Identification and Characterization of Two Splice Variants of Human Diacylglycerol Kinase η*

Diacylglycerol kinase (DGK) participates in regulating the intracellular concentrations of two bioactive lipids, diacylglycerol and phosphatidic acid. DGKη (η1, 128 kDa) is a type II isozyme containing a pleckstrin homology domain at the amino terminus. Here we identified another DGKη isoform (η2, 135 kDa) that shared the same sequence with DGKη1 except for a sterile α motif (SAM) domain added at the carboxyl terminus. The DGKη1 mRNA was ubiquitously distributed in various tissues, whereas the DGKη2 mRNA was detected only in testis, kidney, and colon. The expression of DGKη2 was suppressed by glucocorticoid in contrast to the marked induction of DGKη1. DGKη2 was shown to form through its SAM domain homo-oligomers as well as hetero-oligomers with other SAM-containing DGKs (δ1 and δ2). Interestingly, DGKη1 and DGKη2 were rapidly translocated from the cytoplasm to endosomes in response to stress stimuli. In this case, DGKη1 was rapidly relocated back to the cytoplasm upon removal of stress stimuli, whereas DGKη2 exhibited sustained endosomal association. The experiments using DGKη mutants suggested that the oligomerization of DGKη2 mediated by its SAM domain was largely responsible for its sustained endosomal localization. Similarly, the oligomerization of DGKη2 was suggested to result in negative regulation of its catalytic activity. Taken together, alternative splicing of the human DGKη gene generates at least two isoforms with distinct biochemical and cell biological properties responding to different cellular metabolic requirements.

phatidic acid (PA) (1). By numerous studies, DG and PA have been well recognized as lipid second messengers. DG is known to be an activator of conventional and novel protein kinase Cs (PKCs), chimaerins, Unc-13, and Ras guanyl nucleotide-releasing protein (2)(3)(4), and PA has been reported to modulate the activities of phosphatidylinositol-4-phosphate kinase, Ras GTPase-activating protein, Raf-1 kinase, atypical PKC, and many other important enzymes (5,6). DGK thus appears to participate in various physiological events through modulating the balance between two bioactive lipids, DG and PA, in microenvironments within the cells.
It is now recognized that DGK represents a large enzyme family. The isoforms differ remarkably from each other with respect to their structures, the modes of tissue expression, and enzymological properties (7)(8)(9)(10). To date, nine mammalian DGK isozymes (␣, ␤, ␥, ␦, ⑀, , , , and ), containing in common two or three characteristic zinc finger structures and the catalytic region, are subdivided into five groups according to their structural features (7)(8)(9)(10). Interestingly, the occurrence of alternative splicing was recently identified for four mammalian DGK genes (␥ (11), (12), ␤ (13), and ␦ (14) isoforms). Thus, the list of DGK isoform members is still growing. The occurrence of alternative splicing in multiple DGK genes further adds to the complexity of the DGK gene family members and is probably essential for the action of DGKs regulating a wide range of cellular functions.
DGK␦ and DGK are closely related to each other and are thus classified together into the type II DGK subfamily (7)(8)(9)(10). However, DGK␦ has a sterile ␣ motif (SAM) at the carboxyl terminus (15), whereas DGK contains no SAM domain (16). The SAM domain, which is ϳ70 amino acids long, was first described as a module that is present in a small group of yeast sexual differentiation and Drosophila polyhomeotic proteins (17). The SAM domain was subsequently found to occur in a wide range of proteins. Proteins containing SAM domains include the Eph family of receptor tyrosine kinases (18,19), serine/threonine kinases (20), Src homology 2 domain-containing adaptor proteins (21), ETS transcription factors (22), p53 (21), and many other proteins. Such an extended occurrence suggests that SAM is an evolutionally conserved protein-protein interaction domain that is involved in the regulation of numerous developmental processes among diverse eukaryotes. We recently demonstrated by gel filtration and co-immunoprecipitation analyses that DGK␦ formed homo-oligomeric structures in intact cells and that the SAM domain was critically involved in the oligomerization (23). Moreover, the alternative splicing products of DGK␦ (DGK␦1 and DGK␦2) were shown to form hetero-oligomers via their SAM domains (14).
In the course of genome data base search for the human DGK gene, we identified in silico a potential splice variant of DGK with a SAM domain at the carboxyl terminus. Here we cloned the novel DGK isoform generated through the alternative splicing. The two splice products of DGK were revealed to possess biochemical and cell biological properties distinct from each other. Moreover, we found that the new DGK2 isoform formed hetero-oligomers with other type II members, DGK␦1 and DGK␦2.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-COS-7, human embryonic kidney (HEK) 293, and HepG2 cells were maintained in DMEM (Sigma) containing 10% fetal bovine serum at 37°C in an atmosphere containing 5% CO 2 . Cells were transiently transfected with cDNAs using Effectene transfection reagent according to the instructions from the manufacturer (Qiagen, Tokyo, Japan). 24 or 48 h after transfection, cells were used for further analysis.
Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR)-Total RNA was isolated from HepG2 or HEK293 cells using Isogen (Nippon Gene, Tokyo, Japan) according to the protocol from the manufacturer. Reverse transcription into cDNA was achieved using the SuperScript preamplification system (Invitrogen) according to the instructions from the manufacturer. cDNAs from human normal tissues and tumor-derived cells were purchased from BD Biosciences, Tokyo, Japan. PCR amplification was performed with Takara Ex Taq (Takara Biomedicals, Tokyo, Japan) using gene-specific oligonucleotide primers as follows: a DGK common forward primer (nucleotide positions 3092-3116, 5Ј-TGAATAAAGCCAACCCAAGGTGCCC-3Ј), and a common reverse primer (nucleotide positions 3601-3627, 5Ј-TTTAATTCCCTGGAGAA-TTCGCTTCAC-3Ј). PCR conditions were as follows: 94°C for 4 min; 37 cycles of 94°C for 30 s, 68°C for 2 min; and 68°C for 5 min. For normalization, human glyceraldehyde phosphate dehydrogenase (GAPDH) mRNA was simultaneously amplified (25 cycles). PCR products amplified were separated by agarose gel electrophoresis and stained with ethidium bromide.
Plasmid Constructs-cDNAs coding for DGK-⌬S1,2 (amino acid residues 1-1147), DGK2-⌬PDZ (amino acid residues 1-1217) and DGK2-SAM (amino acid residues 1144 -1220) were generated by PCR from a human DGK2 cDNA clone. DGK2-W1151G cDNA clone was generated by replacing the Trp-1151 in the SAM domain with Gly using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). For subsequent subcloning into expression vectors, primers were designed such that the resulting DGK2 cDNA fragments contained a 5Ј-BamHI restriction site and a 3Ј-stop codon followed by a 3Ј-SalI site. The resulting DGK2 fragments were fused in frame (the BglII/SalI site) with green fluorescent protein (GFP) in the pEGFP-C3 vector (BD Biosciences) and with 3ϫFLAG tag in the p3ϫFLAG-CMV-7.1 (Sigma). Plasmids encoding the SAM domain of DGK2 fused to maltosebinding protein (MBP, 43 kDa) were obtained by inserting the SAM cDNA fragments into the pMAL-c2X vector (New England BioLabs, Beverly, MA).
Expression and Purification of MBP Fusion Proteins-XL1-Blue cells (Stratagene) were transfected with pMAL-c2X constructs, and MBP fusion proteins were expressed and purified according to the procedure recommended by the manufacturer (New England BioLabs). In this case, expression of fusion proteins was induced by 0.3 mM isopropyl-1thio-␤-D-galactoside at 37°C for 3 h. The cells were then harvested and lysed by freeze-thaw cycles in the lysis buffer (20 mM Tris-HCl (pH 7.4), 200 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml pepstatin A, 5 g/ml leupeptin). Fusion proteins were purified using an amylose resin affinity column.
Size Exclusion Chromatography-Gel filtration was performed on a Superose 12 HR 10/30 column (Amersham Biosciences) equilibrated with the column buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl) and run with the fast protein liquid chromatography system equipped with a UV monitor (Amersham Biosciences) at a flow rate of 0.5 ml/min at room temperature. The column was calibrated with gel filtration molecular mass standards (Sigma) as previously described (23). For analysis of the MBP fusion proteins, aliquots (100 l) of the purified samples (40 M) were loaded onto the column.
Co-immunoprecipitation Analysis-COS-7 cells (ϳ1 ϫ 10 7 cells/ 60-mm dish) co-expressing 3ϫFLAG-tagged DGK2 with GFP⅐DGK1 or GFP⅐DGK2 were harvested in buffer A supplemented with 1% Nonidet P-40. The mixture was centrifuged at 100,000 ϫ g for 30 min at 4°C to give cell lysates. Cell lysates (400 l) were precleared with protein A/G PLUS-agarose (10 l, Santa Cruz Biotechnology, Santa Cruz, CA). Anti-GFP (2 l, Living Colors full-length A.v. polyclonal antibody, BD Biosciences) was added to the precleared lysates to immunoprecipitate GFP fusion proteins. After 1 h at 4°C, protein A/G PLUS-agarose beads (5 l) were added and further incubated at 4°C for 1 h. After the agarose beads were washed five times with wash buffer (50 mM Tris-HCl (pH 7.4), 1 mM EDTA, 0.1% Triton X-100, 0.6 M KCl), immunoprecipitated proteins were extracted with 50 l of SDS sample buffer and then subjected to SDS-PAGE analysis.
