Conversion of lysophospholipids to cyclic lysophosphatidic acid by phospholipase D.

Phospholipase D from Streptomyces chromofuscus hydrolyzes lysophosphatidylcholine or lysophosphatidylethanolamine in aqueous 1% Triton X-100 solution. In situ monitoring of this reaction by P NMR revealed the formation of cyclic lysophosphatidic acid (1-acyl 2,3-cyclic glycerophosphate) as an intermediate which was hydrolyzed further by the enzyme at a functionally distinct active site to lysophosphatidic acid (lyso-PA). Synthetic cyclic lyso-PA (1-octanoyl 2,3-cyclic glycerophosphate) was found to be stable in aqueous neutral solutions at room temperature. It was hydrolyzed by the bacterial phospholipase D to lyso-PA at a rate which was approximately 4-fold slower than the rate of formation of cyclic lyso-PA. The addition of 5-10 mM sodium vanadate could partially inhibit the ring opening reaction and thus increase substantially the cyclic lyso-PA accumulation. Cyclic lyso-PA may act as a dormant configuration of the physiologically active lyso-PA or may even possess specific activities which await verification.

Phospholipase D from Streptomyces chromofuscus hydrolyzes lysophosphatidylcholine or lysophosphatidylethanolamine in aqueous 1% Triton X-100 solution. In situ monitoring of this reaction by 31 P NMR revealed the formation of cyclic lysophosphatidic acid (1-acyl 2,3-cyclic glycerophosphate) as an intermediate which was hydrolyzed further by the enzyme at a functionally distinct active site to lysophosphatidic acid (lyso-PA). Synthetic cyclic lyso-PA (1-octanoyl 2,3-cyclic glycerophosphate) was found to be stable in aqueous neutral solutions at room temperature. It was hydrolyzed by the bacterial phospholipase D to lyso-PA at a rate which was approximately 4-fold slower than the rate of formation of cyclic lyso-PA. The addition of 5-10 mM sodium vanadate could partially inhibit the ring opening reaction and thus increase substantially the cyclic lyso-PA accumulation. Cyclic lyso-PA may act as a dormant configuration of the physiologically active lyso-PA or may even possess specific activities which await verification.
Activation of phospholipases has been implicated in a wide range of signal transduction pathways (1). With regard to phospholipase D (PLase D), 1 the current information as to the molecules generated by PLase D activation and their mode of action is still scarce (reviewed in Ref. 2).
The first step in the lytic activity of phospholipase D is the formation of a phosphoryl enzyme intermediate which is analogous to the acyl enzyme in the action of common esterases (3,4). This intermediate is generally cleaved by the ambient water molecules with the net hydrolysis of one phosphoester bond. However, alcoholic hydroxyl residues can compete for the cleavage of the phosphoryl enzyme intermediate to yield a phosphodiester product (5,6). If a hydroxyl group is positioned appropriately in the substrate it can also compete for the phosphoryl enzyme yielding a cyclic product. Cyclic products of phospholipase action have been detected in the action of phosphatidylinositol-specific phospholipase C (PLase C) on phosphatidylinositol yielding 1,2-cyclic inositol phosphate as the initial product (7,8), and in the PLase C hydrolysis of phosphatidylglycerol to 1,3-cyclic glycerophosphate (9). Similar cyclization takes place in the hydrolysis of glycerophosphorylcholine or glycerophosphorylethanolamine by glycerophosphinicocholine diesterase, where 1,2-cyclic glycerophosphate is formed (10,11). All of these cyclic phosphates are relatively stable at neutral pH but can be hydrolyzed by specific phosphodiesterases to form the respective phosphate monoesters (12,13). It seems, therefore, that when the phosphoryl residue attached to the enzyme includes a free hydroxyl group, the formation of a cyclic phosphodiester by intramolecular transphosphorylation can take place. In analogy to the above reactions with a hydroxyl group, it has been suggested (9) that phospholipids with a free primary amine group (i.e. phosphatidylethanolamine and phosphatidylserine) may yield a cyclic phosphoramidate intermediate upon the action of PLase C. These putative five-membered cyclic phosphoramidates are expected to be hydrolyzed spontaneously (9).
