Modulation of the Mammalian Target of Rapamycin Pathway by Diacylglycerol Kinase-produced Phosphatidic Acid*

The protein known as mammalian target of rapamycin (mTOR) regulates cell growth by integrating different stimuli, such as available nutrients and mitogenic factors. The lipid messenger phosphatidic acid (PA) binds and positively regulates the mitogenic response of mTOR. PA generator enzymes are consequently potential regulators of mTOR. Here we explored the contribution to this pathway of the enzyme diacylglycerol kinase (DGK), which produces PA through phosphorylation of diacylglycerol. We found that overexpression of the DGKζ, but not of the α isoform, in serum-deprived HEK293 cells induced mTOR-dependent phosphorylation of p70S6 kinase (p70S6K). After serum addition, p70S6K phosphorylation was higher and more resistant to rapamycin treatment in cells overexpressing DGKζ. The effect of this DGK isoform on p70S6K hyperphosphorylation required the mTOR PA binding region. Down-regulation of endogenous DGKζ by small interfering RNA in HEK293 cells diminished serum-induced p70S6K phosphorylation, highlighting the role of this isoform in the mTOR pathway. Our results confirm a role for PA in mTOR regulation and describe a novel pathway in which DGKζ-derived PA acts as a mediator of mTOR signaling.

The protein known as mammalian target of rapamycin (mTOR) regulates cell growth by integrating different stimuli, such as available nutrients and mitogenic factors. The lipid messenger phosphatidic acid (PA) binds and positively regulates the mitogenic response of mTOR. PA generator enzymes are consequently potential regulators of mTOR. Here we explored the contribution to this pathway of the enzyme diacylglycerol kinase (DGK), which produces PA through phosphorylation of diacylglycerol. We found that overexpression of the DGK, but not of the ␣ isoform, in serum-deprived HEK293 cells induced mTOR-dependent phosphorylation of p70S6 kinase (p70S6K). After serum addition, p70S6K phosphorylation was higher and more resistant to rapamycin treatment in cells overexpressing DGK. The effect of this DGK isoform on p70S6K hyperphosphorylation required the mTOR PA binding region. Down-regulation of endogenous DGK by small interfering RNA in HEK293 cells diminished serum-induced p70S6K phosphorylation, highlighting the role of this isoform in the mTOR pathway. Our results confirm a role for PA in mTOR regulation and describe a novel pathway in which DGK-derived PA acts as a mediator of mTOR signaling.
Diacylglycerol kinase (DGK) 1 comprises an evolutionarily conserved family of lipid kinases that phosphorylate diacylglycerol (DAG) to produce phosphatidic acid (PA). These enzymes contain at least two N-terminal cysteine-rich domains and a conserved catalytic domain. The DGKs have additional functional domains that were used to classify them into five subgroups (I-V; for an extended review, see Refs. 1 and 2). Their structural diversity, as well as their distinct tissue expression and specific intracellular localization, may confer on each DGK isoform the ability to regulate distinct DAG and PA pools and thus to participate in different signaling complexes.
DAG is a second messenger whose levels increase transiently in response to hormones and growth factors. DAG is an allosteric activator of protein kinase C (PKC) (3) as well as nonkinase proteins such as GTPase-activating proteins and guanine nucleotide exchange factors for the Ras family of GTPases (4,5). Termination of the DAG-derived signal must be rapid and requires DGK activity. DGK␣ and are the best-studied isoforms; both function as DAG signaling terminators during T-cell receptor (TCR) activation.
Class I DGK␣ has two Ca 2ϩ binding motifs (1) and is expressed in spleen, thymus, skeletal muscle, lung, testis, and peripheral T cells (6). During TCR-induced T-cell activation, the enzyme translocates to the plasma membrane, where it down-regulates DAG levels and thus diminishes Ras protein activation (6). DGK␣ membrane translocation is rapid and transient and requires both tyrosine kinase activity and an increase in intracellular Ca 2ϩ levels. This isoform also diminishes the response elicited by an ectopically expressed muscarinic type I receptor (7). DGK belongs to class IV and has nearly ubiquitous expression. It has a PDZ binding motif, four ankyrin repeats, and a myristoylated alanine-rich PKC substrate (MARCKS) homology domain, which overlaps with a bipartite nuclear localization signal (NLS) (8,9). Like the ␣ isoform, DGK acts as a negative regulator of TCR signaling; its overexpression thus attenuates activation of the extracellular signal-regulated kinase (ERK) pathway in both T lymphocytes after TCR binding (10) and gonadotrope cells stimulated with gonadotrophin-releasing hormone (11). DGK-deficient T cells are hyper-responsive to TCR stimulation; accordingly, mice lacking the enzyme have a more robust immune response than wild-type mice (12). DGK membrane translocation is a sustained event that requires phosphorylation of the MARCKS domain by PKC (13) and dissociation from the cytoskeleton induced by phosphorylation of the C-terminal region (14).
Our previous data indicate that DGK-generated PA is necessary for IL-2-regulated G 1 to S-phase transition in IL-2-dependent cell lines (23,24) and for CD4 ϩ /CD8 ϩ cell survival during thymic development (25). DGK-derived PA is also essential for the stabilization and activity of the transcription factor HIF-1␣ during onset of the hypoxic response (26). These two processes, proliferation and hypoxic response, involve mTOR kinase activity (27)(28)(29). PA binds an mTOR region * This work was supported by Grant BMC2001-1066 from the Spanish Ministry of Science and Technology, Grant 08.37002272002 from the Comunidad de Madrid, and Grant G037179 from the Instituto de Salud Carlos III. 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.
