Sphingosine-1-phosphate (S1P) and lysophosphatidic acid (LPA) are phospholipid growth factors which are present in normal serum and plasma. These lipids elicit diverse responses from a wide range of cell types, including enhanced cell survival, cell proliferation, induction of cytoskeletal changes and chemotaxis (reviewed in 1234. Some of these responses reflect activation of G protein-coupled receptors of the endothelial differentiation gene (Edg) family. The Edg receptor family is classified into two clusters based on ligand selectivity: S1P_1/2/3/4/5 (formerly Edg1/5/3/6/8) specifically respond to S1P whilst LPA_1/2/3 (formerly Edg2/4/7) respond to LPA 5. Members of the S1P receptor family display higher sequence similarity to each other (approximately 40% identity) than to members of the LPA receptor family (approximately 30% identity). These homologies, coupled with observed differences in the structure of S1P and LPA receptor genes, suggest that these receptor families evolved from distinct ancestral genes. The S1P receptors contain a conserved glutamic acid residue present within the third TM that corresponds to glutamine in the LPA receptors. Interaction between distinct functional groups present on S1P and LPA with this residue was shown for the S1P₁ and LPA₁ receptors using computational modelling techniques 67 and was demonstrated as the basis for the ligand preference displayed by the receptors. Experimental characterisation confirmed that replacement of glutamic acid with glutamine in S1P₁ changed ligand specificity from S1P to LPA, and the reciprocal mutation in LPA₁ resulted in recognition of both LPA and S1P 7.

In the present study, the role of this residue in determining ligand selectivity for the S1P₄ receptor was examined. Phylogenetic analysis of the Edg family of receptors indicates that S1P₄ is more closely related to other S1P receptors than receptors which respond to LPA. However, S1P₄ lies on the edge of the S1P family cluster and has been shown to bind S1P with lower affinity than other S1P receptors and hence it has been suggested that S1P is not the true endogenous agonist of this receptor 8. We therefore decided to investigate whether replacement of this residue (E^3.29(122)) with glutamine conferred LPA-responsiveness to the S1P₄ receptor and hence determine the role of this residue in this lower-affinity S1P receptor. To achieve this, we expressed wild type and E^3.29(122)Q mutant S1P₄ receptors in CHO-K1 cells and studied responses to lysophospholipids using a [³⁵S]GTPγS binding assay. Since CHO-K1 cells respond to LPA, we utilised fusion proteins constructed between the S1P₄ receptor and a pertussis toxin-insensitive Gα_i1(C³⁵¹I) G protein. Expression of these proteins in CHO-K1 cells followed by treatment with pertussis toxin prior to harvest allowed elimination of any signal due to stimulation of endogenous LPA receptors. Within this study, we also examined how the length of the LPA acyl chain affected potency at the mutant S1P₄ receptor, using a panel of naturally occurring LPA analogues. Computational models of complexes between the wild type or mutant S1P₄ receptor and S1P and LPA species were used to provide a molecular interpretation of the experimental findings.

Human HA-S1P₄ was mutated at position 122 to replace the naturally occurring glutamic acid with glutamine. The mutant and wild type receptors were stably expressed in CHO-K1 cells as in-frame GPCR-G protein fusions with pertussis toxin-insensitive Gα_i1(C³⁵¹I). Western blotting was used to detect expression of these fusion proteins. Membranes from HA-S1P₄-Gα_i1(C³⁵¹I)- or HA-S1P₄(E^3.29(122)Q)-Gα_i1(C³⁵¹I)-transfected cells contained a polypeptide with an apparent molecular mass of approximately 110 kDa, which reacted with anti-HA and anti-SG1 antibodies (Figure 1A and 1B) and was consistent with expression of the GPCR-G protein fusion. Confirmation of comparable cell-surface expression of these proteins was obtained via FACS analysis using an anti-HA antibody directly conjugated with fluorescein (Figure 1 Panel III).

The response of HA-S1P₄-Gα_i1(C³⁵¹I) and HA-S1P₄(E^3.29(122)Q)-Gα_i1(C³⁵¹I) to S1P was assessed using membranes from cells transfected to express these proteins and treated with pertussis toxin prior to harvest. S1P promoted dose-dependent increase in [³⁵S]GTPγS binding to membranes containing HA-S1P₄-Gα_i1(C³⁵¹I) with an EC₅₀ of 355 nM ± 155 nM (n = 3); in contrast, membranes from HA-S1P₄(E^3.29(122)Q)-Gα_i1(C³⁵¹I)-expressing cells demonstrated severely impaired response to S1P (Figure 2A). The EC₅₀ for S1P stimulation of the HA-S1P₄-Gα_i1 fusion protein (355 ± 155 nM) was not statistically significantly different from that obtained using the unfused HA-S1P₄ receptor (439 ± 187 nM, Figure 2B) and compared favourably with published values for this receptor in HEK293T cells from two different research groups of 270 nM 9 and 790 nM 10.

