New Era of Diacylglycerol Kinase, Phosphatidic Acid and Phosphatidic Acid-Binding Protein

Diacylglycerol kinase (DGK) phosphorylates diacylglycerol (DG) to generate phosphatidic acid (PA). Mammalian DGK consists of ten isozymes (α–κ) and governs a wide range of physiological and pathological events, including immune responses, neuronal networking, bipolar disorder, obsessive-compulsive disorder, fragile X syndrome, cancer, and type 2 diabetes. DG and PA comprise diverse molecular species that have different acyl chains at the sn-1 and sn-2 positions. Because the DGK activity is essential for phosphatidylinositol turnover, which exclusively produces 1-stearoyl-2-arachidonoyl-DG, it has been generally thought that all DGK isozymes utilize the DG species derived from the turnover. However, it was recently revealed that DGK isozymes, except for DGKε, phosphorylate diverse DG species, which are not derived from phosphatidylinositol turnover. In addition, various PA-binding proteins (PABPs), which have different selectivities for PA species, were recently found. These results suggest that DGK–PA–PABP axes can potentially construct a large and complex signaling network and play physiologically and pathologically important roles in addition to DGK-dependent attenuation of DG–DG-binding protein axes. For example, 1-stearoyl-2-docosahexaenoyl-PA produced by DGKδ interacts with and activates Praja-1, the E3 ubiquitin ligase acting on the serotonin transporter, which is a target of drugs for obsessive-compulsive and major depressive disorders, in the brain. This article reviews recent research progress on PA species produced by DGK isozymes, the selective binding of PABPs to PA species and a phosphatidylinositol turnover-independent DG supply pathway.

With respect to PA production by DGK, DGKζ was demonstrated to enhance the activities of mTOR [61] and PIP5KIα [62] through increases in PA levels ( Figure 1). It was reported that DGKα and DGKγ activated PKCζ [63,64] and β2-chimaerin [65], respectively, probably via PA production. DGKδ was recently revealed to activate Praja-1, the E3 ubiquitin ligase acting on the serotonin transporter in the brain, through PA production [40,66]. It was demonstrated that creatine kinase-muscle type (CKM) functionally associated with DGKδ and was activated by PA [67,68].
In this review, we shed light on PAs, especially the diversity of PA molecular species, produced by DGK isozymes and on PABPs, especially those that possess selectivity for PA molecular species. We will also touch on a PI turnover-independent upstream pathway of DGK that was recently found. Moreover, a new PA probe, which is reliable and widely applicable, will be briefly mentioned.

PA Molecular Species Produced by DGK Isozymes Except for DGKε
DGKε (type III) was purified and it clearly showed selectivity for 18:0/20:4-DG in vitro ( Figure 3) [70]. Moreover, cDNA-cloned DGKε selectively phosphorylated 18:0/20:4-DG in vitro [69]. Rodriguez de Turco et al. reported that knockout (KO) of DGKε indeed disturbed PI turnover [71]. Interestingly, DGKε is strongly expressed in Purkinje cells of the cerebellum and pyramidal cells of the hippocampus and regulates seizure susceptibility and long-term potentiation [71], which are governed by PI turnover. Taken together, these results indicate that DGKε is an essential component of PI turnover and exerts its physiological functions as a component of PI turnover in the brain (Figure 3).
On the other hand, nine other isozymes, except for DGKε, fail to show selectivity for 18:0/20:4-DG in vitro [72][73][74][75][76]. Therefore, we questioned whether these nine isozymes indeed utilize PI turnover-derived 18:0/20:4-DG species in cells and organs. However, it has been difficult to quantitatively determine small changes in PA species levels caused by KO and silencing of a DGK isozyme in physiological and pathological events. Liquid chromatography-mass spectrometry (LC-MS) is a powerful tool to detect different molecular species of phospholipids in cells. However, PA was difficult to detect with high accuracy and reproducibility in the general LC conditions because of ion suppression by other major phospholipids, phosphatidylcholine (PC) and sphingomyelin [77]. Therefore, we optimized mobile phases using silica column LC to separate PA from major phospholipids and confirmed that PA species were quantitatively and reproducibly detected in LC-MS [77]. Then, we determined the PA molecular species produced by DGK isozymes in cells and organs and found that a variety of PA species other than 18:0/20:4-PA were generated by DGK isozymes. Intriguingly, these results do not support the dogma that all DGK isozymes utilize DG derived from PI turnover and instead support a new view that DGK isozymes, except for DGKε, utilize DG species derived from pathways independent of PI turnover, as described below ( Figure 3).

