Lysophosphatidic Acid-induced Mitogenesis Is Regulated by Lipid Phosphate Phosphatases and Is Edg-receptor Independent*

Lysophosphatidic acid (LPA) is an extracellular signaling mediator with a broad range of cellular responses. Three G-protein-coupled receptors (Edg-2, -4, and -7) have been identified as receptors for LPA. In this study, the ectophosphatase lipid phosphate phosphatase 1 (LPP1) has been shown to down-regulate LPA-mediated mitogenesis. Furthermore, using degradation-resistant phosphonate analogs of LPA and stereoselective agonists of the Edg receptors we have demonstrated that the mitogenic and platelet aggregation responses to LPA are independent of Edg-2, -4, and -7. Specifically, we found that LPA degradation is insufficient to account for the decrease in LPA potency in mitogenic assays, and the stereoselectivity observed at the Edg receptors is not reflected in mitogenesis. Additionally, RH7777 cells, which are devoid of Edg-2, -4, and -7 receptor mRNA, have a mitogenic response to LPA and LPA analogs. Finally, we have determined that the ligand selectivity of the platelet aggregation response is consistent with that of mitogenesis, but not with Edg-2, -4, and -7.

Lysophosphatidic acid (LPA) is an extracellular signaling mediator with a broad range of cellular responses. Three G-protein-coupled receptors (Edg-2, -4, and -7) have been identified as receptors for LPA. In this study, the ectophosphatase lipid phosphate phosphatase 1 (LPP1) has been shown to down-regulate LPAmediated mitogenesis. Furthermore, using degradationresistant phosphonate analogs of LPA and stereoselective agonists of the Edg receptors we have demonstrated that the mitogenic and platelet aggregation responses to LPA are independent of Edg-2, -4, and -7. Specifically, we found that LPA degradation is insufficient to account for the decrease in LPA potency in mitogenic assays, and the stereoselectivity observed at the Edg receptors is not reflected in mitogenesis. Additionally, RH7777 cells, which are devoid of Edg-2, -4, and -7 receptor mRNA, have a mitogenic response to LPA and LPA analogs. Finally, we have determined that the ligand selectivity of the platelet aggregation response is consistent with that of mitogenesis, but not with Edg-2, -4, and -7.
The structurally simple phospholipid lysophosphatidic acid (LPA, 1 1-acyl, 2-hydroxy-sn-glycerol-3-phosphate) has been shown in the past decades to mediate an array of biological responses that is anything but simple. LPA has been shown to signal such vital cellular events as mitogenesis (1), platelet aggregation (2), tumor cell invasion (3), and escape from apoptosis (4). LPA is present in serum at low (1-20 M) micromolar concentrations (5) and is produced and released by stimulated platelets (6). The autocrine activation of platelet aggregation, coupled with LPA's mitogenic effects on smooth muscle cells (7) and fibroblasts (8), suggests a functional role for LPA as a wound-healing hormone (see review (9)). Furthermore, LPA accumulates to high concentrations in ovarian cancer ascitic fluid and is mitogenic to ovarian cancer cells, implicating LPA as a marker and mediator of ovarian cancer progression (10).
Recently, specific G-protein-coupled receptors for LPA have been identified within a cluster of eight related receptors termed the Edg (Endothelial differentiation gene) cluster. To date, eight members of the Edg cluster have been described. Edg-2 (15,19), -4 (20), and -7 (21,22) have been shown to be LPA receptors and Edg-1 (23,24), -3, -5 (25), -6 (26,27), and -8 (28) are receptors for the structurally related phospholipid sphingosine 1-phosphate. The LPA receptors differ with respect to distribution and G-protein coupling. Edg-2 is widely expressed, with highest mRNA levels appearing in brain (20), reflecting expression in Schwann cells and oligodendrocytes (29,30). Edg-4 is most highly expressed in testis and leukocytes (20), whereas Edg-7 is very highly expressed in kidney and prostate (21,22). When stably transfected into the LPA-unresponsive cell line RH7777 (Rat Hepatoma), Edg-2 mediates inhibition of cAMP accumulation, whereas Edg-4 and Edg-7 mediate calcium mobilization (22). Because only the inhibition of cAMP accumulation is pertussis toxin-sensitive, this suggests that, in this system, Edg-2 is G i -linked, whereas Edg-4 and -7 are G q -linked. However, in Jurkat T cells (31) or rat oligodendrocytes (32,31) Edg-2 mediates calcium mobilization via G␣ i , while Edg-4 utilizes G␣ i and G␣ q . Thus, LPA receptor coupling may be dependent on the system of expression. The cloning of specific receptors for LPA represents a major advancement in the study of LPA signaling and the physiologic significance of receptor activation; however, endogenous cellular responses to LPA have not yet been mapped to specific endogenous receptors due to a lack of selective ligands.
Van Corven and colleagues first described a pertussis toxinsensitive mitogenic response to LPA in Rat 1 cells (8). In contrast to LPA's effects on calcium mobilization, cAMP accumulation, and cytoskeletal rearrangement, the mitogenic response has a half-maximal concentration in the mid-micromolar range in fibroblasts, ovarian cancer cell lines, and cultured vascular smooth muscle cells (7,8,33). This discrepancy has been suggested to reflect degradation over the long time course of the mitogenesis experiments (24 h or more) as compared with the assays for higher potency responses (seconds to minutes). LPA also has micromolar EC 50 with respect to platelet aggregation, although this response is measured in an assay of only a few minutes duration and is therefore presumably not greatly affected by degradation (34 -36).