Structure activity relationships were determined by the [³⁵S]GTPγS assay for S1P₄ or its E^3.29(122)Q mutant using S1P and LPA species with 14:0, 16:0 and 18:1 acyl chains at a single (10 μM) concentration. Of the lysophospholipids tested, only S1P induced a strong response over basal levels (approximately 48% ± 5%) in membranes containing HA-S1P₄-Gα_i1 (Figure 3A), whilst 18:1 LPA did not stimulate a statistically significant response; weak stimulation of [³⁵S]GTPγS binding was observed with 14:0 and 16:0 LPA (approximately 11% ± 3% above basal in each case). In contrast, membranes containing HA-S1P₄(E^3.29(122)Q)-Gα_i1(C³⁵¹I) showed at least weak response to each ligand (Figure 3A). The weakest agonist at the E^3.29(122)Q mutant S1P₄ receptor was 18:1 LPA, which produced only 14% ± 2% stimulation over basal. S1P and 16:0 LPA gave approximately 21% ± 4% and 19% ± 4% stimulation over basal response, respectively. The best agonist for the E^3.29(122)Q mutant was 14:0 LPA, which gave a 40% ± 2% enhancement over basal levels. The stimulation promoted by 14:0 LPA was statistically different from that produced by 18:1 LPA (p < 0.01). Sensitivity of the receptor to stimulation by each form of LPA, and particularly 14:0 LPA, was markedly increased after introduction of the E^3.29(122)Q mutation and indicated that this position was important in influencing HA-S1P₄ ligand preference.

These results indicate that introduction of the E^3.29(122)Q mutation in the S1P₄ receptor confers LPA-responsiveness, and that a short form of LPA was a more effective agonist than the intermediate and longer forms, when tested at this single concentration. Dose response curves were constructed for ligand-induced activation of the E^3.29(122)Q S1P₄ mutant by the 14:0 and 18:1 forms of LPA as well as S1P (Figure 3B). An EC₅₀ could only be determined for the 14:0 form of LPA as S1P and 18:1 LPA caused minimal stimulation at only the highest concentration tested. The EC₅₀ value for activation of HA-S1P₄(E^3.29(122)Q)-Gα_i1(C³⁵¹I) was calculated to be 3.8 ± 1.4 μM. However, since a plateau of maximal stimulation was not achieved, interpretation of this EC₅₀ value needs caution. This result clearly showed that 14:0 LPA was a weak agonist of HA-S1P₄(E^3.29(122)Q)-Gα_i1(C³⁵¹I) and hence confirmed the involvement of residue 122 in S1P₄ ligand preference. Similar results were obtained using a second CHO-K1 clone expressing this fusion protein (not shown).

Computational modelling of the S1P complex with the wild type S1P₄ receptor identifies the best S1P binding site within the TM with the phosphate group at the extracellular end (Figure 4A). Ion pairs appear between the phosphate group of S1P and two cationic amino acids, R^3.28(121) and K^5.38(202). An additional ion pair occurs between the cationic ammonium of S1P and E^3.29(122). Hydrophobic residues from TM2, TM3, TM5 and TM6 line the binding pocket and surround the alkyl chain of S1P.

The best complex of 14:0 LPA in the E^3.29(122)Q S1P₄ mutant receptor model has striking similarity to the best complex of S1P in the wild type S1P₄ receptor model (Figure 4B). Both models demonstrate ion pairing between the phosphate group and two cationic amino acids, R^3.28(121) and K^5.38(202). Each ligand interacts with the amino acid at position 3.29(122), S1P by an ion pair with the carboxylate of the wild type glutamate and 14:0 LPA by a hydrogen bond with the mutated glutamine. Multiple hydrophobic residues surround the nonpolar tails of the lipid ligands. The superimposition of the two complexes (Figure 4B) also demonstrates that the ligands occupy almost identical volumes. Common interactions and overlap volumes are qualitatively consistent with the experimental findings that these ligands give similar 48% and 40% maximal stimulation over basal for S1P at the wild type and 14:0 LPA at the mutant receptor, respectively.