DGKα
DGKα, which is a type I isozyme, was first identified by cDNA cloning [78,79]. This isozyme possesses a recoverin homology (RVH) domain and tandem repeats of two Ca 2+ -binding EF-hand motifs [80,81] (Figure 2). Several lines of evidence suggested that the Ca 2+ -induced dissociation of the intramolecular interaction between the EF-hand motifs and the C1 domains of DGKα is the key event that regulates the activity of the enzyme [82][83][84].
In addition to being expressed in cancer cells, DGKα is highly expressed in T cells [78]. In contrast to cancer cells, DGKα facilitates the immune nonresponsive (nonproliferation) state known as T cell clonal anergy [49,93,94]. T cell anergy induction represents the main mechanism by which advanced tumors avoid immune action [95]. Indeed, DGKα limits the antitumor immune response by tumor-infiltrating CD8 + T cells [96]. Therefore, the inhibition of DGKα activity is thought to enhance T cell activity, which governs cancer immunity [44,85,97,98]. We recently found that palmitic acid  Figure 3) [99]. Intriguingly, the profile in starved T cells (palmitic acid-and/or palmitoleic acid (16:1)-containing PA species) [99] is different from that in starved melanoma cells (palmitic acid-containing PA species) [92]. Therefore, DGKα generates distinct PA species in different cells, and the differences in the PA molecular species may account for the opposing functions of DGKα in cancer and T cells.
DGKα-selective inhibitors would be dual effective compounds (i.e., ideal cancer therapy candidates) because, as described above, they attenuate cancer cell proliferation and simultaneously enhance immune responses, including anticancer immunity [100]. Indeed, a DGKα-selective inhibitor, CU-3, induced both cancer cell apoptosis and T-cell activation [91,92]. Other DGKα-selective inhibitors, ritanserin [101] and analogs of Amb639752 (11 and 20) [102], were recently reported as well. These compounds are expected to be able to become ideal cancer drugs.

