Mammalian Lipid Phosphate Phosphohydrolases*

Phosphatidate phosphohydrolase (PAP) was first identified as being involved in glycerolipid synthesis (1). A second PAP activity (PAP-2) was subsequently characterized in mammalian cells based upon a lack of requirement for bivalent cations and insensitivity to inhibition by N-ethylmaleimide (2). This review will concentrate on the Type 2 phosphohydrolases that hydrolyze a variety of lipid phosphates (3, 4). Because the precise biological functions of these phosphohydrolases are not known, we propose to rename the PAP-2 family (Table I) as lipid phosphate phosphohydrolases (LPPs).

distinct cDNA that encodes for yet another hLPP and which we designate LPP-3 ( Fig. 1; Table I). hLPP-3 appears to be the human homolog (13) of the rat Dri42 gene product (17), because these two proteins share 94% amino acid sequence identity.
All tissues that have been examined express LPP activity against PA. The specific activity in rats is low in skeletal muscle and heart and highest in brain, lung, kidney, and spleen (6). mRNAs for LPP-1/1a are expressed widely and to a high extent in prostate, aorta, bladder, uterus, kidney, lung, and heart (13,14,18). hLPP-1a may be the predominant isoform expressed in heart, whereas hLPP-1 appears predominant in kidney, lung, placenta, and liver (14). The expression of mRNA for LPP-2 was more restrictive, being found mainly in brain, pancreas, and placenta (15). mRNA for LPP-3 was expressed relatively uniformly in all human tissues examined (13). mLPP-1 was expressed predominantly in plasma membranes (4,7), in agreement with the distribution reported for LPP activity in liver and adipose tissue (2,3). The Dri42 protein resides in the endoplasmic reticulum of epithelial cells of intestinal mucosa (17). By contrast, Kai et al. (13) concluded that hLPP-3 should have a post-Golgi localization based on the sensitivity of its oligosaccharide chain to endo-␤-glycosidase (13).
Results of experiments with G6Pase and chloroperoxidase demonstrate that the conserved amino acids in each domain ( Fig. 1) are involved in the coordination and hydrolysis of the phosphate ester, which probably occurs through a phosphohistidine intermediate. Thirty-seven percent of mutant alleles in people with glycogen storage disease 1a contain a single point mutation in the codon for Arg-83 in G6Pase (31). This arginine and histidines 119 and 176 of G6Pase are absolutely required for catalytic activity (32,33), and the two histidines are conserved within the phosphatase superfamily. In Curvularia inaequalis chloroperoxidase, Lys-353, Arg-360, Ser-402, Gly-403, Arg-490, and His-496 coordinate vanadate in the active site of the enzyme (36, 37). All of these residues are con-served in the phosphatase superfamily. Additionally, His-404, the conserved histidine in Domain 2 of chloroperoxidase (Table II), is believed to be the hydrogen donor in the reaction mechanism (36,37) as is the equivalent histidine (His-119) in hG6Pase (33). The conserved histidine (His-176) of hG6Pase is proposed to be the phosphoryl acceptor during catalysis (33). Chloroperoxidase exhibits both vanadate-dependent peroxidase activity and vanadateinhibitable phosphatase activity. Moreover, the peroxidase activity of chloroperoxidase is inhibited by phosphate, and the phosphatase activity of G6Pase is inhibited by vanadate (38). These results indicate that the binding of phosphate and vanadate is mutually exclusive and may occur at the same site (20). These results suggest that the three-dimensional architecture of the active sites of chloroperoxidase, G6Pase, and the LPPs is conserved and that this family of enzymes shares a similar catalytic mechanism. However, mutational studies have yet to be performed for the LPPs to verify this proposal.
The hydrophobicity plots (39) of the LPPs and Dri42 are almost superimposable, suggesting that their three-dimensional structures are similar. There are six putative membrane-spanning regions that are boxed in Fig. 1, and this is compatible with their being integral membrane proteins. Barilà et al. (17) concluded that the hydrophobic transmembrane domains 1, 3, and 5 had a signal/ anchor function and that membrane insertion of Dri42 was achieved co-translationally by the action of a series of alternating insertion and halt transfer signals, resulting in the exposure of both termini to the cytosolic side. All LPPs contain a consensus N-glycosylation site in the loop between transmembrane regions 3 and 4 (Fig. 1). This is the only glycosylation site that was demonstrated in LPP-1 (7), LPP-3 (13), and Dri42 (17). Glycosylation can be blocked in cells with tunicamycin (7), and the LPP isoforms are sensitive to N-glycanase (6,7,13,17) and sialidase (13). The presence of the glycosylation site restricts the topology of LPPs such that the loop containing this site should be extracellular (at the plasma membrane) or luminal (in endoplasmic reticulum membranes). This conclusion is consistent with the topology of Dri42 (17) and G6Pase (33). There is no obvious preprotein leader sequence that targets the processing of the LPP proteins via traditional routes, and the mechanism of targeting is unknown. This is also true for chloroperoxidase (28) and G6Pase (30). Fig. 1 also identifies several putative phosphorylation sites for the LPPs.

