Insulin Controls Subcellular Localization and Multisite Phosphorylation of the Phosphatidic Acid Phosphatase, Lipin 1*

Brain, liver, kidney, heart, and skeletal muscle from fatty liver dystrophy (fld/fld) mice, which do not express lipin 1 (lipin), contained much less Mg2+-dependent phosphatidic acid phosphatase (PAP) activity than tissues from wild type mice. Lipin harboring the fld2j (Gly84 → Arg) mutation exhibited relatively little PAP activity. These results indicate that lipin is a major PAP in vivo and that the loss of PAP activity contributes to the fld phenotype. PAP activity was readily detected in immune complexes of lipin from 3T3-L1 adipocytes, where the protein was found both as a microsomal form and a soluble, more highly phosphorylated, form. Fifteen phosphorylation sites were identified by mass spectrometric analyses. Insulin increased the phosphorylation of multiple sites and promoted a gel shift that was due in part to phosphorylation of Ser106. In contrast, epinephrine and oleic acid promoted dephosphorylation of lipin. The PAP-specific activity of lipin was not affected by the hormones or by dephosphorylation of lipin with protein phosphatase 1. However, the ratio of soluble to microsomal lipin was markedly increased in response to insulin and decreased in response to epinephrine and oleic acid. The results suggest that insulin and epinephrine control lipin primarily by changing localization rather than intrinsic PAP activity.

Brain, liver, kidney, heart, and skeletal muscle from fatty liver dystrophy (fld/fld) mice, which do not express lipin 1 (lipin), contained much less Mg 2؉ -dependent phosphatidic acid phosphatase (PAP) activity than tissues from wild type mice. Lipin harboring the fld 2j (Gly 84 3 Arg) mutation exhibited relatively little PAP activity. These results indicate that lipin is a major PAP in vivo and that the loss of PAP activity contributes to the fld phenotype. PAP activity was readily detected in immune complexes of lipin from 3T3-L1 adipocytes, where the protein was found both as a microsomal form and a soluble, more highly phosphorylated, form. Fifteen phosphorylation sites were identified by mass spectrometric analyses. Insulin increased the phosphorylation of multiple sites and promoted a gel shift that was due in part to phosphorylation of Ser 106 . In contrast, epinephrine and oleic acid promoted dephosphorylation of lipin. The PAP-specific activity of lipin was not affected by the hormones or by dephosphorylation of lipin with protein phosphatase 1. However, the ratio of soluble to microsomal lipin was markedly increased in response to insulin and decreased in response to epinephrine and oleic acid. The results suggest that insulin and epinephrine control lipin primarily by changing localization rather than intrinsic PAP activity.
Lipin is a M r ϳ100,000 Mg 2ϩ -dependent phosphatidic acid phosphatase (PAP) 2 (1). This enzyme generates pools of diacylglycerol utilized in the synthesis both of TAG and of the phospholipids, phosphatidylserine, phosphatidylcholine, and phosphatidylethanolamine (2,3). Lipin is encoded by the Lpin1 gene, 3 which is mutated in mice with fatty liver dystrophy (fld) (4). Consequently, lipin was known to have an important role in lipid metabolism before the PAP activity of the protein was discovered.
The fld allele arose spontaneously as a result of a complex genetic rearrangement of the Lpin1 gene (4). Thus, fld/fld mice express neither lipin 1a nor lipin 1b, a slightly larger protein, which arises from alternative mRNA splicing (4,5). Within a few days of birth fld/fld mice develop hypertriglyceridemia and hepatic steatosis (6). These abnormalities resolve after a few weeks, but several long lasting derangements develop, including glucose intolerance, insulin resistance (7), and defective peripheral myelination (8). A striking abnormality of the fld/fld mice is the absence of normal adipose tissue (7). Another mutant Lpin1 allele, designated fld 2j , encodes a protein identical to wild type lipin, except that Arg replaces Gly at position 84 (4). The phenotype of fld 2j /fld 2j mice is indistinguishable from that of fld/fld mice. The effect of the Gly 84 3 Arg mutation on the enzymatic activity of lipin has not been described.
The proteins encoded by Lpin1 are expressed in a wide variety of tissues, with the highest levels found in adipose tissue (4,5). Lipin 1b increases markedly as fat cells differentiate, and it is the prevalent isoform in mature adipocytes (9). Two other isoforms, lipin 2 and lipin 3, are encoded by different genes (4). Conservation of sequence among the lipin isoforms is highest in domains, referred to as NLIP and CLIP, which are located in the NH 2 -and COOH-terminal regions, respectively, of the proteins (4). These domains are found in lipin proteins from other species, including the Saccharomyces cerevisiae lipin (1, 10), Pah1p, and the Schizosaccharomyces pombe protein, Ned-1 (11). Imbedded in the CLIP domain is a signature sequence of the HAD superfamily of hydrolases (12). In HAD family phosphatases, the phosphate group from the substrate is transferred to Asp (corresponding to Asp 712 in lipin 1b) in the HAD motif I consensus sequence, ⌽⌽⌽⌽DXXGT⌽ (where ⌽ represents a bulky hydrophobic residue and X is any residue) (13).
