithelial cells 18 . However, this peripheral localization of ppMLC appears to be different from that seen in the present study, but it is similar to that of circumferential actin filaments, which are associated with mature cell-cell adhesion 5 . Furthermore, pMLC and ppMLC were shown to localize in different regions of stress fibers in MDCK II cells; pMLC localized in the region where stress fibers were stretching, while ppMLC localized in the region where stress fibers were contracting 18 . The present study thus demonstrate that the thrombin-induced barrier disruption is a new situation where pMLC and ppMLC exhibit differential localization.
In addition to the subcellular localization, pMLC and ppMLC exhibited different kinetics and sensitivities toward kinase inhibitors. These observations suggest that pMLC is not a prerequisite for ppMLC, and that they are differently regulated. It is conceivable that the distinct localization of pMLC and ppMLC is attributable to the distinct localization of the responsible kinases. The Ca 2+ -calmodulin-dependent MLC kinases, MLCK, phosphorylates MLC with a preference for S19 to T18 6,10 . Rho-kinase, ZIP kinase and integrin-linked kinases phosphorylate MLC with no preference between T18 and S19 [31][32][33][34][35] . In the present study, the level of ppMLC was almost completely abolished by Rho-kinase inhibitors, while pMLC was resistant in the early phase and sensitive in the later phase following thrombin stimulation. The Rho-kinase-dependent ppMLC at the cell periphery appears to be consistent with the fact that RhoA is activated at the plasma membranes 15 . Rho-kinase contributes to MLC phosphorylation either by directly phosphorylating MLC 31,32 or by inhibiting MLC phosphatase activity via the phosphorylation of MYPT1, a regulatory subunit of MLC phosphatase, or CPI-17, an inhibitory protein of MLC phosphatase [13][14][15] . However, precisely how pMLC and ppMLC are regulated during the time course of barrier disruption and the identity of the responsible kinases remain to be determined.
The observations with the mutants of MLC suggest that mono-phosphorylation is functionally sufficient for peripheral actin bundle formation and barrier disruption. The double mutation of T18 and S19 to A was required for the inhibition of both events. This observation gives further support for the correlation between peripheral actin bundle formation and barrier disruption. It also suggests that the localization at the cell periphery, but not the degree of phosphorylation, is essential for the role of ppMLC in barrier disruption. The effects of pMLC and ppMLC on the myosin ATPase activity are quantitatively different 11 formation of myosin filaments and the sliding velocity of actin filaments are quantitatively similar to those of ppMLC 21,23 . These effects of pMLC are consistent with the observations with mutant MLC in the present study.
In the present study, actin stress fiber formation was observed in the late phase (>10 min) during thrombin-induced barrier disruption. However, in the cells examined on early culture days or in the absence of The effects of Rho-kinase inhibitors, 10 μ M Y27632 and 1 μ M H1152, on the level of pMLC at 20 min after thrombin stimulation. The inhibitors were applied 10 min after thrombin stimulation, and the levels of pMLC were evaluated 10 min later, i.e., 20 min after thrombin stimulation, with Phos-tag SDS-PAGE. The data are the means ± SEM (n = 5). *P < 0.05, according to an ANOVA followed by the Turkey-Kramer post-hoc test. (c) A schematic presentation of the proposed sequential events that occur during endothelial barrier disruption. extracellular Ca 2+ , actin stress fiber formation was readily observed as early as 3 min after thrombin stimulation. The degree of maturity of the cell-cell adhesion appears to determine which pattern of actin filaments is induced as an initial event. This is consistent with the notion that the arrangement of actin filaments is subject to the state of VE-cadherin-mediated cell adhesion, the degree of cell density and the nature of the extracellular matrix in cell culture 5 . Subconfluent cultures with an interrupted VE-cadherin lining exhibit many stress fibers 5 . With increasing cell density, the stress fibers disappear and the circumferential actin bundles develop in association with the development of the VE-cadherin lining. Highly confluent endothelial cultures with continuous VE-cadherin lining are characterized by circumferential actin bundles with fewer stress fibers 5 . The notion that the stress fiber formation is a prerequisite for barrier disruption is based on the observations obtained mainly with human umbilical vein endothelial cells or other human endothelial cells 1,2,5 . Although the inter-endothelial lining of VE-cadherin can be observed at confluence in these cells, it is possible that the maturity of the cell-cell contact remains poor in human endothelial cells. Therefore, stress fiber formation might be preferentially observed during barrier disruption in these cells. In PAECs, the developmental state of the cell-cell contact was heterogeneous on the early culture days. In the areas of a low cell density, the inter-endothelial VE-cadherin lining was poorly developed (Fig. 5b,c). In the areas of a relatively high cell density, the inter-endothelial VE-cadherin lining was well developed. However, thrombin induced stress fiber formation in both cases (Fig. 5b,c). Therefore, the continuous inter-endothelial lining of VE-cadherin does not necessarily indicate the maturity of the cell-cell adhesion. In the cells with the mature cell-cell contacts, thrombin induced the formation of peripheral actin bundles, thereby generating concentric force 29 , which then loosened the inter-endothelial cell-cell adhesion. Once the cell-cell adhesion was loosened, the peripheral actin bundles could be rearranged into stress fibers. This scenario is consistent with the observations of the present study (Fig. 8c).
