Introduction
Chemical agents provide powerful tools for dissecting complex physiological functions mediated through diverse receptor subtypes. In particular, selective agonist and antagonist pairs that are active in vivo have the distinct advantage of enabling acute, reversible modulation of molecular function while circumventing the developmental compensations that can arise in gene deletion studies. We have targeted this approach to the signaling pathway mediated by S1P and have thereby showed that in vivo–active reversible chemical tools can be used to address a series of mechanistic and therapeutic questions.
S1P is a pleiotropic autocrine and paracrine signaling lipid1 that mediates graded rheostat control of numerous physiological functions through a family of G protein–coupled receptors. Small variations in ligand concentration are amplified by selective high-affinity receptors to acutely regulate vital functions such as heart rate2, 3, vascular and stromal barrier integrity4 and lymphocyte egress5. The functioning of S1P receptors in the maintenance and modulation of biological barrier activity is of profound biological importance and has therapeutic implications4, including prevention of transplant rejection and treatment of multiple sclerosis and perhaps adult respiratory distress syndrome as well6.
Among the five known high-affinity S1P receptors, the S1P1 receptor subtype is expressed by many cell types, including endothelial, blood and cardiac cells1, 6. Genetic deletions of S1P1 have revealed the receptor's essential endothelial functions in arterial smooth muscle cell migration; absence of S1P1 results in embryonic lethality7 and abnormalities in the development of the immune system8, 9. These abnormalities relate both to cell egress from the thymus and to development of abnormal surface phenotypes such as elevated expression of the lymphocyte differentiation antigen CD69 (a retention signal for lymphoid egress8, 10, 11, 12) on thymocytes and B cells. However, mechanistic interpretations derived from such gene deletion studies may be clouded by developmental compensations and anomalies.
Chemical approaches have also provided novel insights into the functioning of the S1P system. Reverse pharmacology using compounds of unknown molecular interaction that induce disappearance of lymphocytes from the blood and that proved to be S1P receptor agonists first identified the S1P1 receptor as a target capable of inducing clinically useful immunosuppression2, 3, 13. Such agonists induce reversible inhibition of lymphocyte egress (known as lymphocyte sequestration) from the thymus, lymph nodes and Peyer's patch. Within 1 h of administration, nanomolar concentrations of agonists induce14 and maintain immunosuppression15, 16 and promote maturation of late medullary thymocytes14, 17, 18. This observation, together with the organ and microanatomical selectivity of lymphocyte sequestration by agonists13, led researchers to propose the existence of a stromal (endothelial) site for S1P1 regulation of lymphocyte egress5. In support of this hypothesis, we showed in a recent two-photon imaging study of excised, perfused lymph node that agonist-dependent inhibition of lymphocyte trans-sinusoidal egress in the medulla is reversed by antagonism19. In contrast, S1P1 deletion studies have been interpreted as suggesting that agonists inhibit lymphocyte egress by 'functional antagonism', that is by causing a downregulation of lymphocyte-intrinsic S1P1 receptors that mediate chemotaxis along an S1P gradient from node to lymph8, 9.
The complex multisystem biology through which S1P controls these diverse and essential functions via S1P1 receptors is at present a topic of debate1, 20 and is best resolved in vivo. We have characterized selective, reversible chemical tools that acutely alter the set point of S1P-S1P1 rheostats21, and we have used them to discriminate between mechanisms that are controlled by basal, unperturbed levels of S1P1 receptor-ligand tone and those that are induced by pathophysiologically or pharmacologically evoked excursions in S1P-S1P1 occupancy and response. Here, we characterize a potent, selective, in vivo–active chiral antagonist of S1P1 that inhibits both basal and induced S1P-S1P1 responses. Using this antagonist in conjunction with a similarly effective, potency-matched, in vivo–active agonist2, 22, we are now able to acutely modulate S1P1 activities that maintain vascular barrier function at physiological S1P concentrations in skin and lung capillaries, and we can also regulate the pharmacologically evoked S1P1 agonism that is required for microanatomically selective lymphocyte sequestration in lymph nodes and for lymphocyte maturation in the thymic medulla.
Results
Chiral S1P1 antagonists with in vivo stability
Previously characterized antagonists of S1P1 are limited by low potency19, chemical lability23, poor solubility and lack of in vivo activity. We hypothesized that replacing the phosphate ester with phosphonate could potentially create a potent and selective antagonist of S1P1 having enhanced biological properties, provided that we paid careful attention to the chiral center and the head group (which has a double negative charge) as a prerequisite both for retention of potency in competition for S1P and for adequate formulation.