Activity measurements in hepatocytes and adipocytes indicate that the subcellular distribution of the Mg 2ϩ -dependent PAP is controlled by hormones and fatty acids (14,15). Incubating adipocytes with insulin increased the proportion of soluble phosphatase activity, whereas incubating the fat cells with norepinephrine or fatty acids, such as palmitic or oleic acid, increased the proportion of activity that fractionated with intracellular membranes upon centrifugation (15). The membrane-associated phosphatase has been proposed to be the form that is active in TAG synthesis (16); however, there is not a strict relationship between rates of TAG synthesis and membrane association of the phosphatase in adipocytes (15). Fatty acids directly inhibit PAP activity in vitro (17,18). It would be premature to conclude that the activities that have been measured in different subcellular fractions are due to lipin, since other PAP enzymes exist. Overexpressed green fluorescent protein-tagged lipin has been detected in both the nucleus and cytoplasm of HEK 293 cells (4); however, information on the subcellular localization of the lipin protein is limited.
The mechanisms involved in the control of lipin are still poorly defined. In previous 32 P-labeling experiments, we discovered that lipin was phosphorylated in a rapamycin-sensitive manner in response to insulin in rat adipocytes (5), indicating that lipin was a target of the mTOR signaling pathway. The present study was conducted to investigate control of the activity and phosphorylation of lipin.

EXPERIMENTAL PROCEDURES
Animals-Mice homozygous for the fld allele were generated by mating heterozygous BALB/cByJ-fld/ϩ mice (Jackson Laboratories). Five-month-old wild type and fld/fld progeny from this cross were used in experiments. Where indicated, some of the mice were fasted overnight. Otherwise, the mice were allowed free access to food and water.
Subcellular Fractionations-3T3-L1 adipocytes were fractionated by slight modification of the method described by Piper et al. (21). For each treatment, a 10-cm-diameter plate of cells was rinsed in PBS before the cells were homogenized as described above except in 1 ml of Buffer B (0.25 M sucrose, 50 mM NaF, 1 mM EDTA, and 50 mM Tris-HCl, pH 7.4). The homogenates were centrifuged at 16,000 ϫ g for 20 min, and the pellets containing plasma membranes, mitochondria, and nuclei were retained. The supernatants were centrifuged at 175,000 ϫ g for 1 h. The high speed supernatants containing soluble proteins were collected, and the pellets containing microsomes were suspended in 1 ml of Buffer B.
Measurements of PAP Activities-[ 32 P]Phosphatidic acid was purified by thin layer chromatography after using [␥-32 P]ATP and Escherichia coli diacylglycerol kinase to phosphorylate 1,2dioleoyl-sn-glycerol (22). PAP activity was determined by using a modification of the method of Carman and Lin (22) to measure [ 32 P]phosphate release from [ 32 P]phosphatidic acid. Briefly, samples of cell or tissue extracts (5 l), subcellular fractions (5 l), or immune complexes (10 l) were added to reaction mixtures (100 l) containing 1 mM Triton X-100, 10 mM ␤-mercaptoethanol, 0.1 mM [ 32 P]phosphatidic acid (10,000 cpm/nmol), 50 mM Tris-maleate (pH 7.0), and either no added Mg 2ϩ or 2 mM MgCl 2 as described below. After 20 min at 30°C, the reactions were terminated by adding 0.5 ml of 0.1 N HCl in methanol. [ 32 P]phosphate was extracted by vigorous mixing after adding 1 ml of chloroform and 1 ml of 1 M MgCl 2 . The [ 32 P]phosphate in 0.5 ml of the aqueous phase was determined by scintillation counting.
With tissue extract samples (see Fig. 1) or subcellular fractions of cells (see Fig. 7), Mg 2ϩ -dependent PAP activity was determined by subtracting activity measured in the absence of MgCl 2 from activity measured in the presence of 2 mM MgCl 2 .
To determine PAP activities attributable to overexpressed lipin proteins in extract samples (see Fig. 2), reaction mixtures were supplemented with 2 mM MgCl 2 , and the activities from cells transfected with a control plasmid were subtracted from the activities from cells overexpressing lipin. For lipin immune complexes (see Fig. 6), PAP activities were routinely measured only in the presence of 2 mM MgCl 2 . Results of control experiments revealed that PAP activity in these complexes was negligible without MgCl 2 . 4 Results of time course and dilution experiments verified that, under the conditions of the assays, the amount of [ 32 P]phosphate released from [ 32 P]phosphatidic acid was linearly related to the time of incubation and the amount of extract or immune complexes. 4 In some experiments, corrections were made for the amount of lipin protein in immune complex assays. This was accomplished by dividing PAP activity by the relative amount of lipin estimated from lipin immunoblots.
Generation of cDNA Constructs Encoding Lipin Proteins-cDNA encoding lipin 1b (accession number AF412811) flanked by EcoRI sites was generated by PCR using PFU Turbo DNA polymerase (Stratagene) and appropriately designed primers. The PCR product was digested with EcoRI and introduced into the EcoRI site of pBluescript to generate pBluescript lipin 1b . Next a DNA fragment encoding a triple HA epitope tag was excised from pKH3 (23) by using EcoRI and SalI and inserted between the EcoRI site upstream of the lipin coding region and the SalI site of pBluescript lipin 1b . The resulting pBluescript HA-lipin 1b was then cut with NotI and KpnI, and the fragment containing HA-lipin was introduced between the NotI and KpnI sites of pcDNA3 to generate pcDNA3 HAlb . cDNAs encoding lipin proteins having point mutations were generated by oligonucleoti-de-directed mutagenesis of pcDNA3 HAlb by using a QuikChange II kit (Stratagene). Deletion constructs were generated by using PCR and appropriate primers. All constructs were cassette-subcloned, and the mutated fragments were sequenced and found to be free of errors.