The degradation of LPA may be attributed to the activity of integral membrane enzymes lipid phosphoric acid phosphatases (LPPs) (also named type 2 phosphatidic acid phosphatases (PAP2)) (reviews: Refs. 37,38). The human clones LPP1 (PAP2A) and LPP3 (PAP2B) were cloned by Kanoh in 1997 (39), and our group subsequently cloned a third isoform, LPP2 (PAP2C) (40). In addition to phosphatidic acid, all three enzymes exhibit broad substrate specificity, including the lipidsignaling mediators sphingosine 1-phosphate and LPA (39 -42). The three isoforms of LPP act as ectophosphatases by hydrolyzing exogenous LPA (44), suggesting a potential role in the regulation of LPA signaling by degrading LPA at the cell surface. The K m for LPP-mediated hydrolysis of LPA is several orders of magnitude higher concentration than that required for most LPA-signaling responses (40) (see Fig. 5A). Therefore, it is not clear if the activity of the LPPs has a physiologically significant effect on the activity of LPA at its receptors, because LPA is maximally efficacious at concentrations well below the affinity of the phosphatases for LPA. However, the phosphatases may significantly affect LPA receptor activation despite their lower affinity, if the concentration of phosphatase is much higher than the receptor concentration. Furthermore, the high K m of the phosphatases suggests that they will exhibit first order reaction kinetics under physiologic concentrations of LPA.
This study was initiated with the goal of determining the role of the phosphatases in the regulation of LPA signaling; specifically, we wanted to test the hypothesis that the markedly lower potency of LPA with respect to mitogenesis was simply a result of LPP-mediated hydrolysis. Brindley has previously reported that in one-point assays the mitogenicity of LPA is diminished in LPP1-overexpressing fibroblasts (44). We wanted to characterize this interaction further and demonstrate that LPP1 lowers the potency of LPA by hydrolyzing the phosphoester bond and thus destroying the ligand for LPA receptors. To this end, we synthesized two LPA analogs with phosphonate head groups, which should render them resistant to hydrolysis by the LPPs. If our hypothesis is true, then the potency of the phosphonate compounds relative to LPA should be greater in mitogenic assays than in assays with a shorter incubation period. Also, increases in the activity of LPP enzymes should diminish the activity of LPA but have no effect on the potency of the phosphonates. In the course of testing this hypothesis, we discovered a striking difference in the ligand selectivity of the mitogenic response as compared with that of the three cloned LPA receptors. The mitogenic response to LPA and LPA analogs is consistent with the platelet aggregation response but inconsistent with activation of Edg-2, -4, or -7.

EXPERIMENTAL PROCEDURES
Synthesis of ␣-Hydroxy and ␣-Keto Ethanolamide Phosphonic Acids-The synthesis of the ␣-hydroxy and ␣-keto propanolamide phosphonic acids will be described in detail in a manuscript currently being prepared on a series of phosphonate compounds. 2 Briefly, oleoyl chloride was acylated using 1-aminopropanol and pyridine followed by Dess-Martin oxidation. The resulting aldehyde was reacted with di-tbutylphosphonate in the presence of NaOH to generate the protected ␣-hydroxy derivative. Treatment with trifluoroacetic acid provided WLS31, whereas oxidation with pyridinium chlorochromate followed by trifluoroacetic acid deprotection provided WLS60.
Transfection and Membrane Preparation-Plasmid DNA was transfected into HEK 293 T cells using a calcium phosphate transfection protocol (45). Briefly, a DNA mixture containing 25 g of DNA and 0.25 M CaCl was added to HEPES-buffered 2 mM Na 2 HPO 4 . Subconfluent monolayers of HEK 293 T cells were poisoned with 25 mM chloroquine, and the DNA precipitate was then applied to the cells. After 4 h, the monolayers were washed with phosphate-buffered saline and refed media (90% 1:1 Dulbecco's modified essential media (DMEM):F-12 ϩ 10% fetal bovine serum). The cells were harvested 48 -72 h after addition of the DNA by scraping in HME buffer (in mM: 20 HEPES, 5 MgCl 2 , 1 EDTA, pH 7.4) containing 10% sucrose on ice, and disrupted using a Dounce homogenizer. After centrifugation at 800 ϫ g, the supernatant was diluted with HME without sucrose and centrifuged at 100,000 ϫ g for 1 h. The resulting pellet was rehomogenized and centrifuged a second hour at 100,000 ϫ g. This crude membrane pellet was resuspended in HME with sucrose, aliquoted, and snap-frozen by immersion in liquid nitrogen. The membranes were stored at Ϫ70°C. Protein concentration was determined spectroscopically by Bradford protein assay.
Radiolabeling Lipids-Phosphorus-32-labeled lysophosphatidic acid was prepared by reacting monoolein (NuChek Prep) with [␥-32 P]ATP in the presence of diacylglycerol kinase (Sigma) as described by Walsh et al. (46). The method used here differs from the published protocol in that the reaction size was increased 10-fold, and the reaction was allowed to proceed for 18 h. The labeled products were purified by normal-phase high pressure liquid chromatography. Using a Varian Microsorb MV, 4.6 mm x 250 mm, 5-m silica column, lipids were eluted using the following gradient profile: isocratic solvent A (HCCl 3 /MeOH/ H 2 O/NH 4 OH: 77/21.65/.85/.5) from 0 to 5 min, linear transition from solvent A to solvent B (HCCl 3 /MeOH/H 2 O/NH 4 OH: 59/34.5/6/.5) from 5 to 15 min, and isocratic solvent B from 15 min onward. The collected fractions were then washed with 0.1% HCl and MeOH, and the organic phase was retained and dried.