We synthesized the R (2a) and S (2b) enantiomers of 3-amino-4-(3-hexylphenylamino)-4-oxobutylphosphonic acid (Fig. 1a), which are useful competitive antagonists of S1P1, to attempt to solve issues of selectivity, stability and (especially in vivo) efficacy. Compound 2a is a full competitive antagonist and is equipotent to the corresponding phosphate ester23. It shifted the effector concentration for half-maximum response (EC50) for the physiological ligand Compound 1 in a GTP-
S binding assay from 0.8 nM to 400 nM when in molar excess (Fig. 1b), with a Ki of 
70–80 nM (Fig. 1c). In contrast, the S enantiomer Compound 2b was less favored, having a Ki of 4.6 
M (Fig. 1b), and was therefore unlikely to have in vivo activity. We observed the same potency trend in competitive antagonism of the S1P1-selective agonist SEW2871 (Compound 3; Fig. 2a)2, that is Compound 2a (Ki of 18 nM) was substantially more potent than Compound 2b (Ki of 2.84 M; Fig. 2b). The antagonistic effect of Compound 2a was selective for S1P1 receptors, and Compound 2a showed neither agonism nor antagonism at 10 M in GTP-S assays for S1P2, S1P3 and S1P5 receptors (data not shown; no data are available for S1P4 receptors). These results are in contrast to those for antagonists that have an eight-member acyl chain, which showed strong antagonist activity on S1P3 and S1P1 receptors. Compound 2a was equipotent on mouse and human S1P1 receptors, which share extracellular and transmembrane domain sequence identity.
Figure 1: Compound 2a is a selective antagonist for human S1P1 (hS1P1) receptor activation by the physiological agonist S1P.
(a) Structures of 2a and 2b. (b) Shift in EC50 for S1P activation of hS1P1 receptor by 2a and 2b. We tested CHO cell membranes expressing stably transfected hS1P1 receptors for antagonism in a GTP-S binding assay. Activation of hS1P1 receptors by 1 alone (closed squares) or in the presence of 10 M 2a (inverted triangles) and 2b (circles) is expressed normalized to percentage GTP-S binding induced at maximal S1P concentration. The EC50 value was 0.8 nM for 1 alone; it was 398 nM and 50 nM in the presence of 2a and 2b, respectively. (c) We derived Ki values for S1P activation of the hS1P1 receptor from antagonist competition curves; they were 77 nM for 2a (inverted triangles) and 4.63 M for 2b (circles).
Figure 2: Compound 2a is an antagonist for proximal and downstream signals from hS1P1 receptor activation.
(a) EC50 shift by 2a and 2b of 3 activation of hS1P1 receptors. Curves show GTP-S binding as a percentage of maximal responses induced by 3 for 3 alone (closed squares) and in the presence of S1P antagonists 2a (1 M; inverted triangles) and 2b (10 M; circles). EC50 values were 50 nM for 3 alone and 10 M and 600 nM, respectively, in the presence of molar excess 2a and 2b. (b) Ki values for 3 activation of hS1P1 receptors derived from antagonist competition curves for 2a (inverted triangles) and 2b (circles) were, respectively, 18 nM and 2.89 M. (c) We serum starved CHO cells stably transfected with S1P1 for 16 h, then we incubated them with 2a at 10 M for 1 h and stimulated them with either 1 or 3 at the indicated concentrations for 5 min. We performed western blot analyses of phospho-Akt, total Akt, phospho-ERK1/2 and total ERK1/2 levels as described in Methods (n 
3). Data are representative autoradiograms of one of three independent experiments. 2a is a competitive inhibitor of Akt and ERK phosphorylation in response to S1P1 activation by either 1 or 3.
In addition to competitively inhibiting receptor-proximal signaling events, such as Gi activation induced by Compound 1 or Compound 3 receptor binding, Compound 2a fully inhibited downstream events including ligand-induced receptor internalization of green fluorescent protein (GFP)-tagged S1P1 (Supplementary Fig. 1 online). We also assessed the effect of Compound 2a on intracellular signals downstream of G protein activation that is dependent upon S1P1 activation by measuring extracellular signal–regulated kinase (ERK) and Akt phosphorylation separately in response to S1P and the selective S1P1 agonist Compound 3 (Fig. 2c). Compound 1 and Compound 2a induction of the S1P1-dependent accumulation of phospho-ERK and phospho-Akt were competitively inhibited by Compound 2a. We observed complete inhibition of the response to selective agonist Compound 3 and >70% inhibition of responses to nonselective ligand Compound 1, results reflecting some basal expression of endogenous S1P receptors in the S1P1-transfected Chinese hamster ovary (CHO) cell line. We saw no suppression of S1P3-dependent ERK or Akt phosphorylation and no S1P3 receptor agonism with Compound 1 in CHO cells stably expressing S1P3 (Supplementary Fig. 2 online).
The phosphonate Compound 2a is more stable in vivo than its phosphate ester23. We could not detect the phosphate ester 5 min after intravenous injection in mice, whereas plasma concentrations of Compound 2a were 268 
98 nM at 5 h after a 10 mg kg-1 injection. We determined pharmacokinetic parameters of Compound 2a in cannulated rats (n = 3) by LC-MS and found a half-life of 73 3 min, a small steady-state volume of distribution of 0.46 liter kg-1 and a modest plasma clearance of 4.63 ml min-1 kg-1. We maintained an adequate in vivo receptor blockade for 5–6 h following a single injection of 2a. We therefore evaluated the ability of this selective S1P1 antagonist to alter vascular and lymphoid responses regulated by S1P1 agonism under conditions of physiological S1P1 tone and in the presence of an exogenous agonist.
