Erythrocyte-derived sphingosine-1-phosphate stabilizes basal hydraulic conductivity and solute permeability in rat microvessels

Abstract

Exogenous sphingosine-1-phosphate (S1P), a lipid mediator in blood, attenuates acute microvascular permeability increases via receptor S1P1 to stabilize the endothelium. To evaluate the contribution of erythrocytes as an endogenous source of S1P to the regulation of basal permeability, we measured permeability coefficients in intact individually perfused venular microvessels of rat mesentery. This strategy also enabled the contributions of other endogenous S1P sources to be evaluated. Apparent permeability coefficients (P_S) to albumin and α-lactalbumin and the hydraulic conductivity of mesenteric microvessels were measured in the presence or absence of rat erythrocytes or rat erythrocyte-conditioned perfusate. Rat erythrocytes added to the perfusate were the principal source of S1P in these microvessels. Basal P_S to albumin was stable and typical of blood-perfused microvessels (mean 0.5 × 10⁻⁶ cm/s) when erythrocytes or erythrocyte-conditioned perfusates were present. When they were absent, P_S to albumin or α-lactalbumin increased up to 40-fold (over 10 min). When exogenous S1P was added to perfusates, permeability returned to levels comparable with those seen in the presence of erythrocytes. Addition of SEW 2871, an agonist specific for S1P1, in the absence of red blood cells reduced P_S^BSA (40-fold reduction) toward basal. The specific S1P1 receptor antagonist (W-146) reversed the stabilizing action of erythrocytes and increased permeability (27-fold increase) in a manner similar to that seen in the absence of erythrocytes. Erythrocytes are a primary source of S1P that maintains normal venular microvessel permeability. Absence of erythrocytes or conditioned perfusate in in vivo and in vitro models of endothelial barriers elevates basal permeability.

sphingosine-1-phosphate (S1P), a lipid mediator found in blood plasma at 0.1–4 μM, where it is carried by serum albumin and HDL, has been demonstrated to modulate a range of cardiovascular functions including blood vessel development and maturation, immune cell migration, and vascular permeability (12, 22, 26, 27, 44, 58). Many investigations of the action of S1P to modulate vascular permeability have been carried out using exogenous S1P added to the bathing solutions in vitro (24, 40) or the perfusate in isolated microvessels (5, 50, 63). However, the role of endogenous sources of S1P to modulate vascular permeability is an area of active investigation. Previous studies have focused on platelets as a source of numerous vascular permeability mediators, including S1P. Endothelial monolayer studies have shown that platelet-conditioned media reduces monolayer permeability possibly through the action of platelet-derived lipids (23, 49) including S1P (55). Low platelet count in either clinical or experimental situations can induce increased vascular permeability in various organs (7, 25), and repletion with platelet-rich plasma reduces permeability toward normal (43, 45). However, platelets are known to increase and decrease permeability via a variety of mechanisms involving interactions with endothelium and leukocytes (28, 56), and the specific contributions of S1P have not been determined.

Recent investigations clearly identify red blood cells (RBCs) as a major source of S1P in plasma (9, 11, 26). S1P is synthesized from sphingosine via the action of two sphingosine kinases (Sphk1 and Sphk2) (9). Plasma levels of S1P are regulated in part by export from the erythrocytes via a novel ATP-dependent transporter (38). The observation that infusion of erythrocytes from wild-type mice into mice with selective knockout of Sphk1 and Sphk2 in hematopoietic cells reduced the sensitivity of the mice to inflammatory mediators such as platelet activating factor (PAF) supports a major role of RBCs as a source of plasma S1P (12). Although investigations in perfused lung with and without RBCs suggest that exogenous S1P can reduce permeability when RBCs are absent (57), no quantitative evaluation of change in baseline permeability in the presence and absence of S1P specifically derived from red cells has been made. As a step toward this goal, we devised experiments to measure vascular permeability in venular microvessels individually perfused in the presence or absence of RBCs or RBC-conditioned perfusates.

We have exploited methods in our laboratory to cannulate and perfuse rat mesenteric microvessels and measure their permeability properties. In these vessels previous investigations have demonstrated that exogenous S1P protects against acute increases in the hydraulic conductivity of the microvessel wall induced by bradykinin or PAF (5, 50, 63). The protective action is associated with ligation of the S1P1 receptor by the exogenous S1P and results in peripheral localization of cortactin and maintenance of continuous vascular endothelial cadherin (VE-cadherin) and occludin. To measure hydraulic conductivity the experiments were carried out in the presence of a small number of RBCs (hematocrit 1.3%) added to the perfusate as flow markers to measure transvascular fluid exchange (the Landis-Michel micro-occlusion method) (35, 47, 48). Thus the first series of investigations reported here were to evaluate the contribution of S1P from the red cell flow markers to the regulation of vascular permeability as measured by hydraulic conductivity. It was assumed previously that these marker RBCs had no functional action on vascular permeability under the conditions of the experiments. After we found that the red cells appeared to be a major source of S1P to stabilize baseline hydraulic conductivity in our experiments, we extended our investigations to measure solute permeability coefficients to albumin and to α-lactalbumin in individually perfused microvessels. Because this approach did not require RBCs as markers of changed permeability, we could investigate directly the specific contribution of RBC-derived S1P to the regulation of baseline permeability. We could also investigate whether RBCs had to be continuously present in the perfusate, or whether RBC-conditioned perfusates also maintained permeability.