DGKζ
DGKζ (type IV) contains a MARCKS (myristoylated alanine-rich C kinase substrate) phosphorylation site domain and four ankyrin repeats ( Figure 2) [72,103]. Topham et al. [104] demonstrated that the nuclear-localization signal of DGKζ is located in a MARCKS phosphorylation site domain and that PKCs α and γ regulate the mode of DGKζ localization by phosphorylation of the domain. DGKζ-mediated synaptic conversion of DG to PA is required for the maintenance of dendritic spines [105]. Moreover, DGKζ, syntrophin, and Rac1 form a ternary signaling complex that controls neurite outgrowth in N1E-115 neuroblastoma cells [106]. Previous reports showed that the level of PA was increased during neuronal differentiation [107,108]. However, it has not been revealed what PA molecular species are produced. Recently, 16:0/16:0-PA and, to a lesser extent, 14:0/16:0-PA and 16:0/18:0-PA, were found to be exclusively generated during differentiation of Neuro 2A neuronal cells in a DGKζ-dependent manner ( Figure 3) [109]. DGKζ1 ( Figure 2), but not DGKζ2, was physically associated with RasGRP1 and attenuated RasGRP1 activity by DG consumption [59]. Therefore, in addition to DGKα, DGKζ acts as a suppressor of T cell functions and its inhibitors are expected to be useful for cancer immunotherapy [44,85,98,110]. DGKs ζ and α appear to share the same function (inhibition of RasGRP1 activity in T cells and consequently attenuation of T cell activity). Indeed, the combination of the inhibition of DGKα and DGKζ additively or synergistically induces activation of T cells [111]. However, it is still unclear what PA species are generated by DGKζ in T cells.
DGKδ is strongly expressed in the skeletal muscle [112]. DGKδ regulates glucose transport [54,56,118] and contributes to exacerbating the severity of type 2 diabetes (T2D) [54,56]. It was recently found that, in response to high glucose-stimulation, 16 Interestingly, we recently demonstrated that myristic acid (14:0) increased the expression of DGKδ and enhanced glucose uptake in C2C12 myotube cells [118,119]. Moreover, chronic oral administration of myristic acid improved hyperglycaemia by decreasing insulin-responsive glucose level in Nagoya-Shibata-Yasuda mice, a spontaneous model for studies of obese T2D [120]. These results indicate that myristic acid is a potential candidate for the prevention and therapy of T2D and its related diseases.
It is known that docosahexaenoic acid (DHA, 22:6, ω-3) deficiency occurs during aging and dementia and that the deficiency impairs memory and learning, exacerbates anxiety and depression, and promotes age-related neurodegenerative diseases, including Alzheimer's disease [214]. DHA is asserted to increase membrane fluidity, strengthen antioxidant activity, and plays anti-inflammatory roles [214]. However, all these effects chemically/physically, nonselectively, and indirectly affect the brain functions. On the other hand, DHA-containing PA biologically, selectively, and directly activates Praja-1 E3 ubiquitin-protein ligase and, consequently, reduces the amount of SERT protein [215], which attenuates the serotonergic system and is the target of anti-depression and anti-OCD drugs [216,217]. Therefore, it is possible that DHA incorporated into PA (and chemical compounds mimicking 18:0/22:6-PA) biologically, selectively, directly, and most effectively protect the brain dysfunctions listed above.

DGKκ
DGKκ, which is a type II isozyme, possesses a PH domain at the N-terminus (Figure 2) [229]. DGKκ, but not other type II DGKs, is tyrosine-phosphorylated at Tyr-78 in the N-terminal, κ-isoform-specific extension through the Src family kinase pathway in response to oxidative stress [229]. Moreover, the stress inhibits DGKκ activity.

DGKθ
DGKθ (type V) has three C1 domains, a glycine/proline-rich region, a Ras association domain and a PH domain-like region (Figure 2) [74]. GWAS suggested that single nucleotide polymorphisms (rs1564282 and rs11248060) of DGKQ (DGKθ gene) are associated with a higher risk of Parkinson's Disease [231,232]. DGKθ is highly expressed in the cerebellum and hippocampus in the adult rat brain [74]. Intriguingly, overexpression of DGKθ mainly increases the amount of 18:1/18:1-PA in mouse primary hepatocytes (Figure 3) [233]. It is interesting that the PA species strongly binds to α-synuclein (see Table 1), which is associated with the pathogenesis of Parkinson's Disease (see Section 3.1) [42].

Molecular Species Selectivity of PABP
A number of proteins such as protein kinases, lipid kinases, protein phosphatases, lipid phosphatases, phospholipases, G-proteins, G-protein regulators, and phosphodiesterases have been identified as PABPs to date (Table 1) [76]. Therefore, in many cases, the screening for detecting PABPs was performed with a major PA species, 16:0/18:1-PA. Moreover, because the molecular species selectivity of PABPs has not attracted attention so far, PA species mixtures were used to detect PABPs in many cases. However, as described previously, we recently found that several DGK isozymes generate diverse PA species. It is possible that the general screening with 16:0/18:1-PA and PA mixtures missed some PABPs that are selective for minor PA species. Therefore, we recently started a comprehensive screening for PABPs in the skeletal muscle and brain using several minor PA species, including 16:0/16:0-PA, which is generated by DGKδ in myoblast cells [76] and DGKζ in neuronal cells [109], and 18:0/22:6-PA, which is produced by DGKδ in the brain [66]. As a result, we found several new PABPs that have different selectivities to PA species. In addition to PA-selective PABPs discovered by us, there are only several such PABPs. Intriguingly, these PABPs do not exhibit the selectivity to 18:0/20:4-PA, which is derived from the PI turnover, indicating that they interact with PI turnover-independent PA species.
As described previously, DGKθ, which has been reported to be associated with the risk of Parkinson's Disease [231,232], preferentially produced 18:1/18:1-PA [233]. Interestingly, the content of PA increased in aged male mice (12-14 months old), but that of PS decreased with age [240]. Aging is the greatest risk factor for developing sporadic Parkinson's Disease [241]. Moreover, Parkinson's Disease incidence is 1.5 times higher in men than women [242,243]. Therefore, it is possible that 18:1/18:1-PA produced by DGKθ enhances the pathogenesis of Parkinson's Disease (Figure 3).