Electroporation of 3T3-L1 Adipocytes and Transfection of 293T Cells-Adipocytes (4 -6 days postdifferentiation) were removed from the culture dishes (150-mm diameter) by incubating with trypsin (0.05%) for 5 min. The cells were collected by centrifugation (1000 ϫ g for 5 min), rinsed once with Buffer C, and suspended in Buffer C (350 l/dish), which contained 120 mM KCl, 0.15 mM CaCl 2 , 2 mM EGTA, 5 mM MgCl 2 , 2 mM ATP, 5 mM glutathione, 10 mM potassium phosphate, and 25 mM HEPES, pH 7.6. Cells (350 l) were added to cuvettes (4 mm) containing 50 g of the appropriate pcDNA3 plasmid and then electroporated with a single pulse from a Gene Pulser (Bio-Rad) set at 280 V and 960 microfarads. Transfection efficiency judged by electroporating cells with a vector encoding green fluorescent protein was 40 -50%. 293T cells (10-cm plate for each vector) were transfected with pcDNA3 vectors as described previously (19).
Expression of HA-lipin in 3T3-L1 Adipocytes by Adenoviralmediated Gene Transfer-Viruses for expressing HA-lipin and A106 HA-lipin were prepared using procedures described by He et al. (24). Briefly, cDNAs encoding the HA-tagged lipin 1b proteins were excised with KpnI and NotI from the pBluescript constructs and inserted between the KpnI and NotI sites in the shuttle vector, pAdTrackCMV. The resulting plasmids were cotransformed with the adenoviral backbone plasmid, pAdEasy-1, into BJ5183 bacteria. Recombined plasmids were selected and transfected into HEK 293 cells to generate virus, which was amplified and then purified by CsCl gradient centrifugation to create a high titer viral stock. 3T3-L1 adipocytes were infected essentially as described by Kotani et al. (25). The efficiency of infection judged by expression of green fluorescent protein, which is also encoded by the HA-lipin virus, was ϳ50%. The adipocytes were used 40 -48 h after infection. Virus encoding ␤-galactosidase was used as a control.
Antibodies-The two lipin antibodies, LAb-1 and LAb-2, prepared by immunizing rabbits with peptide conjugates have been described previously (5). The peptide for LAb-1 corresponded to the sequence between Ser 409 and Gly 430 in mouse lipin 1b, and that for LAb-2 corresponded to the last 11 amino acids in the protein. Monoclonal antibodies to the HA epitope were purified from culture medium from 12CA5 hybridoma cells.
To generate antibodies that specifically recognize lipin phosphorylated in Ser 106 , rabbits were immunized with the following phosphopeptide coupled to keyhole limpet hemocyanin: Cys-Tyr-Leu-Ala-Thr-Ser(P)-Pro-Ile-Leu-Ser-Glu (where Ser(P) represents phosphorylated Ser). To select the phosphospecific antibodies, the serum was first incubated with the nonphosphorylated peptide that had been coupled to SulfoLink resin (Pierce). Next, the fraction that did not bind to the nonphosphorylated peptide resin was incubated with the phosphopeptide coupled to SulfoLink. After washing the column, the phosphospecific antibodies were eluted at pH 2.7, immediately neutralized, and affinity-purified using protein A-agarose.
Immunoprecipitations-Adipocyte extract samples (800 l) were incubated with lipin antibodies (2 g) bound to protein A-agarose beads (15 l) or with the anti-HA antibody, 12CA5 (2 g), bound to protein G-agarose beads (15 l), at 4°C for 12 h with constant mixing. As a control for specificity, rabbit or mouse nonimmune IgG was substituted for the lipin antibodies or 12CA5, respectively. The beads were then washed four times with Buffer A supplemented with 1% Triton X-100.
Electrophoretic Analyses-Protein samples were subjected to electrophoresis in 8.75% polyacrylamide gels in the presence of SDS (30). Relative amounts of 32 P were determined using a PhosphorImager (Amersham Biosciences). Immunoblots were prepared as described previously (5), except with different antibodies, and signals were detected by using either film or a CCD camera system (Fujifilm LAS3000). Band intensities were determined by using the volume integration function of Image-Quant (Amersham Biosciences).
Mass Spectrometry-After SDS-PAGE and staining with Coomassie Blue, the band containing lipin was sliced from the gel. In-gel digestion of the protein with trypsin was performed as described previously (26). An aliquot of the resulting peptides was analyzed by nanoflow HPLC coupled to microelectrospray ionization on an LCQ Deca ion trap mass spectrometer (Thermo Electron Corp., San Jose, CA) (27). MS/MS spectra acquired were searched against a human-rat-mouse data base (National Center for Biotechnology Information, available on the World Wide Web at www.ncbi.nlm.nih.gov/) using SEQUEST (28). Peptide sequence and phosphorylation site assignments and were validated by manual interpretation of the MS/MS spectra.

RESULTS
Fatty Liver Dystrophy Mutations-Mg 2ϩ -dependent PAP activities were measured in extracts of brain, heart, kidney, lung, skeletal muscle, and liver from wild type and fld/fld mice that had been fed ad libitum (Fig. 1). Most tissues from fld/fld mice contained dramatically lower levels of Mg 2ϩ -dependent PAP activity than the corresponding tissues from wild type mice (Fig. 1A). An exception was the liver, where the activity from fld/fld mice was approximately half of that measured in wild type liver. Thus, enzymes other than lipin account for much of the hepatic Mg 2ϩ -dependent PAP in fed mice. Although differing amounts of Mg 2ϩ -independent activity were observed in different tissues (Fig. 1B), the Mg 2ϩ -independent activities from the fld/fld tissues were not significantly different from the wild type activities.