The action of PLase D on phospholipids forms a phosphatidyl enzyme intermediate which can either react with water to yield phosphatidic acid (PA) or with an alcohol (5,6) to yield a new phospholipid (e.g. phosphatidylethanol if ethanol is added as the alcohol). This transphosphatidylation reaction raises the possibility that the action of PLase D on lysophospholipids may yield cyclic lyso-PA by intramolecular transphosphorylation with the hydroxyl on carbon 2 of the glycerol backbone as schematically presented in Fig. 1. The present work provides evidence that cyclic lyso-PA is formed upon the interaction of bacterial PLase D with lysophosphatidylcholine (lyso-PC) or lysophosphatidylethanolamine (lyso-PE).
1-Octanoyl 2,3-cyclic glycerophosphate was prepared by reacting ␣-octanoyl glycerol (Avanti Chemicals) with phosphorus oxychloride as described (15). It was assumed that the transacylation in this reaction did not exceed 10% (16), and therefore the prepared lysophospholipid was essentially of the ␣ configuration.
Phospholipase D-Two preparations of PLase D from Streptomyces chromofuscus of relatively high specific activity (approximately 500 units/mg of protein) obtained from Sigma and from Boehringer Mannheim GmbH were used in this study. They were found essentially identical in their activity profile. The enzyme activity was routinely assessed by monitoring remaining substrate content in aliquots removed at discrete time points. The aliquots were quenched by the addition of tetrahydrofuran, and the lipid species were separated by TLC on silica gel plates with running solvent as above. Each spot was scraped from the TLC plate and the phosphate content determined (17).
NMR-Spectra were recorded at 28°C either with a Varian 500 NMR spectrometer or with a Bruker AMX-400 NMR spectrometer. 31 P( 1 H) NMR spectra were acquired at 161.97 MHz (on the Bruker instrument) or 202.404 MHz (on the Varian instrument). Spectral parameters included an 80-ppm sweep width and 128 or 256 transients. Qualitative spectra presentation was obtained at 70°pulses using 4-s relaxation delay and composite pulse proton decoupling. For quantitative evaluation, 60°pulse, 7-s relaxation delay, and composite pulse * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
proton decoupling were employed. These conditions complied with the complete relaxation delay (5T 1 ) of the nucleus with the longest relaxation time among the 31 P nuclei in our experiments, which was 6.6 s. A 5% error in this method was estimated based on deviation in reproducibility of a set of seven independent experiments. The assignment of resonances was based on their chemical shift with respect to phosphocreatine at Ϫ2.5 ppm, which served as an internal or external reference.
The in situ enzymic experiments were carried out in D 2 O. Samples in 10-mm NMR tubes contained 30 -40 mg of synthetic lyso-PC dispersed in ϳ1.5 ml of 100 mM borate buffer, pD ϭ 8, with 10 mM calcium chloride. An appropriate amount of PLase D was added, depending on the overall reaction time desired.

RESULTS AND DISCUSSION
Preliminary screening has indicated that bacterial PLase D, unlike cabbage PLase D, can cleave the headgroup of lyso-PC and lyso-PE in the presence of ether or 1% Triton X-100. Fig. 2 presents the comparative degradation profiles of PC and lyso-PC solubilized in Triton X-100 mixed micelles, pH 8.0, as assayed by TLC and determination of the phosphorus content of each lipid spot. Under these conditions lyso-PC is a substrate for bacterial PLase D, but it is not as rapidly hydrolyzed as PC, in qualitative agreement with the findings of Imamura and Horiuti (18). In an analogous assay with dipalmitoyl-PE and 1-palmitoyl-PE (lyso-PE), similar profiles and rates of degradation were obtained. It should be noted that the enzyme "lysophospholipase D" (19,20) acts selectively on L-␣-etherlinked lysophospholipids and not on the common ester-linked L-␣-lysophospholipids which were studied here.