Protein Phosphorylation Analysis in Vivo-COS-7 cells (35-mm dish) expressing 3ϫFLAG-tagged DGK isoforms were used 48 h post-transfection. Transfected COS-7 cells were washed four times with Trisbuffered saline, incubated for 1 h in DMEM without sodium phosphate and sodium pyruvate (Invitrogen) containing 0.1% bovine serum albumin, and then labeled for 2 h with [ 32 P]orthophosphate (0.3 mCi/ml, Amersham Biosciences) in phosphate-free DMEM/0.1% bovine serum albumin. The labeled cells were subsequently incubated in phosphatefree DMEM/0.1% bovine serum albumin containing 1 M 12-O-tetradecanoylphorbol 13-acetate (TPA) or 0.1% Me 2 SO for 1 h. Cells were harvested in Buffer A containing 1% Nonidet P-40 and then centrifuged at 12,000 ϫ g for 20 min at 4°C to give the cell lysates. Cell lysates (300 l) were precleared with protein A/G PLUS-agarose (10 l). Anti-FLAG (M2; Sigma) antibody (2 g) was added to precleared lysates to immunoprecipitate 3ϫFLAG-tagged DGK isoform proteins. After 1 h, protein A/G PLUS-agarose (5 l) was added followed by a 1-h incubation at 4°C. After washing the agarose beads five times with buffer A containing 1% Nonidet P-40, immunoprecipitated proteins were extracted with 50 l of SDS sample buffer and then subjected to SDS-PAGE analysis. The radioactive signal was visualized using a BAS1800 bio-imaging analyzer (Fuji Film, Tokyo, Japan).
Western Blot Analysis-Cell lysates and immunoprecipitates were separated on SDS-PAGE (7.5 or 15%). The separated proteins were transferred to a polyvinylidene difluoride membrane (Bio-Rad) and blocked with Block Ace (Dainippon Pharmaceutical, Tokyo, Japan). The membrane was incubated with anti-FLAG M2 monoclonal antibody or anti-GFP monoclonal antibody (B-2; Santa Cruz Biotechnology) in Block Ace for 1 h. The immunoreactive bands were visualized using peroxidase-conjugated anti-rabbit or -mouse IgG antibody (Jackson ImmunoResearch Laboratories, West Grove, PA) and ECL (Amersham Biosciences).
Fluorescence Microscopy-For immunofluorescence microscopy, COS-7 cells were grown on poly-L-lysine-coated glass coverslips and transiently transfected with expression plasmids containing DGK cDNAs amino-terminal fused with GFP. After 24 h, cells were serumstarved for 3 h and then incubated in DMEM with or without 500 mM sorbitol for 30 min. The cells were then fixed in 3.7% formaldehyde. When non-GFP fusion proteins were needed to be detected, cells were permeabilized in phosphate-buffered saline containing 0.1% Triton X-100 and 1% bovine serum albumin. Coverslips were incubated successively with the primary antibodies and the appropriate rhodaminecoupled anti-species antibodies. The coverslips were mounted using Vectashield (Vector Laboratories, Burlingame, CA). For living cell imaging, COS-7 cells were transiently transfected with the expression plasmids containing DGK cDNAs and were observed in glass-bottom chambers (Asahi Techno Glass, Tokyo, Japan). When fluorescence microscopy was carried out, the medium was changed to a mixture of DMEM and F-12 (1:1) containing 15 mM Hepes (pH 7.2) without phenol red supplement (Invitrogen). After 24 h of the transfection, cells were serum-starved for 3 h and then incubated in DMEM with or without 1 M TPA, 500 mM sorbitol, or 0.5 mM H 2 O 2 . Cells were examined using inverted confocal laser scanning microscopy (Zeiss LSM 510). Images were processed using Adobe Photoshop 7.0 (Adobe Systems, San Jose, CA).

Identification and Characterization of a Novel DGK Iso-
form-In the course of performing a genome data base search (accession numbers of human genomic clones: AL157932, AL139328, and AC018379) using as a probe the nucleotide sequence encoding the human DGK␦ SAM domain (15), we found a potential splice variant of human DGK (Fig. 1A). We tentatively designated the human homolog of hamster DGK cloned previously by Klauck et al. (16) and the newly identified clone as DGK1 and DGK2, respectively. As shown in Fig. 1, A and B, DGK2 utilizes all of the 30 exons, whereas DGK1 skips exon 29. The open reading frames of DGK1 and DGK2 encode 1164-and 1220-amino acid proteins with calculated molecular masses of 128 and 135 kDa, respectively (Fig. 1C). Interestingly, DGK2 has a SAM domain encoded by exons 29 and 30 at the carboxyl terminus ( Fig. 1, A-C). Because the phase of open reading frame used in the exon 30 of DGK1 is different from that used by DGK2 (Fig. 1A), the carboxylterminal portions (amino acids 1148 -1164 for DGK1 and 1148 -1220 for DGK2) have no similarity with each other. The sequence of DGK2 SAM domain (amino acids 1148 -1220) shows high homology (78.1% identity and 87.7% similarity) to that of DGK␦ (15) (Fig. 1D). Moreover, we noted that the last three residues (Ser-Glu-Val) of DGK2 are a potential target sequence, S/TX⌽ (⌽, a hydrophobic amino acid, usually Val, Ile, or Leu; X, unspecified amino acid), for class I PDZ domains (24 -26).