L-Lactate Dehydrogenase (LDH) A
LDHA in skeletal muscle is an energy-metabolizing enzyme critical for tumor-related anaerobic respiration [247]. LDHA was already reported to bind to acidic phospholipids such as PS and cardiolipin (CL), at acidic pH [248]. However, at physiological pH (7. It was reported that LDHA is upregulated in human tumors, including glioblastoma [249][250][251]. The Warburg effect, which is the anaerobic metabolism by tumor cells even under well-oxygenated conditions, has been suggested to be an adaptive mechanism to maintain the biosynthetic requirements of uncontrolled proliferation [252]. LDHA is a key enzyme of the Warburg effect [247,253,254]. Indeed, silencing/genetic disruption of LDHA inhibited tumor growth in vitro and in vivo [255][256][257]. It is noteworthy that arachidonic acid (20:4)-and DHA (22:6)-containing DG were decreased within tumor regions [258]. Therefore, it is likely that a decrease of the PA molecular species containing PUFA, arachidonic acid or DHA, cannot attenuate the activity of LDHA in tumor cells. It is possible that chemical compounds that mimic 18:0/20:4-PA and 18:0/22:6-PA can be drugs against tumor cell growth.

CKM
CKM is also an energy metabolizing enzyme and has long been known to be correlated with T2D [259][260][261][262]. We recently identified CKM by screening using 16

Seipin
Seipin, which is an integral membrane protein in the ER, is important for lipid droplet formation.

PA Probe
As described previously, PAs plays important physiological roles as second messengers. Therefore, tracking the localization and dynamics of intracellular PA is essential for understanding a wide variety of physiological and pathological events regulated by PA. Several PA-binding domains (PABDs), such as Spo20p-PABD [38,186] and PDE4A1-PABD [37,38], are often used as PA probes [265][266][267]. However, they exhibit their own subcellular localization to the plasma membrane (Spo20p-PABD) and Golgi apparatus (PDE4A1-PABD) in a cell stimulation-independent manner (a cell stimulation-induced PA generation-independent manner) [15,38]. The cell stimulation-independent localization disturbs their functions as PA probes and, consequently, makes them relatively difficult to apply. Therefore, a reliable and widely applicable PA probe that can be used for any cell stimulation and cell type has not been sufficiently developed to date.
In this context, α-synuclein N-terminal region (α-synuclein-PABD) is useful for PA sensing in living cells [43]. The region does not exhibit its own subcellular localization to cell membranes such as the plasma membrane and Golgi apparatus in a cell stimulation-independent manner, in contrast to PA sensors developed so far. It was confirmed that α-synuclein-PABD was able to sense physiologically produced, endogenous PA in phagosomes [268]. Moreover, it is interesting to note that the probe detected PA at the peripheral regions (close to the plasma membrane) of neuronal growth cones [268].