METHODS

Animal preparation.

Experiments were carried out on male rats (Sprague-Dawley, 350–450 g; Hilltop Laboratory Animals) anesthetized with pentobarbital (100 mg/kg body wt sc) and maintained with additional pentobarbital (30 mg/kg sc) as determined using toe-pinch reflex. At the end of experiments animals were euthanized with saturated KCl. Animal protocols (no. 16158) were approved by the Institutional Animal Care and Use Committee of the University of California, Davis. Rats were placed on a tray with a heating pad to maintain normal body temperature. A midline abdominal incision (2 to 3 cm) was made, and the mesentery was gently positioned over a cover glass or quartz pillar for microscopic observation. The mesentery was continuously suffused with Ringer's solution (35°C to 37°C). Experiments were performed on straight nonbranched segments of venular microvessels typically of 25–35 μm diameter. Before cannulation experimental vessels had brisk blood flow and were generally free of leukocytes sticking or rolling.

Measurement of vessel wall solute apparent permeability.

A detailed description of the methods and assumptions used to measure fluorescent solute flux (J_s) crossing the walls of single perfused vessels and to calculate apparent permeability (P_s) has been published (13, 29). The apparent permeability coefficient includes both diffusive and convective components of the solute flux between blood and tissue. The present methods, using the same assumptions, are based on the analysis of charge-coupled device images of fluorescent solute as the solute fills a perfused vessel, permeates the vessel wall, and accumulates in the immediately surrounding tissue. We calculate intensity of fluorescence (I_f) within a region of interest (ROI) defined in the images to include a segment of vessel and tissue (usually 100–200 μm along the vessel axis and about 100 μm to either side of the vessel) as the vessel initially fills with fluorescent solute (ΔI_f0) and during accumulation over 30–60 s (dI_f/dt)_i. For a vessel radius r and surface area S, and assuming a constant concentration difference across the wall ΔC, we calculate apparent P_s from the following expressions: 1) P_s = J_s/(SΔC) and 2) J_s/(SΔC) = (1/ΔI_f0) (dI_f/dt)_i (r/2).

Specifically, the mesentery was observed via transillumination using a 20× (0.5 numerical aperture) long-working distance lens (Zeiss Axiovert 100M microscope). One side of a double-barreled (Θ) pipette (16) was filled with the perfusate containing BSA (10 mg/ml) in Ringer's solution (BSA/Ringer, washout), and the other side filled with the same solution additionally containing fluorescently labeled solute, usually Alexa Fluor 555-labeled BSA (0.5–1 mg/ml). With use of a micro-manipulator the Θ pipette was used to cannulate the venule in the direction of normal blood flow. Each side of the Θ pipette was connected to an independently adjustable water manometer. With pressure raised to enable downstream flow (about 25–30 cmH₂O) from one side of the pipette, pressure on the other side was set to yield no flow through the pipette (balance pressure). Thus the solutions on the two sides could be alternately perfused through the venule as either pure washout solution or solution containing fluorescent solute. When the perfusate was switched from washout to fluorescent solute the fluorescence filled the vessel giving a step increase in fluorescence intensity (ΔI_f0) in the field of view. During the subsequent 30–60 s fluorescent solute escaped into the tissue and the fluorescence increased with time (dI_f/dt)_i. Fluorescent excitation was accomplished with a light switch (Lambda DG4; Sutter Instruments). Images were captured using a high-sensitivity camera (Hamamatsu EM CCD C9100) every 500 ms (30-ms exposure) and saved for offline analysis. Excitation and collection of emission images was controlled using Simple PCI (Hamamatsu). Offline analysis to measure fluorescent intensities in defined ROIs was accomplished with ImageJ (National Institutes of Health). Essential calibration and controls have been extensively discussed (13, 29). Briefly, auto-gain functions on the camera must be off, excitation levels and camera sensitivity (gain) must be low enough that no saturation is present in the field of view, and the total fluorescence intensity in the image ROI must be linearly related to the amount of solute in the ROI.

Hydraulic conductivity measurements.

Hydraulic conductivity (L_p) was measured to characterize vessel wall permeability. Experiments were based on the modified Landis technique, which measures the volume flux of water crossing the wall of a microvessel perfused via a glass micropipette following downstream occlusion of the vessel (6, 35, 47, 48). The initial transcapillary water flow per unit area of the capillary wall (J_v/S₀ was measured at predetermined capillary pressures of 30 to 60 cmH₂O. Microvessel L_p was calculated as the slope of the relation between (J_v/S₀ and applied hydraulic pressure. L_p was estimated from each occlusion with the assumption that the net effective pressure determining fluid flow was equal to the applied hydraulic pressure minus 3.6 cmH₂O, the approximate oncotic pressure contributed by the BSA in all perfusates. Experimental reagents were added to the perfusate and delivered via the micropipette continuously during L_p measurement.

In one set of experiments, we measured P_s and L_p in the same microvessel using the same pipette. We cannulated the microvessels with a Θ pipette containing RBCs in both the washout and dye perfusates. After fluorescence images were collected for measuring P_s, we collected white light video using a second camera (Pulnix TM-7CN) mounted on a different port on the microscope. The pressure was set to the same value for both sides of the Θ pipette, and the vessels were occluded to visualize the RBCs and determine water flux across the vessel wall.