In mammals, there are two homologous S6K proteins, S6K1 and S6K2. S6K1 consists of the p70S6K and p85S6K isoforms; the former is localized mainly in cytoplasm, whereas the latter is found in the nucleus due to the presence of an NLS. These isoforms are generated by use of different translational start codons of the same messenger; their sequences are thus identical, with the exception of a short region in the N terminus (33). S6K2 comprises two isoforms; both also have an NLS and are therefore nuclear enzymes (32, 34). p70S6K is the beststudied S6K; it is activated through a coordinated sequence of phosphorylations that begins at the auto-inhibitory region in the kinase C-terminal domain. The p70S6K hydrophobic domain is then phosphorylated, leading to phosphorylation of the T-loop by the phosphoinositide-dependent kinase (PDK) 1 (35). mTOR is proposed to phosphorylate the auto-inhibitory and the hydrophobic motif of p70S6K (36,37).
Here we explored the role of DGK in the mTOR pathway. We determined DGK activity during mitogenic stimulation or hypoxic treatment in vitro and observed that this activity was down-regulated in serum-starved cells and increased after serum stimulation or hypoxia. Moreover, DGK overexpression induced dose-dependent phosphorylation of the hydrophobic motif of p70S6K, an effect that requires the PA binding motif of mTOR. Our results suggest the existence of a pathway in which DGK-derived PA can modulate mTOR activity.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-The human embryonic kidney cell line HEK293 was cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (37°C, 5% CO 2 ). For transfection, cells were plated in 6-well plates. When cells reached 40 -50% confluence (24 h), transfection was carried out using the Jet-PEI reagent (PolyTransfection) with the following amounts of DNA: 0.2 g of EE-p70S6K, 1.5 g of pEF-Bos-GFP or pEF-Bos-GFP-DGK-CT, 0.5-1.5 g of pEF-Bos-GFP-DGK or pEF-Bos-HA-DGK␣, and 0.8 g of the mTOR constructs. The following day, the still actively growing cells were serum-starved (20 h) and then stimulated with 20% fetal bovine serum for 30 min before cell lysis. Where indicated, rapamycin (1-20 nM; Calbiochem) or R5949 (30 M, Calbiochem) was added to medium 30 min before stimulation.
Jurkat cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 2 mM glutamine.
Plasmid Constructs-The DGK constructs used have been described previously. The different versions of rat DGK were cloned in the pEF-Bos vector as GFP C chimeras (13), and murine DGK␣ was hemagglutinin (HA)-tagged (38). Human p70S6K was cloned in pTM1, so the protein was EE epitope-tagged (39). FLAG-tagged wild-type mTOR and mutant mTOR (3ATOR) encoding plasmids were kindly provided by Dr. J. Chen (University of Illinois, Urbana-Champaign, IL) (22).
Western Blot-Serum-starved and stimulated HEK293 cells were washed twice with cold phosphate-buffered saline (PBS) and lysed in p70 buffer (10 mM Hepes, pH 7.5, 15 mM KCl, 1 mM EDTA, 1 mM EGTA, 10% glycerol, 0.2% Nonidet P-40, 1 mM dithiothreitol, 50 mM NaF, 10 g/ml each leupeptin and aprotinin, 1 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, and 20 mM ␤-glycerol phosphate) by incubation for 20 min at 4°C with gentle rocking. The lysates were centrifuged (15,000 ϫ g, 15 min, 4°C), and supernatants were quantified and separated by SDS-PAGE using high resolution acrylamide mix and then transferred to nitrocellulose. The phosphorylation state of p70S6K and expression of the different constructs were evaluated using the appropriate antibody and the ECL detection kit (Amersham Biosciences). Where indicated, values corresponding to the phospho-Thr 389 bands were quantified by analysis of the films with the ImageJ Program, and the values were normalized against that of the corresponding p70S6K bands (pThr 389 /p70 ratio). Relative phosphorylation ratio was normalized to that obtained in either quiescent or serum-stimulated control cells as indicated. Antibodies employed were anti-HA (Covance); mouse polyclonal anti-tubulin; mouse monoclonal anti-vimentin (Sigma); mouse monoclonal anti-GFP; rabbit phospho-antibodies against the pan-PKC site, the pan-PDK1 site, and Thr 389 and Ser 421 /Thr 424 of p70S6K (Cell Signaling); mouse monoclonal anti-histones (Chemicon); mouse anti-human transferrin receptor (Zymed Laboratories Inc.); and mouse anti-Cdk2 and anti-p70S6K (Santa Cruz Biotechnology). Mouse monoclonal antibody mixture against DGK␣ was kindly provided by Dr. W. J. van Blitterswijk (Netherlands Cancer Institute, Amsterdam, the Netherlands) (40). Rabbit polyclonal anti-DGK antibody was a generous gift of Dr. M. K. Topham (University of Utah, Salt Lake City, UT) (14).
