The Pleckstrin Homology Domain of Diacylglycerol Kinase η Strongly and Selectively Binds to Phosphatidylinositol 4,5-Bisphosphate*

Type II diacylglycerol kinase (DGK) isozymes (δ, η, and κ) have a pleckstrin homology domain (PH) at their N termini. Here, we investigated the lipid binding properties of the PHs of type II DGK isozymes using protein-lipid overlay and liposome binding assays. The PH of DGKη showed the most pronounced binding activity to phosphatidylinositol (PI) 4,5-bisphosphate (PI(4,5)P2) among the various glycero- and sphingolipids including PI 3,4,5-trisphosphate, PI 3,4-bisphosphate, PI 3-phosphate, PI 4-phosphate, and PI 5-phosphate. Moreover, the PI(4,5)P2 binding activity of the DGKη-PH was significantly stronger than that of other type II DGK isozymes. Notably, compared with the PH of phospholipase C (PLC) δ1, which is generally utilized as a cellular PI(4,5)P2- probe, the DGKη-PH is equal to or superior than the PLCδ1-PH in terms of affinity and selectivity for PI(4,5)P2. Furthermore, in COS-7 cells, GFP-fused wild-type DGKη1 and its PH partly translocated from the cytoplasm to the plasma membrane where the PLCδ1-PH was co-localized in response to hyperosmotic stress in an inositol 5-phosphatase-sensitive manner, whereas a PH deletion mutant did not. Moreover, K74A and R85A mutants of DGKη-PH, which lack the conserved basic amino acids thought to ligate PI(4,5)P2, were indeed unable to bind to PI(4,5)P2 and co-localize with the PLCδ1-PH even in osmotically shocked cells. Overexpression of wild-type DGKη1 enhanced EGF-dependent phosphorylation of ERK, whereas either K74A or R85A mutant did not. Taken together, these results indicate that the DGKη-PH preferentially interacts with PI(4,5)P2 and has crucial roles in regulating the subcellular localization and physiological function of DGKη. Moreover, the DGKη-PH could serve as an excellent cellular sensor for PI(4,5)P2.

It has recently been demonstrated, using DGK␦ knock-out mice and RNA interference, that DGK␦ regulates the epidermal growth factor (EGF) receptor pathway in lung and skin epithelial cells (20) and insulin receptor signaling in skeletal muscle cells (21,22) by modulating PKC activity. Moreover, a female patient with a disrupted DGK␦ gene who exhibits seizures and a psychiatric disorder was found (23).
We recently reported that DGK is required for the Ras/B-Raf/C-Raf/MEK/ERK signaling cascade to be activated by EGF in HeLa cells, which are derived from cervical cancer (24). Importantly, DGK regulates the recruitment of B-Raf and C-Raf from the cytosol to membranes and controls their heterodimerization. Moreover, the study demonstrated that DGK activates C-Raf, but not B-Raf, in an EGF-dependent manner. The data show that DGK is a novel key regulator of the Ras/B-Raf/C-Raf/MEK/ERK signaling pathway. In addi-tion, Nakano et al. (25) reported that depleting DGK in lung cancer cell lines harboring a mutant EGF receptor reduced their growth on plastic and in soft agar and augmented the effects of afatinib, an EGF receptor inhibitor. In addition to cancer cells, DGK is also highly expressed in the brain (13,16,26). It is interesting to note that a genome-wide association study recently indicated that the gene encoding DGK is implicated in the etiology of bipolar disorder (27,28). Moreover, it was reported that DGK was highly expressed in the brain of bipolar disorder patients (29).
DGK is abundantly expressed in the testis (14,30). A genome-wide association study indicated a potential relationship between DGK and hypospadias (31).
As described above, type II DGKs are physiologically and pathologically important. However, the binding targets and functions of their PHs are still poorly understood. In this study, we investigated the lipid binding properties of the PHs of DGK␦, -, and -using protein-lipid overlay and liposome binding assays. We revealed that the PH of DGK strongly and highly selectively binds to phosphatidylinositol (PI) 4,5-bisphosphate (PI(4,5)P 2 ). The DGK␦-PH also, but to a lesser extent, selectively associated with PI(4,5)P 2 . However, the PH of DGK showed only weak binding activity to PI(4,5)P 2 .