Consistent with the previously observed effect on increasing lipin protein (29), fasting increased Mg 2ϩ -dependent PAP activity in livers from wild type animals by ϳ2-fold (Fig. 1A). In contrast, fasting did not change Mg 2ϩ -dependent PAP activity in livers of fld/fld mice. Thus, lipin appears to account for most, if not all, of the Mg 2ϩ -dependent PAP activity induced in the liver by fasting.
To investigate the effect of the Gly 84 3 Arg mutation found in fld 2j mice, HA-tagged forms of wild type and mutant lipin proteins were overexpressed in HEK 293 cells, and Mg 2ϩ -dependent PAP activities were measured in cell extracts. Approximately equal amounts of lipin 1b and lipin 1a were expressed in the cells, and the PAP activities of these proteins were very similar ( Fig. 2A). Less of the R84 lipin accumulated compared with the wild type lipin 1b, but even after correcting for the expression levels, the PAP activity of the mutant lipin was only ϳ20% that of the wild type enzyme ( Fig. 2A). By comparison, the activity of the lipin protein in which the catalytic Asp in the HAD domain was mutated to Glu was negligible.
Removing the COOH-terminal 15 amino acids from lipin did not significantly affect the PAP activity (Fig. 2B, 1-909). However, removing an additional 37 amino acids (Fig. 2, 1-872) resulted in a striking reduction in activity. Deleting the entire CLIP domain by COOH-terminal truncation abolished PAP activity (Fig. 2, 1-641), which was expected, since this truncated protein lacks the HAD motif that is involved in catalysis. Interestingly, deleting the NLIP domain (Fig. 2, ⌬106) also dramatically decreased PAP activity. Thus, both NH 2 -and COOHterminal regions of lipin are required for full PAP activity.
Phosphorylation of certain sites in lipin decreases the mobility of the protein in SDS-PAGE, explaining, at least in part, the multiple electrophoretic forms of the proteins that were detected (Fig. 2, A and B). R84 exhibited less electrophoretic heterogeneity than the other lipin proteins, and the average mobility of the mutant protein was higher ( Fig. 2A), suggesting that it was not as highly phosphorylated as the wild type lipin. Lipin proteins with COOH-terminal truncations of up to 283 amino acids still exhibited electrophoretic heterogeneity indicative of multisite phosphorylation (Fig. 2B). Consistent with their smaller size, the average mobilities of the truncated proteins were higher than wild type lipin.
Sites in Both the NH 2 -and COOH-terminal Regions of Lipin Are Phosphorylated-Endogenous lipin from 3T3-L1 adipocytes also exhibited electrophoretic heterogeneity, indicative of multisite phosphorylation (Fig. 3A). Incubating 32 P-labeled adipocytes with insulin increased the phosphorylation of lipin, as evidenced both by a gel shift and by an increase in 32 P-labeled lipin (Fig. 3A). The effects of insulin were attenuated by rapamycin, implicating mTOR in the hormonal response. Inhibiting PI 3 kinase, which signals upstream of mTOR (30), with wortmannin also blocked the effects of insulin. These results indicate that lipin phosphorylation is controlled by insulin in a similar manner in 3T3-L1 adipocytes and in primary rat adipocytes, where the hormone was previously shown to increase lipin phosphorylation (5).
To facilitate the analyses of lipin phosphorylation, adenoviral vectors were developed to allow overexpression of epitopetagged forms of lipin in adipocytes. Insulin stimulated the phosphorylation of overexpressed HA-lipin (see Fig. 5B). Indeed, when adjusted for the amount of lipin protein estimated by immunoblotting, the amounts of 32 P incorporated into the endogenous and overexpressed lipin were very similar. 4 As a first step in identifying regions of lipin phosphorylated in response to insulin, 32 P-labeled HA-lipin from adipocytes was incubated briefly with trypsin, and the resulting fragments were immunoprecipitated with HA antibodies or the lipin peptide antibodies, LAb-1 or LAb-2 (Fig. 3B). Several 32 P-labeled peptides were immunoprecipitated by the three antibodies, but as would be expected from the different locations of the epitopes, the labeled peptides exhibited distinctly different mobilities when subjected to SDS-PAGE. A fragment of appar-  ent M r ϳ30,000 was the predominant 32 P-labeled species in HA immunoprecipitates. Considering the size of the fragment and the location of the HA tag at the NH 2 terminus of lipin, the M r ϳ30,000 fragment contains the NH 2 -terminal region of lipin, including the conserved NLIP domain. A fragment of apparent M r ϳ45,000 was the major 32 P-labeled species recovered with LAb-2. Since LAb-2 binds the COOH-terminal 11 amino acids of lipin, this M r ϳ45,000 fragment includes the conserved CLIP domain.
Based on 32 P labeling of the fragments recovered with the three antibodies, the effect of insulin on increasing phosphorylation was most pronounced in the NH 2 -terminal region of lipin. When corrected for the amounts of peptide present, the results indicate that the phosphorylation of the NH 2 -terminal sites were increased by ϳ2-fold in response to the hormone, whereas the COOH-terminal sites were modestly affected (Fig.  3C). Likewise, the effects of rapamycin on decreasing phosphorylation were much more pronounced on the NH 2 -terminal region of lipin than on sites in the COOH-terminal region of the protein.
Mass Spectrometric Analysis of Phosphorylation Sites-To identify phosphorylation sites, HA-lipin peptides generated by exhaustive proteolysis with trypsin were resolved by reverse phase HPLC and analyzed by mass spectrometry. Phosphorylated peptides were identified from the mass contributed by phosphate (80 atomic mass units). The MS/MS spectrum derived from a representative phosphopeptide is shown in Fig.  4A. In this case, fragments of the peptide that contained Ser(P) 106 (pS in sequence YLATpSPILSEGAAR) had the mass predicted from the amino acid composition plus 80 atomic mass units. Fragments differing by 98 atomic mass units result from the loss of phosphoric acid during fragmentation. Thus, Ser 106 is a site of phosphorylation in lipin.