Diacylglycerol Kinase Activity Assays-HEK293 cells were washed twice and scraped with cold PBS. Cells were centrifuged (3000 ϫ g, 5 min, 4°C), and the pellet was frozen, resuspended in lysis buffer (50 mM Tris-HCl, pH 7.4, 50 mM NaF, 10 g/ml each leupeptin and aprotinin, 1 mM phenylmethylsulfonyl fluoride, and 1 mM sodium orthovanadate), sonicated, and centrifuged (800 ϫ g, 10 min, 4°C). The supernatant was assayed for total protein quantification. An aliquot of each sample was reserved to evaluate DGK expression levels by Western blot. A sample of 25 g of each lysate was mixed with 5 l of DAG-octyl-glucoside micelles and 10 l of kinase mix (10 mM ATP, 10 mM MgCl 2 , and [␥-32 P]ATP). The phosphorylation reaction was incubated (10 min, room temperature). Radiolabeled lipids were extracted with chloroform/ methanol (1:1, v/v), dried under a nitrogen stream, dissolved in 20 l of chloroform/methanol (2:1, v/v), and analyzed by TLC using a chloroform/methanol/4 M ammonia (9:7:2, v/v/v) system with DiC18PA as control. The TLC was exposed for 3 h or overnight to detect overexpressed or endogenous DGK activity, respectively.
Fractionation Analysis-Subcellular fractionation was carried out as described previously (14), with minor modifications. Briefly, the HEK293 cell pellet was resuspended in ice-cold lysis buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 1 mM dithiothreitol, 50 mM NaF, 10 g/ml each of leupeptin and aprotinin, 1 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, and 20 mM ␤-glycerol phosphate), and cells were lysed by passage 30 times through a 30gauge needle. Lysates were centrifuged (14,000 ϫ g, 10 min, 4°C). The pellet corresponded to the crude nuclear fraction. Supernatant was centrifuged (100,000 ϫ g, 1 h, 4°C); the supernatant corresponded to the cytosolic fraction, and the pellet was resuspended thoroughly in 1% Triton X-100 containing lysis buffer to extract membrane-associated proteins. This suspension was centrifuged (100,000 ϫ g, 1 h, 4°C). The supernatant corresponded to the membrane fraction. The pellet corresponding to the cytoskeletal fraction was resuspended in SDS-PAGE loading buffer. The purity of the fractions was assessed by Western blot analysis with antibodies against vimentin, human transferrin receptor, histones, and Cdk2 as cytoskeletal, membrane, nuclear, and nuclear/ cytosolic markers, respectively.
Immunofluorescence and Confocal Microscopy-The day before transfection, HEK293 cells were seeded on poly-L-lysine-coated coverslips. After serum stimulation, cells were washed with ice-cold PBS, fixed for 10 min with 4% formaldehyde, washed three times with 150 mM Tris-HCl, pH 7.4, and permeabilized for 10 min with PBS and 0.2% Triton X-100. Coverslips were washed twice with PBS, blocked with PBS and 1% bovine serum albumin, and washed three times with PBS. Rabbit anti-DGK or mouse anti-HA primary antibodies and appropriate Cy3-conjugated goat anti-IgG were used to stain overexpressed DGK. Preparations were examined by confocal microscopy (TCS-NT; Leica).
Small Interfering RNA-The 21-nucleotide sequence for human DGK (nucleotides 2290 -2310) was selected as the target sequence. A BLAST search showed no significant similarity to any other sequence in the database. siRNA was synthesized chemically (Ambion), and transfection was performed using Oligofectamine (Invitrogen) with 100 nM siRNA. The next day, cells were divided and plated into fresh medium. Two days later, cells were starved for 24 h and then serum-stimulated. A mock siRNA transfection was performed simultaneously using GFP (nucleotides 238 -268) as a target.

DGK Activity Increased in Response to Mitogenic Stimuli
and Hypoxia-We predicted that the dual action of DGK might be integrated in mTOR activity. Mitogenic agents produce DAG, which is used by DGK, thus diminishing the mitogenic stimulation of mTOR, whereas conversion of DAG to PA positively stimulates mTOR. As stated above, DGK␣ and are the best-studied isoforms. DGK␣ is proposed to participate in attenuation of the TCR response and IL-2-dependent T-cell proliferation (6,7,(23)(24)(25); the isoform has been assigned a role as down-regulator of the TCR response (10 -12). We explored the role of these two isoforms in mTOR regulation using HEK293 cells, a cell line widely used to dissect the mTOR pathway. DGK␣ is highly expressed in T cells and has limited tissue distribution (1), whereas DGK displays a broad expression pattern (9). In accordance with this observation, analysis of total extracts of HEK293 cells by immunoblot showed no DGK␣ expression, whereas Jurkat T cells showed high DGK␣ levels ( Fig. 1A, top panel). In HEK293 cells and the Jurkat T cell-line, anti-DGK antibody revealed high protein expression (Fig. 1A, bottom panel). In addition, this antibody recognized a nonspecific band of ϳ130 kDa.
We evaluated endogenous DGK activity in HEK293 cells under two conditions known to modify PA levels and mTOR activity, addition of a hypoxia mimetic agent (CoCl 2 ) and serum addition. DGK activity was measured using micelles formed with the detergent octylglucoside (41,42). Activity diminished after serum deprivation and increased after serum addition to cells, and it was also elevated in response to CoCl 2 addition (Fig. 1B, left panel). Cells were then transfected with a construct encoding a chimeric GFP-DGK protein, and activity was assayed under the same conditions. The transfected DGK activity pattern was similar to that of the endogenous enzyme; DGK activity was down-regulated after starvation and increased after serum addition or during hypoxia-mimetic conditions (Fig. 1B, right panels).