To confirm that the two DGK clones encode active enzymes, we subcloned each of the cDNAs into an expression vector, p3ϫFLAG-CMV-7.1, which was subsequently transfected into COS-7 cells. We detected in the transfected cells proteins corresponding to the predicted molecular masses of 128 kDa (DGK1) and 135 kDa (DGK2) upon Western blotting using anti-FLAG antibody (not shown). Both enzymes were confirmed to possess DGK catalytic activities when the cell homogenates were assayed in vitro (not shown). Interestingly, DGK1 showed an apparently 4-fold higher activity than that of DGK2, suggesting the possibility that the SAM domain exerts inhibitory effects on the catalytic activity (see Fig. 7).
Differential Modes of Expression and Induction of DGK Isoforms-To investigate the expression patterns of DGK isoforms, panels of cDNAs synthesized from human normal and tumor tissues were analyzed by RT-PCR. As shown in Fig. 2, the DGK1 transcript was detectable in most of the normal tissues and all of tumor-derived cells examined, with exceptionally low expression in the lung and skeletal muscle. On the other hand, the DGK2 transcript was detected only in testis, kidney, and colon. Compared with DGK1, the expression of DGK2 in tumor-derived cells appeared to be relatively low, although the lack of antibody against human DGK hindered us from confirming the protein levels of isoforms in the tissues.
Differential tissue distribution patterns of DGK1 and DGK2 mRNAs led us to hypothesize that the expressions of these isoforms are regulated under distinct mechanisms. Because Klauck et al. (16) reported that glucocorticoid, triamcinolone acetonide (TAA), increased the expression level of the hamster DGK (a counterpart of human DGK1), we first tested the effect of TAA. When HepG2 cells, which are known to express glucocorticoid receptors (27), were treated with 10 nM TAA for 72 h, the level of DGK1 mRNA was markedly increased (Fig. 3A). In contrast, the DGK2 mRNA was moderately decreased. We recently found that expression of DGK␦2, another type II DGK isoform, was induced by treating cells with epidermal growth factor (EGF) and TPA, whereas the levels of mRNA and protein of DGK␦1 were suppressed by phorbol ester treatment (14). Thus, we next examined effects of TPA (100 nM, 24 h) and EGF (10 ng/ml, 48 h) on the expression levels of DGK isoforms. As shown in Fig. 3B, the mRNA level of DGK2 was not much affected by TPA whereas that of DGK1 was significantly increased. We next examined the effect of EGF on the expression of cellular DGK enzymes. For this purpose, we used HEK293 cells because HepG2 cells failed to respond to EGF. EGF had no detectable effects on the mRNA levels of DGK1 and DGK2 (Fig. 3C). The results collectively demonstrated that the induction patterns of mRNA levels of DGK1 and DGK2 were clearly different from each other. Moreover, the expression patterns of DGK are distinct from those of the closely related isoforms (DGK␦1 and DGK␦2).
Homo-and Hetero-oligomer Formation of DGK Isoforms-We next examined biochemical and cell biological properties of DGK1 and DGK2. We have recently found that DGK␦1 and DGK␦2 form oligomeric (at least tetrameric) structures in vitro and in vivo and that the SAM domain plays a critical role in the oligomer formation (14,23). To study the multimeric nature of the DGK2 SAM domain in vitro, MBP⅐DGK2-SAM purified was analyzed using a Superose 12 HR column. As already observed for DGK␦ (23), MBP⅐DGK2-SAM (40 M) was eluted at the tetramer position (ϳ240 kDa) (Fig. 4A). In the case of EphB2 receptor (28), an aromatic residue, Tyr-912, at 8th position in the SAM domain (corresponding to Trp-1151 in DGK2) was reported to be critical for dimer formation. On the other hand, the SAM domain of the EphA2 receptor, in which the critical Tyr is replaced with Gly, did not form dimer structure (29). Moreover, when the corresponding residue (Trp-1101) of DGK␦1 was mutated to Gly, the mutant did not form the multimeric structures (23). Trp-1151 in the DGK2 SAM domain was thus substituted by Gly to see whether this Trp is critical for the interaction. In contrast to the wild-type SAM, MBP⅐DGK2-SAM-W1151G (40 M) was eluted at the monomer position (ϳ60 kDa) (Fig. 4A). These data, essen- tially consistent with those obtained for the DGK␦ SAM domain (23), indicate that the DGK SAM domain is capable of forming at least tetrameric structure and that Trp-1151 in the SAM domain is essential for the oligomer formation.