Assay for LPP (PAP2) Activity-Assays for phosphatase activity were carried out according to the protocol described by Kanoh et al. (42), using a constant Triton X-100:labeled lipid ratio of 50:1. The final specific activity of the lipids was adjusted to 10 -50 Ci/mol. Briefly, 10 g of membrane protein was prewarmed to 37°C in a reaction mixture containing 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 1 mg/ml fatty acid-free bovine serum albumin. Reactions were started by the addition of the radiolabeled lipid-Triton X-100 mixture, incubated at 37°C for 5 min, and stopped by the addition of 100 l of 0.1 N HCl in methanol. 200 l of CHCl 3 and 200 l of 1 M MgCl 2 were then added, and the reactions were vortexed and centrifuged briefly, and a fraction of the aqueous layer was added to scintillation fluid for counting.
Intact Cell LPP Assays/Ectophosphatase Assays-The ectophosphatase activity of LPP-overexpressing cells was determined with an assay derived from the assay described above. HEK 293 cells were suspended (via trypsinization) in HKRB (in mM: 20 HEPES, 103 NaCl, 4.8 KCl, 0.5 CaCl 2 , 1.2 MgSO 4 , 1.2 KH 2 PO 4 , 25 NaHCO 3 , 15 glucose, pH 7.4) at a concentration of 10 6 cells/ml. LPA was added to each 50-l reaction in 0.1% fatty acid free bovine serum albumin. The reactions were allowed to proceed for 10 min at 37°C and stopped with the addition of acidified methanol. The reactions were spun at 12,000 ϫ g for 5 min to remove any cell debris. The supernatant was extracted with chloroform, and the radioactivity in the aqueous phase was determined with scintillation counting.
Calcium and cAMP Assays-Calcium mobilization assays were performed as described previously by us (13). Briefly, intercellular calcium fluxes were measured on cell populations (2-4 ϫ 10 6 cells) that had been loaded with the calcium-sensitive fluorophore, INDO-1. Responses, which were measured in a temperature-controlled fluorimeter (Aminco SLM 8000C, SLM Instruments, Urbana, IL), are reported as the fraction of the maximal response (i.e. response to 75 M digitonin). Calcium mobilization assays were performed in HKRB.
GTP␥S Binding Assay-GTP␥S binding experiments were performed as described earlier (22). Ligand-mediated GTP␥S binding to G-proteins was measured in GTP binding buffer (in mM: 50 HEPES, 100 NaCl, 10 MgCl 2 , pH 7.5) using 25 g of a membrane preparation from HEK 293 T cells transiently transfected with DNAs encoding G␣ i2 C351F (this mutant G-protein is resistant to pertussis toxin inactivation), G␤ 1 , and G␥ 2 along with the receptor being tested. Ligand was added to membranes in the presence of 10 M GDP and 0.1 nM [ 35 S]GTP␥S (1200 Ci/mmol) and incubated at 30°C for 30 min. Bound GTP␥S was separated from unbound using the Brandel harvester (Gaithersburg, MD) and counted with a liquid scintillation counter.
Mitogenesis Assay-Mitogenesis was estimated by measuring DNA synthesis using a tritiated thymidine incorporation assay. Nearly (70%) confluent monolayers were grown in multiwell plates (48 and 96 wells) in DMEM-F12 (HEK 293 cells), DMEM (MDA-MB231 cells), or MEM supplemented with pyruvate and glutamine (RH7777 cells) plus 10% fetal bovine serum. Cell lines stably transfected with LPP1 or Edg receptor were grown in the appropriate media plus 800 g/ml G418. Cells were deprived of serum for 30 h, and then treated with drug or vehicle for 22 h. 1 Ci/ml of [ 3 H]thymidine was added 4 h before harvest. Incorporated thymidine was harvested by first fixing the cells in cold methanol, washing with phosphate-buffered saline, and precipitating DNA with cold 10% trichloroacetic acid. Precipitated material was either counted directly by adding scintillant and counting in a Packard Top counter (for 96-well plates) or solubilized with 0.2 N NaOH, 0.1% SDS, neutralized with HCl, and added to scintillant in scintillation vials for counting (for 48-well plates).
Platelet Aggregation Assay-Platelets were isolated from fresh human blood from donors (47). Briefly, blood was added to acid-citratedextrose to prevent coagulation, and centrifuged at low speed to prepare platelet-rich plasma. Indomethacin and apyrase were added to the platelet-rich plasma, and this preparation was centrifuged at high speed to sediment platelets. Platelets were resuspended in buffer (in mM: 140 NaCl, 0.34 Na 2 HPO 4 , 2.9 KCl, 10 HEPES, 12 NaHCO 3 , 5 glucose, 2 MgCl 2 ) and their aggregation was measured in a Lumiaggregometer (Chrono-Log, Havertown, PA). Prior to addition of LPA and analogs, 0.16 M ADP was added.