Serum Stimulation Induced Phosphorylation of DGK MARCKS and C-terminal Domains-
The previous experiments showed that DGK activity is serum-regulated; we therefore tested whether the enzyme was phosphorylated in response to serum addition. Reports from our group showed that in T lymphocytes, the DGK C-terminal domain has a restrictive role in membrane translocation (13). This region contains a MAPK consensus phosphorylation site ( Fig. 2A) that is phosphorylated by ERK overexpression (14). We used an antibody raised against the p70S6K Ser 421 /Thr 424 , a region that constitutes a canonical MAPK phosphorylation motif, to study MAPK-mediated, serum-dependent phosphorylation of FIG. 1. DGK activity increases in response to mitogenic stimuli and hypoxia. A, total lysates of HEK293 cells were immunoblotted with specific anti-human DGK␣ or DGK antibodies. Jurkat T-cell lysates were used as a positive control. Bands corresponding to the DGK detected are indicated (arrowheads). DGK antibody recognized a 110-kDa doublet that corresponds to this isoform; the antibody also detected a nonspecific band that migrated at ϳ130 kDa. B, HEK293 cells were transiently transfected with GFP or GFP-DGK constructs and cultured with 10% fetal bovine serum (exponential growth); some were serum-starved for 20 h and serum-stimulated for 30 min or cultured in the presence of CoCl 2 to mimic hypoxic conditions. DGK activity was evaluated (see "Experimental Procedures"). PA produced was separated by TLC and visualized by autoradiography. Lysates of transfected cells were immunoblotted with anti-GFP antibody to estimate transfected DGK expression levels. Results are representative of three independent experiments.

FIG. 2. Serum-dependent DGK phosphorylation.
A, schematic representation of the DGK enzyme. In addition to the two cysteine-rich domains and the catalytic domain characteristic of the DGK family, DGK has four ankyrin repeats, a PDZ-binding motif, a MARCKS domain (with a canonical PKC phosphorylation site), and a consensus MAPK phosphorylation motif. B, HEK293 cells were transfected with GFP, GFP-DGK, or GFP-DGK-CT constructs, and phosphorylation was determined after serum starvation and serum addition. After serum stimulation, the phospho-antibody against Ser 421 /Thr 424 of p70S6K recognized full-length DGK, but not the mutant version that lacks the C-terminal region (top panel). GFP-DGK expression levels were evaluated using an anti-tag antibody (bottom panel). C, HEK293 cells were transiently transfected with GFP or GFP-DGK and serum-stimulated. Nuclear (N), cytosol (C), membrane (M), and cytoskeletal (Ck) fractions were prepared as described (see "Experimental Procedures"). Subcellular localization of GFP-DGK (top panel) and the phosphorylation state of the MAPK motif (middle panel) or the MARCKS domain (bottom panel) were determined by immunoblot of the fractions with anti-GFP antibody, phospho-antibody against Ser 421 /Thr 424 of p70S6K, and a pan-PKC phosphorylation site antibody, respectively. Bands corresponding to GFP-DGK are indicated (arrowheads). Purity of the subcellular fractions was assessed by determining expression of typical nuclear, membrane, and cytoskeletal protein as described in Fig. 5B (data not shown). Results shown are representative of at least three independent experiments.
DGK. This antibody recognized GFP-DGK in a serum-dependent manner (Fig. 2B), but failed to recognize a DGK mutant in which the C-terminal region had been deleted, confirming that the serum-dependent phosphorylation takes place at the C-terminal end of the protein (Fig. 2B).
DGK also contains a MARCKS domain with a canonical PKC-dependent phosphorylation site ( Fig. 2A). Phosphorylation at this site is required for plasma membrane translocation in T lymphocytes (13) and prevents nuclear localization in COS cells (8). We used a pan-PKC phosphorylation site antibody to detect serum-dependent DGK phosphorylation, but no phosphorylation was observed in total cell lysates (data not shown).
We tested whether DGK phosphorylation correlated with a specific subcellular localization. Analysis of different subcellular fractions showed that the main phosphorylation at the C-terminal end corresponded to cytosolic GFP-DGK, whereas cytoskeletal and membrane fractions showed a lesser degree of phosphorylation (Fig. 2C, middle panel). These data concur with observations in muscle cells, in which this motif appears to negatively regulate DGK association with the cytoskeleton (14). As for PKC-dependent phosphorylation, the antibody detected GFP-DGK only in the cytosolic fraction (Fig. 2C, bottom  panel). These experiments suggest that the DGK pool recovered in the cytosolic fraction is phosphorylated at both the PKC and MAPK motifs after serum addition.
DGK Overexpression Increased p70S6K Phosphorylation in the Hydrophobic Motif-Serum addition, which is known to stimulate mTOR, also modulated DGK; we thus studied the effect of DGK overexpression on mTOR activity by examining mTOR-dependent phosphorylation of Thr 389 in p70S6K transfected into quiescent or serum-stimulated cells. Thr 389 is found in the linker or hydrophobic motif of p70S6K and phosphorylated in vitro by mTOR, and it is the major rapamycin-sensitive site (36,43).
In cells overexpressing DGK, both basal and serum-induced phosphorylation of transfected p70S6K were greater than that seen in control cells (Fig. 3A, middle panel, solid arrowhead). In cells overexpressing DGK, we also detected phosphorylation of a faster migrating band (Fig. 3A, middle panel, open  arrowhead), which probably corresponds to endogenous p70S6K. To evaluate expression levels of transfected and endogenous p70S6K, we used a commercial antibody generated against the p70S6K C terminus, a region also present in the p85S6K isoform; it can thus be considered an anti-S6K1 antibody. This antibody recognized two bands, the slow-migrating band of transfected EE-p70S6K and a fast-migrating endogenous p70S6K band. Because transfected EE-p70S6K migrates with an apparent mass greater than the predicted mass of 70 kDa and nearer to 80 kDa, antibody recognition of EE-p70S6K may have masked p85S6K isoform recognition.