To address homo-oligomer formation of DGK2 in vivo, we carried out co-immunoprecipitation analysis using the lysates of COS-7 cells co-expressing GFP-tagged and 3ϫFLAG-tagged DGK2 proteins. When GFP-tagged DGK2 was immunoprecipitated with anti-GFP antibody, 3ϫFLAG-tagged DGK2 was co-immunoprecipitated (Fig. 4B). Reciprocally, when GFPtagged DGK2 was immunoprecipitated with anti-GFP antibody, 3ϫFLAG-tagged DGK2 was co-immunoprecipitated. On the other hand, apparently due to lack of the SAM domain, homo-oligomer formation of DGK1 was not observed (Fig. 4C).
To assess further the contribution of the SAM domain to the oligomer formation, we performed co-immunoprecipitation experiments using DGK mutants DGK2-W1151G and DGK-⌬S1,2 lacking the SAM domain. As shown in Fig. 4D, in comparison with wild-type DGK, significantly less 3ϫFLAG-DGK-⌬S1,2 and 3ϫFLAG-DGK2-W1151G were found in the anti-GFP immunoprecipitated complexes. These results indicate that the SAM domain contributes critically to the oligomer formation of DGK2. We noted, however, that a faint but detectable band of 3ϫFLAG-tagged DGK2-W1151G was still co-immunoprecipitated with the GFP-tagged counterpart (Fig.  4D), suggesting that, in addition to Trp-1151, the other residues of the domain weakly contribute to the formation of oligomer. It was previously reported that DGK␦1-oligomer was dissociated to a monomer by TPA stimulation. However, such disassembly of DGK2-oligomer was not detected (Fig. 4E).
To examine whether DGK2 and the DGK␦ isoforms, possessing in common a SAM domain, form hetero-oligomer structures in vivo, we performed experiments using COS-7 cells co-expressing GFP-tagged DGK2 and 3ϫFLAG-tagged DGK␦1 or DGK␦2. When GFP-tagged DGK2 was immunoprecipitated with anti-GFP antibody, 3ϫFLAG-tagged DGK␦1 or ␦2 was co-immunoprecipitated (Fig. 5, A and B). We confirmed that DGK1 without SAM domain was not co-immunoprecipitated with other DGKs. 2 To confirm that DGK2, DGK␦1, and DGK␦2 indeed form oligomers consisting of all of them, we co-expressed GFP⅐DGK2, 3ϫHA-DGK␦1, and 3ϫFLAG-DGK␦2 in COS-7 cells. As shown in Fig. 5C, all of the three DGK isoforms were co-immunoprecipitated. Moreover, the input/immunoprecipitation ratios remained almost invariable among the isozymes tested, suggesting that these DGKs interact with each other with a similar affinity. Taken together, it is strongly suggested that all of DGK2, DGK␦1, and DGK␦2 can participate in the formation of hetero-oligomers in vivo.

Phosphorylation of DGK Isoforms by TPA Stimulation-
Because we have recently demonstrated that DGK␦1 was phosphorylated upon TPA cell stimulation (23), protein phosphorylation of DGK1 and DGK2 was investigated. As shown in Fig.  6A, DGK1 was clearly phosphorylated in TPA-stimulated cells, whereas DGK2 was not. This result suggested that because the DGK1-specific region at the carboxyl terminus contains four serine residues and one threonine (Fig. 1A), this region might have served as a phosphorylation site. Alternatively, the DGK2-specific region (the SAM domain) might be involved in the suppression of phosphorylation. To address these possibilities, we constructed a DGK mutant lacking both of the DGK1-and DGK2-specific sequences (DGK-⌬S1,2). Subsequent phosphorylation experiments showed that wild-type DGK1 and DGK-⌬S1,2 were phosphorylated to a similar extent (Fig. 6B). The result indicates that the DGK1specific region does not contain the phosphorylation site and that the SAM domain quite likely exerts inhibitory effect(s) on the phosphorylation.
Catalytic Activity of DGK Isoforms-When expressed in COS-7 cells, DGK1 was already noted to give ϳ4-fold higher activity than that of DGK2 (Fig. 7B). We therefore attempted to delineate the mechanism underlying the different catalytic efficiencies of the two DGK isozymes. Because DGK1 was phosphorylated in TPA-stimulated cells (Fig. 6A), we first examined the effects of phosphorylation on the catalytic activities of the DGK isoforms. When COS-7 cells expressing DGK1 or DGK2 were treated with TPA, the DGK activity measured in vitro was not affected (Fig. 7A). We therefore concluded that the catalytic activity of DGK isoforms was not significantly regulated by protein phosphorylation.