␣-Hydroxy and ␣-Keto
Phosphonates Are LPP Inhibitors-To examine the effect of LPP hydrolysis on LPA's potency in mitogenesis, we developed two LPP-resistant LPA analogs by mimicking the phosphate group with an ␣-substituted phosphonate ( Fig. 1). The direct carbon-to-phosphorous bond should render phosphonates resistant to degradation by the phosphatases. In fact, the ␣-hydroxy (wls-31) and ␣-keto (wls-60) phosphonates were not only resistant to LPP-mediated hydrolysis, they also acted as inhibitors of LPP activity. When added to a reaction of LPP enzyme and LPA substrate, the phosphonates significantly blocked the activity of each of the three LPP isoforms. Preincubation of the phosphonates with the enzyme preparation for 30 min before addition of substrate did not diminish the inhibitory effect, confirming that the compounds are not substrates for the enzyme simply acting as competitive inhibitors ( Fig. 2A). The K m values were determined for each LPP isoform using radiolabeled LPA as substrate in the presence and absence of the phosphonates and used to calculate K i values (Fig. 2B). This analysis revealed that both of the phosphonates are competitive inhibitors of each of the three LPP isoforms with mid-micromolar affinities (Fig. 2C). There is a significant difference in the affinity of the phosphonates for the different LPP isoforms: LPP3 Ͼ LPP1 Ͼ LPP2. It is also noteworthy that the K i values for the ketone compound (wls-60) are very nearly half the values for the hydroxy compound (wls-31) at each isoform. This may reflect the fact that the ketone is a discrete molecule, whereas the hydroxyl compound is racemic. If only one enantiomer of the hydroxyl compound is active, it would be equipotent with the ketone compound. Alternatively, the charge on the head group of the ketone compound may be more consistent with that of LPA than the hydroxy compound, as the pK a of the second dissociable proton of the ketone compound is closer to that of LPA.
Activity of the Phosphonates at LPA Receptors-Unfortunately, these phosphonate compounds cannot be used in the most obvious experiment, blocking the LPP activity with the phosphonates and determining if there is an effect of LPA potency, because both wls-31 and wls-60 are themselves potent agonists at the LPA receptors. To compare the potency of these compounds with that of LPA at the three cloned LPA receptors, we used a GTP␥S binding assay (22). Membranes from cells overexpressing the three heterotrimeric G-protein subunits along with a receptor are incubated with ligand, GDP, and [ 35 S]GTP␥S. The amount of GTP␥S bound to the G-proteins is determined by filtering and liquid scintillation counting (see "Experimental Procedures" for details). These assays measure G-protein activation directly, and do not reflect multiple levels of amplification, as do many measures of receptor activation. Therefore, the dose-response curves determined by this assay are often right-shifted with respect to assays of more downstream events (e.g. calcium mobilization). GTP␥S binding doseresponse curves show that wls-31 and wls-60 are one to two log orders less potent than LPA at Edg-2 and Edg-4 and have negligible activity at Edg-7 at concentrations up to 10 M (Fig.  3). wls-60 is more potent than wls-31 at both Edg-2 and Edg-4. The activity seen with wls-60 at Edg-7 is consistent with the amount of activity in untransfected cells (data not shown); however, all dose-response curves, including LPA, are strongly right-shifted at Edg-7 (21,22), so it is possible that the phosphonates would have a similar potency relative to LPA at Edg-7 as they do at Edg-2 and -4, if the curves were extended beyond 10 M (21,22). We also determined the activity of these compounds in intact cells by measuring calcium mobilization in nontransfected HEK 293 cells and MDA MB-231 cells. In both cell types, dose-response curves reveal rank order potency distribution nearly identical to that seen at Edg-2 and Edg-4 in the GTP␥S assay, with all of the curves shifted 1-2 log orders to the left reflecting amplification (data not shown). Therefore, the relative potency of the phosphonates with respect to LPA that we observed in the GTP␥S binding assay reflects the relative affinities of the compounds at endogenous receptor sites in intact cells.
LPPs Attenuate LPA-driven Mitogenesis-Having characterized the activity of the phosphonates with the LPPs and the LPA receptors Edg-2, -4, and -7, we next used the compounds to probe the interaction between the LPPs and LPA receptors in the mitogenic response. The first question we addressed was whether the phosphonates had a higher potency relative to LPA in mitogenesis compared with short-term (GTP␥S binding LPA (1-oleoyl, 2-hydroxyl-sn-glycerol-3phosphate); wls-31/␣-hydroxy phosphonate (N-oleoyl ␣-hydroxy propanolamide phosphonic acid); wls-60/␣-keto phosphonate (N-oleoyl ␣-keto propanolamide phosphonic acid); MHEPA (N-palmitoyl methylenehydroxy ethanolamide phosphoric acid). and calcium mobilization) assays. To address this question, we used HEK 293 cells, which respond to LPA with a pertussis toxin-sensitive tritiated thymidine incorporation (data not shown). These cells express Edg-2 and Edg-7 (22). Tritiated thymidine incorporation dose-response curves for LPA and the phosphonates wls-31 and wls-60 show that indeed the trend predicted by our hypothesis was observed (Fig. 4). The phosphonates were 3-4 log orders more potent relative to LPA in the mitogenic assay versus the GTP␥S assay.
