Mammalian Diacylglycerol Kinases, a Family of Lipid Kinases with Signaling Functions*

Many intracellular signaling pathways are initiated by a simple reaction, the hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP2) 1 (1), which results in a transient rise in the amounts of diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP3). The initial signaling has two predictable components: IP3 binds to intracellular receptors to initiate calcium release from intracellular stores (2), and DAG functions as an allosteric activator of protein kinase C (PKC) (3). Both the polar product, IP3, and the lipid messenger, DAG, are converted to inactive products to return the cell to the basal state; the pathways that regulate inositol phosphate levels were reviewed in the fourth minireview in this series (22). In addition to activating PKC, DAG participates in other cellular events. For example, it is a potent activator of the guanine nucleotide exchange factors vav (4) and Ras-GRP (5), indicating a potential role for DAG in regulating Ras and Rho family proteins. In addition to these signaling roles, DAG occupies a central position in the synthesis of major phospholipids (phosphatidylcholine and phosphatidylethanolamine (6)) and triacylglycerols. Thus, to maintain cellular homeostasis, intracellular diacylglycerol levels must be tightly regulated. This is illustrated by evidence that inappropriate accumulation of diacylglycerol contributes to cellular transformation. For example, cell lines that overexpress PLC g have a malignant phenotype (7). Also, cells transformed with one of several oncogenes have elevated DAG levels (8–11), and growth factors that are proto-oncogenes stimulate this pathway. Most of the evidence for this pathological effect centers on excessive and/or prolonged activation of PKC, which is a common feature of the transformed state, both in tumors and in cell cultures (12). PKC function was identified, in part, by virtue of being the target for phorbol esters; these tumor promoters function in the same way as DAG to activate PKC but persist because they are not metabolized (or at least this happens very slowly). Thus, these observations have led to the hypothesis that prolonged elevation of DAG functions as a tumor promoter, the equivalent of an endogenous phorbol ester. This review focuses on a family of enzymes, the diacylglycerol kinases (DGKs) that phosphorylate diacylglycerol to phosphatidic acid (PA) (Fig. 1), which also has signaling functions; it stimulates DNA synthesis (13, 14) and modulates the activity of several enzymes including phosphatidylinositol 5-kinases (PI-5-K) (reviewed by Rameh and Cantley (84), first article in this series), PAK1 (15), PKCz, and Ras-GAP (16). Although the bulk of the signaling “pool” of PA (it, too, is an intermediate in phospholipid synthesis) is thought to derive from the action of phospholipase D (16), DGKs likely contribute to it as well. Thus, DGKs catalyze a reaction that removes DAG and would terminate the PKC-mediated signal but yield a product, PA, that has other functions both in signaling and phospholipid synthesis. The net result on cellular events is, therefore, difficult to predict, but the potential outcomes all support the conclusion that DGKs occupy an interesting niche.

Many intracellular signaling pathways are initiated by a simple reaction, the hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP 2 ) 1 (1), which results in a transient rise in the amounts of diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP 3 ). The initial signaling has two predictable components: IP 3 binds to intracellular receptors to initiate calcium release from intracellular stores (2), and DAG functions as an allosteric activator of protein kinase C (PKC) (3). Both the polar product, IP 3 , and the lipid messenger, DAG, are converted to inactive products to return the cell to the basal state; the pathways that regulate inositol phosphate levels were reviewed in the fourth minireview in this series (22).
In addition to activating PKC, DAG participates in other cellular events. For example, it is a potent activator of the guanine nucleotide exchange factors vav (4) and Ras-GRP (5), indicating a potential role for DAG in regulating Ras and Rho family proteins. In addition to these signaling roles, DAG occupies a central position in the synthesis of major phospholipids (phosphatidylcholine and phosphatidylethanolamine (6)) and triacylglycerols. Thus, to maintain cellular homeostasis, intracellular diacylglycerol levels must be tightly regulated. This is illustrated by evidence that inappropriate accumulation of diacylglycerol contributes to cellular transformation. For example, cell lines that overexpress PLC ␥ have a malignant phenotype (7). Also, cells transformed with one of several oncogenes have elevated DAG levels (8 -11), and growth factors that are proto-oncogenes stimulate this pathway. Most of the evidence for this pathological effect centers on excessive and/or prolonged activation of PKC, which is a common feature of the transformed state, both in tumors and in cell cultures (12). PKC function was identified, in part, by virtue of being the target for phorbol esters; these tumor promoters function in the same way as DAG to activate PKC but persist because they are not metabolized (or at least this happens very slowly). Thus, these observations have led to the hypothesis that prolonged elevation of DAG functions as a tumor promoter, the equivalent of an endogenous phorbol ester.
