pS1Pless mice also showed increased responses to leak-inducing challenges. Hindpaw thickness increased 41% in pS1Pless, compared with 25% in control mice after local injection of histamine (Figure ​(Figure1).1). Increase after local serotonin injection was 58% in pS1Pless versus 44% in controls (Figure ​(Figure1).1). Basal leak and response to VEGF in a Miles assay were also increased (Supplemental Figure 1; supplemental material available online with this article; doi: 10.1172/JCI38575DS1).

In a model of passive systemic anaphylaxis (PSA), all control mice survived antigen challenge, but 44% of pS1Pless mice died (Figure ​(Figure2A).2A). Within 90 seconds of antigen administration, hematocrit increased 10% over baseline in control mice and 23% in pS1Pless mice (Figure ​(Figure2B).2B). Lung wet weight and wet/dry weight ratio also increased more in pS1Pless mice than in controls (Figure ​(Figure3A),3A), and histological analysis of lungs showed increased peribronchial fluid (Figure ​(Figure3C).3C). These findings are all consistent with increased extravasation of plasma in pS1Pless mice.

The increased sensitivity of pS1Pless mice to PSA could be due to elevated sensitivity to leak-inducing mediators or to their enhanced production. Rises in plasma histamine levels during PSA were similar in control and pS1Pless mice (Supplemental Figure 2), militating against increased production. Conversely, responses to exogenous histamine and to platelet-activating factor (PAF), major mediators of anaphylaxis in mice (18, 19), were greatly exaggerated in pS1Pless mice. After i.v. PAF injection, survival was severely impaired in pS1Pless mice (Figure ​(Figure2C),2C), and hematocrit again increased significantly more in pS1Pless mice than in wild-type (Figure ​(Figure2D).2D). Similar results were seen after histamine injection (Figure ​(Figure2H).2H). The rapid increase in hematocrit after PAF injection was not prevented by splenectomy and hence did not result from mobilization of erythrocytes from spleen; PAF challenge caused Evans blue extravasation in multiple organs including heart, lung, skin, liver, and gut; and repletion of blood volume with i.v. saline prevented death after PAF treatment (Supplemental Figure 3). Thus, death of pS1Pless mice after PAF injection was likely caused by widespread and exaggerated extravasation of plasma with depletion of intravascular volume, and the increased response of pS1Pless mice to PSA was likely due to their enhanced responsiveness to leak-inducing inflammatory mediators.

To more directly examine pS1Pless mice for altered control of vascular leak, we assessed extravasation of fluorescent microspheres via interendothelial cell gaps in blood vessels in the trachea (Figure ​(Figure44 and Supplemental Figure 4), which features stereotyped and readily visualized vascular anatomy (20). No extravasation of microspheres was detected in the absence of PAF challenge. In mice challenged with PAF, microspheres accumulated in the walls of the postcapillary venules that run longitudinally along the tracheal cartilage rings. Collections of microspheres were more numerous and were generally larger in pS1Pless mice than in control mice (Figure ​(Figure44 and Supplemental Figure 4); the area occupied by microspheres in low-power images of tracheas from PAF-treated pS1Pless mice (determined using NIH ImageJ software) was about 3.4- ± 0.5-fold higher than that in PAF-treated littermate controls (n = 4–5; P = 0.001). The extravasated microspheres were located at the cell-cell junctions, often at large interendothelial cell gaps that were widespread in tracheal preparations from PAF-treated pS1Pless mice but not PAF-treated controls. Analysis of lung tissue by transmission electron microscopy also suggested exaggerated formation of interendothelial cell gaps as well as increased hemoconcentration and vasoconstriction in PAF-treated pS1Pless mice (Supplemental Figure 5).

The results described above demonstrate that Sphk function in Mx1-Cre–sensitive cell types is required to maintain normal vascular integrity and to prevent lethal responses to leak-inducing inflammatory mediators in mice. Mx1-Cre is widely used to trigger excision of floxed alleles in hematopoietic lineages and hepatocytes, and (Mx1-Cre–induced) pS1Pless mice lacked S1P in serum (Supplemental Figure 6) and in platelet-poor plasma (15), consistent with efficient hematopoietic excision and with the notion that the pS1Pless leak phenotype is due to loss of a plasma S1P–dependent function. However, Mx1-Cre–mediated gene excision in endothelial cells has been reported (17), and it was possible that vascular leak in pS1Pless mice might represent an endothelial cell–autonomous phenotype associated with loss of endothelial Sphk activity. We did not detect expression of EYFP or lacZ reporter alleles for Cre-mediated excision in endothelial cells in skin, lung, or trachea from adult mice using our Mx1-Cre induction protocol (Supplemental Figure 7). However, some excision of the floxed Sphk1 allele was detected by PCR and Southern blot analysis of DNA from endothelial cells immunopurified and cultured from the skin of neonatal pS1Pless mice. We therefore used a more direct and functional test of whether vascular leak in pS1Pless mice might be attributable to an endothelial cell–autonomous phenotype as a result of loss of endothelial cell S1P production, by determining whether restoration of S1P to plasma might rescue the pS1Pless phenotype.