Next, we asked if increases in LPP1 activity affect LPA-induced mitogenesis to a greater extent than phosphonateinduced mitogenesis. If LPP enzymes down-regulate LPA signaling by hydrolyzing the phosphoester bond of LPA, then overexpression of the enzymes should lower the potency of LPA in mitogenic assays but the potency of the phosphonates should be unaffected. We developed a stable line of HEK 293 cells overexpressing LPP1 enzymes, resulting in a severalfold increase in ectophosphatase activity, as measured with exogenous albumin-bound radiolabeled LPA as substrate (Fig. 5A). Several clonal lines of LPP1-overexpressing HEK 293 cells were isolated, but the expression of the gene appeared to have a negative effect on growth (i.e. LPP1 transfected cells grew more slowly than nontransfected cells). Additionally, the level of phosphatase activity diminished over time in many of the lines; presumably LPP1 expression was selected against due to the negative effects on growth. Therefore, the following experiments were performed in a single clonal line chosen because of its stable increase in ectophosphatase activity. Dose-response curves for mitogenesis were determined for LPA in normal and LPP1-overexpressing HEK 293 cells (Fig. 5B). The activity of LPA in the phosphatase-overexpressing cells was ablated completely. Curves for the LPP-resistant phosphonates were also determined in LPP1-overexpressing cells, and both compounds were still of similar potency as seen in plain HEK 293 cells (Fig.  5C). These results suggest that the LPP1 enzymes down-regulate LPA signaling by hydrolyzing the phosphoester bond and thereby lower the effective concentration of ligand available for receptor binding. In contrast to the sensitivity of LPA-induced mitogenesis to LPP overexpression, when GTP ␥ S binding curves were determined for Edg-2, -4, or -7, increases in LPP1 activity of the same magnitude (Fig. 6B) had no effect on LPA's potency or efficacy (Fig. 6A). This may reflect a unique sensitivity of the mitogenic response to regulation by the LPPs due to the lower potency of LPA in mitogenesis or to the longer duration of the assays. The Ligand Selectivity of the Mitogenic Response Is Inconsistent with That of Edg-2, -4, and -7-As discussed above, the potency of the phosphonates relative to LPA increased in the long term mitogenic assay compared with the short term GTP␥S assay at Edg-2 and Edg-4 as predicted by our hypothesis. However, the degree to which it increased (3-4 log orders) was much greater than predicted. If this shift is due solely to degradation of LPA by the LPPs, then greater than 99.9% of the LPA added at the beginning of the assay must be degraded during the course of the incubation. To test this possibility, [ 32 P]LPA was added to a parallel mitogenesis experiment under identical incubation conditions. At the end of the assay period, the reaction (including cell monolayer) was extracted with acidified methanol and chloroform, and the radioactivity retained in the organic phase (reflecting intact LPA) was counted. These experiments revealed that at concentrations at the upper limit of the mitogenic assay and approaching the K m of the LPPs (ϳ150 M), no more than 90% of the LPA is destroyed during the 24-h incubation with cells (Fig. 7A). These data suggest that the shift of the rank order potencies of the phosphonates and LPA in mitogenesis cannot be accounted for solely by LPP hydrolysis. Receptor activation in response to LPA, wls-31, and wls-60 was determined by measuring GTP␥S binding to membranes from HEK 293 T cells transiently transfected with DNA encoding Edg-2, -4, or -7 along with DNA-encoding G-proteins. Membranes were incubated with drug and GTP␥ 35 S for 30 min at 30°C and filtered to separate bound from free radioactivity. Basal GTP␥S binding has been subtracted from the data, and the data are normalized to maximal binding with the highest dose of LPA. Typical raw values for basal and maximal binding are 3,500 and 6,400 dpm at Edg-2; 3,500 and 12,800 dpm at Edg-4; and 4,700 and 15,000 dpm at Edg-7, respectively. Data points were determined in duplicate, and these curves are representative of two separate experiments; error bars reflect standard error. A possible explanation for this observation is that the mitogenic response does not proceed through Edg-2, -4, or -7, as we had presumed. Our laboratories have been engaged in defining the structure-activity profiles of the three cloned LPA receptors using synthetic LPA analogs to assign endogenous LPA responses to specific receptors. As part of this effort, we have developed a 2-palmitoyl LPA wherein the alkyl moiety is attached to the second carbon of the glycerol backbone by an amide linkage, N-palmitoyl-2-methylenehydroxy ethanolamide phosphoric acid (MHEPA) (Fig. 1). MHEPA is structurally related to N-acyl serine phosphoric acid (NASPA) (also equivalent to carboxy ethanolamide phosphoric acid), an LPA analog that was first described by Hanahan's group (51) and that our group later demonstrated is a stereoselective agonist of LPA responses (14). L(s)-NASPA is a potent LPA receptor agonist, whereas D(r)-NASPA is inactive. In MHEPA, the chiral center of NASPA is maintained as is the stereoselectivity. In assays of calcium mobilization, cAMP accumulation (data not shown), and GTP␥S binding at Edg-2, -4, and -7 (Fig. 7B), MHEPA is inactive in the R configuration but is a potent agonist in the S-form; such ligand stereoselectivity is common and observed in most receptor-mediated signaling. To determine whether this stereoselectivity was echoed by the mitogenic response, we tested both enantiomers in thymidine incorporation assays. In stark contrast to their activity at Edg-2, -4, and -7, the R-form of MHEPA was at least as active as the molecule in the S conformation in mitogenic assays. Dose-response curves for these compounds revealed that R-MHEPA is a full agonist of the mitogenic response and is slightly less potent than LPA (Fig. 7C). We tested LPA, both conformations of MHEPA, and wls-60 in MDA MB-231 cells for their potency in mitogenic assays to determine whether the distinct ligand selectivity was cell type-specific. As seen in Fig. 7D, the rank order potencies of LPA and LPA analogs is the same in these cells as in HEK 293 cells. The R-form of MHEPA is fully active with respect to mitogenesis (as is the S-form, data not shown), and the wls-60 is considerably more potent than R-MHEPA and LPA. The difference in ligand selectivity between the mitogenic response FIG. 5. LPP1 overexpression down-regulates LPA induced mitogenesis. A, the ectophosphatase activity of a clonal cell line overexpressing LPP1 was determined using intact HEK 293 cells with albumin-bound exogenous radiolabeled LPA as substrate. Reactions were incubated for 5 min, and the media were extracted and counted as described under "Experimental Procedures" to determine phosphatase activity. B, the nontransfected HEK 293 cells and the LPP1-overexpressing cell line were assayed for tritiated thymidine incorporation into the acid-insoluble fraction in response to LPA as described under "Experimental Procedures." The response is ablated in the LPP1-overexpressing line. Each data point was determined in triplicate, and these curves are representative of three separate experiments; error bars reflect standard error. C, the LPP1-overexpressing cell line was also assayed for wls-31-and wls-60-induced tritiated thymidine incorporation. These compounds are high potency mitogens irrespective of increased LPP1 activity. Each data point was determined in triplicate, and these curves are representative of two separate experiments; error bars reflect standard error. and the response of the individual LPA receptors suggests that the activation of Edg-2, -4, and/or -7 is not sufficient to account for the mitogenic response to LPA.