LPPs and Signal Transduction
The identification of LPP in plasma membranes led to the hypothesis that it is involved in cell signaling (3,4). This could involve an intracellular site of action by dephosphorylating PA formed by PLD or DAG kinase (4). PA activates the NADPH oxidase system in neutrophils and stimulates protein kinases, phosphatidylinositol 4-kinase, phospholipase C-␥, and the Ras-Raf-MAP kinase pathway (3,40). 3 PA is also involved in stimulating the formation of the actin cytoskeleton (41) and in microvesicle budding from Golgi membranes (42). The LPPs could therefore attenuate signaling by PA while producing DAG, which might activate protein kinase Cs. Evidence was provided that LPP can dephosphorylate PA generated by PLD using ras-transformed fibroblasts, which have low LPP activity compared with control fibroblasts. Stimulation of PLD produced an increased formation of PA relative to DAG in ras-transformed compared with control fibroblasts (43). Furthermore, PA accumulated as a function of time in culture for the ras-transformed fibroblasts compared with control cells (44). Overexpression of hLPP-1 or hLPP-1a in ECV304 endothelial cells decreased PA concentrations by 50% (14). LPP-1 mRNA 3 To save space reviews cited rather than original articles. Conserved phosphatase domains (1, 2, and 3) are highlighted in yellow. Red-colored amino acids within those domains highlight residues that are believed to be essential for catalytic activity. Blue highlighted residues indicate potential protein kinase C, casein kinase II and cAMP-dependent protein kinase phosphorylation sites. The magenta-colored asparagine residue is a consensus N-linked glycosylation site. Accession numbers for the sequences are shown in Table I. hLPP-1 (accession number AF014402) was used in the sequence alignment. expression was also diminished in human colon tumor tissue compared with matching tissue from normal colon (14). It may also be significant that HIC-53, which codes for a putative protein having extensive sequence homology to LPP-1, has been described as ras-recision gene (35 Sphingolipids regulate cell differentiation and proliferation (40,45), 3 and they "cross-talk" with the glycerolipid signaling pathway (3,40). 3 Ceramides are formed by the agonist-induced activation of sphingomyelinase. They cause apoptosis and inhibit PA production by blocking PLD activation (40). 3 Little is known about the physiological effects of C-1-P, although it may regulate some aspects of synaptic vesicle function and a C-1-P phosphatase was located in synaptic vesicles and in plasma membranes (46). Exogenous C-1-P stimulates cell division in fibroblasts (40). 3 A major fate of ceramide is conversion to sphingosine, a compound that directly inhibits protein kinase C (40). 3 Sphingosine also decreases protein kinase C activity through inhibition of PAP-1 and LPP activity with consequent decreases in DAG formation (1)(2)(3)40). 3 Sphingosine also increases PA concentrations by stimulating DAG kinase and PLD activities in some cells (40). 3 Some effects of sphingosine are mediated by conversion to S-1-P, a potent activator of PLD, Ca 2ϩ mobilization, and cell proliferation by the activation of MAP kinase (40). 3 The ability of the LPPs to hydrolyze PA, lyso-PA, C-1-P, or S-1-P indicates that these enzymes can potentially attenuate signaling by these lipids while simultaneously generating other signals through the formation of DAG, ceramide, and sphingosine.