DGK-induced p70S6K phosphorylation is mTOR-mediated because it was inhibited by rapamycin. DGK-induced phosphorylation was reduced by addition of the DGK inhibitor R5949 (Fig. 3A). This inhibitor, albeit much more effective for class I DGK, diminished the in vitro activity of endogenous and overexpressed enzymes when added to intact cells (Fig. 3B).
DGK-induced p70S6K phosphorylation was dependent on protein levels. At higher DGK concentrations, we observed intense p70S6K phosphorylation even in the absence of serum (Fig. 4A, middle panel, solid arrowhead). As before, phospho-Thr 389 antibody also recognized a faster-migrating band, whose pattern was similar to that of transfected EE-p70S6K and corresponded to endogenous p70S6K (Fig. 4A, middle panel,  open arrowhead). The antibody also detected an intermediate migrating band, whose pattern differed from that of the other bands. Because the p70S6K antibody did not recognize this band, it is not p85S6K, but it may correspond to S6K2 isoforms or another abundant kinase with a hydrophobic domain similar to that of the S6K family.
The precise role of PA on mTOR modulation is not well known, although it has been suggested that PA competes with the FKBP12-rapamycin complex for mTOR binding (22). Some cells lines that overexpress phospholipase D (PLD) are reported to have greater rapamycin resistance (44); this prompted us to determine whether DGK overexpression also conferred rapamycin resistance. Although endogenous p70S6K phosphorylation was more resistant to rapamycin action in DGK-overexpressing cells, it remained susceptible (Fig. 4B, open  arrowheads), suggesting that DGK-dependent p70S6K phosphorylation acts via mTOR.
The phospho-Thr 389 antibody again recognized a slow-migrating doublet, whose response pattern to serum addition was similar to that of p70S6K. However, phosphorylation of these bands was resistant at the rapamycin doses tested, and DGK overexpression promoted only partial phosphorylation in the absence of serum. Reprobing the membranes with the p70S6K1 antibody confirmed that the fast-migrating band recognized by the phospho-Thr 389 antibody was p70S6K and suggested that part of the phosphorylated slow-migrating bands corresponded to p85S6K. Nonetheless, we cannot rule out that these bands may be S6K2 isoforms or a kinase with a hydrophobic domain similar to that of S6K.
p70S6K Is Phosphorylated in Different Compartments by DGK Overexpression-Most DGKs are cytosolic enzymes, and translocation from cytosol to other subcellular compartments appears to be an activation mechanism for this enzyme family. Analysis of different cell fractions with anti-DGK antibody showed serum-dependent translocation of both endogenous DGK and GFP-DGK to the membrane (Fig. 5A, top panel). Endogenous DGK was mainly cytosolic, and it migrated as several bands. These bands have been described in other cell lines to correspond to distinct enzyme phosphorylation states (14). After serum stimulation, a small fraction translocated to membrane (Fig. 5, top panel, open arrowhead). GFP-DGK behaved similarly, although there was a small but detectable amount of the enzyme in the membrane fraction under starvation conditions. In serum-treated cells, DGK translocation to the membrane was increased, whereas the enzyme was reduced in cytoskeletal and cytosolic fractions (Fig. 5, top panel,  solid arrowhead). In the presence of GFP-DGK, the endogenous enzyme translocated more efficiently to the membrane, suggesting enzyme heterodimerization.
The DGK nuclear pool showed no marked changes after serum addition, although it is important to consider that the fractionation method used did not render completely pure nuclei (Fig. 5B). In addition, weakly bound proteins in the nucleus may exit this compartment during sample processing. The DGK antibody recognized a higher molecular mass protein in the nuclear fraction that did not translocate after serum addition.
We determined the phosphorylation state of p70S6K in the different subcellular compartments. In cells overexpressing DGK, serum-dependent phosphorylation of both transfected and endogenous p70S6K at the hydrophobic motif was greatly increased in the cytosolic and membrane fractions (Fig. 5A,   middle panel, solid and open arrowheads). Strikingly, DGK overexpression also increased phosphorylation of a nuclear p70S6K pool, even in quiescence, to an extent similar to that observed in control cells after serum addition. The effect of DGK correlates with an increase in the nuclear p70S6K pool, suggesting a certain enrichment of this fraction. This might be due to an accumulation of p70S6K in either active perinuclear ribosomes or nucleoli.
DGK-dependent p70S6K Phosphorylation Required the mTOR PA Binding Region-The results suggested that DGKmediated PA generation increases p70S6K phosphorylation due to direct action on mTOR kinase. An alternative hypothesis is that of PA-dependent phosphatase inhibition, which would result in p70S6K hyperphosphorylation. To distinguish between these two possibilities, we examined DGK-mediated p70S6K phosphorylation in cells expressing either wild-type mTOR or a mutant (3ATOR) with diminished PA binding capacity (22). DGK-induced phosphorylation of p70S6K in Thr 389 was enhanced in cells expressing wild-type TOR kinase, but not in cells expressing 3ATOR (Fig. 6). These results confirmed that DGK-generated PA acts directly on the FRB region of mTOR.