DG-Providing Pathway Upstream of DGK
How do DGK isozymes produce distinct PA species? DGK isozymes, except for DGKε, have no DG species selectivity in vitro, implying that there are different upstream DG supply pathways and/or DG pools, which are independent of PI turnover and provide various DG species to each DGK isozymes. Thus, it is speculated that DG supply pathway(s) upstream of DGK provide certain DG species.
Unlike myoblast cells [76], DG species (18:0/22:6-PA) utilized by DGKδ in the brain are not 16:0containing DG (Figure 3) [66]. In addition to DGKδ, PA species produced by DGKα in melanoma and T cells are also different from each other ( Figure 3) [92,99]. The results imply that DGK isozymes utilize distinct DG-supplying pathways in different organs/tissues/cells and/or in response to different cell stimuli. DGKδ was found to interact with SMSr via the SAM domain. However, only DGKδ1, δ2, and η2 have the SAM domain ( Figure 2). Thus, other DGK isozymes lacking the SAM domain would utilize other DG-providing pathways instead of SMSr. It is urgently needed to explore other DG supply enzymes/pathways.

Physiological Implication of Diversity of PA Molecular Species and PABPs
Unlike DGBPs, which have the common DG-binding domain (the C1 domain), obviously common PA-binding motifs, like the C1 domain, have not been identified in PABPs ( Table 1). The lack of communality may generate the high diversity of PABPs, which have different selectivity to PA species.
PA is the simplest glycerophospholipid. Hydrophilic head groups of PI (phosphate + inositol ring), CL (phosphate + phosphatidylglycerol (PG)), PG (phosphate + glycerol), PS (phosphate + serine), PE (phosphate + ethanolamine) and PC (phosphate + choline) are considerably larger than PA (phosphate alone). PA forms a cone-like molecular shape, rather than the cylindrical shape typical of other glycerophospholipids [16,274]. The shape of PA likely generates void space surrounding PA molecules. Taken together, it is speculated that PABP can easily access the fatty acid moieties of PA. If this is the case, fatty acid composition of PA would be more physiologically significant than those of other phospholipids. In contrast, DGBPs do not show obvious DG species selectivity, exemplified by PKC [275]. Because DG has only hydroxy group as the hydrophilic head, the lipid is deeply embedded in the lipid bilayer membrane. Thus, DGBPs would have difficulties accessing the fatty acid moieties of DG. However, to prove the hypothesis that PABP can easily access the fatty acid Puzzlingly, SMSr shows only slight CPE synthase activity [269]. However, it is interesting to note that, in addition to CPE synthase activity, SMSr protein, which was expressed using the baculovirus-insect cell system and highly purified, generated DG through the activities of PA phosphatase (PAP) and PI-PLC in vitro ( Figure 4) [Murakami, C. and Sakane, F. unpublished work]. These activities were much stronger than the CPE synthase activity. Moreover, SMSr as PAP prefers SFA and/or MUFA-containing PA species (16:0/16:0-PA and 16:0/18:1-PA) but not PUFA-containing PA species (18:0/20:4-PA or 18:0/22:6-PA). Therefore, these results further support that the supply of DG by SMSr (PAP and PI-PLC) is independent of PI turnover.
Unlike myoblast cells [76], DG species (18:0/22:6-PA) utilized by DGKδ in the brain are not 16:0-containing DG (Figure 3) [66]. In addition to DGKδ, PA species produced by DGKα in melanoma and T cells are also different from each other ( Figure 3) [92,99]. The results imply that DGK isozymes utilize distinct DG-supplying pathways in different organs/tissues/cells and/or in response to different cell stimuli. DGKδ was found to interact with SMSr via the SAM domain. However, only DGKδ1, δ2, and η2 have the SAM domain ( Figure 2). Thus, other DGK isozymes lacking the SAM domain would utilize other DG-providing pathways instead of SMSr. It is urgently needed to explore other DG supply enzymes/pathways.

