Metabolic Insights into Phospholipid Function Using Gene-targeted Mice*

Few human diseases have been described that are due to a defect in phospholipid biosynthesis. Phospholipids are required for pro-viding the essential milieu of biological membranes and are impor- tant precursors of signaling molecules. One would anticipate that a null mutation in an enzyme required for the biosynthesis of a phospholipid would not be compatible with life, particularly if that mutation resulted in complete elimination of that phospholipid. However, for several of the major mammalian phospholipids ( e.g. phosphatidylcholine (PC), 1 phosphatidylethanolamine (PE), and phosphatidylserine (PS)) more than one biosynthetic pathway operates (1). Consequently, the possibility exists that humans might have deficiencies in phospholipid biosynthesis that would not be detectable if alternative enzymes that catalyzed the same reaction were active or if alternative pathways were available for making that phospholipid. In the last two decades, advances in gene targeting in mice have made it possible to determine whether or not a particular enzyme or protein is essential for mammalian life. This article will summarize the current state of knowledge about induced mutations in murine genes involved in phospholipid biosynthesis and intracellular transport. To date, gene-targeting experiments have revealed that mice can survive without certain phospholipid biosynthetic enzymes; in each case, an alternative pathway or enzyme exists for making that phospholipid.Themajorbiosynthetic routes for the biosynthesis of PC, PE, and PS were elucidated in the 1950s largely by Eugene Kennedy and co-workers (2). In subsequent decades, attempts were made

In addition to its presence in the plasma membrane, scramblase-1 is also present in the nucleus and participates in signaling pathways related to cell proliferation (19). Consistent with this finding, mice with targeted deletion of scramblase-1 exhibit perinatal granulocytopenia but have no defect in phospholipid scramblase activity in the plasma membrane (20). A flippase function for scramblase-1 cannot, however, be ruled out because other scramblase family members might compensate for a loss of phospholipid scrambling activity contributed by scramblase-1.
Scramblase-3-deficient mice have also recently been generated (21). In contrast to scramblase-1, which is expressed primarily in blood cells, scramblase-3 is also highly expressed in muscle and fat cells (18). In scramblase-3 knock-out mice, phospholipid scramblase activity is normal, but the animals accumulate abdominal fat and develop insulin resistance and dyslipidemia (21). In addition, primary adipocytes and bone marrow-derived macrophages from these mice are engorged with neutral lipids, suggesting a role for scramblase-3 in normal development and/or function of these cells. Thus, although the mechanism of action of scramblases has not yet been defined, studies with scramblase knock-out mice do not provide evidence that these proteins function physiologically in transbilayer movement of phospholipids.