To explore further the possibility that mitogenesis in response to LPA is distinct from Edg receptor activation, we assayed the mitogenic response to LPA in rat hepatoma RH7777 cells. These cells are devoid of LPA-mediated calcium mobilization and inhibition of cAMP accumulation responses (22,48), and do not express Edg-2, -4, and -7 mRNA as judged by reverse transcription-polymerase chain reaction. 3 Nevertheless, RH7777 cells show a dose-dependent incorporation of tritiated thymidine when treated with LPA, R-MHEPA (and S-MHEPA, data not shown), and wls-60 (Fig. 7E). Although the curves reflect significant error due to poor adhesion of the cells, the rank order potency of these compounds was similar to that seen in HEK 293 and MDA MB231 cells. We also examined the mitogenic response to LPA in clonal RH7777 cell populations that had been transfected with Edg-2, -4, or -7 DNAs (22) and found that each of these lines had mitogenic responses to LPA similar to those in nontransfected cells (data not shown). Together, these data suggest that the mitogenic response to LPA and synthetic LPA analogs is independent of Edg-2, -4, or -7 in these three cell lines.
Platelet Aggregation Has a Ligand Selectivity Similar to the Mitogenic Response-Like the mitogenic response to LPA, platelet aggregation requires micromolar concentrations (even though the assay lasts only 1-2 min) and lacks stereoselectivity in response to NASPA/carboxy ethanolamide phosphoric acid and related chiral LPA analogs when tested at a single high concentration (36). We determined platelet aggregation dose-  7. The mitogenic response to LPA and LPA analogs has a ligand selectivity distinct from that of Edg-2, -4, and -7. A, the increased potency of the phosphonates relative to LPA cannot be accounted for by LPP hydrolysis; ϳ10% of LPA is intact following a mitogenic assay. Phosphorous-32-labeled LPA was added to a nearly confluent monolayer of HEK 293 cells under mitogenic assay conditions (see "Experimental Procedures"). At the end of the assay period, the entire reaction was extracted with acidified methanol/chloroform, and the radioactive material remaining in the organic phase was determined as a measure of intact LPA. Data points were determined in triplicate; error bars reflect standard error. B, S-MHEPA but not R-MHEPA is active at Edg receptors. GTP␥S binding to membranes from HEK 293 T cells overexpressing each subunit of the heterotrimeric G-proteins along with Edg-2, -4, or -7 is shown in response to LPA, R-MHEPA, and S-MHEPA (10 M each) (see "Experimental Procedures" for details). C, the mitogenic response to LPA analogs lacks the stereoselectivity observed at Edg receptors. The mitogenicity of the LPA analog MHEPA in both R and S conformations was determined in HEK 293 cells by measuring incorporation of tritiated thymidine. Both the R and S isomers had activities similar to that of LPA. Data points were determined in triplicate, and these curves are representative of two separate experiments; error bars reflect standard error. D, the ligand selectivity of the mitogenic response observed in HEK 293 cells is mirrored by MDA MB231 cells. Tritiated thymidine incorporation dose-response curves were determined for LPA, wls-60, and R-MHEPA in MDA MB231 cells. Data points were determined in quadruplicate, and these curves are representative of two separate experiments; response curves using the keto phosphonate (wls-60), the R and S MHEPA compounds, and LPA to determine whether the ligand selectivity correlated to that of mitogenesis (Fig. 8). The platelet response has a pharmacology that closely matches that of the mitogenic response, i.e. wls-60 and LPA are nearly equipotent and both enantiomers of MHEPA are active. The difference in the relative potencies of LPA and wls-60 in mitogenesis versus platelet aggregation is consistent with the amount of degradation that was observed during the course of the mitogenic assay; because only 10% of LPA is intact throughout the mitogenic assays, the degradation-resistant phosphonate should be one order of magnitude higher in potency relative to LPA with respect to mitogenesis than the shorter assays of platelet aggregation. These data suggest that the platelet aggregation response to LPA is inconsistent with the pharmacology of Edg-2, -4, or -7, whereas it is consistent with the mitogenic response. DISCUSSION The results of the study reported here are remarkable in several aspects. First, we have described a novel pair of compounds, the N-acyl ␣-hydroxy and ␣-keto propanolamide phosphonic acids, which are subtype-specific inhibitors of LPP enzymes and are degradation-resistant agonists at Edg-2 and Edg-4. Second, we have shown that overexpression of the LPP1 enzyme attenuates the mitogenic response to LPA, although it does not affect LPA potency or efficacy in GTP␥S binding assays with Edg LPA receptors. Third, we present several lines of evidence that the mitogenic response to LPA is independent of Edg-2, -4, and -7 in three cell lines. Finally, we have shown that the pharmacology of the platelet aggregation response corresponds to that of the mitogenic response, but not to the known LPA receptors. Together, these results suggest that there exists a low affinity LPA-signaling pathway that is distinct from Edg-2, -4, and -7, which mediates mitogenesis and platelet aggregation and is regulated by LPP enzymes.