It is possible that the DGK activity is positively affected by the DGK1-sepecific sequence (amino acids 1148 -1164). Conversely, the activity may be negatively regulated by the DGK2-sepecific sequence (the SAM domain, amino acids 1148 -1220). To assess these possibilities, we determined catalytic activity of a mutant lacking both the DGK1-and the DGK2-specific sequences, DGK-⌬S1,2. As shown in Fig. 7B, DGK-⌬S1,2 exhibited almost the same activity with that of wild type DGK1, thus suggesting that the SAM domain negatively regulates the activity.
We next attempted to see whether the SAM domain inhibits the DGK2 activity directly or indirectly through mediating enzyme oligomerization. The point mutant (DGK2-W1151G) with a considerably impaired ability to form homo-oligomer (see Fig. 4D) gave DGK activity more than 2-fold higher than that of the wild-type enzyme (Fig. 7B). Although the activity of DGK2-W1151G was lower than that of wild-type DGK1 and DGK-⌬S1,2, this is probably because of the weak interaction still occurring between the mutant proteins (see Fig. 4D). On the other hand, the addition of excess amounts of MBP⅐DGK2-SAM or MBP⅐DGK2-SAM-W1151G invariably failed to affect the catalytic activity of the mutant (DGK-⌬S1,2) lacking the SAM domain (Fig. 7C). The result strongly suggests that the SAM domain does not directly inhibit the activity. Taken together, we infer that oligomerization through the SAM domain is, at least in part, responsible for the suppressed catalytic activity of DGK2.  8. Subcellular localization of DGK1 and DGK2. A, COS-7 cells were plated in glass-bottom chambers and transfected with expression plasmids encoding GFP alone, GFP⅐DGK1, or GFP⅐DGK2 as indicated. After 24 h, cells were incubated for 5 min in the presence of 500 mM sorbitol. After replacement of the medium by sorbitol-free medium, cells were further incubated for 5 min. A p38 MAPK inhibitor (SB203580) or a mitogen-activated ERK activating kinase inhibitor (U0126) was added 30 min prior to the stimulation. Cells were examined using inverted confocal laser scanning microscopy (Zeiss LSM 510). A representative of three repeated experiments is shown. Bar ϭ 10 m. B, COS-7 cells were plated in glass-bottom chambers and transfected with expression plasmids encoding GFP⅐DGK-⌬S1,2, GFP⅐DGK2-⌬PDZ, GFP⅐DGK2-W1151G, GFP⅐DGK2-SAM, or GFP⅐DGK2-SAM-W1151G as indicated. After 24 h, cells were incubated for 5 min in the presence Intracellular Localizations of DGK Isoform Proteins-We have recently found in HEK293 cells TPA-induced translocation of DGK␦1 from cytoplasmic vesicles to the plasma membrane (23). However, no translocation of DGK isoforms was detected when HEK293 and COS-7 cells were treated with TPA. 2 We next examined effects of stress stimuli on the intracellular localization of DGK. As shown in Fig. 8A, both DGK1 and DGK2 expressed in COS-7 cells were diffusely distributed in the cytoplasm. In response to osmotic shock (500 mM sorbitol), both isoforms were rapidly (within less than 5 min) translocated from the cytoplasm to punctate vesicles in more than 95% of COS-7 cells expressing the enzymes. Essentially the same results were obtained in cells exposed to oxidative stress (0.5 mM H 2 O 2 ). 2 A similar translocation pattern was also observed in the cDNA-transfected HEK293 cells. 2 Because more than 90% of DGK2 became associated with the Triton X-100 insoluble cytoskeletal fraction in the presence of 500 mM sorbitol, we could not assess the relationship between oligomer formation and enzyme translocation. In the presence of sorbitol, DGK2 (and DGK1) 2 markedly co-localized with an endosome marker protein, RhoB (30) (Fig. 8C).
Osmotic shock and oxidative stress are known to activate p38 mitogen-activated protein kinase (MAPK), c-Jun-NH 2 -terminal kinase/stress-activated protein kinase, and extracellular signal-regulated kinase (ERK) (31,32). When p38 MAPK inhibitor, SB203580, was added, the translocation induced by sorbitol was significantly inhibited (Fig. 8A). On the other hand, an inhibitor of mitogen-activated ERK activating kinase, U0126, did not exert such effects. Essentially the same results were obtained with DGK2. 2 The results strongly suggest that translocation was mainly regulated through the p38 MAPK pathway.
Interestingly, after removal of sorbitol, DGK1 was rapidly (within less than 5 min) relocated back to the cytoplasm in about 95% of cells expressing the isoform (Fig. 8A). On the other hand, DGK2 remained associated with the vesicles in more than 90% of DGK2-transfected cells. Even 15 min after the wash out, DGK2 still associated with the vesicles in about 70% of transfectants. The DGK1 mutant lacking the DGK1specific sequence (DGK-⌬S1,2) showed the same translocation pattern as observed for wild-type DGK1 (Fig. 8B), indicating that the region does not act as a repulsive element in endosomes in the absence of sorbitol. Moreover, because the DGK2 mutant lacking the PDZ binding sequence (DGK2-⌬PDZ) showed the same translocation pattern as observed for wild-type DGK2, the SAM domain, but not the PDZ binding sequence, is considered to be essential for sustained endosomal localization.