In contrast to the complexes of 14:0 LPA with E^3.29(122)Q S1P₄ and S1P with wild type S1P₄, the remaining complexes show much less common volume (Figure 4C). Most complexes exhibit the phosphate interactions described for 14:0 LPA with E^3.29(122)Q S1P₄ and S1P with wild type S1P₄. Of particular interest is the observation that the best complexes generated by Autodock for the 18:1 LPA species with wild type S1P₄ has a very high positive van der Waals interaction energy, > 3000 kcal/mol, compared to values well under 200 kcal/mol for every other complex studied. In the best complexes found for 16:0 LPA and 18:1 LPA in both constructs, the terminal six to eight carbons of the hydrophobic tails fold into L-shaped conformations quite different from the extended conformations observed in the S1P complex with wild type S1P₄ or the 14:0 LPA complex with the E^3.29(122)Q S1P₄ mutant. The terminal carbons in several complexes curl out of the receptor between TM5 and TM6 (Figure 4C) due to the restricted length of the binding pocket. These results suggest that the complete lack of S1P₄ activation in response to 18:1 LPA is likely due to failure to form a complex. The strongest complexes formed, S1P with wild type S1P₄ and 14:0 LPA with the E^3.29(122)Q S1P₄ mutant, have complementary interactions with the residue at position 3.29(122). These strong complexes give the most robust activation. Weak complexes are formed for other combinations due to mismatched interactions with position 3.29(122) or excessive length of the hydrophobic tail. The presence of hydrophobic tails of 16:0 or 18:1 LPA between transmembrane domains may additionally impair the conformational change necessary for full agonist responses.

Parental CHO-K1 cells respond to LPA in functional assays, reflecting expression of endogenous LPA₁ (G. Holdsworth, et al., manuscript in preparation). For this reason, fusion proteins between wild type or mutant HA-S1P₄ and the pertussis toxin-insensitive Gα_i1(C³⁵¹I) G protein were used in these studies. Expression of these proteins in CHO-K1 cells followed by treatment with pertussis toxin prior to harvest allowed elimination of any signal due to stimulation of endogenous LPA receptors. McAllister et al. 11 (.(adopted a similar approach for studies of the LPA₁ receptor.

We examined the role of residue E^3.29(122) in controlling S1P₄ ligand selectivity using functional and computational methods. This residue, which is conserved throughout the S1P receptors, has been shown to control ligand specificity for the related S1P₁ receptor 7. Introduction of the E^3.29(122)Q mutation severely affected the response of S1P₄ to S1P: in dose-response experiments S1P caused minimal stimulation at only the highest concentration of ligand used. This is in agreement with published observations for activation of the equivalent S1P₁ mutant 7. 14:0 LPA was able to induce dose-dependent stimulation of S1P₄(E^3.29(122)Q) with an EC₅₀ of approximately 3.8 μM but only promoted minimal stimulation of the wild type S1P₄ receptor. The modelled complexes of 14:0 LPA with E^3.29(122)Q S1P₄ and S1P with wild type S1P₄ demonstrate nearly identical volumes occupied by the two ligands and very similar interactions between these ligands and their respective receptors. Of particular importance are amino acid residues at positions 3.28(121), 3.29(122) and 5.38(202), which either ion pair with the phosphate or interact with the 2-amino or 2-hydroxyl group in S1P and 14:0 LPA, respectively. The importance of interactions with amino acids at positions 3.28 and 3.29 has been previously noted for the S1P₁ 67 and LPA_1,2,3 712 receptors.

The S1P₄ receptor exhibits marked constitutive (agonist-independent) activity (Figure 5) which was unaffected by the introduction of the E^3.29(122)Q mutation (data not shown). This indicates that the mutation perturbs S1P recognition without affecting the ability of the receptor to spontaneously adopt an active conformation. Similar observations have been reported for the β₂AR, where a mutation in the sixth transmembrane domain abolished agonist activation but not constitutive activity 13.