Physiological Implication of Diversity of PA Molecular Species and PABPs
Unlike DGBPs, which have the common DG-binding domain (the C1 domain), obviously common PA-binding motifs, like the C1 domain, have not been identified in PABPs ( Table 1). The lack of communality may generate the high diversity of PABPs, which have different selectivity to PA species.
PA is the simplest glycerophospholipid. Hydrophilic head groups of PI (phosphate + inositol ring), CL (phosphate + phosphatidylglycerol (PG)), PG (phosphate + glycerol), PS (phosphate + serine), PE (phosphate + ethanolamine) and PC (phosphate + choline) are considerably larger than PA (phosphate alone). PA forms a cone-like molecular shape, rather than the cylindrical shape typical of other glycerophospholipids [16,274]. The shape of PA likely generates void space surrounding PA molecules. Taken together, it is speculated that PABP can easily access the fatty acid moieties of PA. If this is the case, fatty acid composition of PA would be more physiologically significant than those of other phospholipids. In contrast, DGBPs do not show obvious DG species selectivity, exemplified by PKC [275]. Because DG has only hydroxy group as the hydrophilic head, the lipid is deeply embedded in the lipid bilayer membrane. Thus, DGBPs would have difficulties accessing the fatty acid moieties of DG. However, to prove the hypothesis that PABP can easily access the fatty acid moieties of PA, 3D structures and molecular dynamics simulations of PABPs associated with PA molecule are further needed.
The results recently obtained suggest that DGK-PA-PABP axes can potentially construct a large and complex signaling network. DGK isozymes generate various PA species. Moreover, several PA species-selective PABPs, which regulate their related functions, have been found, and the list of PA species-selective PABPs is still growing. In addition to DGK, phospholipase D (PLD) [276] generates PA as a signaling lipid through the hydrolysis of PC (Figure 1). It has been reported that many PABPs are controlled by PLD-dependent PA [11,[14][15][16][17]. Although DGK and PLD commonly generate PA, the profiles of PA species would be distinct from each other. PLD employs only PC as a substrate. On the other hand, for example, DGKδ can utilize DG species derived from PA, PI, and PE through PAP, PI-PLC and CPE synthase activities of SMSr [Murakami, C. and Sakane, F. unpublished work]. Therefore, it is likely that the variation of PA species produced by DGK is higher than that by PLD. Lysophosphatidic acid acyltransferase (LPAAT) also generates various PA species (Figure 1), which are basically utilized as precursors of various phospholipids. Interestingly, there are several LPAAT isozymes that can add different fatty acids to LPA [277][278][279]. PLD-and LPAAT-derived PA species, which can also bind to PA species-selective PABP, together with DGK-derived PA species would confer complexity to the PA molecular species-signaling network. The network consists of various PA producing enzymes including DGK isozymes, various PA molecular species and various PABPs and may regulate a wide variety of physiological functions and pathogenesis.
In mammals, yeasts and plants, different PA species are enriched. As previously described, in mammalian myoblast cells, 34 [282,283]. Therefore, it is likely that these organisms utilize different PA species for their cellular signaling systems and that different PA species construct distinct PA molecular species-signaling networks.

Conclusions
In addition to DG, PA is a versatile lipid second messenger. It was recently demonstrated that DGK isozymes selectively generate various PA species, which are independent of PI turnover, in isozyme-dependent and cell/stimulation-dependent manners. Moreover, there are a number of PABPs and several of them exhibit PA species selectivity. In addition, the lists of DGK isozyme-derived PA species and PABPs, especially PA species-selective PABPs, are still growing. Because PA species selectivity of only a small part of identified mammalian PABPs has been determined, the selectivity of other PABPs should be re-evaluated to explore the functions of PA species/PABPs in more detail. Most likely, many of them would show their own PA selectivity. Therefore, the recent progress in DGK and PABPs allows us to speculate that the DGK-PA-PABP axes may configure a massive network that is more complex and larger than we expect.
However, there are still many questions concerning PA and PABPs. For example, why does a variety of PA species and PABPs exist? Do PABPs most efficiently recognize fatty acid compositions? What are the upstream DG supply pathways for DGK isozymes lacking the SAM domain? Hence, we may still be in the dark in terms of PA molecular species, their generating pathways, and their molecular functions. However, there is no doubt that diversities of PA species and PABPs are key to exploring molecular mechanisms of a variety of physiological and pathological events regulated by DGK isozymes (and PLD/LPAAT).

Conflicts of Interest:
The authors declare no conflict of interest. Sporulation-specific protein 20 T2D

Abbreviations
Type 2 diabetes