LPPs could also act as "ecto-enzymes," which regulate signaling by exogenous lyso-PA and S-1-P that may be secreted as autocrine or paracrine mediators. For example, newly formed lyso-PA in thrombin-activated platelets is released and is believed to stimulate wound repair (47). 3 In addition, lyso-PA production through activation of secretory phospholipase A 2 may represent a proinflammatory pathway (48). Lyso-PA acts through specific receptors (49) 3 and increases tyrosine kinase activities, phospholipase C-␥, Ca 2ϩ mobilization, arachidonate release, PLD, MAP kinase, focal adhesion kinase, and stress fiber formation (3,40,47). 3 Exogenous PA, lyso-PA, and S-1-P decrease adenylate cyclase activity through a pertussis toxin-sensitive mechanism (3,47). 3 Overexpression of mLPP-1 in rat fibroblasts increases the ability of intact cells to dephosphorylate exogenous 32 P-labeled lyso-PA and C-1-P and release 32 P i in the culture medium. 4 These results confirm that at least a portion of the LPP-1 activity can function as an ecto-enzyme. This conclusion could identify LPP-1 as the ecto-PAP (9, 50) or ecto-lyso-PAP (51) reported previously. The definition of the "ecto-phosphohydrolase" in work published so far relies on the assumption that the transmembrane movement of lipid phosphates is slow (unless specifically catalyzed by translocases). 32 P i from labeled PA and lyso-PA is also released quickly into the culture medium, whereas intracellular P i is not rapidly secreted. Work from Dr. D. English 5 (30,32) were used to generate the proposed consensus domains. Numbers flanking the outside of domains 1 and 3 refer to the numbering of the first and last amino acid in those domains, respectively, from the linear sequence of the protein. The intervening numbers refer to the number of amino acids between the adjoining domains. Blue lettering indicates amino acids that are conserved in the consensus sequences, as indicated. TM, transmembrane domain, determined using TMPredict (36). X, any amino acid; b, branched chain amino acid; h, hydrophobic amino acid (I, L, F, M, V, W, Y); n, small neutral amino acid (A, C, G, S); ⌽, aromatic amino acid (F, H, Y, W). Ca 2ϩ mobilization and consequent induction of actin polymerization (52). Thus, ecto-LPP could degrade exogenous bioactive lipid phosphates and by doing so generate new signals to direct cell migration. The latter situation is exemplified by the wunen gene product, which dephosphorylates PA. 6 Expression of the wunen product in the gut of Drosophila embryos transforms a permissive environment into a repulsive one and guides germ cells to the mesoderm (34).
There is relatively little information concerning LPP regulation. Kai et al. (13) demonstrated that the mRNA for LPP-3, but not LPP-1, increased up to 3-fold after treating quiescent HeLa cells with epidermal growth factor. Recent work has identified LPP-1 as being an androgen-regulated gene in human prostatic adenocarcinoma cells (18). Expression of Dri42 in rat intestinal mucosa is increased during epithelial differentiation (17).
The involvement of mammalian LPPs in signaling functions is supported by the properties of other homologs. Mastoparan stimulates the production of PA, which is then converted to DGPP by PA kinase (53,54). Yeast DGPPases dephosphorylate DGPP rather than PA because the specificity constant (V max /K m ) is about 10 times higher for DGPP than for PA (11,24,55). DGPP could signal in its own right or serve as the precursor of PA while preventing PA hydrolysis (53,54). An LPP-l gene, distinct from that of DGPPase, has also been identified in Saccharomyces cerevisiae, and the encoded phosphatase exhibited the following substrate preference: PA Ͼ lyso-PA Ͼ DGPP (27). Deletion mutants of LPP-1 and DGP-Pase showed decreased concentrations of phosphatidylinositol with the greatest decrease in the double mutant. PA concentrations were also increased in the DGPPase and LPP1/DGPPase double mutant (27). Yeast deletion mutants of DGPPase (24), LPP1 (27), and phytosphingosine phosphate phosphatases (also known as lipid-binding proteins 1 and 2) (26) demonstrate normal growth, cell morphology, and mating, but lipid-binding protein mutants have dramatically enhanced survival to heat shock (56). Overexpression of lipid-binding protein 1 (also called Lcb3) confers resistance to the inhibition of cell growth by sphingosine (25).

Summary and Conclusions
We are at an exciting point where various LPP isoforms have been identified. The regulation of LPP homologs has not been investigated in detail. Another area of uncertainty is the full extent of the LPP family and whether other LPPs exist with relatively little sequence homology. The exact subcellular localization for each mammalian LPP is not yet known. Therefore, we do not know the extent to which different lipid phosphates have access to the LPP isoforms. The identification of lyso-PA and S-1-P as physiological extracellular messengers and our observations that LPPs can dephosphorylate these exogenous substrates imply a role for the LPPs in modulating extracellular signaling. Equally we need to understand the roles of the LPP isoforms in dephosphorylating intracellular signals from lipid phosphates. These studies can now be undertaken to elucidate the functions of each LPP isoform in regulating signal transduction.