This review focuses on a family of enzymes, the diacylglycerol kinases (DGKs) that phosphorylate diacylglycerol to phosphatidic acid (PA) (Fig. 1), which also has signaling functions; it stimulates DNA synthesis (13,14) and modulates the activity of several enzymes including phosphatidylinositol 5-kinases (PI-5-K) (reviewed by Rameh and Cantley (84), first article in this series), PAK1 (15), PKC, and Ras-GAP (16). Although the bulk of the signaling "pool" of PA (it, too, is an intermediate in phospholipid synthesis) is thought to derive from the action of phospholipase D (16), DGKs likely contribute to it as well. Thus, DGKs catalyze a reaction that removes DAG and would terminate the PKC-mediated signal but yield a product, PA, that has other functions both in signaling and phospholipid synthesis. The net result on cellular events is, therefore, difficult to predict, but the potential outcomes all support the conclusion that DGKs occupy an interesting niche.

The Diacylglycerol Kinase Gene Family
DGKs have been identified in bacteria (17), Drosophila melanogaster (18 -20), Caenorhabditis elegans, 2 and plants (21), but this review will focus on the mammalian isoforms. However, special note should be made of the DGK from Escherichia coli; it is the best studied member of a family of prokaryotic DGKs that have little structural relationship to the eukaryotic enzymes. Because there is no evidence that there is a signaling function for DAG in bacteria, the DGKs presumably serve exclusively for the synthesis of complex lipids. The E. coli DGK has also established a technological niche as a reagent to determine DAG levels (10). Mammalian DGK activities have been identified in multiple cell types and a wide range of tissues, indicating their functional significance (23). Nine isoforms have been cloned (Fig. 2), and all of them have a conserved catalytic domain that likely functions similarly to the protein kinase catalytic domains by presenting ATP as the phosphate donor. One interesting feature of DGKs ␦ and (and one Drosophila DGK (19)) is that their catalytic domains are bipartite, indicating that the two modules may act cooperatively. All of the DGK catalytic domains have at least one presumed ATP binding site with the consensus GXGXXG that is also found in protein kinases (24). A mutation of the third glycine to aspartate abolishes activity of DGKs ⑀, , 2 and a Drosophila DGK (20). However, this ATP binding motif differs from that of the protein kinases, where there is an essential lysine 14 -23 amino acids downstream of the glycines (24). All of the known DGKs have a lysine in a similar position, but site-directed mutagenesis of this amino acid in DGKs ␣ (25, 26), ⑀, or 2 does not alter activity, indicating that the ATP binding pockets of DGKs have a different conformation than the protein kinases.
All DGKs have at least two cysteine-rich regions homologous to the C1A and C1B motifs of PKCs. DGK has three. These regions are also found in several other proteins and in most cases are thought to bind diacylglycerol or phorbol esters. However, C1 domains in several proteins clearly do not bind DAG and in some cases serve as sites of protein-protein interaction (27). Originally, these domains in DGKs were thought to present diacylglycerol for phosphorylation. However, no one has conclusively demonstrated that this occurs. In fact, Sakane et al. (26) observed that DGK␣ was still active without its C1 domains and that DGKs ␣, ␤, and ␥ all failed to bind phorbol esters. We have observed, however, that DGK lacking either of its two C1 domains was inactive. 2 Thus, it is unclear precisely what role these regions perform in DGKs, but in at least some isoforms they are essential for activity. Hurley et al. (28) recently examined the sequence homology of 54 C1 domains, including those of six DGKs (␣, ␤, ␥, ␦, ⑀, and ). They proposed that except for the C1A domains of DGKs ␤ and ␥, all other DGK C1 domains may not bind DAG. The different functions of the proteins must be considered, however; when PKCs bind DAG, they essentially exclude it from an attacking phosphoryl group, which is sensible as the DAG is functioning as an allosteric activator. In contrast, DGKs must present DAG for phosphoryl transfer, which suggests that they might bind it differently; thus, an altered C1 conformation might serve such a purpose. * This minireview will be reprinted in the 1999 Minireview Compendium, which will be available in December, 1999. This is the fifth article of five in "A Thematic Series on Kinases and Phosphatases That Regulate Lipid Signaling." ‡ Howard