In conclusion, the present study proposes that ppMLC and peripheral actin bundle formation play specific roles as initial events that occur during the thrombin-induced barrier disruption. The peripheral localization but not the degree of phosphorylation is suggested to be essential for the function of ppMLC as an initial event. This initial event loosens the inter-endothelial junctions, causing not only an increase in paracellular permeability, but also predisposing actin filaments to rearrange into stress fibers during the sustained phase of barrier dysfunction (Fig. 8c). Thus, the uncovered initial event, i.e., peripheral ppMLC and actin bundle formation, could be a missing event that connects the transition of the circumferential actin filaments associated with mature junctions to the actin stress fibers formed during endothelial barrier disruption.
Measurement of the trans-endothelial electrical resistance. The endothelial barrier function was evaluated by assessing the trans-endothelial electrical resistance (TEER). PAECs were plated at 1.5 × 10 4 cells in 0.32-cm 2 culture inserts with polyethylene terephthalate membranes (0.4-μ m pore size, 1.6 × 10 8 pores cm −2 ) in 24-well culture plates (BD Falcon, Tokyo, Japan). TEER measurements were conducted in Hepes-buffered saline (HBS: 10 mM Hepes at pH 7.4, 135 mM NaCl, 5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 5.5 mM D-glucose) at 37 °C with an EVOM 2 voltohmmeter equipped with a STX2 electrode (World Precision Instruments, Sarasota, FL USA). The cells were equilibrated in HBS (1 ml in the lower chamber and 0.4 ml in the upper chamber) for 10 min, and then were subjected to the experimental protocol. All reagents were applied to the upper chamber by changing the whole solution. The change in the TEER from the value obtained in HBS just prior to agonist stimulation (Δ TEER) was expressed as an absolute value of the resistance normalized by the surface area of the culture insert membrane (Ω cm −2 ), unless otherwise specified. There were no significant changes in the TEER during the 10-min equilibration period in HBS.
Preparation of cell lysates for the analysis of myosin light chain phosphorylation. The cells were plated at 3 × 10 5 cells in 60-mm culture dishes. On the experimental day, the cells were rinsed three times with pre-warmed phosphate-buffered saline (PBS; 136.9 mM NaCl, 2.7 mM KCl, 8.1 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 ), and then equilibrated for 10 min and subjected to the experimental protocols in HBS at 37 °C. At each time point, the cells were rinsed once with ice-cold PBS containing 0.1 mM EDTA to remove residual extracellular Ca 2+ . After the wash buffer was thoroughly removed, the cells on dishes were lysed in 70 μ L ice-cold lysis buffer containing 50 mM Hepes, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM dithiothreitol, 10 μ g ml −1 leupeptin, 10 μ g ml −1 aprotinin, 10 μ M 4-amidinophenylmethanesulfonyl fluoride (p-APMSF), 20 μ M NaF, 1 mM Na 3 VO 4 , 50 μ M calpain inhibitor I and 5 μ M microcystin-LR. The lysates were immediately collected using a cell scraper, transferred to a microcentrifuge tube and snap-frozen in liquid N 2 . The lysates were kept at − 80 °C until use. The lysates were then thawed on ice for 20 min, and centrifuged at 12,000 rpm for 20 min on microcentrifuge (Type 1120; Kubota, Tokyo, Japan) in a cold room. The protein concentrations of supernatants were estimated with a Coomassie (Bradford) protein assay kit (Thermo Scientific Pierce, Rockford, IL, USA). The lysates were prepared for SDS-PAGE by adding an appropriate volume of 6x Laemmli's sample buffer and β -mercaptoethanol to a final concentration of 5%.

Phos-tag and conventional SDS-PAGE, and the immunoblot analysis of MLC phosphorylation.