Preparation RBCs and RBC-conditioned perfusates.

RBCs from each rat collected into a heparin-containing syringe by tail vein puncture were washed by dispersal in Ringer (0.2 ml blood per 15 ml Ringer) and then gently centrifuged in a table top clinical centrifuge (about 1,000 rpm, 7 min) to loosely pellet the RBCs. The supernatant was discarded and the cells washed two more times with the same protocol, yielding a pellet of RBCs washed of heparin and plasma constituents and largely devoid of platelets. The cells were washed one more time in 1.5 ml Ringer to pellet in a microcentrifuge tube (3 min, 4,000 g, table top centrifuge Eppendorf 5415C). RBCs were added to solutions (1.3% vol/vol) to act as flow markers for L_p experiments or to be present in P_s experiments where indicated. To generate RBC-conditioned perfusates, RBCs were added (1.3% vol/vol) to BSA/Ringer and let stand for 20 min, then separated by pelleting the cells (3 min, 8,200 g, table top centrifuge Eppendorf 5415C) and removing the supernatant. The conditioned BSA/Ringer was used to make up all RBC-conditioned perfusates used in that rat. To test for residual platelets, samples of prepared RBCs were analyzed by automated complete blood count (Siemens ADVIA 120 Hematology System) with parameters specific for rat blood cells.

Preparation of 30% plasma perfusate.

For each rat, 0.72 ml of blood was collected from the jugular vein into a syringe containing 0.16 ml heparin (1,000 units/ml). Plasma was separated from cells by centrifuging (4 min, 16,000 g, table top centrifuge), then diluted with BSA/Ringer to 30% by volume. The 30% plasma was allowed to sit for at least 20 min and filtered through a 0.2-μm syringe filter before making perfusates.

Solutions and reagents.

Mammalian Ringer's solution was composed of (in mM) 132 NaCl, 4.6 KCl, 2 CaCl₂, 1.2 MgSO₄, 5.5 glucose, 5.0 NaHCO₃, and 20 HEPES and Na-HEPES. The ratio of acid-HEPES to Na-HEPES was adjusted to achieve pH 7.40–7.45. With the exception of the 30% plasma perfusate, all perfusates were mammalian Ringer's solution additionally containing BSA (essentially fatty acid free; Sigma A0281) at 10 mg/ml (BSA/Ringer's solution). Stock solution of S1P1 receptor antagonist W-146 (Avanti Polar Lipids,) was prepared at 5 mM with 2% 2-hydroxypropyl-β-cyclodextrin in Ringer; W-146 was diluted to working concentration (10 μM and 0.004% 2-hydroxypropyl-β-cyclodextrin in BSA/Ringer) with or without added fluorescent solute immediately before each experiment. The S1P1-specific agonist SEW 2871 (Tocris Bioscience) was prepared in EtOH (20 mM) and diluted immediately before use to working concentration (10 μM).

The majority of experiments in this study were done using fatty acid-free BSA (Sigma A0281; 66 kDa, 3.6 nm Stokes radius) labeled with Alexa Fluor 555 (emission max 570 nm) according to manufacturer (Invitrogen) instructions to produce BSA-555 with a labeling efficiency of 3–5 mol of dye per mole of protein and then processed to reduce free dye concentration as described previously (42). Fluorophore labeling can alter the physical and chemical properties of the proteins, including size and net charge, potentially altering the permeability characteristics of the tracers with respect to the native proteins. Although the present studies did not directly address these issues, the 3–5 mol of dye per mole of protein would yield tracer molecules having higher molecular weights by 4–7 kDa and having more negative net charge. Both changes could be expected to decrease the permeability of the tracers relative to that of the native proteins (2, 8, 53). Nonetheless, the use of consistent, well-characterized tracers enabled comparison of macromolecule transport in control versus test protocols. Additional experiments compared results acquired with the following fluorescent solutes: rat serum albumin (RSA; Sigma A2018; 65 kDa) labeled with Alexa Fluor 555, α-lactalbumin (Sigma L6010; 14,178 Da, 2.0 nm Stokes radius) labeled with tetramethylrhodamine-isothiocyanate (TRITC), and α-lactalbumin labeled with Alexa Fluor 555. All fluorescently labeled solutes were frozen in small aliquots and thawed before use. Previous experience with Alexa Fluor 555 labeling has shown that free dye can continue to dissociate even during frozen storage, so all Alexa Fluor 555-labeled proteins were ethanol precipitated at −20°C for 1 h before use (42). The labeled protein was collected via centrifugation (6 min, 16,000 g) and redissolved in BSA/Ringer. Concentration of labeled solutes in final perfusates varied typically from 0.5 to 1 mg/ml.

Measurement of S1P in perfusates.

S1P concentration in RBC-conditioned perfusate was measured using an ELISA kit in accordance with manufacturer's instructions (K-1900; Echelon Bioscience).

Analysis and statistics.

Representative data were graphed as individual values for both L_p and P_s measurements. To establish values for each vessel to be used in statistical comparisons the final 2–4 L_p measurements during each period were averaged. Final 2–4 P_s values were averaged during each period to establish a given value for each vessel. For the time course averages of Fig. 5B, values of both L_p and P_s were averaged on two to three values nearest the chosen times shown in Fig. 5B for each vessel and those values averaged across vessels. Throughout, averaged L_p or P_s values were reported as mean ± SE. Statistical tests (indicated in the figure legends) were performed assuming significance for probability levels <0.05.