Cell Culture and cDNA Transfection-COS-7 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (Wako Pure Chemical Industries) supplemented with 10% fetal bovine serum (Biological Industries), 100 units/ml penicillin, and 100 g/ml streptomycin (Wako Pure Chemical Industries) at 37°C in an atmosphere containing 5% CO 2 . COS-7 cells were seeded in 60-mm dishes at a density of 2.5 ϫ 10 5 cells/dish. cDNA was transfected into COS-7 cells by electroporation with a Gene Pulser Xcell TM electroporation system (Bio-Rad) according to the manufacturer's instructions.
Protein-Lipid Overlay Assay-One hundred picomoles of the indicated lipids were spotted onto a nitrocellulose membrane (Bio-Rad). The membranes were subjected to blocking with 1% skim milk in Tris-buffered saline, pH 7.4, for 1 h at room temperature. After the blocking, 10 ml of 3% fatty acidfree bovine serum albumin in Tris-buffered saline, pH 7.4, containing lysates from cells expressing GST fusion protein of interest (final concentration, 20 nM) or 80 l of the cell lysates containing DsRed monomer or 3xFLAG fusion protein were added to the membranes. The membranes were then incubated for 30 min at room temperature and then at 4°C overnight. The membranes were incubated with an anti-GST antibody, anti-FLAG antibody, or anti-RFP antibody followed by incubation with peroxidase-conjugated goat anti-mouse IgG antibody or goat anti-rabbit IgG antibody. Finally, lipid-bound proteins were visualized using an ECL Western blotting detection system. Quantitative densitometry was performed using ImageJ software.
Liposome Binding Assay-The liposome preparation contained 1 mg/ml total lipid with the following composition: 100% (w/w) PC, 95% (w/w) PC and 5% (w/w) PS or 95% (w/w) PC and 5% (w/w) PA, and 95% (w/w) PC and 5% (w/w) PI(4,5)P 2 . The combined dried lipid mixture was resuspended in liposome buffer (100 mM NaCl, 1 mM dithiothreitol, and 20 mM HEPES, pH 7.4). Liposome formation was induced by 1-min sonication at 4°C using a Branson Sonifier 450. For sedimentation assays, the cell lysates expressing GST alone and GST-DGK-PH were ultracentrifuged at 100,000 ϫ g for 30 min. One hundred microliters of the cell lysates (100,000 ϫ g supernatant) were mixed with 100 g of liposomes in 100 l of the liposome buffer, incubated for 60 min at 4°C, and ultracentrifuged at 100,000 ϫ g for 60 min at 4°C. The supernatant and pellet were analyzed by SDS-PAGE followed by immunoblotting. Quantitative densitometry was performed using ImageJ software.
DGK Activity Assay-The octyl glucoside mixed micellar assay for DGK activity was performed as described previously (12). In brief, the assay mixture (50 l) contained 50 mM MOPS, pH 7.4, 50 mM octyl glucoside, 1 mM dithiothreitol, 100 mM NaCl, 20 mM NaF, 10 mM MgCl 2 , 1 mM EGTA, 5 mM phosphatidylserine, 1.5 mM diacylglycerol, and 1 mM [␥-32 P]ATP (10,000 cpm/nmol; ICN Biomedicals). The reaction was initiated by adding cell lysates (5 g of protein) and continued for 5 min at 30°C. Lipids were extracted from the mixture, and phosphatidic acid separated by thin layer chromatography was scraped and counted by a liquid scintillation spectrophotometer.
Confocal Laser Scanning Microscopy-Confocal laser scanning microscopy was carried out as described previously (34). Briefly, COS-7 cells were grown on poly-L-lysine-coated glass coverslips and transiently transfected with expression plasmids containing DGK1 or its mutant cDNAs that were N-terminally fused with AcGFP and a DsRed monomer-tagged PLC␦1-PH (35). After 48 h, the cells were serum-starved with DMEM (0.1% fetal bovine serum) for 3 h and incubated with sorbitol in DMEM (final concentration, 500 mM) for 30 min to induce hyperosmotic stress. The cells were then fixed in 3.7% formal-dehyde. For immunofluorescence microscopy, COS-7 cells were permeabilized with 0.1% Triton X-100 in phosphate-buffered saline for 10 min at room temperature. Cells were incubated in phosphate-buffered saline containing 1% bovine serum albumin for 10 min at room temperature as a blocking step. The cells were then incubated in 1% bovine serum albumin in phosphate-buffered saline containing anti-FLAG monoclonal antibody for 30 min at room temperature. After being washed twice with phosphate-buffered saline, cells were incubated with the Alexa Fluor-conjugated secondary antibodies (Invitrogen). The coverslips were mounted using Vectashield (Vector Laboratories). Fluorescence images were acquired using an Olympus FV1000-D confocal laser scanning microscope.