Hughes Medical Institute physician postdoctoral fellow. § To whom correspondence should be addressed: Huntsman Cancer Inst., University of Utah, 15 N. 2030 East, Rm. 4220, Salt Lake City, UT 84112. Tel.: 801-585-3773; Fax: 801-585-6345; E-mail: stephen.prescott@hci. utah.edu. 1 The abbreviations used are: PIP 2 , phosphatidylinositol 4,5-bisphosphate; DAG, diacylglycerol; DGK, diacylglycerol kinase; PA, phosphatidic acid; PKC, protein kinase C; PLC, phospholipase C; IP 3 , inositol 1,4,5-triphosphate; PI, phosphatidylinositol; IL-2, interleukin-2; PH, pleckstrin homology; PLD, phospholipase D; PI-5-K, phosphatidylinositol 5-kinase; MARCKS, myristoylated alanine-rich C kinase substrate; GDI, guanine nucleotide dissociation inhibitor. 2  In addition to their catalytic and C1 domains, most DGKs have structural motifs that form the basis for dividing them into five subtypes and that likely play regulatory roles (Fig. 2). Type I DGKs (29 -31) have calcium binding EF hand motifs at their N termini, making these isoforms calcium-responsive to slightly different extents (32). Diacylglycerol kinases with pleckstrin homology (PH) domains at their N termini are defined as type II (33,34). Takeuchi et al. (35) found that the PH domain of DGK␦ could bind phosphatidylinositols (PI). DGK␦ also has at its C terminus a region homologous to the EPH family of receptor tyrosine kinases. The function of this domain is unclear. DGK⑀ (36) has the simplest structure in that it does not have any identifiable regulatory domains and is the sole member of group III to date. Paradoxically, although it is the simplest in structure, it is the only DGK known to have substrate specificity; it strongly prefers diacylglycerol substrates with an arachidonoyl group at the sn-2 position. This preference initially suggested that DGK⑀ might be the component in the phosphatidylinositol cycle that accounted for the enrichment of PI with arachi-donate (37); however, this isoform has a limited tissue distribution and therefore cannot serve this function broadly. The defining characteristic of type IV DGKs (38,39) is that they have ankyrin repeats at their C termini. One enzyme of this family, DGK, undergoes tissue-specific alternative splicing that results in an enzyme with an elongated N terminus; it is found predominantly in muscle (40). Both DGK and have a region homologous to the phosphorylation site domain of the MARCKS protein. Finally, DGK defines group V (41); it has three cysteine-rich domains and a PH domain, as well as a region that is structurally similar to those in other proteins that have been implicated as mediating association with Ras, although this point is controversial (42). Thus, the complexity and diversity of the DGK family strongly suggest that the DGKs perform multiple roles in cellular functions.

Regulating Kinase Activity
Because DGKs influence cellular DAG and PA levels, control of their activity is essential; and this may be an important mechanism by which different functions are segregated, e.g. DAG that is used for complex lipid synthesis contrasted with DAG used as a signal. Activation of Type I DGK activity by calcium represents a potentially elegant way to increase DGK activity in situations in which DAG is elevated, because calcium usually increases in parallel because of the simultaneous signaling by IP 3 . This mechanism would attenuate the DAG signal, returning that arm of the signaling cascade to a quiescent state. Other than this effect, however, relatively little is known about how the activity is regulated. Several groups have observed compounds that can modulate DGK activity in cellular homogenates. These include arachidonic acid (43), vitamin E (44), sphingosine (45), 15-hydroxyeicosatetraenoic acid (46), ceramide (47), and several fatty acids (48,49). In most cases it is not clear that these observations are physiologically relevant, but Reddy and co-workers (50) observed that dietary fatty acids induced marked alterations in DGK activity in colon tumors. Also, Walsh et al. (51) found that PIP 2 was a potent inhibitor of arachidonoyl-specific DGK activity.
Several investigators have examined post-translational modifications that may regulate DGK function. Schaap and co-workers (52) and Kanoh et al. (53) have shown that DGK␣ is phosphorylated by PKC isoforms in vitro and in vivo. These findings were consistent with previous observations that suggested that PKC regulated the activity of DGK in cellular homogenates (54); however, neither group could identify a functional consequence resulting from the PKC-mediated phosphorylation.