Including Ser 106 , 15 phosphorylation sites were unambiguously assigned based on the MS/MS spectra of the phosphopeptides (Fig. 4B). The location of the sites with respect to other features in lipin 1b is shown in Fig. 4C. Four phosphopeptides were detected in which the position of the phosphorylated residue could not be assigned. In these peptides, two potential sites were closely spaced (Ser 281 and Thr 282 , Ser 353 and Ser 356 , Ser 647 and Ser 648 , Ser 720 and Thr 722 ). Although the MS/MS spectra definitively placed phosphate in the pair, we were unable to determine which residue in the pair was phosphorylated. Thus, results of the mass spectrometric analyses of lipin indicated that there were at least 19 phosphorylation sites.
Insulin-stimulated Phosphorylation of Ser 106 -Several phosphorylation sites were identified in the NH 2 -terminal region of lipin. Mutational analyses of lipin were conducted to determine which of these NH 2 -terminal sites represented the major site of insulin-stimulated phosphorylation found in the M r ϳ30,000 peptide generated by limited trypsin proteolysis. Results with ⌬106 lipin provided an important clue. Deleting the first 106 amino acids not only increased the electrophoretic mobility of lipin but also abolished the shift in electrophoretic mobility produced by insulin (Fig. 5A). This finding suggested that the site responsible for the gel shift resided in the first 106 amino acids of lipin. Ser 106 was the only site identified by mass spectrometry in this region, and introducing a Ser 106 3 Ala mutation clearly limited the gel shift produced by insulin (Fig. 5A), indicating that Ser 106 is phosphorylated in response to the hormone.

. Peptide mapping by limited trypsin proteolysis indicates that the NH 2 -terminal region of lipin is phosphorylated in response to insulin.
A, 3T3-L1 adipocytes were incubated for a total of 135 min in medium supplemented with 0.2 mCi/ml [ 32 P]phosphate. During this time, the cells were incubated as follows: no additions, 10 milliunits/ml insulin (INS) for the last 15 min, 20 nM rapamycin (RAP) for the last 75 min, rapamycin for 60 min followed by insulin plus rapamycin for the final 15 min, 100 nM wortmannin (Wm) for the last 75 min, and wortmannin for 60 min followed by wortmannin plus insulin for the last 15 min. 32 P-Labeled lipin was immunoprecipitated and subjected to SDS-PAGE. A phosphor image and a lipin immunoblot are shown. B and C, 2 days after infection with an adenoviral construct encoding HA-tagged lipin, 3T3-L1 adipocytes were incubated for a total of 2 h in medium supplemented with 0.2 mCi/ml [ 32 P]phosphate. During this time, the cells were incubated as follows: no additions (CON), 10 milliunits/ml insulin (INS) for the last 15 min, 20 nM rapamycin (RAP) for the last 1 h, or rapamycin for 30 min followed by insulin plus rapamycin for the final 30 min (RAP ϩ INS). After preparing extracts, HA-lipin was immunoprecipitated by using HA antibodies. The immune complexes were washed twice (1 ml/wash) with a solution containing 1 mM dithiothreitol and 10 mM sodium phosphate, pH 7.4, and suspended in 25 l of the same solution. After adding trypsin (100 ng), the samples were incubated at 37°C for 4 min. The digestions were terminated by incubating the samples at 95°C for 5 min after adding 50 l of a stop solution containing 2% SDS, 1 mM dithiothreitol, 10 g/ml soybean trypsin inhibitor, and 10 mM sodium phosphate, pH 7.4. The volume of each sample was adjusted to 1 ml by adding stop solution (minus SDS) that had been supplemented with 1.2% Triton X-100, and fragments were immunoprecipitated with antibodies to HA, LAb-1 (1), or LAb-2 (2). B, samples were subjected to SDS-PAGE, and an autoradiogram of the 32 P-labeled fragments was prepared. Binding sites of HA, LAb-1, and LAb-2 in lipin are depicted in the model. C, the relative amounts of 32 P recovered in the M r ϳ30,000 NH 2 -terminal fragment (HA-IP) and the M r ϳ45,000 COOH-terminal fragment (LAb-2 IP) were determined by phosphorimaging. The results are expressed as percentages of the control values and are means Ϯ S.E. from three experiments.
The phosphorylation of Ser 106 was confirmed by immunoblotting lipin with a phosphospecific antibody directed against the site. Insulin caused a severalfold increase in reactivity of this antibody with lipin (Fig. 5B). The antibody failed to react with A106 lipin or with lipin that had been dephosphorylated with PP1 C (Fig. 6A), providing additional support for the specificity of the antibody for Ser(P) 106 . Incubating cells with rapamycin abolished the reactivity of lipin with the phosphospecific antibody (Figs. 5B and 6A), indicating that phosphorylation of Ser 106 is dependent on the mTOR signaling pathway.
The finding that insulin decreases the electrophoretic mobility of most of the lipin protein (e.g. see Fig. 5A) indicates that Ser 106 is highly phosphorylated. However, this finding does not mean that other sites are not phosphorylated in response to insulin. To estimate the contribution of Ser 106 phosphorylation to the overall effect of insulin, 32 P-labeling experiments were conducted after overexpressing HA-tagged forms of wild type and A106 lipins. Insulin increased the 32 P labeling of both lipin proteins (Fig. 5B). After correcting for the amount of lipin protein recovered, 32 P incorporated into wild type lipin was found to be increased ϳ2-fold by insulin (Fig.  5C). The -fold increase in A106 phosphorylation was only slightly lower, indicating that sites other than Ser 106 are phosphorylated in response to insulin.