Fig. 5.

Paired measurements of unstable P_s to α-lactalbumin (α-lact) compared with stable L_p. A: representative data from 1 vessel in which P_s to α-lactalbumin (perfusate contained no RBCs) was increasing over 25 min. The vessel was recannulated, and a low and stable L_p was measured using RBCs as flow markers (note L_p scale at right). When the vessel was recannulated to measure P_s to α-lactalbumin (no RBCs), the permeability was again unsteady and rising for 20 min. B: averaged values show that P_s in the absence of RBCs was unsteady and rising when measured both before and after measurement of L_p, where L_p was measured in the presence of RBCs (n = 6).

RESULTS

Inhibition of receptor S1P1 in the presence of RBCs increases L_p.

We have consistently found that the L_p of rat venular microvessels is stable with a mean value close to or less than 1 × 10⁻⁷cm/(s × cmH₂O). Data from a representative experiment (Fig. 1A) demonstrates that in microvessels with stable normal L_p, measured using RBCs as flow markers, W-146 (10 μM; a competitive antagonist specific for the S1P1 receptor) increased L_p within 1 to 2 min when added to the perfusate. Hydraulic conductivity was restored when W-146 was removed from the perfusate. The vehicle used for W-146 did not modify L_p. Summary data in Fig. 1B from seven microvessels with a baseline L_p (measured with RBCs in the perfusate) of 0.70 ± 0.12 × 10⁻⁷ cm/(s × cmH₂O) show that mean L_p increased about 10-fold over about 15 min when each vessel was perfused with W-146 in the perfusate in addition to RBCs; L_p after recovery was 0.68 ± 0.10 × 10⁻⁷ cm/(s × cmH₂O). These results demonstrate that, in the absence of exogenous S1P, but with RBCs present, an endogenous source of S1P is sufficient to maintain stable baseline permeability. To further investigate the contribution of RBCs to this S1P source, we exploited our methods to measure albumin solute permeability coefficient (P_s^BSA) with and without red cells added to the perfusate, and using red cell-conditioned media as the perfusate as described below.

Fig. 1.

Inhibition of sphingosine-1-phosphate 1 (S1P1) using W-146 in presence of red blood cells (RBCs) reversibly increases hydraulic conductivity (L_p). A: representative data from an experiment showing that the stable L_p measured in the presence of RBCs (RBCs1) is increased when the antagonist (W-146) to the principal S1P receptor in these microvessels (S1P1) is present with the RBCs. Normal permeability is restored when the W-146 is removed and the vessel is perfused again with the RBCs alone (RBCs2). B: averaged results of 7 such experiments show strong effect of W-146 to increase L_p. Addition of 0.004% 2-hydroxypropyl-β-cyclodextrin (vehicle for W-146) had no effect on L_p (n = 4). *P < 0.05, different from initial-RBCs group, repeated-measures 1-way ANOVA with Bonferroni's post-test.

P_s^BSA increases in absence of RBCs or RBC-conditioned perfusate.

When RBCs were present in the perfusate (1.3% hematocrit) the mean P_s^BSA was 0.7 ± 0.2 × 10⁻⁶ cm/s and remained stable for periods of more than 10 min (representative experiment in Fig. 2A). In the same vessels, removing the RBCs caused P_s^BSA to increase about 20-fold within 10 min (mean of 7 vessels was 15.3 ± 6.3 × 10⁻⁶ cm/s). To test whether RBCs must be physically present we extended the protocol in three of the seven vessels. After P_s had increased from the stable state with RBCs present to the elevated state in the absence of RBCs, we recannulated the vessels using only RBC-conditioned perfusate (BSA/Ringer exposed to 1.3% hematocrit for 20 min and the RBCs removed by centrifugation). In each microvessel the RBC-conditioned perfusate restored the P_s^BSA to the low, stable state (mean 0.8 ± 0.1 × 10⁻⁶ cm/s). Reperfusion without RBCs again led to increasing P_s^BSA. Figure 2B summarizes results. These results demonstrate that the action of RBCs is not likely to be some physical plugging of sites of increased permeability on the microvessel wall, but an action of one or more substances derived from erythrocytes. These results also demonstrate that other possible sources of S1P (from the endothelial cells themselves or from other cells close to the perfused microvessel) do not contribute sufficient S1P to maintain normal permeability in the absence of RBCs. In additional experiments we demonstrated the action of RBCs to maintain both low L_p and low P_s^BSA in the same vessels. In each vessel, with RBCs present L_p and P_s^BSA were alternately measured using the same perfusate (no recannulation). The P_s^BSA was 0.7 ± 0.2 × 10⁻⁶ cm/s (n = 6); the mean L_p was 0.5 ± 0.1 × 10⁻⁷ cm/(s × cmH₂O).

Fig. 2.

Permeability coefficient to BSA (P_s^BSA) increases in absence of RBCs or RBC-conditioned perfusate. A: representative data from a vessel in which a low stable P_s was first measured in the presence of RBCs, then a rising P_s was seen in the absence of RBCs. With the use of RBC-conditioned perfusate, the P_s returned to low and stable values. This 1 vessel was recannulated a final time returning to perfusate containing no RBCs, and the P_s increased rapidly. B: averaged values from 7 vessels show strong stabilizing effect of either RBCs or RBC-conditioned perfusate on P_s^BSA (n = 7, with and without RBCs; n = 3, RBC conditioned). *P < 0.05, different from with-RBCs group, Kruskal-Wallis 1-way ANOVA with Dunn's post-test.