The phosphonate compounds characterized in this study have potential to be useful tools to study the interaction of LPPs and LPA receptors. They are the first useful inhibitors of LPPs. It has been shown that LPP activity is blocked by propranolol and sphingosine, but the IC 50 values are quite high. Furthermore, high dose propranolol is nonspecific, and sphingosine is thought to inhibit by binding substrate (49). The lack of specific inhibitors and the ubiquity of the enzymes has made it difficult to investigate the physiologic role of the LPPs. The ideal inhibitor for investigating the endogenous role of the LPPs would specifically suppress the activity of the phosphatases, but lack LPA receptor agonist activity. The compounds reported here are not yet ideal due to their activity at the Edg receptors and in mitogenic and platelet aggregation assays, but they might be useful as lead structures for optimization.
Indeed, the phosphonate compounds used in this study are best described as nonhydrolyzable or LPP-resistant LPA agonists. In GTP␥S binding assays, they are active at Edg-2 and -4 with potencies ϳ2-fold lower than that of LPA, and these rank order potencies are reflected in intact cells with endogenous levels of receptor and G-protein in calcium mobilization assays, confirming the GTP␥S binding assay as a model for cellular receptor activation. Ours is not the first report of activity of phosphonate compounds in assays of LPA receptor activation. In 1991, Proll and Clark reported that an LPA analog with a phosphonate head group inhibits cAMP accumulation (50), but Moolenaar's group later showed that the same compound had little activity with respect to calcium mobilization (12). We have synthesized and tested a similar compound in a GTP␥S binding assay at each Edg/LPA receptor and found that it is an agonist at Edg-2 but has very little activity at Edg-4 or Edg-7. 4 Our data do not support our hypothesis that LPP1-mediated degradation of LPA is solely responsible for the lower potency of LPA in mitogenic assays. Instead, we have shown three lines of evidence that suggest that the mitogenic response to LPA is mediated by a pathway or mechanism that is distinct from activation of Edg-2, -4, and -7 and has different ligand selectivity, including an intrinsically lower affinity for LPA. First, we have shown that the potency of the phosphonate compounds wls-60 and wls-31 is increased in mitogenic assays to a degree that cannot be accounted for by degradation of LPA during the mitogenic assays. Second, R-enantiomers of compounds that are inactive in a variety of Edg-2, -4, and -7 signaling responses (GTP␥S binding, calcium mobilization, inhibition of adenylyl cyclase, and calcium-induced chloride currents in Xenopus oocytes) are full agonists in mitogenic assays. Third, RH7777 cells that do not express Edg-2, -4, and -7 have mitogenic responses to LPA and LPA analogs similar to those of cells that express the receptors. As additional evidence that LPA-induced mitogenesis is Edg receptor-independent, we have observed LNCaP cells that express functional Edg-2 (transfected), -4, and -7 LPA receptors but lack a mitogenic response to LPA, even though they undergo mitogenesis in response to other stimuli such as epidermal growth factor (data not shown).
Several studies on LPA-induced platelet aggregation have revealed inconsistencies with Edg receptor-mediated responses to LPA. The platelet response lacks stereoselectivity (36), requires micromolar concentrations of LPA (36,(51)(52)(53), and displays distinct ligand selectivity with respect to 1-O-alkyl LPA (ether linkage) (34). Gueguen and colleagues (36) have proposed, based on these observations, that there is an LPA reerror bars reflect standard error. E, RH7777 cells that do not express Edg-2, -4, and -7 exhibit a mitogenic response to LPA and LPA analogs with the same ligand selectivity as is seen in Edg-expressing cells. Thymidine incorporation was measured in response to LPA, wls-60, and R-MHEPA in RH7777 cells. Data points were determined in quadruplicate, and these curves are representative of three separate experiments; error bars reflect standard error.
FIG. 8. The platelet aggregation response to LPA has ligand selectivity that is consistent with the mitogenic response but not with Edg-2, -4, and -7. Dose-response curves were determined for LPA, wls-60, and both isomers of MHEPA. Aggregation was measured in an aggregometer in response to each drug following a subthreshold dose of ADP. Each data point was determined in triplicate or quadruplicate; error bars reflect standard error. ceptor in platelets that is pharmacologically distinct from Edg-2, -4, and -7; our data support their idea. Furthermore, we have shown that the ligand selectivity of the platelet response is consistent with the pharmacology of the mitogenic response, suggesting that these responses are mediated by a common pathway.