It is possible that the oligomer formation through the SAM domain is responsible for the sustained localization. Alternatively, the domain itself may directly stabilize the membrane association. As shown in Fig. 8B, DGK2-W1151G failed to exhibit sustained endosomal association, indicating that the enzyme oligomerization is involved at least in part in the stabilization of membrane association. The SAM domain alone (DGK2-SAM) was translocated from the cytoplasm to endosomes upon sorbitol treatment, but different from the intact enzyme, DGK2-SAM was rapidly dissociated from the membranes by removal of sorbitol. Moreover, the W1151G mutation in the SAM domain (DGK2-SAM-W1151G) did not affect the transient nature of the enzyme translocation. Because the same point mutation introduced into full-length DGK2 markedly reduced the sustained endosomal localization, the SAM domain alone cannot directly stabilize the membrane association. Taken together, the oligomer formation is indicated to mainly contribute to the sustained endosomal localization of DGK2. DISCUSSION The present study added a new DGK isoform to the growing list of the mammalian DGK gene family (altogether nine independent genes and five alternative splicing products at present). Interestingly, although alternative splicing of the DGK␦ gene replaces the most amino-terminal sequence (14), that of the DGK gene generates a DGK isoform having the carboxylterminal addition, including the SAM domain and a PDZ-binding motif. Exchanges of distinct functional domains at separate sites give a great diversity of structure and function to the type II DGKs. As expected from the alternative splicing of the DGK␦ gene (14), DGK1 and DGK2 were expressed in different manners. Moreover, we revealed that skipping only a single exon (exon 29) of the DGK gene resulted in generating two splice variants having many different biochemical and cell biological properties, such as catalytic activity, protein phosphorylation, and subcellular localization.
Klauck et al. (16) reported that the hamster DGK1 was glucocorticoid (TAA)-inducible. In this report, we confirmed that TAA significantly increased the DGK1 mRNA level in HepG2 cells. Interestingly, the mRNA level of the newly identified DGK2 was, if anything, decreased by the TAA treatment. Although TPA also enhanced the DGK1 expression, but not DGK2, EGF failed to affect the expressions of both isoforms. These results suggest that choices of the splice sites occur in highly inducer-specific manners. However, because of the general lack of understanding of how cells choose and change particular splice sites (33), regulatory mechanisms of alternative splicing of the DGK gene are still unclear. DGK1 showed a much broader tissue distribution, whereas the expression of DGK2 was relatively tissue-specific. The results suggest that DGK1 plays a more general role in regulating cellular DG and PA levels when compared with DGK2.
DGK1 showed a molecular activity in vitro ϳ4-fold higher than that of DGK2. A DGK␦1 mutant lacking the SAM domain also had about 4-fold higher activity in vitro than that of wild-type DGK␦1. 3 Thus, it is likely that the molecular activities of type II DGK isoforms lacking the SAM domain are generally higher than those of the isoforms equipped with this domain. In this study, we inferred that oligomer formation through the SAM domain rather than the direct effects of this domain resulted in the inhibition of the catalytic activity. Thus, the amino-terminal part of DGK2 oligomerized through the SAM domain may interfere with the catalytic action. Alternatively, it remains possible that the amino-terminal portion of monomeric DGK interacts with an unknown activating factor for the DGK activity. 3 S. Imai, F. Sakane, and H. Kanoh, unpublished observation. of 500 mM sorbitol. After replacement of the medium by sorbitol-free medium, cells were further incubated for 5 min. Cells were examined using inverted confocal laser scanning microscopy (Zeiss LSM 510). A representative of three repeated experiments is shown. Bar ϭ 10 m. C, COS-7 cells were plated on glass coverslips and co-transfected with pEGFP⅐RhoB and p3ϫFLAG-CMV-DGK2 or p3ϫFLAG-CMV vector alone as indicated. After 24 h, cells were incubated for 30 min in the presence of 500 mM sorbitol and then fixed. DGK2 was labeled with anti-FLAG and rhodamine-conjugated secondary antibodies. GFP (green) and rhodamine (red) fluorescence was visualized by confocal microscopy. Overlap of DGK2 and RhoB is shown in yellow. Arrows indicate punctate vesicles where co-localization of DGK2 and RhoB was apparent. Cells were examined using inverted confocal laser scanning microscopy (Zeiss LSM 510). Representative cell images reproducibly obtained for repeated experiments are given. Bar ϭ 10 m.