S1P in conditioned perfusate.

We measured S1P concentration in the RBC-conditioned perfusate and found the mean S1P concentration was 0.34 ± 0.02 μM (n = 8). This concentration is within the range (0.1 to 5 μM) for which we have previously demonstrated that S1P rapidly activates the small GTPase Rac1 in cultured endothelial cells; it is also in the range of measured values (0.1 to 4 μM) reported for normal plasma (26).

Residual platelets in prepared RBCs.

Samples of RBCs that were prepared for perfusion and for making conditioned perfusate were analyzed for the presence of platelets. The perfusates were very poor in platelets, having about 1 × 10³ platelets per μl of fluid (0.9 ± 0.2 × 10³/μl; n = 4), close to 0.1% of the normal platelet count (880 × 10³/μl) in rat blood (17).

S1P1 inhibition increases P_s^BSA in presence of RBC-conditioned perfusate.

We extended this strategy to test whether the S1P1 receptor antagonist, W-146, blocked the action of RBC-conditioned perfusate to stabilize barrier function when P_s^BSA was measured. In all eight vessels tested, the P_s^BSA was stable when RBC-conditioned perfusate alone was present (mean 0.5 ± 0.1 × 10⁻⁶ cm/s). Addition of W-146 to the RBC-conditioned perfusate increased P_s^BSA in all vessels (mean 13.6 ± 4.8 × 10⁻⁶ cm/s at 10 min). Figure 3A shows a typical experiment. The increase in P_s^BSA in the presence of W-146 above the value with RBC-conditioned perfusate alone was similar to that with perfusate containing no RBCs. After P_s^BSA was increased in the presence of W-146, we successfully recannulated seven of the eight microvessels with the perfusate containing RBC-conditioned perfusate alone (i.e., without W-146). Permeability was restored to low values in all cases (mean 1.3 ± 0.8 × 10⁻⁶ cm/s); summary data are shown in Fig. 3B.

Fig. 3.

S1P1 receptor-specific antagonist W-146 reversibly inhibits effect of RBC-conditioned (RBC-cond) perfusate on P_s^BSA, and S1P1-specific agonist SEW 2871 mimics effect of RBCs. A: representative data from a vessel show the stabilizing effect of RBC-conditioned perfusate on P_s^BSA and that P_s rises rapidly when the vessel is reperfused with perfusate additionally containing W-146 (10 μM), a S1P1-specific antagonist. The modulating effect of RBC-conditioned perfusate is returned during the subsequent recannulation in the absence of W-146. B: averaged data from 7 experiments showing final values for each condition demonstrate that the modulating effect of RBC-conditioned perfusate is blocked by antagonism of the S1P1 receptor. *P < 0.05, different from initial RBC-conditioned group, Kruskal-Wallis 1-way ANOVA with Dunn's post-test. C: representative data showing that perfusion with SEW 2871 (10 μM) reduces P_s similar to RBCs or RBC-conditioned media. D: averaged data from 5 experiments showing strong effect of S1P1-specific agonist to maintain low P_s in absence of RBCs or conditioned media. *P < 0.05, different from no-RBCs group, repeated-measures ANOVA with Bonferroni's post-test.

S1P1 agonist SEW 2871 maintains normal P_s^BSA.

To test that activation of receptor S1P1 maintains low permeability we used the S1P1-specific agonist SEW 2871. A representative experiment is shown in Fig. 3C, and averaged values are in Fig. 3D. After a stable low P_s in the presence of RBCs (P_s^BSA = 0.5 ± 0.2 10⁻⁶ cm/s; n = 5) was established, perfusion in the absence of RBCs induced a high and unstable P_s^BSA (P_s^BSA = 42 ± 16 × 10⁻⁶ cm/s). After recannulation and perfusion with solution containing SEW 2871 (10 μM) in the absence of RBCs, P_s^BSA in the same vessels returned to a low value (P_s^BSA = 1.0 ± 0.2 × 10⁻⁶ cm/s). In three additional experiments, after P_s^BSA increased in the absence of RBCs (mean 28.8 ± 11.3 × 10⁻⁶ cm/s), we added exogenous S1P (1 μM) to the perfusate. The S1P containing perfusate restored P_s^BSA to values (mean 0.9 ± 0.5 × 10⁻⁶ cm/s) comparable with those measured in the presence of RBCs or RBC-conditioned perfusate. Thus the actions of S1P1 receptor antagonist, the S1P1-specific agonist, and exogenous S1P at a concentration previously shown to stabilize the endothelial barrier (5, 50) are consistent with the idea that S1P is the RBC-derived product that stabilizes the endothelial barrier.

Rat plasma maintains low P_s^BSA.