mTOR-dependent p70S6K Phosphorylation at the Hydrophobic Motif Is Specific for DGK Isoform-DGK overexpression resulted in hyperphosphorylation of the p70S6K hydrophobic motif; we thus analyzed whether this effect could be exerted by other DGK isoforms. HEK293 cells were transfected with plasmids encoding tagged DGK␣ or DGK. When we examined the subcellular localization of these isoforms, DGK␣ was found in the cytosol, whereas DGK displayed a punctuate pattern in cytosol and nucleus (Fig. 7A). To assess the role of these DGKs on p70S6K phosphorylation, we examined serum-dependent mTOR-mediated p70S6K phosphorylation after transfection using increasing cDNA concentrations of the plasmids. DGK overexpression increased p70S6K phosphorylation, whereas DGK␣ diminished phosphorylation at the Thr 389 residue (Fig.  7B). These results suggest that the distinct subcellular distribution of these DGK isoforms and/or the differences in their domain composition are determinants of their capacity to modulate mTOR activity.
DGK siRNA Reduced p70S6K Phosphorylation in HEK293 Cells-To assess the role of DGK in the mTOR pathway, we attenuated endogenous DGK protein levels in siRNA experiments. We selected a region between the catalytic domain and the first ankyrin repeat to knock down DGK expression. Interference was maximal at 4 days post-transfection, producing a 60% decrease in endogenous DGK levels (Fig. 8). Serum stimulation induced a marked increase in p70S6K and p85S6K phosphorylation at Thr 389 in mock siRNA-transfected HEK293 cells, whereas knocked-down DGK cells showed reduced S6K phosphorylation. The same effect was detected, to a lesser extent, in the phosphorylation of Ser 421 /Thr 424 of S6K, another mTOR putative phosphorylation region in S6K. This effect is more evident in the p85S6K isoform (Fig. 8).
mTOR is regulated by nutrients and by the phosphatidylinositol 3-kinase/Akt pathway (45), whose activation leads to mTOR phosphorylation in Ser 2448 (46). We thus evaluated modulation of the phosphatidylinositol 3-kinase/Akt pathway using anti-phospho-mTOR(Ser 2448 ) and anti-phospho-AKT-(Ser 473 ) antibodies. We found no great difference in phosphorylation of these proteins between control and DGK knockeddown cells (Fig. 8), suggesting that DGK-promoted p70S6K phosphorylation lies downstream of the phosphatidylinositol 3-kinase/Akt pathway.
Because DGK generates PA at DAG expenses, we evaluated the effect on DAG-dependent responses of a reduction in DGK protein. We examined the activation state of PKC family members using a pan-phospho-PKC antibody (pan-PDK1 site) that recognizes phosphorylation by PDK1 in the PKC activation loop. PKC is phosphorylated by PDK1 when it is membranebound, and PDK1-dependent phosphorylation is required to render PKC catalytically competent (3). Because DGK produces PA by phosphorylating DAG, the absence of DGK causes DAG accumulation. This elevation in DAG levels must lead to receptor-independent membrane translocation of DAG-sensi-tive PKC isoforms (classical and novel), even under starvation conditions, thus favoring PDK1-dependent phosphorylation. The pan-PDK1 site antibody recognized several bands corresponding to different PKC isoforms in serum-stimulated con- B, the blots shown in A were re-blotted with antibodies against a nuclear/cytosolic (cdk2) protein and typical nuclear (histones), cytoskeletal (vimentin), and membrane (human transferrin receptor; hTfR) proteins to asses the efficiency of the fractionation procedure.
FIG. 6. DGK-dependent p70S6K phosphorylation requires the mTOR PA-binding region. HEK293 cells were cotransfected with GFP-DGK, EE-p70S6K, and FLAG-tagged wild-type mTOR (TORwt) or a mutant (TOR 3A) version that has diminished PA binding capacity. p70S6K phosphorylation was evaluated using phospho-antibodies against Thr 389 of p70S6K. Expression levels of the overexpressed proteins were measured using appropriate antibodies. p70S6K phosphorylation at Thr 389 , quantified and normalized as described in the Fig. 3 legend, is shown at the bottom of this figure; the phosphorylation ratio in control cells in the absence of serum was taken as 1.00, and the values obtained at the different conditions were expressed relative to this. In the table, the ratio of DGK-induced phosphorylation of p70S6K in Thr 389 is expressed independently and relative to the values obtained in the absence of TOR expression or cotransfection with either wild-type (wt) or mutant (3A) TOR. Similar results were obtained in three independent experiments.