We examined the binding activities of the PHs of DGK, -␦, and -compared with other glycerolipids and sphingolipids. In this experiment, their lipid binding activities were detected after longer exposure. Compared with PA, the DGK-PH bound to CL to a lesser extent (Fig. 2, C and D). PG, PE, SM, Sph, and C1P exhibited either subtle or no detectable binding activities. In contrast, the PH of DGK␦ more strongly bound to PG and CL than to PA and showed either subtle or no detectable binding activities to PE, SM, Sph, and C1P (Fig. 2, C and D). The binding activities of the PH of DGK to these lipids were not detectable (Fig. 2, C and D). Taken together, the PH of DGK exhibited the most pronounced binding activity to PI(4,5)P 2 among the various glycero-and sphingolipids. Although the binding activity is weaker than that of the DGK-PH, the PH of DGK␦ also selectively interacts with PI(4,5)P 2 .
Characterization of Selective Binding Activity of the DGK-PH to PI(4,5)P 2 -Next, we more quantitatively determined the affinity of the PH of DGK for PI(4,5)P 2 . Various concentrations of PI(4,5)P 2 , PI(3,4)P 2 , and PI(3,4,5)P 3 were spotted on a nitrocellulose membrane. As shown in Fig. 3, the intensity of the spot of the DGK-PH on 6.3 pmol of PI(4,5)P 2 was almost equivalent to those of 100 pmol of PI(3,4)P 2 and PI(3,4,5)P 3 . Therefore, the result indicates that the binding affinity of the DGK-PH for PI(4,5)P 2 is markedly stronger than for PI(3,4)P 2 and PI(3,4,5)P 3 . Moreover, in the same experiment, the intensity of the spot of the DGK-PH on 3.1 pmol of PI(4,5)P 2 was almost equivalent to that of 100 pmol of PI(4,5)P 2 of the DGK␦-PH (data not shown).
The effects of the tag and expression system were examined next. The DGK-PH was tagged with DsRed monomer protein instead of GST and expressed in mammalian COS-7 cells instead of E. coli (Fig. 4A). It was confirmed that DsRed monomer-DGK-PH was expressed with the expected molecular mass (38 kDa). As shown in Fig. 4, B and C, the DsRed monomer-DGK-PH expressed in COS-7 cells strongly bound to PI(4,5)P 2 . Similar to the GST-DGK-PH, the binding affinity  (1.6, 3.1, 6.3, 13, 25, 50, and 100 pmol) of PI(3,4)P 2 , PI(4,5)P 2 , PI(3,4,5)P 3 , PA, PC, and PS were spotted onto a nitrocellulose membrane. The membrane was incubated with GSTtagged DGK-PH. Lipid-bound DGK-PH was detected by immunostaining with anti-GST antibody. A representative result of three independent experiments is shown. It was confirmed that lipid-bound GST alone was not detectable in this experiment (data not shown).
In addition to the protein-lipid overlay assay, we confirmed the binding of the DGK-PH to PI(4,5)P 2 using a liposome binding assay. GST alone was not recovered in the precipitate fraction of PI(4,5)P 2 liposome, indicating that GST alone was not associated with PI(4,5)P 2 (Fig. 5, A and B). However, more than 60% of the GST-DGK-PH was recovered in the precipitate fraction of PI(4,5)P 2 liposome (Fig. 5, A and B). In contrast, only ϳ10, 20, and 40% of the DGK-PH was detected in PC, PS, and PA liposomes, respectively. Similar to the lipid-protein overlay assay, these results indicate that the DGK-PH more strongly binds to PI(4,5)P 2 than to PC, PS, and PA.

FIGURE 5. Binding of DGK-PH to PI(4,5)P 2 -containing liposomes.