Mice That Lack Phosphatidylethanolamine N-Methyltransferase Are Viable but Metabolism of Lipoproteins and Homocysteine Is Abnormal
Phosphatidylethanolamine N-methyltransferase (PEMT) catalyzes the methylation of PE to PC (22,23). Although PEMT activity is detectable in non-hepatic tissues, the activity is usually less than 1% of that in the liver. Thus, PEMT is essentially a liver-specific enzyme. The PEMT gene was selected as a target for gene ablation in mice because the liver can use an alternative route for making PC (the CDP-choline pathway) that might compensate for loss of PEMT. Disruption of murine Pemt was the first example in which an enzyme of phospholipid biosynthesis was eliminated in an intact animal (24). Pemt Ϫ/Ϫ mice are viable and do not exhibit any obviously abnormal phenotype. In the livers of Pemt Ϫ/Ϫ mice, the amount of PC is the same as in Pemt ϩ/ϩ mice. The CDP-choline pathway can apparently compensate for the lack of PEMT because the membrane-associated activity of CTP:phosphocholine cytidylyltransferase (CT), which under most metabolic states catalyzes the regulated and rate-limiting step in the CDP-choline pathway for PC biosynthesis (25), is increased by 60%.
PC synthesis via the CDP-choline pathway requires the input of an exogenous source of choline. Best and Huntsman (26) first described choline as an important dietary component in 1932. Therefore, an obvious question was: how would Pemt Ϫ/Ϫ mice fare if they were fed a choline-deficient diet? Choline deficiency in rodents has frequently been used as an experimental model for studies on PC function. Rats fed a choline-deficient diet can survive for at least a year but eventually develop hepatic steatosis and, often, hepatic cancer (27,28). When Pemt Ϫ/Ϫ mice were fed a choline-deficient diet for 3 days, the PC content of the liver decreased by 56%, the level of hepatic triacylglycerol (TG) increased by 3-6-fold, and liver failure ensued ( Fig. 1) (29). In contrast, when Pemt ϩ/ϩ mice were fed the same diet for 3 days there was no obvious liver pathology or TG accumulation. The damage to the liver and the decreased PC levels, caused by simultaneously eliminating dietary choline and PEMT activity, were reversed when the mice were subsequently fed a choline-supplemented diet (30). The specificity of the requirement for choline in Pemt Ϫ/Ϫ mice is remarkable. Dimethylethanolamine, a choline analog that contains two, rather than three, methyl groups can be readily incorporated in vivo into the phospholipid phosphatidyldimethylethanolamine (31). When fibroblasts are cultured in the presence of dimethyleth-anolamine instead of choline, the cells grow normally implying that in these cells phosphatidyldimethylethanolamine can substitute for PC (32,33). Thus, the expectation was that dimethylethanolamine would be able to substitute for choline in Pemt Ϫ/Ϫ mice. However, when the Pemt Ϫ/Ϫ mice were fed dimethylethanolamine instead of choline, liver failure still occurred but 1 day later than in Pemt Ϫ/Ϫ mice fed the choline-deficient diet without supplementation (34).
Whereas choline-deficient Pemt Ϫ/Ϫ mice rapidly develop liver failure, the PC content of other organs is not reduced nor is obvious damage noted to any other tissues (29). Interestingly, the cholinedeficient diet did not reduce the amount of PC secreted into bile of Pemt Ϫ/Ϫ mice (24). A possible reason for the rapid onset of liver failure in choline-deficient Pemt Ϫ/Ϫ mice is that a 20-g mouse normally excretes 23 mg of biliary PC/day (35), whereas the total amount of PC in the liver is ϳ20 mg. Therefore, a mouse secretes the equivalent of its entire hepatic pool of PC into bile each day; some of this PC is reabsorbed by the intestine (35). Therefore, the possibility that the large drain of PC from the liver into bile was responsible for the liver failure was investigated. Pemt Ϫ/Ϫ mice were bred with Mdr2 Ϫ/Ϫ mice to generate a strain of mice that lacked both PEMT and MDR2. 2 As noted above, Mdr2 Ϫ/Ϫ mice have a defect in the secretion of PC into bile. Remarkably, the double knock-out mice usually survive for longer than 3 months when fed a choline-deficient diet. 2 These studies demonstrate that elimination of biliary PC secretion protects Pemt Ϫ/Ϫ mice from liver failure induced by lack of dietary choline and confirm that excretion of PC into bile greatly aggravates the potential for liver damage in these mice.
Because Pemt Ϫ/Ϫ mice do not show an obvious phenotype when fed laboratory chow, the mice were challenged with a high fat/high cholesterol "Western-style" diet for 3 weeks (36). Livers from male, but not female, Pemt Ϫ/Ϫ mice showed a striking 4 -6-fold accumulation of TG and cholesteryl esters compared with Pemt ϩ/ϩ mice but no decrease in the amount of PC (36). Correspondingly, the plasma concentration of TG was 60% lower in male Pemt Ϫ/Ϫ mice than in Pemt ϩ/ϩ mice. The plasma level of apoB100 was also reduced by 50% by PEMT deficiency (36). Based on experiments with intact mice and cultured hepatocytes (36 -38), the reduction in plasma TG and apoB100 was attributed to a defect in hepatic secretion of very low density lipoproteins. The mechanism by which a lack of PEMT inhibits lipoprotein secretion and the basis for the sexual dimorphism remain to be elucidated. A lack of PEMT also reduced the secretion of lipoproteins in low density lipoprotein receptor-deficient mice. Furthermore, when these double knock-out mice were fed a high fat/high cholesterol diet for 12 weeks, the development of atherosclerosis was greatly attenuated compared with that of low density lipoprotein receptor-deficient mice. 3 In mouse plasma, PC and cholesterol are largely present in high density lipoproteins. In Pemt Ϫ/Ϫ mice fed either chow or a high fat/high cholesterol diet, plasma levels of PC and cholesterol were 25-45% lower than in Pemt ϩ/ϩ mice (36). The mechanism responsible for this reduction is not known.
For each PC molecule generated, the PEMT reaction produces 3 molecules of S-adenosylhomocysteine that are subsequently catabolized to homocysteine in the liver. Homocysteine is then converted to either methionine or cysteine or is secreted from the liver into plasma. A high level of plasma homocysteine is an independent risk factor for development of cardiovascular disease (40). Unexpectedly, plasma homocysteine was 50% lower in Pemt Ϫ/Ϫ mice than in Pemt ϩ/ϩ mice (39). Moreover, cultured hepatocytes from Pemt Ϫ/Ϫ mice secreted ϳ50% less homocysteine than did Pemt ϩ/ϩ hepatocytes. These experiments demonstrate that PEMT is a more important source of plasma homocysteine than was previously recognized.
From these studies with Pemt Ϫ/Ϫ mice unexpected roles for PEMT have been discovered in liver viability, regulation of plasma homocysteine levels, and lipoprotein metabolism.   (41). Chinese hamster ovary cells with a temperature-sensitive mutation in CT undergo apoptosis at 40°C (42,43). Thus, it was anticipated that global disruption of the CT gene in mice would be embryonic lethal. However, since a second murine CT gene, Pcyt1b, is present on the X chromosome, the possibility was raised that embryonic lethality would not necessarily occur in CT␣-deficient mice (44). Alternative splicing of the CT␤ gene in mice yields two mRNAs that encode CT␤2 and CT␤3 (45,46). The CT␣ gene, Pcyt1a, has been classically studied and resides on murine chromosome 16. Recent studies from Jackowski's laboratory show that disruption of Pcyt1a in mice is embryonic lethal. 4 Tabas and colleagues (47) used the Cre-lox system specifically to disrupt the Pcyt1a gene in macrophages in mice. Despite the absence of CT␣, cultured peritoneal macrophages derived from these mice appeared normal. CT activity was decreased by 70 -90%, with residual activity ascribed to enhanced expression of CT␤2. Macrophages that are incubated with acetylated low density lipoproteins take up cholesterol, and CT activity is increased (48). The increased CT activity was postulated to protect the macrophages from toxic effects of cholesterol accumulation. This hypothesis was tested by comparing the extent of cell death in wild-type and CT␣-deficient macrophages. After incubation with acetylated low density lipoproteins, 29% of CT␣-deficient macrophages died compared with only 2% of wild-type macrophages (47). Thus, increased CT activity provided protection against cholesterol-induced death of cultured murine macrophages.