Translocation of DGK␦1 from cytoplasmic vesicles to the plasma membrane was dependent on TPA stimulation (23). In addition to DGK␦1, translocations of many DGKs to the plasma membrane or to the nucleus have been reported so far in TPA-treated cells or in activated T-lymphocytes (7)(8)(9)(10). In this report, we demonstrated for the first time that DGK1 and DGK2 were significantly translocated from the cytoplasm to endosomes by stress stimuli. TPA is known to stimulate PKC (2) and subsequently the Raf-1/ERK pathway (34). ERKs are predominantly activated by mitogenic stimuli, and other p38 MAPKs and c-Jun-NH 2 -terminal kinases/stress-activated protein kinases are preferentially activated by non-mitogenic stimuli, including environmental stresses. However, osmotic shock and oxidative stress are known to activate all of p38 MAPKs, c-Jun-NH 2 -terminal kinases/stress-activated protein kinases, and ERKs. Because a p38 MAPK inhibitor (SB203580), but not an inhibitor of mitogen-activated ERK activating kinase (U0126), significantly inhibited the translocation of DGK, this event is considered to be mainly regulated by the p38 MAPK pathway. This concept is supported by the fact that TPA, which activates the ERK pathway, did not markedly enhance the translocation. At present, we cannot deny the possibility that c-Jun-NH 2 -terminal kinases/stressactivated protein kinases also partly contribute to the enzyme translocation. Interestingly, the endosomal association of DGK1 induced by sorbitol was quite transient, whereas that of DGK2 was rather sustained. Using DGK mutants we found that the sustained association of DGK2 with endosomes was largely caused by its oligomer formation through its SAM domain. Such distinct modes of membrane association should reflect their different physiological functions. Many signaling proteins such as phospholipase C␥1, p120 Ras GTPase-activating protein, and c-Raf were found associated with endosomal compartments (35). Because it is known that phospholipase C liberates DG and that the activities of p120 Ras GTPase-activating protein and c-Raf are regulated by PA (7-10), DGK1 and DGK2 may be involved in these signaling pathways on transient or continuous demands, respectively.
DGK1, but not DGK2, was phosphorylated by phorbol ester stimulation. Obvious phosphorylation of the DGK isoforms was not detected by sorbitol stimulation, 2 suggesting that the phosphorylation is probably dependent on PKC. Because the DGK1-specific sequence was not considered to serve as a phosphorylation site, phosphorylation occurs in the region common to the two isoforms. To explore physiological consequence of the isoform-selective protein phosphorylation, identification of phosphorylation site(s) and protein kinase(s) involved is needed.
One of the most remarkable features of the SAM domain of DGK2 is its ability to interact with DGK itself and with the closely related isoforms, DGK␦1 and DGK␦2. Because DGK2, DGK␦1, and DGK␦2 (14) are all co-expressed in lung carcinomas (LX-1 and GI-117), a colon adenocarcinoma (CX-1), a prostatic adenocarcinoma (PC3), pancreatic adenocarcinoma (GI-103), and HepG2 cells, these isoforms can potentially form hetero-oligomer structures at least in these tissues and cells. Because DGK␦ is known to form at least tetramer structures (23), permutations and combinations of the hetero-tetramer yield 3 4 (ϭ81) variations. What is the physiological implication of the DGK hetero-oligomers being greatly diversified? This question is the target of future investigation.
In addition to the DGK gene, alternative splicing was found in the DGK␤ (13), DGK␥ (11), DGK␦ (14), and DGK (12) genes to date. DGK␤ isoforms differ at their carboxyl termini through alternative splicing and one variant was associated with the plasma membrane, whereas the other isoform was predomi-nantly localized in the cytoplasm (13). Active DGK␥ is predominantly expressed in human retina, but an inactive form lacking 25 amino acids within the catalytic domain is detected in most human tissues and cells, such as kidney, testis, HepG2 cells, and HL-60 cells (11). The expression of DGK␦2 was induced by treating cells with EGF and TPA, whereas the levels of mRNA and protein of DGK␦1 were suppressed by the phorbol ester treatment (14). Moreover, DGK␦1 was translocated through its pleckstrin homology domain from the cytoplasm to the plasma membrane in response to phorbol ester stimulation, whereas DGK␦2 remained in the cytoplasm even after stimulation. In the case of DGK, the initial 54 amino acids are replaced with a 262-amino acid fragment (12). DGK1 is a more widely expressed form, whereas a DGK2 protein was predominantly detected in muscle tissues. Inhibitory effects on Ras guanyl nucleotide-releasing protein and subcellular localization of the DGK isoforms were different from each other (36). In plants, alternative splicing of the tomato DGK gene generates an isoform having the calmodulin-binding domain near the carboxyl terminus (37). Calcium recruited only the calmodulin-binding DGK from soluble to membrane fractions. Therefore, alternative splicing may be a common mechanism in regulating DGK functions in different tissues and cells. The marked divergence of DGK isoforms elicited by alternative splicing may reflect their physiological importance and the needs to respond to a variety of signaling pathways operating under distinct regulatory mechanisms.