To test if freshly prepared rat plasma, assumed to contain endogenous S1P, could maintain normal basal permeability, we investigated the action of dilute rat plasma (30% vol/vol in BSA/Ringer perfusate) in comparison with RBC-conditioned perfusate. Dilute rat plasma without RBCs maintained normal P_s^BSA, but cannulation and perfusion with perfusate without either RBCs or plasma increased permeability (Fig. 4A) in a manner similar to the results seen in Fig. 2. Figure 4B summarizes results from six vessels showing that rat plasma maintained basal permeability. Finally, in two vessels where we compared permeability of BSA555 with permeability of RSA555 when using dilute rat plasma as the perfusate, we found that the permeability remained low for both solutes showing that the use of BSA as test solute, rather than rat serum albumin, was not a cause of increased permeability (not shown).

Fig. 4.

Addition of 30% plasma to perfusate yields low stable P_s^BSA. A: representative data from 1 vessel initially perfused with 30% rat plasma in addition to the BSA/Ringer (no RBCs) showed a very stable P_s^BSA. When the vessel was perfused with BSA/Ringer without RBCs or plasma, the P_s rose rapidly. B: averaged data from 6 experiments show strong effect of 30% plasma to stabilize P_s^BSA similar to the results with either RBCs or RBC-conditioned perfusate. *P < 0.05, different from 30% plasma group, paired t-test.

A common pathway for water and solutes.

To test whether RBC-derived S1P acted on a pathway common to water and solutes we measured both P_s and L_p on the same vessels. In these experiments we used α-lactalbumin with molecular radius 2.0 nm, about half the size of albumin (3.6 nm) as a second solute. It is known to cross the wall via the intercellular pathway in these microvessels (19, 20). P_s^α-lact measured within 2 to 3 min of beginning perfusion without RBCs in the perfusate was five times higher than the values measured over the same period for P_s^BSA, as expected for this smaller solute. P_s^α-lact then increased continuously over a period of 20 min, similar to the results for albumin (representative experiment in Fig. 5A). After P_s^α-lact had increased, the same microvessels were re-perfused with RBCs (without labeled solute) to measure L_p; the vessels were restored to a normal, low, and stable permeability state with a mean L_p value of 0.6 ± 0.6 × 10⁻⁷ cm/(s × cmH₂O). A second perfusion to measure P_s^α-lact after L_p, again with no RBCs present, resulted in renewed increasing permeability. In all vessels the initial P_s^α-lact after measuring L_p was lower than that just before measuring L_p. Averaged data from six vessels are shown in Fig. 5B.

Further control experiments.

The results in Figs. 4 and 5 also demonstrate that differences in the experimental design (consecutive occlusions to measure L_p vs. continuous perfusion to measure P_s) do not modify the action of RBCs to modulate permeability. In addition we note that during some of the experiments in Fig. 4 we occluded the vessels directly before measuring permeability for the same periods as in the measurement of L_p. Such transient occlusion did not modify the increase in permeability even though S1P from sources other than RBCs could accumulate in the vessel lumen during the occlusion. In summary, the results conform to the hypothesis that S1P derived from RBCs in the perfusate is required to maintain a normal stable permeability state to both water (as measured by L_p) and solutes (as represented by P_s to serum albumin and α-lactalbumin).

DISCUSSION

Our results conform to the hypothesis that RBCs are a primary source of the endothelial barrier stabilizing agent S1P acting to maintain baseline permeability. Under conditions where the platelet count was reduced to about 0.1% of the normal level, RBCs or RBC-conditioned perfusate maintained normal permeability. The results also showed that the S1P1-specific antagonist (W-146) blocked the permeability stabilizing effect of RBCs or RBC-conditioned perfusate and that exogenous S1P and the S1P1-specific agonist (SEW 2871) maintained normal permeability in the absence of RBCs. When combined with additional recent observations on the relative contribution of platelets to plasma S1P levels under noninflammatory conditions, these results contribute to the shift in understanding of a significant contribution from cells other than activated platelets as sources of plasma S1P to the modulation of vascular function, particularly the permeability barrier. Specifically S1P, derived from membrane components, is phosphorylated in, stored in, and transported from the RBC (9, 26, 38). The unique feature of our experimental design is the ability to accurately measure the changes in vascular permeability to serum albumin and α-lactalbumin in microvessels under conditions where RBCs are either absent or present in the same microvessels. We used this design to highlight the importance of RBC-derived S1P by showing that when RBCs were absent from the perfusate, permeability to albumin and α-lactalbumin increased even though there were other possible sources of S1P such as the endothelial cells themselves, adjacent tissue cells, and a reduced number of platelets. Furthermore, we demonstrated that the action of RBCs was not a nonspecific action to plug the microvessel wall, because RBC-conditioned Ringer-BSA perfusate and diluted blood plasma were as effective as a perfusate containing RBCs to maintain normal permeability. Finally, our results demonstrated that the S1P1 receptor, previously shown to be present in the walls of these rat mesenteric microvessels (63), was the primary receptor involved in determining the action of S1P from RBCs.