The samples (3.5 μ g proteins) were subjected to SDS-PAGE on 10% polyacrylamide gels containing 30 μ M Phos-tag and 60 μ M MnCl 2 in Phos-tag SDS-PAGE 26,42 . After electrophoresis, the gels were soaked for 30 min at room temperature in transfer buffer (25 mM Tris-hydroxymethyl aminomethane, 192 mM glycine, 20% methanol) containing 2 mM EDTA to remove Mn 2+ , and then in the transfer buffer for 15 min. In conventional SDS-PAGE, the proteins were separated on 12.5% polyacrylamide gels, with no Phos-tag or MnCl 2 , in triplicate for the detection of total MLC, pMLC and ppMLC. The Mn 2+ removal step was omitted in conventional SDS-PAGE. Proteins were then transferred to polyvinylidene difluoride membranes (Merck Millipore). The membranes were post-fixed in PBS containing 0.5% formaldehyde for 45 min 42 , then rinsed once with PBS containing 0.1% Tween-20 (T-PBS), and blocked with 5% non-fat dry milk in T-PBS overnight in a cold room. The membranes were incubated with primary and then secondary antibodies diluted in Can-Get-Signal immunoreaction enhancer solution (Toyobo, Osaka, Japan) for 60-90 min at room temperature. The primary antibodies and their dilutions were: anti-pan MLC antibody (× 2,000), anti-pMLC S19 antibody (× 2,000 -× 5,000) and anti-pMLC T18+S19 antibody (× 2,000). The immune complexes were detected with an ECL Plus or ECL Select detection kit (GE Healthcare, Buckinghamshire, UK).
The chemiluminescence emission was captured with a ChemiDoc XRS-J instrument and was analyzed using the Quantity One software program (Bio Rad, Hercules, CA, USA). The image capture time was adjusted to keep the intensity of image below the maximal saturation level of the system. In the Phos-tag SDS-PAGE analysis, the percentages of pMLC and ppMLC in the total MLC (a sum of the optical densities of non-phosphorylated MLC, pMLC and ppMLC), all of which were detected with the anti-pan MLC antibody, were calculated for each lane. In the conventional SDS-PAGE analysis, the density of the pMLC and ppMLC bands was normalized to the density of the corresponding pan-MLC band, then the levels of pMLC and ppMLC were expressed as a fold-increase from the prestimulation level.
Immunofluorescence staining. The cells were plated at 1.5 × 10 5 cells on glass coverslips (22 × 22 mm, Matsunami, Osaka, Japan) in 35-mm culture dishes. For all experiments, the cells were rinsed three times with pre-warmed PBS, and then equilibrated for 10 min and subjected to the experimental protocols in HBS at 37 °C. At each time point, the cells were fixed with 4% formaldehyde in PBS for 20 min, and permeabilized by a 5-min incubation in 0.05% Triton X100-containing PBS. After 60-min of blocking with 5% bovine serum albumin in PBS, the cells were incubated with the indicated combinations of primary antibodies diluted in Can Get Signal Immunostain (Toyobo) for 60 min at room temperature and then with the appropriate combinations of fluorescent dye-conjugated secondary antibodies for 60 min at room temperature. For staining of the actin filaments, the specimens were incubated with either 5 units ml −1 FITC-or TEXAS-red phalloidin in PBS for 20 min after the immunostaining procedures. The cells were then mounted in ProLong Gold antifade reagent with DAPI (Life Technologies-Molecular Probe). The confocal fluorescence images were captured with a Nikon confocal laser microscope A1 with a 60x objective lens (Tokyo, Japan). The original images were saved in the JPEG2000 format using an imaging software program, NIS-elements AR (Nikon). The images used for presentation were exported from the original files and saved in the TIFF format using a Snapshot function of the software. The intensity of the red, green or blue channels was set to 0 using the Photoshop software (Adobe, San Jose, CA, USA) to show the remaining two colors in TIFF files.

Construction of adenoviral vectors. The reverse transcription-polymerase chain reaction (RT-PCR)
analysis detected the expression of MYL12B, but not MYL12A or MYL9, as phosphorylatable MLC in PAECs. In brief, total RNA was prepared by using an RNeasy Mini Kit (Qiagen, Hilden, Germany), and the RT reaction was performed with a ReverTra Plus kit and random primers (Toyobo). The PCR reaction was performed with a DNA polymerase, KOD plus ver. 2 (Toyobo). The following primers based on the sequences of human MLC were used for the PCR reaction: AgC CAA gAT gTC CAg CAA gC (forward primer for MYL9), TCA ggg ggC Tgg ggT ggC CT (reverse primer for MYL9), CAC CAT gTC gAg CAA AAg AA (forward primer for MYL12A), Tgg AAT TTg AAg TTA TTT CA (reverse primer for MYL12A), CAC CAT gTC gAg CAA AAA gg (forward primer for MYL12B) and TTT TAg CTA AAg TTC TTT CA (reverse primer for MYL12B). The double underlined ATG sequences indicate initiation codons. The reverse primers were designed to include a stop codon or a 3′ region from the stop codon. Therefore, the PCR products covered the full coding region of each isoform. HeLa cells were used as a positive control for the RT-PCR analysis. The expression of all three isoforms were detected in HeLa cells, and the sequences of PCR products were confirmed. Since this preliminary study determined that MYL12B was a major isoform of phosphorylatable MLC expressed in PAECs, this isoform was used in the subsequent investigations.