S1P1 antagonism enhances capillary leak
S1P agonism enhances endothelial barrier function in peripheral capillaries24, lung4, 25 and lymphatic sinus19 through cytoskeletal reorganization and changes to vascular endothelial cadherin junctions26. We found evidence for systemic blockade of S1P1 receptors by Compound 2a (but not Compound 2b) within the vasculature for basal responses induced by both S1P and exogenous S1P agonists.
The target selectivity of Compound 2a was initially demonstrated by in vivo assessment of the reversal of effects evoked by the exogenous S1P1-selective agonist Compound 3 (ref. 2) and is illustrated in Figure 3. Administration of Compound 3 (ref. 2) protected mice from vascular endothelial growth factor (VEGF)-induced skin leakage of Evans blue dye (complexed to plasma albumin) from plasma into the interstitial spaces (Fig. 3a), as quantified in skin-core biopsies and by dye extraction and spectrophotometric absorbance measurements (Fig. 3b). The preferred enantiomer Compound 2a entirely reversed the protective effects of Compound 3 (Fig. 3a,b), whereas Compound 2b did not. Moreover, Compound 2a modestly enhanced skin capillary leakage in response to VEGF even when administered in the absence of Compound 3, suggesting that endogenous S1P1 tone contributes to the maintenance of basal capillary integrity.
Figure 3: S1P1 antagonism reverses protection from VEGF-induced leakage by the selective S1P1 agonist 3.
(a) Representative photograph of the Miles assay in animals treated with S1P1 agonist or antagonist. Skin leakage is visualized as the accumulation of blue dye. (b) Means of Evans blue dye leakage in biopsies (VEGF minus BSA) in treated animals from two pooled experiments. Error bars show 1 s.d. Statistical analyses were performed with ANOVA and the Bonferroni Multiple Comparisons Test. From left to right, n was 5, 5, 9, 11 and 3 mice for the respective groups. P < 0.001 for 3 versus vehicle or 2a, and for 2a versus 3 or the combination of 3 and 2b. The S enantiomer 2b failed to reverse protection by 3 from VEGF leakage. (c) Correlation between lymphopenia13 and protection from VEGF-induced skin leakage across all treatment groups in Figure 3b. Scatter plot shows blood lymphocyte (LY) counts obtained in the experiments of Figure 3b against corresponding VEGF-induced skin leakage. Data are normalized to percentage of maximal leak or lymphopenia. A linear regression line is fitted to the data (slope is 0.97 0.13; r2 = 0.67).
S1P1 blockade induces spontaneous pulmonary edema
To better evaluate the modulation of basal capillary integrity by endogenous S1P-S1P1 tone, we sought an alternative model system in which a greater surface area of capillary endothelium exposed to plasma S1P could be assayed for permeability. This led us to study the lung by measuring leakage of Evans blue dye from the vasculature into the alveolar spaces4. The lung barrier consists solely of pulmonary microvascular endothelium and alveolar epithelial cells, each only one cell thick, separated by a thin extracellular matrix. The tight junctions between epithelial cells are impervious to the movement of small molecules, allowing the airway and vascular surfaces to be independently challenged with the antagonist.
Vascular administration of Compound 2a significantly enhanced basal pulmonary leakage by approximately 4.5-fold over vehicle (P < 0.0001) in a dose-dependent manner in the absence of any exogenous stimulus or ligand (Fig. 4), whereas Compound 2a was inactive. Pulmonary leakage was transient and clearance of Compound 2a from the plasma restored lung capillary integrity. Administration of S1P1-selective agonist together with antagonist Compound 2a inhibited pulmonary leakage, confirming the mechanistic selectivity of pulmonary edema. Leakage is associated with loss of cortical actin organization within endothelial cells and is the subject of ongoing studies. In contrast to the effects of vascular administration, 75 M Compound 2a did not induce measurable leakage on airway epithelial challenge. Thus, a basal level of vascular S1P-S1P1 tone seems to be necessary to prevent pulmonary leakage, and this tone is disrupted by S1P1 antagonism. The lack of activity of Compound 2b demonstrates the essential chirality for antagonist action and controls for the possibility of nonselective pulmonary leakage.
Figure 4: Basal S1P1 tone provides protection from capillary leakage in lung.
Systemic antagonism of S1P1 under basal physiological conditions induced pulmonary leakage. Photographs show Evans blue dye accumulation in lungs 90 min after i.p. injection of either vehicle, 2b (10 mg kg-1) or 2a (10 mg kg-1). Histogram shows mean values of dye accumulation in the lungs of 2a-treated C57BL1/6J mice, normalized to those of mice treated with vehicle alone. Bars indicate means s.d.; * indicates value is significantly different (P < 0.05) from vehicle; # indicates value is significantly different from 2b. n = 11 for vehicle; n = 3 for 2a at 1 and 3 mg kg-1; n = 9 for 2b and 2a at 10 mg kg-1 as determined in three separate experiments for vehicle, 2b and 2a at 10 mg kg-1.