A key observation is that the concentration of S1P in the RBC-conditioned perfusate (0.34 μM) fell within the concentration range (0.1 - 5 μM) where we have previously demonstrated that exogenous S1P acting via the S1P1 receptor stabilizes barrier function of the endothelial monolayers by RacGTPase activation through a G_i/PI3kinase/Akt cascade (5, 41, 46). This activation occurs rapidly (within 1 min). Equally important for these experiments we have demonstrated that the S1P-dependent barrier stabilization decreases within 1 to 2 min after S1P is removed (1). We have also demonstrated that the S1P/Rac dependent stabilization of the endothelial barrier involves peripheral localization of VE-cadherin, cortactin, and enhanced peripheral cytoskeleton and is sufficient to attenuate acute increases in permeability (5, 40). When S1P is added to the perfusate in the presence of an inflammatory mediator such as PAF or bradykinin, the tendency of these agents to cause rearrangement of VE-cadherin and occludin resulting in characteristic small gaps between endothelial cells is attenuated (5, 50). Furthermore, in additional experiments large increases in the permeability of these vessels were measured when Rac activation was inhibited (59, 60). Similar mechanisms, using S1P derived from RBCs, are likely to contribute to endothelial barrier stabilization in the present experiments. Our results show that in the absence of RBCs a stable endothelial permeability barrier cannot be maintained, whereas the presence of RBCs in the same intact tissue environment maintains a stable permeability state.

We note that endothelial cells are not the only targets of S1P-modulated mechanisms that modify the permeability barrier. S1P has been shown to reduce leukocyte binding and transmigration (10, 37, 44). The latter mechanism was not the primary focus of the current investigation. However, we did observe that one effect of adding RBCs to the perfusate was that leukocytes, which rolled slowly on the wall before cannulating the vessels or were stationary at the time perfusion began, generally washed away from the wall within the first min of perfusion. Further investigations are needed to determine whether part of the action of RBC-derived S1P is to attenuate activation of leukocytes that remain on or near the vessel wall during perfusion in the absence of RBCs.

We note that our reported P_s values are apparent values that include both diffusive and convective components of the flux. We have previously described the nonlinear flux equations and that full analysis of solute permeability requires measurement of solute flux at multiple perfusion hydraulic pressures as well as estimation of L_p for each vessel (29). We do know the hydraulic pressure (balance pressure about 20 cmH₂O) under the present conditions and can estimate the relevant variables. Using the typical measured L_p of 0.8 × 10⁻⁷ cm/(s cmH₂O) and reflection coefficient of about 0.9 for albumin in the basal state (3, 35), we estimate that the true diffusive permeability is 0.4–0.5 × 10⁻⁶ cm/s and that convection accounts for 20–30% of the total basal flux. In the high permeability state, where the L_p has increased by about 10-fold, most of the apparent permeability could be accounted for by a convective component of flux (e.g., close to 20 × 10⁻⁶ cm/s would be added to the apparent permeability coefficient).

Implications of these observations: reduced supply of RBC S1P?

Our results highlight the importance of further investigations of the storage and release of S1P by RBCs as a regulatory mechanism promoting a normal low permeability state in microvessels. Recent investigations reveal that the release of S1P from RBCs into plasma is a regulated process with release triggered by interaction of RBCs with HDL and serum albumin (9). S1P bound to serum albumin or to HDL was able to stimulate the S1P1 receptor. One implication of our observations is that a failure of RBCs to maintain a normal supply of S1P in microvessels may contribute to capillary leak syndromes. Our results are consistent with the observation that mice deficient in Sphk1 in erythrocytes have increased capillary leakage (12). Future investigations are warranted to evaluate the possibility that there is reduced S1P production and release when RBCs are damaged due to genetic defects, such as in sickle cell disease and spherocytosis, or in some forms of infectious disease including malaria. Of interest is the observation that children with cerebral malaria have lowered circulating levels of plasma S1P (18).

Maintenance of S1P delivery may be a therapeutic goal in some conditions rather than concern with reduced S1P supply. Apolipoprotein M, a component of HDL, is the principal carrier in plasma of S1P. In sepsis the plasma concentration of apolipoprotein M can fall by more than half (39). This significant decrease in the main carrier for S1P may contribute to reduced availability of active S1P and be associated with increased capillary permeability, which is a major factor in the severity of the acute phase response following sepsis or other causes of inflammation.

Determining the relative contributions from RBCs and platelets to the maintenance of plasma S1P under normal conditions is an area of continuing investigation. There is clear evidence that platelets and platelet-derived medium can be significant sources of S1P or other mediators that maintain basal vascular permeability (49, 55). However, there are conditions where platelet numbers are significantly reduced (58) but S1P levels are not reduced. Furthermore, there are significant differences between S1P generation, storage, and release in RBCs and platelets (33). For example, platelet release of S1P and other mediators often requires activation by other mediators or exposure to injury. Under such activating conditions platelets may also potentiate inflammatory responses, especially those by leukocytes (56). In contrast, apart from the requirement for albumin or HDL to be present for continued S1P release, RBCs do not appear to require stimulation for S1P release. They also do not directly potentiate inflammatory responses. Thus it appears to be important to distinguish normal and injury/inflammatory conditions when evaluating the contribution of platelets relative to that of red cells.

Implications of these observations: other permeability measurements.