FIG. 7. Overexpression of DGK, but not DGK␣, promoted phosphorylation of the p70S6K hydrophobic domain. A, immunofluorescence localization of overexpressed DGK. HEK293 cells were transfected with DGK or DGK␣, fixed, and stained with a mouse anti-HA or a rabbit anti-DGK antibody. DGK␣ showed a uniform pattern in cytosol, whereas DGK showed a punctuate pattern in cytosol and nucleus. B, HEK293 cells were cotransfected with EE-p70S6K and increasing amounts of plasmid encoding GFP, HA-DGK␣, or GFP-DGK. p70S6K phosphorylation in Thr 389 was evaluated after starvation or serum addition. DGK expression levels were evaluated using anti-GFP or anti-HA antibodies. p70S6K phosphorylation at Thr 389 , quantified and normalized as described in the Fig. 3 legend, is shown at the bottom of the figure; the phosphorylation ratio in nontransfected cells in the presence of serum was taken as 1.00, and the values obtained at the different DGK concentrations in the presence of serum were expressed relative to this. Similar results were obtained in three independent experiments. trol cells, but not in starved control cells. As predicted, the knocked-down DGK cells showed PDK1-dependent PKC phosphorylation under starvation conditions. The response to serum was reduced in these cells because phosphorylation of some proteins was lower than that observed in control cells after serum addition (Fig. 8). This decrease probably reflects the PA requirement for activation of certain PKC isoforms, such as ⑀ and (16,17). These differences indicated the dual action of DGK. DISCUSSION The lipid PA has mitogenic properties due, at least in part, to its capacity to modulate the mTOR pathway. Stimulation with serum or factors such as IL-2 induces an increase in cellular PA levels (22,23), which is required for activation of the mTOR targets p70S6K and 4E-BP1 (22). Three different enzymes generate PA: PLD, lysophosphatidic acid acyltransferase, and DGK. PLD is regarded as the main contributor of PA to mTOR signaling (22,47). Nonetheless, other PA-generating enzymes can also contribute to mTOR activation (48); lysophosphatidic acid acyltransferase is thus reported to be elevated in some tumors, and its overexpression leads to cell transformation (49). Here we demonstrate the contribution of DGK to mTORdependent p70S6K phosphorylation, suggesting a role for this enzyme in the regulation of mTOR-dependent pathways.
Serum stimulation leads to PLD activation, which correlates with increased mTOR signaling (22,47). Our experiments here show serum-dependent modulation of DGK and DGK-dependent p70S6K phosphorylation in an mTOR-modulated residue. These results suggest that PLD and DGK act in parallel pathways as PA generators (Fig. 9). Serum is, in fact, a mixture of mitogenic agents that act through G protein-coupled receptors or tyrosine kinase-coupled receptors; PLD activity increases in response to stimulation of both receptor types. Butanol (a widely used PLD inhibitor) nevertheless blocks mTOR signaling elicited by G protein-coupled receptors, such as those for lysophosphatidic acid (50) or phenylephrine (41), but does not block mTOR signaling elicited by tyrosine kinase-coupled receptors, such as that for platelet-derived growth factor (41,50). It is thus possible that some mitogens, while promoting PLD activity, modulate mTOR signaling either by a PA-independent mechanism or, more likely, by activating alternative routes to generate PA.
Lipids such as DAG and PA are generated in membrane domains, where an intimate connection between distinct lipid metabolic pathways is maintained, to produce appropriate spatio-temporal responses. PLD and DGK may operate in parallel pathways, but they may also act as DAG-and PA-generating enzymes in a single pathway. Experiments in yeast have shown that PLD activity is necessary but not sufficient to bypass the "Sec14 phenotype" that is due to mutations in phosphatidylinositol transfer protein (Sec14) function. This PLD effect is PA-independent and cannot be mimicked by DGK overexpression, which, on the contrary, exacerbates the phenotype and abolishes the bypass (51,52). This suggests that the bypass effect of PLD is due to the consecutive action of this enzyme and PA-specific phosphatases, which generate a DAG pool susceptible to phosphorylation by DGK. In mammalian cells, the PA generated at internal membranes, such as Golgi, is produced mainly by PLD action on phosphatidylcholine. This PA can serve either as a messenger, promoting vesicle fission (53)(54)(55), or as a substrate for phosphatases that transform PA to DAG. Because phosphatidylcholine is the most abundant lipid in mammalian membranes, this pathway serves as a robust supplier of DAG (56, 57) that could then be used as a DGK sub- FIG. 8. DGK siRNA diminished p70S6K phosphorylation in HEK293 cells. HEK293 cells were transfected with a control siRNA (Ϫ) or DGK siRNA (ϩ). After 3 days, cells were starved and serumstimulated. The phosphorylation state of p70S6K and of other proteins of the mTOR pathway was evaluated using the antibodies indicated. The quantitation of the bands indicated with arrowheads was performed as described under "Experimental Procedures" and is shown immediately below each blot. The value of expression for DGK in cells transfected with control siRNA was taken as 1.00, and expression of this protein in cells transfected with DGK siRNA is expressed relative to this. Values for the corresponding phosphorylation sites in p70S6K were normalized to p70S6K protein levels. Values for TOR, Akt, and PKC phosphorylation were normalized to tubulin values. When detectable, values in quiescent control cells were taken as 1.00. If they were not detectable, the values were normalized for those obtained after serum stimulation of control cells. Similar results were obtained in at least three independent experiments.
FIG. 9. mTOR pathway modulation by DGK. Mitogenic factors in serum promote DGK translocation and activation. Plasma membrane DGK translocation attenuates the response elicited by some receptors, probably G protein-coupled receptors, whereas translocation to internal membranes produces PA able to bind the mTOR FRB region. mTOR in turns phosphorylates the p70S6K hydrophobic motif. Other PA generator enzymes, such as PLD, also modulate the mTOR pathway.
strate. In agreement with this hypothesis, we found that part of the DGK-driven phosphorylation of p70S6K can be blocked with n-butanol (data not shown), suggesting that PLD can in fact provide a share of the DGK substrate.