Reversal of agonist effects on lymphocyte egress and maturation
During the experiments in which we measured the effects of agonists and antagonists on capillary leak in response to VEGF administration, we obtained simultaneous counts of lymphocytes in peripheral blood to assess the loss of blood lymphocytes (lymphopenia) resulting from lymphocyte sequestration. A scatter plot (Fig. 3c) of the degree of vascular protection from VEGF leakage by S1P1 agonism versus the induction of lymphopenia revealed a positive correlation between these parameters, both of which were reversed upon S1P1 antagonism by Compound 2a (but not by Compound 2b). The selective S1P1 agonist Compound 3 induces reversible lymphocyte sequestration from peripheral blood with a time course matching the plasma half-life of the compound2. Compound 3 induces neither a ligand-activated calcium response nor mitogen-activated protein (MAP) kinase phosphorylation in freshly isolated lymphocytes from spleen, lymph nodes or blood2, but it has allowed the visualization of lymphocyte egress inhibition in lymph node explants and the demonstration of agonist-dependent stromal gating19.
We found no significant alteration in either lymphocyte number (Fig. 5a) or percentage of lymphocytes within total white blood cell counts (differential count; Fig. 5b) after administration of Compound 2a alone in mouse peripheral blood. A single injection of 10 mg kg-1 of Compound 2a at the time of oral administration of Compound 3 at doses up to 20 mg kg-1 abolished the agonist effects (Fig. 5a; P 
4.5 
10-6), which alone was sufficient to induce and maintain lymphopenia for >24 h (ref. 2). Compound 2b was ineffective. Doses of Compound 2a as low as 3 mg kg-1 were sufficient to inhibit the lymphopenia induced by Compound 3.
Figure 5: Systemic S1P1 antagonism reverses agonist-induced sequestration of blood lymphocytes and medullary thymocyte phenotypic maturation.
(a) The S1P1 antagonist 2a does not induce lymphopenia, and indeed it reverses agonist-induced lymphopenia. We administered 2a (10 mg kg-1) i.p. in C57BL6 mice at the time of oral dosing of 3 (20 mg kg-1) or vehicle. We measured blood lymphocyte numbers as described2 at 5 h. The data shown are pooled from four independent experiments with a total n = 12 mice per group. Mean s.d. and P values from Student's t-test are shown. (b) S1P1 antagonist reverses agonist-induced alterations in differential leukocyte count (namely the percentage of lymphocytes, monocytes and neutrophils in the volume of blood counted) under the conditions and methods described in Figure 5a. Representative data (n = 3) from one of four similar experiments is shown. Because S1P agonists sequester lymphocytes (lyts) passing through specific lymphoid microenvironments, neutrophils and monocytes (myeloid cells) that do not traverse the same barriers are not sequestered. The number of myeloid cells remains constant while lymphocytes disappear, and therefore the percentage of myeloid cells rises. (c) S1P1 antagonist reversal of agonist-induced loss of CD69 in late medullary thymocytes. We quantified CD69 expression on single-positive CD4+ thymocytes by FACS at 5 h after administration of 20 mg kg-1 of 3 in the presence or absence of 10 mg kg-1 2a. Mean s.d., n = 3, data from one of two similar experiments is shown.
We obtained further evidence that Compound 2a is able to maintain systemic inhibition of S1P1 activation within lymphoid organs by measuring CD69 expression on single-positive CD4+ thymocytes in thymic medulla. The well-characterized S1P agonists FTY720 (Compound 4)17, its chiral analog AAL(R) (Compound 5)14 and Compound 318 all induce rapid loss of CD69 immunoreactivity on these thymocytes, and flow cytometric analysis within 2 h of dosing reveals quantitative conversion of a CD69-intermediate population to CD69-, a result that is likely to represent a lymphocyte-intrinsic effect of S1P1 agonism. Analysis of CD4+ single-positive thymocytes from mice treated with Compound 3 and/or Compound 2a for 5 h showed that administration of Compound 2a alone has no effect on CD69 expression, whereas treatment with Compound 3 induced an approximately seven-fold expansion of the CD69- CD4+ single-positive population (Fig. 5c). This Compound 3-induced loss of CD69 was almost completely inhibited by Compound 2a pretreatment (Fig. 5c). Compound 2a therefore penetrates the matrix of lymphoid organs and inhibits effects (such as the rapid sequestration of lymphocytes in lymph nodes and Peyer's patch) attributable to activation of S1P1 in both peripheral lymphoid systems and the thymus.
In vivo two-photon imaging of antagonist effects in lymph nodes
We examined the dynamics of lymphocyte arrest and recovery by imaging, within an intravital (in vivo) preparation of the inguinal lymph node (Fig. 6a) of mice, the effects of a potency-matched combination of the agonist Compound 3 and the antagonist Compound 2a. The procedures used for in vivo imaging of the inguinal lymph node (Fig. 6a) were previously described27. In addition, we defined landmarks for regional compartments within the medulla and cortex by in vivo staining using fluorescein isothiocyanate (FITC)-conjugated dextran to outline medullary sinuses (Fig. 6b) and blue-labeled B cells to identify B-cell follicles (Fig. 6c).
