Lysophosphatidic acid (LPA; 1-acyl-3-phosphoglycerol) exerts its biological activity through both extracellular and intracellular targets. Receptor targets include the cell-surface G-protein-coupled receptors LPA1–4 and the nuclear PPAR-γ (peroxisome-proliferator-activated receptor γ). Enzyme targets include the secreted cancer cell motility factor, autotaxin, and the transmembrane phosphatases, LPP1–3 (where LPP stands for lipid phosphate phosphatase). Ion channel targets include the two pore domain ion channels in the TREK family, TREK-1, TREK-2 and TRAAK. Structural features of these targets and their interactions with LPA are reviewed.
Lysophosphatidic acid (LPA; 1-acyl-3-phosphoglycerol) is a bioactive lipid with important roles in physiology and pathophysiology. Our understanding of the biological roles of LPA is growing rapidly. LPA action through its cell-membrane target, the LPA3 receptor, was recently determined to influence embryonic implantation and spacing through knockout studies in mice . LPA action through a second cellmembrane target, LPA1, is involved in the neuropathic pain response . LPA also influences cancer cell motility through feedback inhibition of the serum lysophospholipase D [3,4] and the cancer motility factor [5–8] ATX (autotaxin) . The proliferative and anti-apoptotic effects of LPA additionally affect cortical development and folding [10–14]. Finally, LPA was shown to promote neointima formation through its intracellular target, PPAR-γ (peroxisome-proliferator-activated receptor γ) , highlighting its importance in the early development of cardiovascular disease. Further studies of LPA biological targets promise to reveal additional physiological and pathophysiological roles of LPA.
Early reports of specific biological targets of LPA centred on cell-surface receptors [16–18], but more recent reports identify nuclear receptors [15,19,20], potassium channels  and enzymes  as additional targets of LPA action. Signalling responses mediated by LPA-metabolizing enzymes in the LPP (lipid phosphate phosphatase) family have also been reported [22,23]. Actions of LPA at these targets can additionally be modulated by interactions with serum proteins [15,24].
LPA activates four cell-surface receptors in the GPCR (G-protein-coupled receptor) family [25–28]. Three of the receptors, LPA1/EDG2/vzg-1 (where EDG stands for endothelial differentiation gene) [29,30], LPA2/EDG4  and LPA3/EDG7 [32–34] are members of the EDG receptor family. The fourth receptor, P2Y9/LPA4 , is more closely related to the purinergic receptors. LPA has also been demonstrated to bind to and activate a nuclear receptor, PPAR-γ.
GPCRs are the most extensively studied targets of LPA at this time. The LPA receptors in the EDG subfamily share 45–50% of their amino acids. LPA4/p2y9/GPR23 is an LPA receptor sharing less than 17% amino acid identity with LPA1–3 . Several structural characteristics of these receptors are known based on general GPCR topology [36,37]. These characteristics include an extracellular N-terminus, seven α-helical TMs (transmembrane domains), and an intracellular C-terminus. Comparison of GPCR structures is facilitated by the use of the index numbering system first described by Ballesteros and Weinstein . In their system, residues are noted first by the TM in which they occur and then based on their position in that TM compared with a highly conserved residue assigned to position 50 in the GPCR superfamily. Thus P7.50 is the conserved proline residue in TM7 and N7.49 is the immediately preceding residue. Additional structural features of LPA receptors in the EDG family based on modelling studies have been reported. First, selective recognition of LPA versus sphingosine 1-phosphate is conferred by a conserved glutamine, rather than glutamate, residue at the extracellular end of the third TM, Q3.29 [39,40]. Secondly, a conserved arginine residue immediately preceding this selectivity switch is required for agonist binding to all of the EDG receptors tested so far [41–43]. Nevertheless, mutagenesis studies on EDG receptor family members highlight some key differences among LPA1–3. In particular, mutagenesis studies of S1P1 (R7.34A)  and S1P4 (R7.33A)  demonstrate that a cationic residue in TM7 is important for phospholipid binding and activity at some, but not all, EDG receptors. LPA3 has cationic residues at positions 7.35 and 7.36, and LPA1 and LPA2 have anionic residues at 7.35 and cationic residues at 7.36. Mutations are needed in order to define the importance of these residues in LPA action at these receptors. In fact, recent mutagenesis studies of LPA3 indicate that the cationic residue at position 7.35, but not that at 7.36, is involved in agonist binding [44a]. A cationic residue at position 7.35 is unique to LPA3, thus these results highlight a distinctive aspect of agonist recognition. Additionally, S1P4 mutagenesis  demonstrates the importance of a cationic residue in TM5, which is conserved in all EDG receptors except LPA1. Figure 1 shows aligned segments of LPA1–3 highlighting these similarities and differences. It is intriguing to note that the LPA4 receptor shares none of these amino acids and thus recognizes LPA by a completely different motif. Finally, modelling studies suggest that antagonists of LPA action at LPA receptors occupy a binding site that overlaps the agonist-binding site only through interactions with cationic residues at the top of the agonist-binding pocket [25,45,46]. Modelled antagonist structures occupy a pocket within the extracellular loops.