Erythrocytes are a major source of plasma S1P, and transfusion of pS1Pless mice with wild-type erythrocytes restores S1P levels in plasma (15). To probe whether supplying S1P to plasma might suffice to prevent the pS1Pless leak phenotype, we asked whether erythrocyte transfusions could reverse the increased sensitivity of pS1Pless mice to PSA and PAF. This was indeed the case (compare Figure ​Figure2,2, E and F, with Figure ​Figure2,2, C and B; compare Figure ​Figure3B3B with Figure ​Figure3A).3A). Of note, S1P is highly compartmentalized in tissues — low in interstitial spaces and high in plasma and lymph. Plasma and lymph are themselves distinct S1P compartments supplied by distinct sources; transfusion of pS1Pless mice with wild-type erythrocytes or transplantation with wild-type bone marrow restores S1P specifically to plasma and not to lymph (15). The observation that altered vascular permeability in pS1Pless mice can be rescued by restoring S1P selectively to plasma again suggests a model in which vascular integrity and responses to leak-inducing agents are regulated by plasma S1P acting in trans on the blood vessel wall.

S1pr1 is expressed at relatively high levels in human (21) and mouse (Supplemental Table 1) endothelial cells, and activation of this receptor is barrier protective in endothelial cultures (2). Moreover, it was recently reported that pharmacological S1pr1 antagonism increases basal Evans blue extravasation in lung in a manner similar to that seen in pS1Pless mice (6, 22). The other S1P receptors expressed in mouse endothelial cells, S1pr2 and S1pr3, are found at lower levels than S1pr1 (Supplemental Table 1) and are barrier disruptive in culture studies (2, 23, 24). The concentration of S1P in plasma is approximately 1–3 μM, and it is about 98.5% protein bound (4). This suggests that free plasma S1P concentration is about 15–45 nM, substantially above the approximately 0.5-nM IC₅₀ for S1P inhibition of ³²P-S1P binding to S1pr1 (4). In accord with these values, the concentration of S1P in wild-type plasma is at least 10-fold higher than that needed to downregulate S1pr1 on circulating lymphocytes and is sufficiently decreased in pS1Pless plasma to permit upregulation of S1pr1 on these same cells (15). Thus, endothelial cells are probably exposed to S1pr1-saturating concentrations of S1P in wild-type but not in pS1Pless mice, and a lack of endothelial S1pr1 activation might account for the pS1Pless phenotype. Accordingly, we tested the possibility that pharmacological S1pr1 agonism might reverse the increased sensitivity of pS1Pless mice to leak-inducing challenge. Administration of the S1pr1-specific agonist AUY954 i.v. (25, 26) 2 minutes prior to PAF eliminated the increased sensitivity of pS1Pless mice (compare Figure ​Figure2G2G with Figure ​Figure2C).2C). The same treatment failed to protect wild-type mice against LD₅₀ PAF (data not shown), perhaps because plasma S1P is already present at S1pr1-saturating concentrations in wild types. The possible reduction in sensitivity to PAF of pS1Pless mice compared with control mice suggested in Figure ​Figure2G2G might be due to rapid partial desensitization of S1pr1 by AUY954 in the controls and/or to relative upregulation of S1pr1 in the pS1Pless mice at baseline due to a lack of plasma S1P (resulting in an increased effect of AUY954 in pS1Pless compared with wild-type mice). Taken together with the results described above, these results suggest that S1P in plasma maintains normal vascular integrity and is necessary to prevent lethal responses to leak-inducing inflammatory mediators such as histamine and PAF, and that S1pr1 signaling over minutes might account for these functions.

Some pharmacological S1pr1 agonists can be more effective than S1P itself at downregulating S1pr1, likely by increasing the probability that internalized receptors are delivered to lysosomes instead of recycling to the plasma membrane (27). Such agonists can also activate and downregulate S1pr1 in interstitial spaces in which S1P is normally absent (15, 25). Pharmacological activation and downregulation of S1pr1 on lymphocytes in the lymphoid interstitium is associated with failed lymphocyte egress and reduced peripheral lymphocyte counts (25). AUY954 indeed dramatically reduced lymphocyte counts in wild-type mice within 2 hours of administration, with a nadir between 2 and 8 hours (Supplemental Figure 8). Of note, wild-type mice exposed to AUY954 for 4 hours became sensitive to PAF (Figure ​(Figure5,5, A–C), similar to untreated pS1Pless mice (Figure ​(Figure2,2, C and D), but a 2-minute treatment with AUY954 had no effect (Figure ​(Figure5A).5A). The ability of prolonged but not brief AUY954 exposure to cause a pS1Pless-like phenotype in wild-type mice is likely due to its ability to downregulate S1pr1 and, together with our other results, is consistent with the model that plasma S1P and S1pr1 comprise a system that conditions or regulates responses to leak-inducing challenge.