Our results indicate that in HEK293 cells, the positive effect of DGK-dependent PA generation on the mTOR pathway is isoform-specific because DGK and DGK␣ exerted opposite effects on p70S6K phosphorylation at Thr 389 . The difference in effects of these isoforms might be a consequence of their distinct modular architecture, which would confer on each isoform a unique regulatory response to stimuli and/or different in vivo substrate specificity. In HEK293 cells, which have no endogenous DGK␣, overexpression of this kinase has a negative effect on mTOR-dependent phosphorylation, probably due to phosphorylation of a DAG pool that participates in the mitogenic pathway. In this case, DGK␣ renders a PA that is ineffective or does not have access to mTOR. On the contrary, PA produced by DGK, which has a broad subcellular distribution, increases p70S6K phosphorylation. This suggests that the unique combination of structural domains in DGK allows this isoform to gain access to DAG pools with no mitogenic effect on mTOR and to generate PA that is an effective modulator of mTOR activity. Another explanation for the lack of positive DGK␣ action on mTOR would be a difference in the nature of the PA generated. Other PA-modulated targets are reported to respond specifically to different PA species; for example, the MAPK cascade is elicited more efficiently by saturated PA than by unsaturated PA (58). Whether mTOR has such a preference is not known; nonetheless, our results showed that DGK-derived PA is capable of modulating mTOR.
The DGK isoform specificity suggested by these experiments concurs with the work of Pettit and Wakelam (59), who reported that DGK⑀ overexpression in porcine aortic endothelial cells specifically diminishes polyunsaturated DAG levels, causing redistribution of DAG-responsive PKC isoforms from membrane to cytosol. On the contrary, DGK overexpression caused an overall reduction in DAG levels, but it had no effect on PKC distribution. Moreover, DGK increased PKC⑀ activity, possibly due to the reported PA requirement for PKC⑀ (16). We observed that DGK overexpression in HEK293 cells provoked an increase in the amount of cytosolic protein recognized by a pan-phospho-PKC site antibody (data not shown) and that a decrease in endogenous DGK correlated with a lack of response to serum by some PKC isoforms (probably those that require PA for their activation).
Although no substrate specificity has been shown for DGK␣, studies of this isoform in T lymphocytes showed differential regulation in response to distinct stimuli. TCR activation causes DGK␣ translocation to the plasma membrane, where it phosphorylates phospholipase C-generated DAG and acts as a negative modulator of DAG-dependent responses (6,7). IL-2, a mitogenic factor for T cells that does not stimulate phospholipase C activity, induces DGK␣ translocation to a perinuclear region (23), with concomitant PA production derived not from receptor-generated DAG but from a pre-existing pool of diradylglycerol 1-O-alk-1Ј-enyl-2-acylglycerol (60). Taken together, the data allow us to hypothesize that the net effect of a specific DGK isoform on mTOR is dictated by the sum of factors such as mitogen concentration, the nature of the substrate, and/or the type and spatio-temporal accessibility of the PA. These factors would vary depending on cell type, mitogen, and growth conditions, as well as the presence of different DGK isoforms.
We analyzed whether there was a correlation between a specific DGK localization and p70S6K phosphorylation. Our data indicate that serum induces an increase in both DGK activity and its translocation to the membrane fraction. Nonetheless, DGK promotes p70S6K phosphorylation in any subcellular compartment, including the nucleus. When we examined endogenous p70S6K phosphorylation in serum-deprived cells, we noted that DGK overexpression induced phosphorylation of a slow-migrating band, probably nuclear p85S6K. mTOR and S6K family proteins are found in the nucleus, a surprising location for proteins of the translational machinery; nonetheless, nuclear mTOR and S6K are essential for cell growth and survival (34,61). DGK has an NLS (8), and confocal microscopy analysis of whole cells shows a large amount of the enzyme in the nucleus. Additional studies are needed to determine whether DGK promotes S6K phosphorylation in this compartment or whether phosphorylation of any of the four S6K isoforms can augment their nuclear transport or retention.
These studies describe a novel function for DGK as a PAgenerating enzyme. We and others have observed that, in DGK-overexpressing T lymphocytes, agonist stimulation of endogenous TCR or of an ectopically expressed muscarinic receptor induced a decrease in ERK signaling 2 (10, 13) due to DGK down-regulation of DAG. In these cells, serum stimulation provoked p70S6K Thr 389 hyperphosphorylation. 3 This again suggests that, as for DGK␣, the role of DGK as a DAG attenuator or PA generator depends on the type of stimuli. It remains to be determined whether specific subcellular localization would favor the attenuation of receptor-elicited DAG signals versus PA-dependent mTOR activation.
siRNA experiments further confirm that DGK acts in more than one pathway; these studies also show attenuation of mTORdependent S6K phosphorylation as well as constitutive PKC activation. Analysis of DGK-deficient mice shows that the lack of this enzyme does not induce a strong phenotype because mice are viable, fertile, and appear normal. Peripheral T cells develop normally in these mice but respond more efficiently to TCR stimulation. T-cell proliferation is increased, and knockout mice mount a stronger immune response to choriomeningitis virus infection. After TCR binding, DGK deficiency did not result in complete inhibition of PA production. This indicates that, at least during attenuation of the TCR response, the role of DGK overlaps that of other isoforms, probably DGK␣ (12). In these mice, the effect of the decrease in PA production is not known, but the PA required for T-cell proliferation could be provided by activation of another DGK isoform, or the PA deficiency could be counteracted by the robust TCR response.
Previous reports identified the PA produced during DGK activation as a mitogenic agent (23,62). The present work, however, is the first to provide direct evidence of a DGK positive role on mTOR signaling. Because the mTOR pathway is critical for regulating normal and tumor growth, determining the exact mechanisms that control the activation and possible interrelation of the different PA generators enzymes will be of valuable interest to conceive effective therapeutic strategies.