A, GSTtagged DGK-PH and GST alone were incubated with liposomes containing PC alone and a 9.5:0.5 molar ratio of PC to PS, PA, and PI(4,5)P 2 as indicated. Liposome-binding proteins were recovered by ultracentrifugation. Liposome binding (precipitate (ppt)) and lack of binding (supernatant (sup)) of GST-DGK-PH and GST alone were detected by Western blotting with anti-GST antibody. B, the blots were scanned and quantified using ImageJ software. Relative intensities of liposome binding (precipitate) and lack of binding (supernatant ) of DGK-PH and GST alone are indicated, respectively. The data are shown as the means Ϯ S.D. of three independent experiments. Statistical significance was determined using Student's t test versus PI(4,5)P 2 liposome supernatant (**, p Ͻ 0.01, ***, p Ͻ 0.005) and versus PI(4,5)P 2 liposome precipitate (##, p Ͻ 0.01; ###, p Ͻ 0.005). Error bars represent S.E.
We also examined whether PI(4,5)P 2 can activate DGK. Although 1.0 mol % PI(4,5)P 2 was added to the assay mixture, the activity of DGK1 was not increased (data not shown). Thus, it seems likely that PI(4,5)P 2 is not directly involved in the activation mechanism of DGK in vitro.
Full-length DGK1 and Its PH Co-localize with PLC␦1-PH in Osmotically Shocked Cells-Because the PH of PLC␦1 recognizes and co-localizes with cellular PI(4,5)P 2 (35,36), it is generally used as a cellular PI(4,5)P 2 probe. Moreover, osmotic shock increases the amount of PI(4,5)P 2 in the plasma membrane (38,39). We next observed the co-localization of AcGFPtagged full-length DGK1 with the DsRed monomer-tagged PLC␦1-PH in osmotically shocked cells. Indeed, the PLC␦1-PH was partly localized to the plasma membrane in osmotically shocked cells (Fig. 7). As reported previously (16,32), fulllength DGK1 was translocated to punctate vesicles in response to osmotic shock (Fig. 7B). Interestingly, full-length DGK1 was partly co-localized with the PLC␦1-PH at the punctate regions close to the plasma membrane (Fig. 7B). However, AcGFP alone did not exhibit such co-localization (Fig.  7A). These results indicate that in addition to punctate vesicles in the cytoplasm full-length DGK1 was translocated to subcellular compartments enriched with PI(4,5)P 2 in the plasma membrane.
An AcGFP-tagged DGK mutant containing the PH and the C1 domains (C1Ds) was co-localized with the DsRed monomer-tagged PLC␦-PH at punctate vesicles close to the plasma membrane in osmotically shocked cells (Fig. 7C). However, the C1Ds alone lacking a PH were not co-localized with the PLC␦1-PH (Fig. 7E). In contrast, the AcGFP-tagged DGK-PH alone was translocated to punctate regions in the plasma membrane and markedly co-localized with the DsRed monomertagged PLC␦1-PH there in response to osmotic shock (Fig. 7D). We performed the same experiments using DsRed monomertagged DGK-PH and AcGFP-tagged PLC␦-PH and obtained essentially the same results (Fig. 7G). Taken together, these results indicate that the PH of DGK plays an important role in recognition and co-localization with PI(4,5)P 2 -containing subcellular compartments.
To further evaluate the co-localization of DGK-PH with PI(4,5)P 2 -containing subcellular compartments, we next examined the effect of overexpression of the PI(4,5)P 2 phosphatase synaptojanin. Synaptojanin dephosphorylates the D-5 position phosphate from PI(4,5)P 2 (40). As shown in Fig. 8, AcGFP-DGK-PH was markedly translocated from the cytoplasm to the plasma membrane in response to osmotic shock. However, synaptojanin significantly inhibited the osmotic shock-induced plasma membrane translocation of AcGFP-DGK-PH. The result further indicates that DGK-PH interacted and co-localized with PI(4,5)P 2 in an osmotic shock-dependent manner.