Disruption of Murine CT␣ and CT␤ Genes
The Cre-lox system was similarly used to inactivate CT␣ selectively in murine livers in which Cre expression was governed by the albumin promoter. These mice are viable and fertile (49). Because in the knock-out mice CT activity was only 15% of normal, the large majority of CT activity in the liver is apparently contributed by CT␣. The residual activity was due to a 2-fold induction of CT␤2 as well as CT activity in hepatic cells other than hepatocytes. In addition, PEMT activity was doubled in the CT␣-deficient livers. As a result of the lack of CT␣, the mass of PC in the liver was reduced by 7-25%. Although the supply of PC apparently sufficed to allow the mice to grow and breed normally, the levels of PC, cholesterol, and TG in plasma lipoproteins were markedly lower in hepatic CT␣-deficient mice than in wild-type mice. Correspondingly, plasma levels of apoA1 and apoB100 (but not apoB48) were decreased by ϳ50% in hepatic CT␣-deficient mice. The reduction in plasma TG and apoB100 was likely due to decreased hepatic secretion of very low density lipoproteins (49). Therefore, when PC biosynthesis is limited by deletion of either CT␣ or PEMT, VLDL secretion is impaired. It is not yet clear, however, if the defects in hepatic lipoprotein secretion caused by deficiency of PEMT or CT␣ operate via the same or different mechanisms. The CDP-choline pathway is responsible for production of ϳ70% of hepatic PC (50,51), and the CT␣ isoform apparently catalyzes the majority of this reaction (49). In contrast, PEMT provides only ϳ30% of hepatic PC. Elimination of PEMT does not decrease the overall PC content of the liver, although one cannot discount a reduction in specific subcellular PC pools. On the other hand, the PC content of CT␣deficient livers is significantly reduced by up to 25%. Nevertheless, the magnitude of the reduction in lipoprotein secretion caused by disruption of each of these genes is similar.
CT␤2 is an abundant CT isoform in brain, lung, and gonads (45). Because CT␣ is also present in these tissues, it was reasonable to anticipate that mice lacking CT␤2 might be viable. Jackowski and co-workers (52) demonstrated that this was the case and that lungs and brains of CT␤2-deficient mice are apparently normal. However, CT␤2-deficient females are defective in ovarian follicle development, and male CT␤2-deficient mice show testicular degeneration and reduced fertility. The targeting vector used to generate these mice did not eliminate CT␤3 expression. Thus, it is possible that complete deletion of CT␤ will be embryonic lethal or that the targeted mice will exhibit more severe gonadal pathology and/or additional defects.