Our observations are also important for interpreting the results of investigations on model endothelial barriers. The absence of RBCs, RBC-conditioned media, or other sources of S1P are likely to result in an unstable permeability barrier when artificial perfusates and culture media are used in perfused mammalian microvessels or cultured endothelial cells. Many investigations of individually perfused mammalian microvessels have been carried out using the modified Landis method to measure L_p of the microvessel wall as an indication of changes in permeability (5, 35, 36, 52, 54, 63). In these experiments RBCs (previously isolated from the same animal or from another donor such as hamster or human) are added to the perfusate as flow markers. As in the present experiments, in mammalian microvessels perfused under these conditions, the baseline L_p values are low [less than 1 × 10⁻⁷ cm/(s × cmH₂O)] and are remarkably stable; vessels can be perfused for up to 3 h with stable L_p. Furthermore, where we have investigated the structure of the mammalian microvessel wall we have always used perfusates that contained RBCs as flow markers to measure L_p under the same conditions as the structure was investigated (3–6). Thus the presence of red cells in the perfusate has contributed to the stability and reproducibility of the modified Landis L_p method and our structural investigations in ways that we had previously not appreciated. In this way the new observations increase our confidence in the reliability of all our published results based on measurement of microvessel L_p. We predict, that by using flow markers other than wild-type RBCs [e.g., RBCs isolated from the Sphk-deficient mice described by the Camerer et al. group (12)] and no additional source of S1P, measurements of L_p will be elevated, similar to results in Fig. 1 where the S1P1 receptor was inhibited. On the other hand experiments in individually perfused microvessels to measure solute permeability coefficients, P_s, have been carried out under a variety of experimental conditions without the presence of RBCs.

We have previously measured P_s in frog microvessels using essentially the same methods as in the present experiments and reported low and stable solute permeability coefficients in the absence of RBCs (14, 20, 29). The reasons for stability of the frog vessels may include differences in the S1P signaling in these vessels or the use of albumin in the perfusate that has not been stripped of free fatty acids (see below), but the most obvious difference from the present experiments was that the frog vessels were investigated at 15°-20°C, where the tendency for any increased permeability is reduced. However, we do note that the use of plasma in the perfusion solutions reduced P_s to tracer albumin about three-fold relative to that measured in presence of Ringer with BSA (30). Although the plasma effect in frog microvessels was attributed primarily to an enhanced charge selectivity of the wall, it is possible that S1P from the plasma also played a role.

Our laboratory has not previously reported P_s in mammalian vessels, but there are several reports of stable P_s to α-lactalbumin, albumin, potassium ion, or sodium fluorescein in rat venular microvessels. Those were measured without RBCs and using methods similar to those described above (21, 34, 51, 62). At least one important difference between those experiments and the current observations appears to be the use of fatty acid-free albumin (Sigma A0281) in the present studies. It seems possible that albumins that have not been treated to remove fatty acids may contain S1P or an S1P precursor bound to albumin or some other component. Thus perfusates containing albumin not treated to remove fatty acids may be similar to the RBC-conditioned perfusates used in the present experiments.

In other mammalian vessels, Huxley and colleagues (8, 31, 32) have reported stable P_s to serum albumin (and α-lactalbumin) in coronary arterioles and coronary venules of exercised pigs (Yucatan miniature swine) measured in absence of RBCs or RBC-conditioned perfusates. In contrast with our present results measured at 37°C all those measurements of albumin permeability were made at 15°C (values for males ranged from 0.5 to 0.8 × 10⁻⁶ cm/s, whereas values for females were as high as 2 × 10⁻⁶ cm/s). The rationale given by these investigators for using 15°C was to attenuate changes in vessel diameter due to changes in cell cytoskeleton. As suggested above for frog experiments it is possible that changes in the endothelial cell cytoskeleton (associated with increased permeability) due to removal of S1P are also attenuated at 15°C, perhaps explaining why relatively low P_s values were measured in the absence of RBCs at the lower temperature. Similar to above, another difference from the present results is that these coronary microvessel experiments were conducted with serum albumin (porcine serum albumin) not stripped of fatty acids. Measurements in porcine coronary venules where P_s values to FITC-labeled BSA were on the order of 4–7 × 10⁻⁶ cm/s were measured at 36°C (61). It is clear that future investigation of permeability properties using artificial perfusates should be made under conditions where any source of S1P, best supplied by RBCs, is known.

An equally important question is whether the absence of RBCs in most endothelial cell culture systems contributes to the fact that these cultured endothelial barriers have high baseline permeability. As noted above, the true diffusive serum albumin permeability in our present experiments is close to 0.3–0.4 × 10⁻⁷ cm/s, a value one to two orders of magnitude smaller than that measured in cultured endothelial monolayers (15) and microvessels perfused in the absence of RBCs.

Finally, we note that our investigations provide a bridge between investigations of the control of baseline permeability at the whole organ level in genetically modified mice, and investigations at the cell and molecular level in cultured endothelial cells. Our results are consistent with the conclusion that both increased accumulation of Evans Blue-labeled albumin in lung and trachea and a tendency for paws to become edematous in mice deficient in S1P in plasma reflect at least in part increased baseline permeability (12). Further investigations using more refined methods now available to evaluate the contribution of RBC-derived S1P to maintain normal permeability under a variety of proinflammatory conditions (e.g., diabetes, wound healing, dyslipidemia) in cultured endothelial monolayers, single perfused vessels, and genetic mouse models are needed (15).

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants HL-28607 and HL-44485.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: F.-R.E.C., J.F.C., and R.H.A. conception and design of research; F.-R.E.C., J.F.C., and R.H.A. analyzed data; F.-R.E.C., J.F.C., and R.H.A. interpreted results of experiments; F.-R.E.C. and R.H.A. drafted manuscript; F.-R.E.C., J.F.C., and R.H.A. edited and revised manuscript; F.-R.E.C., J.F.C., and R.H.A. approved final version of manuscript; J.F.C. and R.H.A. performed experiments; J.F.C. and R.H.A. prepared figures.