To confirm the subcellular translocation of AcGFP-DGK-PH, we performed cell fractionation. As shown in Fig. 9, the DGK-PH was translocated to the 100,000 ϫ g precipitate (membrane) fraction in osmotically shocked cells, whereas AcGFP alone was not. It was also confirmed that the PLC␦1-PH was translocated to the membrane fraction in response to hyperosmotic shock (data not shown). The membranes were incubated with COS-7 cell lysates containing 3xFLAG-tagged full-length DGK or its PH deletion mutant. C, the blots were scanned and quantified using ImageJ software. The PI(4,5)P 2 binding level of full-length DGK1 was set to 100%. The data are shown as the means Ϯ S.D. of three independent experiments. Statistical significance compared with the PI(4,5)P 2 binding activity of full-length DGK was determined using Student's t test (**, p Ͻ 0.01; ***, p Ͻ 0.005). Error bars represent S.E. APRIL 8, 2016 • VOLUME 291 • NUMBER 15

Discussion
The lipid binding properties of the PH of DGK have not been clear. In this study, we demonstrated that the DGK-PH is a PI(4,5)P 2 -selective binding domain with high affinity (Figs. 2  and 4). It was also confirmed that the full-length DGK1 strongly and selectively interacts with PI(4,5)P 2 through its PH (Fig. 6). Affinities for various phosphoinositides and phospholipids were as follows: PI(4,5)P 2 Ͼ Ͼ PI(3,4)P 2 Ϸ PI(3,4,5)P 3 Ͼ PA Ϸ PI(3)P Ϸ PI(4)P Ϸ PI(5)P Ͼ Ͼ CL Ͼ Ͼ PG Ϸ PE Ϸ PS Ϸ PC Ͼ PI Ϸ SM Ϸ Sph Ϸ C1P (Figs. 2 and 3). Therefore, these results indicate that the DGK-PH is highly selective for PI(4,5)P 2 .   Compared with the PHs of DGK␦ and DGK, the affinity of the PH of DGK for PI(4,5)P 2 is markedly strong (DGK-PH Ͼ Ͼ DGK␦-PH Ͼ Ͼ DGK-PH) (Fig. 2). Partial formation of the proper conformation may cause low affinity for PI(4,5)P 2 . However, high concentrations of the PHs of DGK␦ and DGK, which probably increase the number of the domains that have the proper conformation, showed no effect on their binding affinities for PI(4,5)P 2 , indicating that the different affinities are intrinsic properties.
It was reported that Lys-30, Lys-32, Arg-40, Ser-55, Arg-56, and Lys-57 in the PLC␦1-PH are important for strong PI(4,5)P 2 -selective binding activity (41). The four amino acid residues Lys-30, Arg-40, Ser-55, and Arg-56 in the PLC␦1-PH are conserved in the DGK-PH (Lys-74, Arg-85, Ser-100, and Lys-101, respectively) (Fig. 10). We confirmed that K74A and R85A in DGK-PH are indeed critical residues for PI(4,5)P 2 binding activity (Fig. 11B). In the DGK␦-PH, the three amino acid residues Lys-62, Arg-73, and Lys-89, corresponding to Lys- ever, in the DGK-PH, only two amino acid residues, Lys-225 and Arg-236, are conserved. Therefore, it is possible that the differences in the sequences of these PHs cause the distinct affinities for PI(4,5)P 2 among them (DGK-PH Ͼ DGK␦-PH Ͼ DGK-PH; see Fig. 2). The presence of Gly at the end of the ␤1 strand of PH is important to confer PI(3,4,5)P 3 binding activity (42,43). Because the amino acid residue at the end of the ␤1 strand of the DGK-PH is Asn, it is reasonable that it does not preferably bind to PI(3,4,5)P 3 (Fig. 2). However, the determination of the three-dimensional structure of the PHs of DGK, -␦, and -is required to analyze their different PI(4,5)P 2 binding properties in more detail.
A previous study demonstrated that the PH of DGK␦ nonselectively interacted with PI(4,5)P 2 because it also strongly bound to PS (37). However, high amounts (ϳ20 times higher) of lipids were used in the overlay assay in the previous report (37). Therefore, it is likely that the selectivity of the PH of DGK␦ for PI(4,5)P 2 was detected in the present study (Fig. 2), which used relatively low amounts of phospholipids. The PH of DGK␦ is important for its phorbol ester-dependent plasma membrane localization (15,44). However, it is still unclear whether PI(4,5)P 2 is involved in the translocation.