Reciprocal Changes in Lipin Phosphorylation Produced by Insulin and Epinephrine Are without Effect on PAP Activity-In contrast to insulin, both epinephrine and oleic acid increased the electrophoretic mobility of lipin (Fig. 6A), indicative of dephosphorylation. Incubating cells with insulin, rapamycin, epinephrine, or oleic acid was without effect on PAP activity measured in immune complexes isolated using lipin antibodies (Fig. 6B). An implication is that changes in the phosphorylation of lipin were insufficient to change PAP activity, at least  (34). NLIP and CLIP are conserved regions identified by Reue and co-workers (4). NLS, location of the putative nuclear localization sequences. b, the 33-amino acid insert that distinguishes lipin 1b from lipin 1a. Brackets indicate pairs in which either residue may have been phosphorylated. The Asp necessary for phosphatase activity is located in the HAD domain located in the CLIP region. LHR and NHR, regions of yeast lipin that have very low homology or no homology, respectively, to mammalian lipins. Note that most of the phosphorylation sites that have been identified in Pah1p are found in the no homology region.
under the conditions of these assays. To investigate further the possible effect of lipin phosphorylation, immune complexes were incubated with recombinant PP1 C , the protein phosphatase 1 catalytic subunit, before measuring PAP activities. After incubating with the phosphatase, lipin collapsed into a single band of relatively high mobility in SDS-PAGE (Fig. 6A), confirming that the electrophoretic heterogeneity of lipin was due to differences in phosphorylation. PP1 C also abolished reactivity with the phosphospecific antibody (Fig. 6A), providing additional evidence of the efficiency of dephosphorylation. However, PP1 C had no effect on PAP activity, supporting the conclusion that the activity of lipin was not affected by changes in its phosphorylation state.
Insulin and Epinephrine Promote Reciprocal Changes in Lipin Localization Correlated with Changes in Phosphorylation-Several agents have been shown to affect the subcellular distribution of Mg 2ϩ -dependent PAP. To investigate the effects of insulin on the subcellular distribution of lipin, cells were homogenized in a detergent-free solution containing 0.25 M sucrose. The homogenates were fractionated by differential centrifugation to yield fractions containing the following: plasma membrane/ mitochondria/nuclei, microsomal membranes, and soluble proteins. Relatively little lipin or Mg 2ϩ -dependent PAP activity was detected in the plasma membrane/mitochondria/nuclear fraction. 4 Insulin markedly increased the amounts of lipin (Fig.  7, A and C) and PAP activity (Fig. 7D) in the soluble fraction. Since the effect of insulin appeared to be too rapid to be explained by synthesis of new lipin protein, it seemed more likely that insulin was promoting a change in the subcellular distribution of lipin. As expected, insulin decreased both lipin protein (Fig. 7, A and C) and PAP activity (Fig. 7D) in the microsomal fraction.
The relative amount of soluble Mg 2ϩ -dependent PAP in control 3T3-L1 adipocytes was lower than that observed in previous experiments in untreated rat adipocytes (ϳ28% versus 56%) (15). Although it would not be unreasonable to expect differences between cultured 3T3-L1 adipocytes and primary fat cells, other factors might also have contributed to the different amounts of soluble activity measured. For example, our FIGURE 5. Insulin-stimulated, rapamycin-sensitive phosphorylation of Ser 106 . A, pcDNA3 constructs encoding HA-tagged forms of wild type lipin (WT), ⌬106, and lipin in which Ser 106 was replaced with Ala (A106) were electroporated into 3T3-L1 cells 48 h before the cells were incubated without or with 10 milliunits/ml insulin for 20 min. Immunoblots of the HA-tagged lipin proteins are presented. B and C, 3T3-L1 adipocytes were infected with adenoviruses encoding ␤-galactosidase (as a control) or HA-tagged forms of wild type (WT) lipin or A106 lipin. After 48 h, the cells were labeled with 32 P and incubated with insulin and rapamycin as described in the legend to Fig. 3. HA lipin proteins were immunoprecipitated, and immune complexes were subjected to SDS-PAGE before proteins were transferred to Immobilon membranes. B, a phosphor image of the 32 P-lipin proteins is shown, along with immunoblots prepared using either phosphospecific antibodies to the Ser 106 site or antibodies to lipin. C, after correcting for differences in the amounts of lipin, the 32 P content of lipin was expressed relative to the maximum. Results are mean values Ϯ S.E. from four experiments. After homogenizing cells in 1 ml of Buffer A plus 1% Triton X-100, lipin was immunoprecipitated from extracts by using LAb-1. Immune complexes were washed four times and suspended in 100 l of a solution containing 50 mM NaCl, 1 mM MnCl 2 , 1 mM dithiothreitol, and 10 mM HEPES, pH 7.4. Samples (50 l) of the suspension were incubated at 30°C for 45 min without or with 1.25 g/ml purified recombinant catalytic subunit of protein phosphatase 1␣ (PP1 C ). Next, the beads were washed with 1 ml of Buffer A plus 1% Triton X-100 and 0.5 M microcystin and then suspended in 100 l of Buffer B. A, samples (40 l) were subjected to SDS-PAGE, and immunoblots were prepared to detect Ser(P) 106 and lipin. B, PAP activities associated with the beads (10-l suspensions) were measured in the presence of 2 mM MgCl 2 . In three experiments, the activity in control cells averaged 18.9 Ϯ 2.6 nmol/min/10-cm plate. After correcting for the relative amounts of lipin protein determined from immunoblots, activities were expressed as percentages of the control. The results are mean values Ϯ S.E. from three experiments.