Figure 6: Imaging the in vivo kinetics of antagonist reversal of agonist-induced medullary arrest of T cells in inguinal lymph nodes by intravital two-photon fluorescence microscopy.
(a) Inguinal lymph node of a BALB/c mouse with fat dissected away for multiphoton imaging. The node is outlined in white, the medullary region is circled in blue and a typical imaging frame is denoted by the black square. Scale bar, 1 mm. (b) Representative two-photon image of the medullary sinus of a 3-treated mouse before antagonist injection and 18 min after antagonist injection. Carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled T cells are green; sinus tissue is red as described. Images show maximum intensity projections through a depth of 10 m, and overlay images were captured at 28-second intervals for 8 min. The sinus space is outlined in white (scale bar, 20 m). (c) Three-color image obtained in the cortical region, showing B cells (blue) in a follicle, T cells (red) and high endothelial venule (green); scale bar, 20 m. (d) Instantaneous velocities (mean s.e.m.) of CD4+ T cells in the medullary region 4 h after SEW2871 treatment. (e) Increase in medullary T-cell velocity after intravenous injection (red line) of 10 mg kg-1 of 2a (Supplementary Video 1). Data are from the same imaging volume as in Figure 6d. (f) Velocity of CD4+ T cells in the cortex of the lymph node 4 h after SEW2871 administration, before and after administration of 2a, showing no alterations in cortical velocities.
We administered agonist Compound 3 before imaging to induce lymphocyte sequestration in the lymph nodes. We observed T lymphocytes arrested in the medullary region outside of empty sinuses (Fig. 6b) with velocities averaging only 3.5 m min-1 (Fig. 6d). A time-lapse view of the sinus areas devoid of T lymphocytes is provided in Supplementary Figure 3 online. Subsequent intravenous injection of Compound 2a induced a very rapid increase in average T-cell velocity to 5.7 m min-1 (Fig. 6e; Supplementary Video 1 online), whereas control injections of vehicle had no effect (data not shown). In five out of six experiments, T-cell velocity in the medullary region recovered fully within 30 s (a single imaging time point) after intravenous administration of the antagonist (Supplementary Video 1; before and after videos of antagonist reversal of agonist arrest in lymph node medulla are shown in Supplementary Video 2 online). T-cell velocities in these intravital experiments, measured in the presence of Compound 3 alone or together with antagonist, correspond closely to values observed previously in an explant preparation19. In marked contrast to the effects in the medulla, corresponding measurements in the T-cell zone of the diffuse cortex showed no substantial changes in T-cell motility with either agonist or antagonist (Fig. 6f; Supplementary Video 3 online). At the boundary between B-cell follicles and the T-cell zone, T cells maintained velocities of 10–11 m min-1, regardless of S1P1 agonist or antagonist administration. Thus, agonist-induced lymphocyte arrest in the lymph node in vivo is restricted to the inhibition of medullary egress alone and is rapidly reversed in the presence of antagonist competition.
Discussion
Here, we demonstrate that a selective chiral S1P1 antagonist (Compound 2a) chosen for its distributional properties and stability in vivo can inhibit both basal and induced functions of the S1P-S1P1 axis once adequately formulated. Unselected data derived from experiments in more than 120 mice show that Compound 2a has robust in vivo activity. We have confirmed the structure-activity relationships previously described23, and it is notable that the intrinsic receptor binding potency is retained after conversion of the phosphate ester23 to phosphonate. This observation suggests that the oxygen lone pair is not essential for receptor interactions. Acyl chain length contributes to selectivity: the octanyl analogs have strong antagonist activity toward both S1P1 and S1P3, and the hexanyl analogs are selective for S1P1. The large difference in affinity between the R and S enantiomers points to a critical orientation of the head group in space, and this provides us with an in vivo inactive control that (excepting the head-group orientation) is chemically and physically identical to the active species. Moreover, the lack of biological actions of Compound 2b the S enantiomer, demonstrates that the effects of Compound 2ain vivo are related to its intrinsic receptor binding potency.
Regulation of vascular integrity by S1P
The systemic antagonism of S1P1 receptors by Compound 2a evoked changes in vascular integrity under basal homeostatic (physiological) conditions and during pharmacological intervention. Protection of skin capillaries from VEGF-induced leakage by S1P1 agonism was fully reversed by S1P1 antagonism. Manipulation of the S1P-S1P1 rheostat by pharmacological antagonism or agonism is of potential medical importance; for example, transient changes in the permeability of biological barriers, such as in the brain vasculature, may enhance delivery of drugs to otherwise inaccessible compartments. However, therapeutic attempts at antagonism need to be approached cautiously, as S1P1 antagonism induces lung capillary leakage and may be implicated in renal reperfusion injury and cancer angiogenesis28, 29, 30.
In the lung, we have shown that Compound 2a enhances leakage of marker dye under basal conditions when applied through the vasculature but not when applied via the airway. Thus, the constitutive integrity of the pulmonary capillary bed seems to depend on tonic plasma concentrations of S1P acting through S1P1 receptors that are accessible exclusively from the vasculature. Furthermore, in conjunction with previous findings, these data illuminate the mechanism by which the spatial distribution of different S1P receptor subtypes having opposing actions to physiological ligands may function as a biological rheostat to regulate lung barrier function. Whereas we show here that S1P1 agonism maintains capillary integrity and that antagonism breaches the pulmonary barrier, our earlier findings demonstrate that S1P3 agonism on pulmonary epithelium causes dissolution of tight junctions in vivo, resulting in acute pulmonary edema4—an effect that is abolished by S1P3 deletion.