Mice Survive without Phosphatidylserine Synthase-2
Mammals use two pathways for PS synthesis (25). PS synthase-1 (PSS1) catalyzes an exchange of serine for the choline head group of PC whereas PS synthase-2 (PSS2) catalyzes a parallel reaction in which serine is exchanged for ethanolamine in PE. PSS2 is highly expressed in testis and brain whereas PSS1 is more ubiquitously expressed (53,54). An obvious question is: are both enzymes required for viability? This question was addressed using mice derived from embryonic stem cells with an insertional mutation in Pss2 that was identified in a gene-trapping screen (55). These mice expressed ␤-galactosidase driven by the Pss2 promoter. The ␤-galactosidase marker showed high expression of Pss2 in Sertoli cells of the testis, brown fat, and Purkinje cells of adult cerebellum. Pss2 Ϫ/Ϫ mice grew normally; however, testis weight was decreased and some males were infertile. Surprisingly, in light of the large (65-95%) reduction of in vitro PS synthase activity in homogenates from Pss2 Ϫ/Ϫ mouse tissues, the phospholipid composition was indistinguishable from that of Pss2 ϩ/ϩ tissues (55). These observations confirm that maintenance of a normal cellular phospholipid composition is an important homeostatic response. Studies with primary hepatocytes show that PS degradation is slowed by PSS2 deficiency, presumably as a mechanism for maintaining normal levels of PS. 5 Future studies are expected to establish whether or not PSS1 is required for embryonic development and if Pss1 Ϫ/Ϫ mice are viable.

Phospholipid Transfer Protein Genes
Phospholipid transfer proteins were discovered in the 1960s and were proposed as candidates for the intracellular transfer of phospholipids between membranes (56). Although these proteins catalyze phospholipid transfer between membranes or vesicles in vitro, there is no persuasive evidence that these proteins perform such a function in vivo.
The PC transfer protein is an abundant cytosolic protein that is highly specific for PC in transfer assays in vitro. One function suggested for this protein was to catalyze the net transfer of PC from the endoplasmic reticulum, a site of PC synthesis, to other intracellular membranes including the bile canalicular membranes. Another function proposed was in PC secretion for lung surfactant (57). Consequently, Borst and colleagues (57) investigated if ablation of the PC transfer protein gene in mice resulted in viable mice and if bile excretion and/or lung surfactant production were compromised. Remarkably, no defect was observed in PC secretion into bile or lung surfactant. Apparently the PC transfer protein is not essential for intermembrane trafficking of PC in vivo, and the physiological function of this protein remains unknown.
Phosphatidylinositol transfer proteins (PITPs) transfer phosphatidylinositol and PC between membranes in vitro, and this function was thought to reflect their role in vivo. Three mammalian PITPs, designated PITP␣, PITP␤, and rdgB␤ have been identified. PITP␣ and PITP␤ share 77% sequence identity (58). Both proteins catalyze PC and phosphatidylinositol transfer between membranes in vitro, and PITP␤ also catalyzes sphingomyelin transfer. To gain insight into the function of these proteins, Bankaitis and colleagues (58) generated mice that lacked PITP␣. Pitp␣ Ϫ/Ϫ mice survived the prenatal period but developed severe neurodegenerative disease as well as intestinal and hepatic steatosis. Moreover, Pitp␣ Ϫ/Ϫ mice were severely hypoglycemic. Nevertheless, elimination of PITP␣ in murine cells produced no obvious defects in bulk phospholipid metabolism (58). However, PITP␣ does appear to function in lipoprotein assembly and/or secretion from the intestine and liver, as well as in maintaining plasma glucose levels. How, or if, these functions relate to a role of PITP␣ in mediating the intermembrane transfer of PC and/or phosphatidylinositol is not clear. Attempts to generate Pitp␤ Ϫ/Ϫ mice and to obtain murine embryonic stem cells that lack both copies of the PITB␤ gene have been unsuccessful, suggesting that PITB␤ is an essential gene for early embryonic murine development (59).
Another putative phospholipid transfer protein is the plasma phospholipid transfer protein that also catalzyes PC transfer between membranes in vitro (60). Plasma PC transfer protein plays an important role in plasma by mediating phospholipid transfer among lipoproteins (60), and ablation of the gene in mice markedly reduced plasma high density lipoprotein levels (61). An additional role for this protein was uncovered because secretion of apoB-containing lipoproteins was decreased in mice lacking plasma phospholipid transfer protein (62) and significant plasma phospholipid transfer activity was detected within the Golgi. These findings suggest that in addition to its role in mediating lipoprotein homeostasis in plasma, the plasma phospholipid transfer protein is involved in adding phospholipids to nascent apoB-containing lipoproteins in the Golgi (62).

Future Directions
Targeting of genes involved in phospholipid biosynthesis and transport has provided novel insights into the functions of these proteins. Surprisingly, however, gene targeting of the so-called phospholipid transport proteins has not provided insights into the functions of these proteins. Many more genes of phospholipid metabolism remain to be ablated so that functions of the corresponding proteins and their phospholipid products can be defined. For example, PE is made by two completely independent pathways located in distinct intracellular organelles (endoplasmic reticulum and the mitochondrion). Are these pathways both required? Do mice need both genes that encode choline kinase activity? Is each of the choline/ethanolaminephosphotransferases required? Can some of these genes be disrupted in specific tissues, but not in others, without compromising viability? Is the biosynthesis of cardiolipin and phosphatidylglycerol required in mammals? Ingenious scientists with the powerful approaches now available should be able to answer these questions within the next few years.