Unlike S1P, which exists as a single species in vivo, the term LPA actually refers to a family of molecules that take the general form 1-o-acyl-2-hydroxy-sn-glyceryl-3-phosphate. Naturally occurring forms of LPA contain acyl chains of differing lengths, with differing degrees of saturation. Investigations into the effect of the length and degree of saturation of the acyl chain of LPA have been undertaken for the LPA receptors 1415, but limited SAR information is available for S1P receptors (22). The LPA-responsive E^3.29(122)Q S1P₄ mutant facilitates structure activity relationship (SAR) studies due to the greater availability of LPA analogs relative to S1P analogs. Comparison of space-filling models of the structures of S1P and three analogues of LPA (Figure 4D) revealed that 14:0 LPA most closely resembled S1P in terms of apparent length. [³⁵S]GTPγS binding assays demonstrated greater agonist activity of 14:0 LPA at the mutant receptor relative to 18:1 or 16:0 LPA. This SAR indicates a length restriction for the S1P₄ agonist binding site. Model complexes of 16:0 and 18:1 LPA contained alkyl chains that fold at the bottom of the binding pocket, defined by a cluster of hydrophobic amino acids. Three of these differ either in position of sidechain branching or size relative to LPA receptors and the other S1P receptors. Position 2.46, I88 in S1P₄, is leucine in LPA_1–3 and other S1P receptors. Residue I6.40(256) is larger than the valine found in the other four S1P receptors, LPA₁ and LPA₃. Finally, I7.51(305) corresponds to the smaller valine in S1P₂ and S1P₃ and the much smaller alanine in LPA₂. These findings provide a molecular explanation for a similar SAR observed using para-alkyl amide analogs of S1P 9. SAR obtained with the S1P₄ mutant are in contrast to that shown by LPA receptors, which exhibit the general trend of 18:1 ≥ 16:0 > 14:0 for potency and maximal stimulation 15.

Since mutation of residue 122 in the S1P₄ receptor from the naturally occurring glutamic acid to glutamine conferred responsiveness to 14:0 LPA and severely affected responses to S1P, our observations support the hypothesis that this conserved residue in the third transmembrane domain of the S1P receptors is involved in ligand recognition. This is in contrast to a recent paper describing models of several GPCRs, including S1P₄, which had been generated using novel first principle methods 16. In this model of S1P₄, interactions between S1P and residues T^7.34(127) and W^7.37(291) and E^7.30(284) were observed. Interaction of E^7.30(284) with the ammonium group of S1P appeared to control ligand selectivity since the other residues appeared to interact with the phosphate group, which is present on both LPA and S1P. It is therefore surprising that none of these residues are conserved throughout the S1P or LPA receptor families. The data presented here support the assertion that glutamic acid residue 3.29 present in the third transmembrane domain of the S1P receptors controls ligand selectivity and suggest that the S1P₄ model described by Vaidehi et al. 16 is inaccurate.

The current study provides new information for the development of more selective S1P receptor agonists. In particular, an S1P analog with its hydrophobic chain extended by either 2 or 4 carbons would be a very poor agonist of the S1P₄ receptor. On the other hand, the activation of the S1P₁-E^3.29(121)Q mutant by 18:1 LPA 7 indicates that a chain-extended S1P analog should retain agonist activity at the S1P₁ receptor. S1P receptor agonists with differing selectivity profiles will be useful tools to more completely map the physiological and pathophysiological roles of these receptors.

These studies confirm that glutamic acid residue 3.29, present in the third transmembrane domain of the S1P receptors is important for the selective recognition of S1P, versus the closely related lipid, LPA. Mutation of E3.29 to glutamine diminished response to S1P and allowed structure activity studies using the diverse available LPA species. The mutant S1P₄ receptor is stimulated most strongly by LPA 14:0 and is not activated by the longer LPA 18:1, in contrast with a previous report on the analogous S1P₁ receptor mutant that responded to LPA 18:1. Thus the S1P₄ receptor ligand binding pocket is shorter in length than the S1P₁ ligand binding pocket.

CHO, Chinese hamster ovary; Edg, endothelial differentiation gene; ERK, extracellularly regulated kinase; FACS, fluorescence activated cell sorter; G protein, guanine nucleotide-binding protein; GPCR, G protein-coupled receptor; HA, haemagglutinin; LPA, lysophosphatidic acid; MAP, mitogen-activated protein kinase; PBMC, peripheral blood mononuclear cell; PTx, pertussis toxin; SIP, sphingosine-1-phosphate; TM, transmembrane domain

G Holdsworth performed and interpreted all studies with experimental S1P₄ fusion proteins and drafted the manuscript. D Osborne performed and interpreted docking studies to generate all mutant complexes with all LPA species and S1P and all wild type complexes with LPA species. TC Pham generated the homology model of the human S1P₄ receptor. J Fells performed docking studies of S1P with the wild type S1P₄ receptor. G Hutchinson and G Milligan participated in the design and coordination of the experimental studies with S1P₄ fusion proteins. A Parrill participated in the design and coordination of the modelling studies and edited the manuscript.