By applying defined and selective chemical agonists (and now antagonists), in conjunction with genetic and compartmental approaches, to study the functions of S1P receptor subtypes, we have developed a model of the integrated spatial control of vascular physiology and pathology in the lung (Fig. 7a). Vascular S1P basal tone is likely to be important in protection from acute pulmonary edema and acute respiratory distress syndrome (ARDS) and may be an independent variable of potential use in determining the subset of patients susceptible to adult respiratory distress syndrome. Selective S1P1 agonists therefore hold promise as therapeutic agents for enhancing lung barrier function as a treatment of pulmonary edema and ARDS31, 32, 33.
Figure 7: Models of the vascular and lymphoid effects of S1P1 agonists and antagonists.
(a) Agonism by the physiological ligand S1P (1) of endothelial cell S1P1 receptors leads to the maintenance of adherens junctions26 in both skin24 and lung capillaries25. Alteration of endothelial junctional integrity by VEGF induces leakage of plasma proteins into interstitial spaces24 and is reversed by pharmacologically selective S1P1 agonist SEW2871. Competition for endogenous 1 activation of S1P1 by 2a induces plasma protein leakage in both skin capillaries and lung. (b) Lymphoid effects of S1P1 antagonist 2a are shown for lymph node (top) and thymus (bottom). Egress from both lymph node and thymus requires migration from basolateral to apical aspect of sinus lining or vascular endothelium in lymph node or thymus, respectively5; S1P agonists inhibit both processes and enhance organization of sinus endothelial junctions in lymph nodes26. Lymph node egress is maintained in the presence of 2a alone, and also in the presence of physiological concentrations of S1P, and it is inhibited by pharmacological agonist 3 (ref. 2). Medullary lymphocyte arrest in nodes that is induced by 3 is reversed by antagonists19, including 2a. Blood lymphocyte numbers are maintained by 2a, whereas agonist induces lymphopenia. In the thymus, 2a prevents the CD69 intermediate CD4+ or CD8+ single-positive thymocyte downmodulation of the surface lectin and retention signal CD69 in response to agonist14, 18. There is no direct evidence for S1P1 regulation of thymic egress through thymic endothelium.
Full size image (95 KB)S1P1-receptor regulation of lymphocyte trafficking
Our in vivo studies using Compound 2a indicate that physiological S1P concentrations do not measurably alter constitutive lymphocyte recirculation, whereas pharmacological elevation in medullary S1P1 tone by the S1P agonist Compound 3 does induce lymphocyte sequestration (Fig. 7b). The critical role of pharmacological S1P1 tone seems to be localized within the medulla of the lymph node, as evident from the rapid restoration of medullary T-cell motility and from the application of antagonist in the continued presence of agonist (Fig. 6e; Supplementary Video 1); by contrast, these agents were without effect, individually or together, in the cortical T-cell zone. Notably, Compound 2a applied alone caused no change in medullary T-cell movement or in blood lymphocyte numbers or expression of CD69 in thymic lymphocytes. However, it is effective in antagonizing (i) agonist-induced lymphopenia that is secondary to inhibition of lymphatic sinus egress2, 19 and (ii) S1P1 agonist–induced phenotypic maturation (CD69 expression) of medullary thymocytes14, 18. Thus, the homeostatic concentration of physiological S1P seems insufficient to modulate these functions, but they are affected by larger pharmacological deviations from the S1P-S1P1 set point.
The actions of the specific antagonist Compound 2a aid in distinguishing between different mechanisms proposed to account for regulation of lymphocyte egress in vivo by S1P1 receptors, namely intrinsic lymphocyte-receptor down modulation8, 34 versus stromal gating of lymphatic sinus barrier1, 19. Extrapolations from S1P1 gene deletion studies have been used to argue that chemical agonists work as 'functional antagonists' to cause downregulation of lymphocyte S1P1 receptors that normally mediate chemotaxis out of the node and into the circulation along a gradient of S1P8, 35. This model predicts that a systemic antagonist should mimic the actions of an agonist, resulting in sequestration of lymphocytes in the node. However, this was not apparent: only the agonist, but not antagonist, induced and maintained lymphopenia (Fig. 5). Thus, we conclude that a gradient of S1P-S1P1 activation within lymphoid organs is not necessarily a point of regulation of constitutive egress35, although we cannot rule out the possibility that a small, fractional priming signal in lymphocytes that is derived from small numbers of S1P1 receptors is biologically relevant. Furthermore, neither S1P1 agonists14, 18 nor S1P1 antagonists replicate the developmental phenotypes of S1P1 deletion (namely the CD69 upregulation in single-positive medullary thymocytes seen in S1P1-null lymphocytes8), and neither naive B cells nor naive T cells express CD69. Thus, acute sequestration of naive lymphocytes by S1P1 agonists cannot be explained by functional antagonism of S1P1 or by a direct interaction between CD69 and S1P1 (ref. 12).