The data shown here do not rule out the possibility that the known LPA receptors play a role in the mitogenic response to LPA in some cells. It is possible that Edg receptor oligomerization or interaction with other modulating proteins can affect the relative potency of receptors for ligands; however, the mitogenic response in the RH7777 cells suggests that at least in some cases the mitogenic response is independent of these receptors. An obvious explanation is that an additional member of the Edg receptor family remains to be identified, but this is becoming less likely as the sequencing of the human genome nears completion with no additional members of the Edg cluster discovered. A more likely explanation is that a G-proteincoupled receptor outside of the Edg cluster (presumably G ilinked due to pertussis toxin sensitivity) is responsible for mitogenesis. In fact, the G-protein-coupled receptor PSP24 has been reported as a LPA receptor and is not within the Edg cluster (54). However, the lack of stereoselectivity and high concentration required for mitogenesis are also consistent with the possibility that these effects are not receptor-mediated but rather are mediated by a physical perturbation of the membrane. Pertussis toxin sensitivity demonstrates that LPA-induced mitogenesis is dependent on the action of a G i -linked protein, but this protein could be several steps downstream of LPA's site of action. Therefore, the nature of the mediator of the mitogenic (and platelet aggregation) response to LPA is unknown.
The data presented here do support the hypothesis that LPPs down-regulate LPA signaling in mitogenic assays by degrading LPA, although this degradation does not account for the lower potency of LPA in mitogenic assays. In HEK 293 cells that have severalfold overexpression of LPP1, the mitogenic response to LPA is ablated. Interestingly, we did not see an effect on the potency or efficacy of LPA-induced GTP␥S binding with similar levels of LPP1 overexpression in membranes expressing Edg-2, -4, or -7. Several differences between the assays for mitogenesis and GTP␥S binding may be responsible for the apparently greater sensitivity of the mitogenic response to LPP1 overexpression: hours versus minutes of incubation time, intact cells versus isolated membranes, and endogenous versus overexpressed receptors. However, the mitogenic response to LPA occurs over a concentration range that is much closer to the K m values of the LPP enzymes than most Edg receptormediated responses, which may allow the LPPs to specifically regulate the lower affinity LPA pathway. Brindley and colleagues (55) recently reported that overexpression of LPP1 attenuated LPA responses that were suggested to be mediated by Edg-2. Because Edg-2 was the only Edg/LPA receptor expressed in their system, it was assumed to be responsible for the responses based on the same presumption we made entering this study: All LPA responses are mediated by one of the three currently cloned and characterized LPA receptors. Therefore, the degree to which LPA signaling through the Edg receptors is affected by LPP activity is not known.
Brindley and colleagues (55) proposed in the same report that high and low potency effects of LPA correlate to low and high concentrations of calcium, respectively; however, there are many reports in the literature of high potency LPA effects (low nanomolar EC 50 levels) in the presence of high calcium ion concentrations. Hopper et al. (56) reported calcium mobilization in response to LPA with an EC 50 of 5 nM in MDA MB-231, and Moolenaar's group (12) has reported an LPA EC 50 of 0.2 nM in A-431 cells, measuring calcium mobilization in a buffer system containing 1 mM calcium. The potency of LPA with respect to inhibition of cAMP accumulation has been measured and reported in two previous papers by us to be 20 nM (13) and 52 nM (14), and these experiments were performed in 1.8 mM calcium. LPA was also shown to have a similar but slightly lower potency (EC 50 ϭ 100 nM) in assays of inhibition of cAMP accumulation in the absence of calcium ions (57). Lafontan and colleagues (58) have reported LPA-induced spreading of adipocytes with an EC 50 of 4 nM in 1.8 mM calcium.
We support, however, Brindley's general proposal that the presence of calcium greatly affects the physical state of LPA and that the physical state of LPA can potentially affect its signaling properties. LPA has been shown to bind calcium ions with high affinity, causing the formation of microprecipitates of LPA at micromolar concentrations under physiologic concentrations of calcium (11). The presence of calcium and other LPA binding molecules such as albumin affects the physical and signaling properties of LPA. Also potentially affecting LPA's signaling properties is its ability to insert into the plasma membrane bilayer. Insertion of lipid monomers into the membrane is not expected to occur to any appreciable extent below the critical micelle concentration; however, the critical micelle concentration of LPA (reported to be ϳ1.3 mM in pure water (8)) has not been determined in physiologic solutions in the presence of biological membranes. Therefore, the degree to which LPA partitions into the bilayer under physiologic conditions is not known. Much work remains to be done to determine the existence of LPA in various physical states (free LPA monomer, calcium aggregates, bound to albumin, or inserted in the bilayer) under physiologic conditions and the signaling capabilities of these different forms of LPA.
In conclusion, the data presented here suggest that the LPA receptors that have been cloned and characterized to date are neither sufficient nor necessary to account for the mitogenic response to LPA. We propose a new model for LPA signaling in which at least one additional mechanism exists that mediates LPA-induced mitogenesis and platelet aggregation, and LPP enzymes negatively regulate this pathway by ligand degradation (Fig. 9). The nature of this mechanism has yet to be defined and is not necessarily a receptor. The lack of a molecular target for this pathway is ironic in that mitogenesis and platelet aggregation are among the first described responses to LPA FIG. 9. The new model for LPA signaling derived from the results presented here includes an Edg-receptor-independent pathway that mediates platelet aggregation and mitogenesis by an unknown mechanism that is down-regulated by LPPs. It is not known if the LPPs significantly down-regulate Edg receptor-mediated LPA signaling. Also, the physiologic cellular response to Edg receptor activation is not known. and are responsible for much of the scientific interest in LPA signaling due to their potential involvement in therapeutically relevant physiologic and pathologic roles for LPA signaling in wound healing and ovarian cancer progression. This model drastically alters our view of LPA signaling and the way in which it is investigated.