Alignment of LPA1–3 receptor TM3, TM5 and TM7
PPAR-γ is a nuclear receptor whose activation depends upon stabilization of the AF-2 (activator function-2) helix near the C-terminus of the ligand-binding domain . Several crystallographic structures of PPAR-γ and its complexes with activating ligands in the thiazolidinedione family have been published [48–52].
LPA is produced in serum upon hydrolysis of lysophosphatidyl choline  by serum lysophospholipase D, recently identified as ATX [3,4]. LPA acts as a mixed-mode feedback inhibitor of this enzymatic process . ATX is a member of the AlkP (alkaline phosphatase) superfamily of enzymes . This superfamily includes phosphatases, phosphomutases, sulphatases, phosphotransferases and the nucleotide pyrophosphatases/phosphodiesterases [54,55]. Overall sequence similarity in this group of proteins is low, i.e. <15% sequence identity. However, crystallographic structures of several members of the AlkP superfamily demonstrate a highly conserved structural core of six to eight β-sheets surrounded by up to ten α-helices [56–60], five of which can be perfectly superimposed (Figure 2). AlkP superfamily enzymes are metalloenzymes, with between one and three metal ions within the catalytic site [56–59]. These metal ions interact with conserved histidine and aspartate residues, which in several members of the family have been verified by mutagenesis as critical for enzyme function . In particular, glutamine replacements of His316, His360 and His475 in ATX abolish both phosphodiesterase and lysophospholipase D activity . Mutagenesis of rat ATX additionally identified Asp309 and Asp169 as additional metal-co-ordinating residues . The metal ion or ions are located near a conserved catalytic residue, which can be a threonine, serine or cysteine residue [56–59]. The catalytic residue is covalently modified (phosphorylated or sulphated) during the catalytic cycle before donating the phosphate or sulphate to an oxygen acceptor (water or alcohol) . Mutation of Thr210 of ATX abolished phosphodiesterase and lysophospholipase D activity [61,64]. Photochemical labelling of ATX with an ATP analogue resulted in modification of a peptide containing Thr210, providing further evidence of its role as the catalytic residue . Inactivated ATX mutants T210A (Thr210→Ala) and T210D failed to promote cancer cell motility, in contrast with wild-type ATX . The demonstrated requirement for catalytically active ATX to stimulate cancer cell motility indicates that ATX inhibition should be explored in cancer treatment.
Superposition of AlkP superfamily members showing a geometrically conserved structural core
LPP enzymes catalyse phosphate hydrolysis of LPA and related phospholipids [66–70]. The LPP family includes three proteins, LPP1–3. LPP1–3 are membrane proteins with six TMs, an extracellular catalytic site and intracellular termini. Catalytic residues are found in the second and third extracellular loops. The catalytic site residues show significant similarity to those found in the soluble enzyme, vanadate chloroperoxidase [71–73]. The role of LPP enzymes in LPA signalling has recently been reviewed .
The influence of intracellular LPA on the two-pore domain potassium (K2P) channels in the TREK subfamily, TREK-1, TREK-2 and TRAAK, is exciting since this is the first report of LPA action on an ion channel . TREK subfamily members, with 45–65% amino acid identity, have a topology consisting of four TMs and two pore domains that occur after the first and third TMs . LPA was found to reversibly open TREK-1 when applied to inside-out, but not whole cell or outside-out, patch configurations . TREK-1 activation by LPA was not additive with acidosis and was nearly eliminated by mutation of the pH-sensitive residue, Glu306, to alanine. While LPA activation of TREK-1 was not dependent on the cytosolic N- and C-termini, a specific interaction of LPA with TREK-1 has not been confirmed. As Chemin et al.  indicate, the LPA-induced K2P channel stimulation might additionally be attributed to an indirect mechanical membrane effect of LPA.
The biological targets currently known to mediate LPA action paint only a part of the picture. Further studies elucidating the identity and roles of proteins modulating LPA activity in biological fluids will provide added depth and clarity to our understanding of LPA action. Methods applicable to the structural characterization of membrane proteins will additionally provide a valuable contrast with our understanding of LPA interactions with its varied biological targets. Currently available structural information can now guide the identification and optimization of target-selective pharmacological tools and therapeutic leads.
Cellular Information Processing: A Focus Topic at BioScience2005, held at SECC Glasgow, U.K., 17–21 July 2005. Edited by F. Antoni (Edinburgh, U.K.), C. Cooper (Essex, U.K.), M. Cousin (Edinburgh, U.K.), A. Morgan (Liverpool, U.K.), M. Murphy (Cambridge, U.K.), S. Pyne (Strathclyde, U.K.) and M. Wakelam (Birmingham, U.K.).
Support from the American Heart Association (0355199B), the NIH (National Institutes of Health; 1 RO1 CA92160-01) and the Chemical Computing Group is gratefully acknowledged.