The aqueous medium consisting of phospholipid (5 mg/ml) dispersed in 1% Triton X-100, 10 mM borate buffer, pH 8.0, with 2 mM Ca 2ϩ (see Fig. 2 and "Materials and Methods") was found suitable for in situ monitoring of the PLase D cleavage products of lysophospholipids by 31 P NMR. However, short chain lyso-PC substrates (e.g. 1-octanoyl-3-glycerolphosphorylcholine) were found to be hydrolyzed by bacterial PLase D without the use of Triton X-100 or ether. Furthermore, for these compounds the lyso-PC concentration used in the assay was below its critical micellar concentration, ensuring that the substrate is essentially monomeric. During the course of the lyso-PC hydrolysis, a new resonance was observed at 18.5 ppm as presented in Fig. 3. Ordinary phosphate esters (e.g. lyso-PA or lyso-PC) are characterized by resonances in the range of ϩ5 to Ϫ6 ppm (21). The chemical shift of the new resonance, quite downfield from that of the substrate (0.5 ppm) and product (5.1 or 0.8 ppm, with or without vanadate, respectively), was consistent with the formation of a five-membered cyclic phosphodiester. Five-membered cyclic phosphates have a downfield shift due to -electron shielding effect (22). Experimental conditions were selected for the complete hydrolysis of the lysophospholipid substrate by PLase D to occur over several hours. This allowed the production of the five-membered cyclic phosphodiester and the more gradual appearance of lyso-PA to be monitored for different substrate systems as a function of time. In all of the incubations of lyso-PC and lyso-PE with PLase D, the formation of a 31 P resonance at 18.5 ppm (i.e. a cyclic diester product) was observed concomitant with the appearance of the lyso-PA resonance. After rapidly reaching a certain intensity which was maintained for a while (consistent with reaching a steady-state concentration), the resonance at 18.5 ppm declined, while that of the phosphate monoester (i.e. lyso-PA) continued to increase in intensity (Fig. 3). The most likely explanation for the unusual downfield shifted resonance was the formation of a cyclic lyso-PA as an intermediate in the PLase D reaction.
The formation of cyclic lyso-PA by the action of PLase D on lyso-PC or lyso-PE was verified with 31 P NMR spectra of synthetic 1-octanoyl 2,3-cyclic glycerophosphate (i.e. cyclic lyso-PA). The phosphorus chemical shift of this material corresponded to 18.5 ppm, essentially the same shift as the resonance observed by 31 P NMR in assays of the action of PLase D on lyso-PC or lyso-PE (Fig. 3). The synthetic 1-octanoyl 2,3-cyclic glycerophosphate was examined for its stability under assay conditions and as a substrate for bacterial PLase D, since in the NMR assays eventually all the cyclic product was converted to lyso-PA. The synthetic cyclic lyso-PA was found to be stable at pH 6 -8 for at least 10 h as was indicated by invariant intensity at ϩ18.5 ppm (Fig. 4). However, in the presence of PLase D, the intensity of this resonance decreased while that of lyso-PA increased, indicating phosphatase activity. The addition of egg PC in 1% Triton X-100 or in a 1:1 mixture with lyso-PC did not affect the rate of this reaction which excluded the possibility that the cyclization and the ring opening take place at the same catalytic site.
The cyclization and the ring opening activities could, in principle, be related to separate entities which copurified in the partial enzyme purification used to generate the PLase D. All the PLase D preparations used in the NMR experiments also exhibited a strong phosphatase activity with the conventional substrate p-nitrophenyl phosphate (data not shown). To explore this further, we have chromatographed PLase D preparations by gel filtration using Sephadex G-75, G-100, G-150, and S-300, DEAE-cellulose-52, and red Sepharose, which is an affinity binder of alkaline phosphatase, in an attempt to dissociate these activities. In all of these attempts the phosphatase activity, as monitored by hydrolysis of p-nitrophenyl phosphate, coincided with the PLase D activity. Therefore, we tentatively conclude that the phosphatase activity is inherent in the bacterial PLase D.
The reaction profiles presented above and the fact that cyclization and phosphatase activities always coincided suggest that PLase D carries out two sequential enzymatic reactions with lysophospholipids: (i) intramolecular cyclization to form cyclic lyso-PA, followed by (ii) hydrolysis to lyso-PA, namely a phosphodiesterase type cleavage.