homogenization buffer was supplemented with protease inhibitors and protein phosphatase inhibitors, including 50 mM NaF, and it had a higher ionic strength than that used in the previous study (15). Based on earlier findings (31), the higher ionic strength would be expected to stabilize binding of Mg 2ϩ -dependent PAP to membranes. Moreover, in contrast to the previous study (15), we did not routinely include albumin in the homogenization buffer. In control experiments, we found that adding 10 mg/ml fatty acid-free albumin to the buffer used to homogenize 3T3-L1 adipocytes increased the relative amount of soluble lipin by only 15 Ϯ 4% (n ϭ 3 experiments). Importantly, removing free fatty acids with albumin did not block the effects of insulin on the distribution of Mg 2ϩ -dependent PAP between the soluble and membrane-bound forms. 4 Rapamycin did not affect the subcellular distribution of lipin or PAP activity in either the absence or presence of insulin. In contrast to insulin, oleic acid and epinephrine decreased the amount of soluble lipin protein (Fig. 7, B and C) and PAP activity (Fig. 7D). Insulin blocked the responses to 330 M oleic acid but was ineffective in counteracting the effects of epinephrine (Fig. 7C). None of the treatments significantly affected the total (microsomal plus soluble) PAP activity. 4 Interestingly, most of the membrane-associated lipin was found in a high mobility band that was essentially absent in the soluble fraction (Fig. 7, A and B). This band presumably represents a hypophosphorylated form of lipin, since its mobility was the same as that of lipin that had been dephosphorylated in vitro (see Fig. 6A). Moreover, the increase in soluble lipin produced by insulin was associated with loss of the hypophosphorylated form (Fig. 7B), suggesting that phosphorylation of lipin releases the enzyme from intracellular membranes. Consistent with this hypothesis, oleic acid and epinephrine increased the electrophoretic mobility of lipin as well as the amount of microsomal lipin (Fig. 7B).

DISCUSSION
Mice homozygous for the fld of the Lpin1 gene represent a useful model for evaluating the role of lipin in vivo (4). The present finding that tissues from fld/fld mice contained markedly less PAP activity than the corresponding wild type tissues indicates that lipin is a major Mg 2ϩ -dependent PAP in mammalian cells (Fig. 1). The lack of PAP activity normally provided by lipin surely contributes to the failure of fld/fld mice to accumulate TAG in fat cells, since PAP catalyzes the penultimate step in TAG synthesis (2). Loss of PAP activity may also partially explain the lipodystrophy of fld 2j / fld 2j mice, since lipin protein harboring the fld 2j mutation (Gly 84 3 Arg) exhibited only a fraction of the PAP activity of wild type lipin when overexpressed in cells (Fig. 2). Previous studies indicated that this mutant form of lipin was excluded from the nucleus (4). Recent evidence that lipin functions to regulate the activities of PGC-1␣ and PPAR␣ in hepatocytes (29) is consistent with the view that loss of nuclear functions of lipin also contributes to phenotypic changes in fld/fld and fld 2j /fld 2j mice (4).
The phosphorylation of lipin was first observed in rat adipocytes, where it was shown to be increased in a rapamycin-sensitive manner in response to insulin (5). The present results demonstrate that insulin-stimulated phosphorylation of lipin occurs in 3T3-L1 adipocytes (Figs. 3, 5, 6, and 7). The results also show for the first time that the phosphorylation of lipin is decreased by epinephrine (Figs. 6 and 7). The reciprocal effects of insulin and epinephrine on lipin phosphorylation are consistent with the opposing actions of these hormones on TAG accumulation in adipocytes. However, changes in phosphorylation produced by the hormones did not significantly alter the intrinsic PAP activity of lipin (Figs. 6 and 7). Moreover, dephosphorylation of lipin in vitro by using PP1 C did not change the PAP activity of the enzyme (Fig. 6), providing additional evidence that activity is not affected by the phosphorylation state of lipin.
Lipin was recovered primarily in the soluble and the microsomal fractions of 3T3-L1 adipocytes (Fig. 7). Insulin decreased the membrane-associated form, accounting for the increase in soluble lipin. Conversely, epinephrine decreased soluble lipin and increased the membrane-associated form. With both insulin and epinephrine, the changes in PAP activity in the two fractions exactly matched the changes in protein. These find- FIGURE 7. Changes in lipin phosphorylation produced by insulin, epinephrine, and oleic acid are associated with redistribution of lipin between soluble and membrane-bound forms. A, 3T3-L1 adipocytes (10-cm plate/treatment) were incubated for 2 h as follows: no additions (C), 10 milliunits/ml insulin for the last 60 min (I), 20 nM rapamycin for 120 min (R), and rapamycin for 60 min followed by insulin plus rapamycin for 60 min (RI). Cells were homogenized in 1 ml of Buffer B, and a pellet fraction containing nuclei, mitochondria, and plasma membranes was prepared by centrifuging the homogenates at 16,000 ϫ g for 20 min. The 16,000 ϫ g supernatant was centrifuged at 175,000 ϫ g for 1 h to obtain a supernatant fraction containing soluble proteins and a pellet fraction containing microsomal membranes, which were suspended in 1 ml of Buffer B. Samples ings indicate that changing subcellular localization is a major mechanism for controlling lipin function, supporting previous conclusions based on measurements of Mg 2ϩ -dependent PAP activity that were made before the lipin protein was identified (14,15).