S1pr1 is G_i coupled (21) and activates Rac, actin reorganization, lamellipodial extension, cell spreading, and junction formation in cultured cells (2, 3, 28). If S1P provides barrier-protective signals via S1pr1 and G_i in vivo, pertussis toxin (PTX) should trigger endothelial leak similar to plasma S1P deficiency and S1pr1 downregulation. PTX has long been known to sensitize mice in anaphylaxis models (29), and a single i.v. injection of PTX indeed resulted in constitutive leak in lung as well as increased hemoconcentration and markedly impaired survival to both PAF and histamine (Figure ​(Figure5,5, D–F, and Figure ​Figure6D)6D) — a phenotype similar to but more severe than that seen in pS1Pless mice (Figure ​(Figure1A1A and Figure ​Figure2,2, C and H). Of note, poly(I:C)-induced Mx1-Cre+:ROSA26 Lox-STOP-Lox PTX S1 mice (30), which express the PTX catalytic subunit in all hematopoietic cells, failed to show sensitization to PAF (Figure ​(Figure5G),5G), suggesting that systemic PTX sensitizes mice to leak by acting on non-hematopoietic cells. These results are consistent with the notion that S1pr1 and other receptors that couple to G_i in endothelial cells limit permeability responses to leak-inducing agents in vivo. They also raised the possibility that, while S1pr1 agonist might not benefit wild-type mice because saturating concentrations of S1P are already present in plasma, agonists for other G_i-coupled receptors might. Par2 (also known as F2rl1) is expressed in microvascular endothelial cells (31) (Supplemental Table 1) and, like S1pr1, activates G_i and Rac in human umbilical venous endothelial (32) and EA.hy926 cells (33). Administration of i.v. SLIGRL, a peptide agonist for Par2, provided striking protection against hemoconcentration and death from high-dose PAF in wild-type mice (Figure ​(Figure6,6, A and B). No effect of SLIGRL was seen in Par2^–/– mice. SLIGRL decreased the barrier-disruptive effects of PAF in monolayers of mouse endothelial cells grown in culture (Figure ​(Figure6C),6C), consistent with SLIGRL acting directly on endothelial cells and consistent with previous results (34). In accord with protection by SLIGRL being mediated by Par2 activation of G_i, SLIGRL was no longer protective in mice treated with PTX (Figure ​(Figure6,6, compare panels D and A). The protective effect of SLIGRL persisted in Sphk1^–/– mice (Figure ​(Figure7,7, A and B), suggesting that protective signaling by Par2 and the S1P system converge on G_i signaling rather than being linked in an Sphk1-dependent PAR to S1P receptor cascade, as has been proposed for activated protein C (9). However, in contrast to AUY954, acute treatment with SLIGRL did not rescue the increased sensitivity of pS1Pless mice to PAF (Figure ​(Figure7C).7C). The difference between AUY954 and SLIGRL in their ability to blunt the pS1Pless phenotype might be due to differences in the onset, duration, magnitude, or location of their actions in vivo or to qualitative or quantitative differences between S1pr1 and Par2 signaling, and S1pr1 signaling might promote or permit the effects of Par2 activation.

Korhonen and colleagues recently demonstrated that G_q/11 signaling in endothelial cells is required for lethal responses to PSA, histamine, and PAF in mouse models (35). Our demonstration that plasma S1P–dependent G_i signaling blunts such responses complements this study and supports the notion that GPCR signaling plays a central role in both triggering and limiting vascular leak.

Reversion of the permeability phenotype in pS1Pless mice by erythrocyte transfusion to restore plasma S1P levels and by acute S1pr1 agonism suggests that endothelial cell–autonomous mechanisms are unlikely to account for this phenotype. However, our results in no way imply that loss of Sphk function in endothelial cells cannot cause important phenotypes. Indeed, global knockout of Sphk1 function, which does not decrease plasma S1P levels enough to permit upregulation of S1pr1 on lymphocytes in plasma but does remove Sphk1 function from endothelial cells, results in a leak phenotype that resembles that seen in pS1Pless mice but that is substantially less dramatic (Figure ​(Figure7A)7A) (7, 36–38). The ability of Sphk1 deficiency alone to confer some increased susceptibility to PAF is consistent with observations in other models (7, 36). How much of this phenotype is due to toxic/metabolic effects of altered sphingosine metabolism in endothelial cells or other cells, loss of a local source of S1P for autocrine or paracrine signaling to S1pr1 or other receptors, or still other mechanisms is unknown. The fact that plasma S1P levels are only modestly reduced in Sphk1^–/– mice (36, 39) (our unpublished observations) suggests that Sphk1 functions independent of supplying S1P to plasma also contribute to vascular integrity.

Angiopoietin 1 from smooth muscle cells and pericytes acting upon endothelial cell Tie2 constitutes another system required for vascular integrity (40). Interestingly, the effect of plasma S1P deficiency on vascular leak was most obvious in postcapillary venules, which do not express Tie2 (41). Thus, the S1P/S1pr1 and Ang1/Tie2 systems may be anatomically separate and complementary, with the former being most important in postcapillary venules where large physiological excursions in permeability occur and are dynamically regulated, and the latter being most important in large vessels and terminal and precapillary arterioles (41) where permeability is constantly maintained at a low level (Figure ​(Figure4,4, Supplemental Figure 4, and data not shown).