Observations of agonist-dependent sequestration of naive lymphocytes in Peyer's patch and lymph nodes, but not in splenic white pulp, suggest involvement of a stromal element functioning as a physical 'gate' to control lymphocyte trafficking over and above intrinsic lymphocytic mechanisms5, 13, 36. This mechanism is supported by two-photon microscopy in lymph node explants19, which shows that lymphocyte movement across putative stromal 'portals' into lymphatic sinuses is constitutive and unaffected by an S1P1-specific antagonist, ceases in the presence of S1P1 agonist and recovers after addition of antagonist, observations consistent with the closing of stromal portals by the agonist. The intravital imaging data now presented, which show a system with fully intact vascular and lymphatic connections (Fig. 6), further validate this interpretation by replicating data for the actions of agonists on medullary (but not cortical) T-cell velocity loss and for antagonist-induced reversal and restoration of putative movement into sinus spaces, although we were unable to directly visualize trans-sinusoidal migration because in vivo labeling of the sinus barrier was not as clear as in the explant preparation.
Chemical biology of S1P signaling
S1P1-mediated signaling provides a clear example of acute perturbations of biochemical activity through selective chemical approaches producing divergent results from gene deletion studies1, 18. In another recent example37, selective, functional chemical probes of granzymes A and B suggest a different hierarchy, and a distinct contribution of granzyme B to the biological mechanism of target-cell lysis by cytolytic T cells, from what had been deduced from prior gene deletion experiments. The possibility of developmental and compensatory changes following gene deletions complicates interpretation of results from such studies and requires that care be taken when extrapolating mechanisms from gene deletion studies to chemical biology and vice versa. However, a synergistic approach using both methodologies is likely to be profitable. Each interrogates different components of the integrated systems physiology by mechanisms that may turn out to be orthogonal.
The chemical tractability of the S1P-S1P1 signaling axis presages rapid progress in therapeutic applications across immune, vascular and perhaps cancer biology. Potent, selective, reversible chemical probes of complex systems function, such as the S1P1 antagonist and agonist combinations shown here, provide new, integrated insights into the organization and regulation of both the basal physiology and induced pathophysiology of such complex systems. Such probes also move fields forward across multiple disciplines by making highly reproducible experimental tools generally available.
Methods
Chemical synthesis.
The detailed synthetic route and full chemical characterization of Compound 2a and Compound 2b are shown in Supplementary Methods online.
S1P receptor agonists.
We purchased 1 from BioMol and 3 (5-(4-phenyl-5-trifluoromethyl-thiophen-2-yl)-3-(3-trifluoromethyl-phenyl)-1,2,4-oxadiazole) from Maybridge.
Cell culture and transfection.
D. Guerini (Novartis Pharma) provided CHO cells stably expressing human S1P receptors (S1P1-CHO and S1P3-CHO cells), and we maintained them in RPMI 1640 (Life Technologies, Gibco) containing 10% FBS, 0.5 mg ml-1 of the antibiotic G418, 50 g ml-1 gentamicin, 100 units ml-1 penicillin, 100 g ml-1 streptomycin and 10 mM HEPES. We transiently transfected the murine S1P1 receptor provided by J. Chun in CHO cells as described2.
Western blotting.
We cultured control CHO cells and CHO cells stably transfected with human S1P1 or S1P3 to 50% confluence on six-well plates in complete RPMI 1640 supplemented with 10% FBS. We serum starved cells for 16 h and stimulated them with S1P or SEW2871 diluted to various concentrations in the serum-free medium with 0.1% fatty acid–free BSA. At 5 min, 1 h, or 4.5 h, we lysed cells in 50 mM Tris, pH 8.0, 125 mM NaCl, 20 mM CHAPS, 2 mM dithiothreitol, 1 mM EDTA, 2 mM Na3VO4, 10 mM NaF, 1 mM PMSF and protease inhibitor cocktail (Roche). We analyzed cell lysates by western blotting after separating them on 10% SDS-PAGE using mouse monoclonal antibody to phospho-ERK1/2 (sc-7383; Santa Cruz Biotechnology) and rabbit polyclonal antibody to phospho-Akt (BD Biosciences). We detected total ERK1 and ERK2 using a rabbit affinity-purified polyclonal antibody to ERK (sc-94; Santa Cruz Biotechnology), and we detected total Akt using a rabbit affinity-purified polyclonal antibody to Akt1 (BD Biosciences). We quantified band intensities corresponding to pERK1, pERK2 and pAkt by image analysis (Kodak 1D Scientific Imaging Systems), and we normalized amounts of pERK1/2 and pAkt for the total amounts of ERK1/2 and Akt.
Membrane preparation and receptor activation assay.
We prepared membranes from CHO cells expressing human S1P1, S1P2, S1P3 and S1P5 for use in ligand and 35S-GTP-S binding studies as previously described2. For 35S-GTP-S binding assays, we added serial dilutions of 1 or 3 to membranes (1–10 g protein per well) that were or were not preincubated for 30 min; we used the concentrations of Compound 2a and Compound 2b (and assayed them) as previously described2.