The EGF receptor signaling pathway, which is regulated by DGK (24), augments PI(4,5)P 2 (45). Osmotic shock also increases the amount of PI(4,5)P 2 at the plasma membrane (38,39). In this study, we also found that the PLC␦1-PH, which is known to be a PI(4,5)P 2 sensor, was localized at the plasma membrane in response to osmotic shock (Fig. 7). DGK-PH(K74A) and -(R85A), which lacked PI(4,5)P 2 binding activities (Fig. 11B), failed to translocate to the plasma membrane and to co-localize with the PLC␦1-PH even in osmotically shocked cells (Fig. 11, C and D). These results strongly suggest that PI(4,5)P 2 was increased in the plasma membrane. The PH of DGK was also co-distributed with the PLC␦1-PH at the plasma membrane. Moreover, depletion of PI(4,5)P 2 by the PI(4,5)P 2 phosphatase synaptojanin significantly reduced the osmotic shock-dependent plasma membrane localization of DGK-PH (Fig. 8). The result further indicates that DGK-PH binds to and is co-localized with PI(4,5)P 2 in cells. Therefore, the PH of DGK could also function as a cellular PI(4,5)P 2 sensor. Compared with the PH of PLC␦1, the PH of DGK is equal to or superior to the PLC␦1-PH in terms of affinity and selectivity for PI(4,5)P 2 (Fig. 4). The PH of PLC␦1 is widely used as a cellular PI(4,5)P 2 detector. Therefore, the PH of DGK could serve as an excellent PI(4,5)P 2 -selective probe with high affinity and selectivity like the PLC␦1-PH.
We have reported that DGK1 was osmotic shock-dependently translocated to punctate vesicles and that the PH and C1Ds are important for the redistribution (16,32). In this study, we observed that the PH alone was translocated to the punctate regions of the plasma membrane where the PLC␦1-PH was colocalized in response to osmotic shock (Fig. 7, D and G), whereas the C1 domains alone, which were distributed to punctate vesicles in the cytoplasm, were not co-localized with the PLC␦1-PH (Fig. 7E). In osmotically shocked cells, full-length DGK and a DGK mutant containing both the PH and the C1Ds were distributed to both PLC␦1-PH-co-localized punctate regions at the plasma membrane and punctate vesicles in the cytoplasm, which do not co-localize with the PLC␦1-PH (Fig. 7, B and C). These results suggest that the PH and C1Ds competitively recruit DGK to subcellular compartments with and without PI(4,5)P 2 , respectively.
We did not detect obvious competition between the PLC␦1-PH and the DGK-PH in both confocal microscopy and cell fractionation. Whether the PI(4,5)P 2 binding competition between these PHs occurs or not is probably dependent on the amount of PI(4,5)P 2 in the plasma membrane. The amount of PI(4,5)P 2 produced in osmotically shocked cells may be greater than the sum of the amounts of the PLC␦1-PH and DGK-PH that are able to access PI(4,5)P 2 in the plasma membrane.
DGK was reported to be implicated in the etiology of bipolar disorder (27)(28)(29). In the brain, DGK is enriched in the dentate gyrus of the hippocampus and the Purkinje cells of the cerebellum (26), which are known to be associated with bipolar disorder (46 -48). A common treatment for bipolar disorder is a mood stabilizer, lithium, which attenuates PI turnover (49). Therefore, DGK may regulate the pathogenesis of bipolar disorder through the PH-dependent binding to PI(4,5)P 2 generated by PI turnover in the dentate gyrus and the Purkinje cells.
Among more than 300 PHs, only the PH of PLC␦1 is an established PI(4,5)P 2 -selective binding domain (50,51). In this study, we added the DGK-PH to the list as a new member. Although the affinity of the DGK␦-PH for PI(4,5)P 2 is relatively lower than that of the DGK-PH, the PH of DGK␦ is an additional PI(4,5)P 2 -selective PH. The PH of DGK could be an excellent cellular sensor for PI(4,5)P 2 that is equal or superior to the PLC␦1-PH. DGK1 has been reported to be involved in EGFdependent cell proliferation (24) and in the pathogenesis of lung cancer (25) and bipolar disorder (27)(28)(29). It will be interesting to determine what role the DGK isozyme, which has the highly PI(4,5)P 2 -selective binding motif, plays in modulating these physiologically and pathologically important events.