An alternative approach for dissecting the cyclization activity from that of the ring opening was the use of phosphatase competitive inhibitors. In a series of 31 P NMR experiments which were carried out under the same conditions as in Fig. 3, we have added ␤-glycerophosphate (5 mM), p-nitrophenyl phosphate (5 mM), and sodium vanadate (2-100 mM). Sodium vanadate at concentrations of 5-10 mM could partially inhibit the phosphatase activity of the PLase D. This was measured by a substantial increase in relative content of cyclic lyso-PA compared with lyso-PA. An example of a reaction profile in the presence of 5 mM sodium vanadate is shown in Fig. 5. Under these reaction conditions there was a clear relative increase in the intensity of the resonance at ϩ18.5 ppm. The presence of sodium vanadate caused a downfield shift and splitting in the resonance of the lyso-PA. One possible explanation for the marked effect of vanadate on the lyso-PA resonances could be the formation of pyrophosphovanadate which is stabilized by the presence of Ca 2ϩ . 51 V spectra recorded in the above system (not shown) displayed an upfield shift similar to that recorded in a pyrophosphovanadate formed between adenosine monophosphate (AMP) and sodium vanadate (23). The two resonances at the region of lyso-PA (Fig. 5) might be related to ␣ and ␤ isomers. Table I represents a summary of the 31 P chemical shifts of the compounds presented in this communication.
Assuming that each of the above PLase D reactions follows a Michaelis-Menten type kinetic scheme with two independent enzyme sites, E 1 and E 2, then the reaction steps may be presented as follows, where A is the starting lysophospholipid (e.g. lyso-PC), B is the putative cyclic lyso-PA, and C is the lyso-PA (in principle in either ␣ or ␤ form). V A and V B are the V max values, while K A and K B are the corresponding K m values. Accordingly (24), The solution for the reduction of A with time is To a first approximation one can assume that along the whole reaction profile A Ͻ K A , which upon integration of Equation 1a-1c leads to a simple first order decline in A.
Assuming further that the same approximation also holds for B (i.e. B Ͻ K B ) then The rapid cleavage of the putative cyclic lyso-PA implied, from the leveling off and decline in the 18.5 ppm band (see Fig. 3), hampered our attempts to apply the kinetic analysis described above. In practice, however, this could be carried out only in the presence of sodium vanadate which acted as an inhibitor of the phosphadiesterase component (see Fig. 5). A typical experimental presentation of the change in A, B, and C with time in the presence of 5 mM sodium vanadate is shown in Fig. 6. The presented values were obtained by pick integration under instrumental setup for quantitative evaluation (see "Materials and Methods"). The profiles comply with Equations 3a, 3b, and 4 and with the implied correspondence C ϭ A o Ϫ A Ϫ B. Approximate analysis of the kinetics described above was carried out by first evaluating V A /K A according to Equations 3a and 3b. For the experimental conditions used in Fig. 6, the corresponding rate of cyclization was around V A /K A ϭ 0.04 min Ϫ1 . V B /K B could be evaluated in an independent experiment (as the one presented in Fig. 5) measuring the hydrolysis of synthetic B, 1-octanoyl 2,3-cyclic glycerophosphate. The value obtained was V B /K B ϭ 0.01 min Ϫ1 . Insertion of these V A /K A and V B /K B values into Equation 4 yielded the change in the cyclic lyso-PA concentration as a function of reaction time. As seen in Fig. 6, the correspondence between the observed and calculated presence of cyclic lyso-PA is reasonable, considering the series of approximations that were applied. Lyso-PA has been recently cited as an important modulator FIG. 5. Degradation profile of lyso-PC by PLase D in the presence of 5 mM sodium vanadate. The conditions and experimental setting were as in Fig.  3. Inset, a spectrum taken for the same reaction carried out with a 10-fold higher enzyme concentration (t ϭ 5 min). The proposed lyso-PA-vanadate structure is as follows:  of cell functions (25,26). It is synthesized either by the action of PLase A 2 on PA or by phosphorylation of ␣-monoglyceride by a specific kinase. Cyclic lyso-PA can in principle be formed in cells by the action of PLase A 2 on membrane phospholipids (27) followed by the action of PLase D which is described here. These consecutive enzyme reactions may impose a serious stringency on the formation of cyclic lyso-PA. Furthermore, the conversion of cyclic lyso-PA to lyso-PA will demand a third enzymatic action. Detection of these yet unexplored reactions will clarify whether cyclic lyso-PA is actually formed in cells and whether it serves as a dormant form of lyso-PA or acts as a modulator of cell functions on its own right.