Oleic acid also decreased soluble lipin and increased lipin in the microsomal fraction. Since epinephrine stimulates lipolysis in adipocytes, an increase in intracellular free fatty acids may be involved in the effects of the hormone on lipin. The interpretation that fatty acids increase the membrane localization of PAP has been called into question by the finding that adding oleic acid to a soluble, membrane-free, preparation of the PAP in vitro causes a portion of the enzyme to sediment upon centrifugation (32). Such an action cannot explain the present finding that the increase in microsomal lipin produced by oleic acid was associated with dephosphorylation of the protein (Fig. 7).
The soluble form of lipin was more highly phosphorylated than the membrane-bound form, as evidenced by its lower electrophoretic mobility in SDS-PAGE (Fig. 7). Thus, the changes in lipin distribution produced by insulin and epinephrine directly correlated with changes in lipin phosphorylation. This correlation suggests that phosphorylation controls the intracellular localization of the protein. However, the results would also be consistent with a model in which changes in phosphorylation occur secondarily to the change in intracellular localization. Such appears to be the case with CTP:phosphocholine cytidylyltransferase, which is dephosphorylated after becoming bound to membranes (33).
It is clear that Ser 106 is not required for the effect of insulin on lipin localization, since rapamycin abolished phosphorylation of Ser 106 (Fig. 5) but did not attenuate the effect of insulin on increasing soluble lipin (Fig. 7). However, these findings with rapamycin do not eliminate a role of phosphorylation in controlling the localization of lipin, since rapamycin did not fully inhibit insulin-stimulated phosphorylation of the protein (Fig.  5C). Indeed, rapamycin was without effect on the insulin-stimulated loss of the highest mobility band that was associated with membrane binding (Fig. 7A). Defining the role of phosphorylation will require identification of other hormonally responsive sites, an objective complicated by the large number of phosphorylation sites in lipin. Mass spectrometric analyses indicated that at least 19 sites contained some phosphate (Fig. 4). An almost equal number were found in the NH 2 -and COOH-terminal halves of lipin, although some regions were enriched in sites, such as the stretch between Ser 281 and Thr 298 , which contained six sites. Many of the sites identified in mouse lipin are present in other mammalian lipin proteins, and three (Ser 106 , Ser 634 , and Ser 720 ) appear to have been conserved from yeast to humans.
None of the highly conserved sites were found to be phosphorylated in a recent analysis of phosphorylation sites of lipin (Pah1p) in S. cerevisiae (34). In this study, the PAP activity of Pah1p harboring Ala mutations in seven (Ser/Thr)-Pro phosphorylation sites exhibited enhanced PAP activity (34). Curiously, the large majority of phosphorylation sites in yeast Pah1p were found in a region of the protein lacking homology with lipins from other species (Fig. 4C). Consequently, how the role of lipin phosphorylation in yeast relates to that in mammals is unclear.
In view of the findings that both yeast (34) and mammalian lipins (Fig. 4) are phosphorylated in several (Ser/Thr)-Pro phosphorylation sites, it is interesting to speculate that one or more of the protein kinases and/or phosphatases acting on lipin proteins in yeasts and mammals have been conserved. Santos-Rosa et al. (10) have presented evidence that the yeast Pah1p is dephosphorylated by a membrane-localized CPD phosphatase complex, Nem1/Spo7. Thus, it is interesting to speculate that Dullard (35), the mammalian homolog of yeast Nem1, might be a lipin phosphatase. Yeast Pah1p is believed to be phosphorylated by the cyclin-dependent kinases Cdc28p and/or Pho85p (10,34). Although there may be instances in which mammalian homologs of Cdc28, such as Cdc2, Cdk2, and Cdk3, phosphorylate lipin, the activities of most cyclin-dependent kinases are relatively low in terminally differentiated adipocytes. Therefore, other kinases, such as mitogen-activated protein kinases (36) or mTOR (37), that have been shown to phosphorylate (Ser/Thr)-Pro sites would seem to be better candidates for phosphorylating lipin in fat cells. In any event, considering the diverse nature of the sites in lipin, the phosphorylation of the protein almost certainly involves multiple protein kinases. Several of the sites in lipin did not appear to conform to any defined consensus motif for phosphorylation, but four sites, including the highly conserved Ser 634 , are good candidates for phosphorylation by protein kinase CK2, and three sites fit the consensus for phosphorylation by GSK3.
Lipin is very highly expressed in adipose tissue (4,5), consistent with its important role in TAG synthesis. However, as discussed elsewhere (32), it is paradoxical that insulin increases TAG synthesis but decreases the membrane-bound phosphatase, which has been proposed to be the more active form in TAG synthesis (16). Perhaps changes in lipin localization have roles related to functions of phosphatidic acid not involving TAG synthesis. Decreasing microsomal lipin would favor accumulation of membrane phosphatidic acid, which has been shown to increase in response to insulin in adipose tissue (38). Phosphatidic acid is important in membrane fusion events, and it has been implicated in controlling the subcellular distribution of many proteins (39). It has also been reported to bind and modify the activity of several important protein kinases and phosphatases, including mTOR, Raf-1, SHP-1, and PP1 C (39). Although rapamycin did not affect lipin localization or PAP activity, the connection between mTOR and lipin phosphorylation is intriguing in view of the potential role of phosphatidic acid in controlling mTOR activity in cells (40,41). Future experiments will be needed to investigate this connection and the possible role of phosphorylation in controlling the function of lipin as a regulator of PPAR␣ and other transcription factors (29).