Assays of capillary leakage.
We assayed pulmonary leakage by measuring accumulation of plasma protein–bound Evans blue dye in the airways of lungs of C57BL1/6J mice treated with vehicle (Compound 2a or Compound 2b) that we solubilized according to previous descriptions38. We injected Evans blue dye intravenously 75 min after administering the compounds, and we perfused the lung vasculature to remove intravascular Evans blue dye before photography and dye quantification as described4. We performed the Miles assay39 on the shaved dorsal skin of Balb/cByJ mice. We treated male mice weighing 30 g with vehicle (10 mM Na2CO3, 2% 
-hydroxypropyl-cyclodextran) or Compound 2a (10 mg kg-1) intraperitoneally (i.p.), and then we orally dosed them with Compound 3 (20 mg kg-1) or its vehicle. After 4.5 h, we injected the mice via the tail vein with 100 l of a 0.5% Evans blue dye solution (Sigma) in saline. We then lightly anesthetized the mice with 2,2,2-tribromoethanol and adminstered murine VEGF165 (100 ng; Calbiochem) or vehicle (PBS in 0.1% BSA) by intradermal injection in triplicate in rows. We assessed the increase in vascular permeability 30 min later by measuring leakage of Evans blue dye into the injected site. After killing the mice, we biopsied the intradermal injection sites (3.5 mm, Sklar Tru-Punch) and incubated these in 100 l of formamide at 58 °C to extract Evans blue dye. We read the absorbance of the extracts at 565 nm in a spectrophotometer.
Induction of lymphopenia in mice.
We dosed C57BL6 mice orally with 10 mg kg-1 of Compound 3 or vehicle (10% DMSO, 25% Tween20, v/v) and collected blood into EDTA tubes (Becton-Dickinson). We used a veterinary autoanalyzer calibrated for mouse blood (H2000, Careside) to make full blood counts at times stated previously2. All animal studies were approved by the Institutional Animal Care and Use Committee. We removed thymi as described and assayed the expression of CD69 on single-positive CD4+ thymocytes by fluorescence flow cytometry as described14, 18.
Intravital imaging of lymphocyte migration by two-photon imaging.
We performed adoptive transfer of CD4+ T cells labeled with 5- and 6-carboxyfluorescein diacetate succinimidyl ester (Molecular Probes) as previously described19, 27 and analyzed the data as described19. We carried out microdissection to remove fat tissue that normally obscures the medullary region, and we performed imaging only in cases in which surrounding blood vessels were intact and rapid blood flow through large and small vessels could be seen under the light microscope. We visualized blood flow using video-rate two-photon imaging after intravenous injection of FITC-conjugated dextrans (Supplementary Video 1). To stain the region of the medullary sinus, we injected dextrans conjugated with 20 kDa rhodamine into the footpad and skin of the ipsilateral flank after securing the imaging chamber and dissecting surrounding fatty tissue. In separate experiments, the rapid (5 min) appearance of fluorescent dextrans in the node after a single footpad injection confirmed that lymphatic circulation was intact. We anesthetized mice that had been orally treated with 10 mg kg-1 of Compound 3 and imaged the inguinal lymph node between 4 and 6 h after dosing with Compound 3. We identified the medullary region using the large Y-shaped blood vessels overlaying the node as a landmark, and we performed medullary imaging out within 200 m of the edge of the larger blood vessels. We obtained three-color images of the cortical region showing B cells in a follicle (labeled with 4-chloromethyl-6,8-difluoro-7-hydroxycoumarin, or CMF2 HC), T cells (labeled with 5-(and-6)-(((4-chloromethyl)benzoyl)amino)-tetramethylrhodamine, or CMTMR) and high endothelial venule (labeled intravenously with 70 kDa FITC-conjugated dextran).
Plasma measurements and pharmacokinetics.
We made plasma measurements in mice 5 h after i.p. injection of 10 mg kg-1 of Compound 2a, and we obtained detailed pharmacokinetic parameters after intravenous injection of 1 mg kg-1 of Compound 2a in cannulated rats (n = 3) as described in detail9, 10.
Statistical analysis.
We performed all statistical analyses using GraphPad Prism 4 software, and we determined EC50 and Ki values by nonlinear regression and Schild analysis. We performed statistical analyses between groups of experimental animals using Anova and the Bonferroni Multiple Comparisons Test. We performed noncompartmental pharmacokinetic analysis of plasma levels using PK Solutions 2.0 software (Summit Research Services).
Note: Supplementary information is available on the Nature Chemical Biology website.
Author contributions
M.G.S., P.J.G.-C., A.D., D.M., E.J. and H.R. participated in the design and execution of the biological experiments. S.-K.W., W.-C.C. and C.-H.W. participated in the synthetic chemistry. M.P.M., S.H.W., I.P., M.D.C. and H.R. contributed to the design and execution of the multiphoton microscopy experiments. H.R., I.P., C.-H.W., M.D.C. and M.G.S. wrote the manuscript.