A consistent observation from several groups is that DGKs shuttle between the cytosol and membrane fractions of cells and that both DGK activity and immunoreactivity become membrane-associated upon exposure to chemotactic factors (55), adrenergic agonists (56), phorbol esters (57), and diacylglycerol (58). These findings raise the distinct possibility that DGKs are regulated by access to specific DAG pools at the membrane, i.e. by moving the enzyme to its substrate. In support of this possibility, van der Bend and co-workers (59) presented experiments indicating that cellular DGK activity was topologically restricted to DAG pools generated after receptor activation rather than upon nonspecific generation of diacylglycerol by PLC treatment of membranes. In agreement with this, Flores et al. (60) recently observed that DGK␣ translocates to the nuclear cytoskeleton upon stimulation of cells via the IL-2 receptor. They went on to show that the PA generated by a DGK (presumably ␣) following IL-2 treatment may be necessary for subsequent cellular proliferation. We found that DGK also translocates to the nucleus and that this localization is negatively regulated by PKC-mediated phosphorylation of the MARCKS phosphorylation site domain, which acts as the nuclear localization signal (61). These observations provide strong evidence that cells regulate DGK activity (and thus DAG levels) by altering the specific activity of DGKs and by translocating them to regions where the DAG has accumulated. Because there is evidence of multiple DGK isoforms within the same cell type, the different isozymes likely perform specialized roles, perhaps by recognizing DAG pools generated in response to different stimuli or that are found in specific membrane compartments.

Function(s) of DGKs
Lessons from Drosophila-The diversity of the DGK family suggests that each isoform may perform a distinct function. However, this has been difficult to study because of a lack of isoform-specific inhibitors, and thus far, there has not been a description of a dominant negative DGK construct. If one turns to Drosophila, however, there has been demonstrated a clear role for activity of one isoform. Drosophila strongly relies on PLC activity to generate IP 3 and DAG for its visual signal transduction (62). Several groups demonstrated that a Drosophila visual mutant, rdgA, was deficient in retinal DGK activity (63), and then Masai et al. (20) cloned the gene responsible for this defect. It was a DGK (dDGK2) with four C-terminal ankyrin repeats and is most similar to the mammalian type IV enzymes. They found a mutation in the ATP binding site that would presumably account for the loss of DGK activity. This group later showed that the rdgA phenotype could be rescued by introducing wild-type rdgA into the mutant strain (64). Together, these studies demonstrate that formation of phosphatidic acid by dDGK2 is necessary for proper visual signaling in Drosophila.
Although there are differences in photoreceptor signaling between Drosophila and mammals (65) these observations may provide insight into a function for some mammalian DGK isoforms. In fact, several groups have found evidence of light-dependent activation of PIP 2 hydrolysis and generation of PA in vertebrate retina (66 -68), and because PKC may participate in vertebrate phototransduction (69) a role for DGK activity in mammalian retina would not be surprising. To date, three mammalian isoforms have been definitively localized to the retina. The first was DGK␥, a type I enzyme. Kai and co-workers (31) found that this isoform is predominantly localized in human retina whereas most other tissues had inactive, truncated mutants. To date, they have not identified a specific function for this protein in visual signaling. Rat DGK⑀ has also been found in both the outer and inner nuclear layers of the retina (70), and we have observed that human and murine DGK⑀ are strongly expressed in the retina. 2 Again, no specific function in vision has been attributed to this isoform. Finally, DGK, a type IV enzyme, was cloned from a cDNA library prepared from retina (39). It is the isoform most similar to dDGK2, and current studies are under way to determine a role for this isozyme in retinal signaling. In conclusion, no one has identified a role for DGK activity in mammalian visual signaling, but several isoforms are strongly and predominantly expressed in the retina, and their functions there have not yet been elucidated.
Lessons from Tissue Distribution-Mammalian DGK isoforms have been observed in a variety of tissues by both Northern and Western blot analysis. With the exception of DGKs ␦ and ⑀, all isoforms identified to date are found in the brain at a level equivalent to or higher than any other tissue. In situ hybridization and Northern blot analysis of specific brain regions have revealed high levels of expression of several isoforms in the hippocampus (␤, ␥, ⑀, , ), cerebellum (␥, ⑀, , ), olfactory bulb (␤, ␥, , ), and the retina as noted above. It is interesting that specific regions of the brain express several DGK isotypes, typically representing many known subfamilies. This is strong evidence that the enzymes within these subfamilies perform distinct functions. In fact, we have observed four DGK isoforms in a glioblastoma cell line: DGKs ␣, ␦, , and . 2 Notably, each isoform is from a different DGK subfamily. Finally, Ding et al. (71) found dramatic evidence of specific association of a DGK with neuronal structures in the developing mouse. Together, these data strongly suggest that DGKs are an integral part of the central nervous system and peripheral neuronal function.
Two other tissues where DGKs appear to be enriched are white blood cells and muscle. Both appear to express multiple DGK isoforms, usually representing different families. Human DGK␣ was purified and then cloned using white cells, spleen, and thymus tissue (29,72). DGKs ␣ and are expressed in HL60 cells, 2 DGK␦ was identified by Northern blotting in leukocytes (33), and mRNA for DGKs and have been observed in the spleen (34) and thymus (73), respectively. Localization in these tissues is of interest because several groups using DGK inhibitors have suggested that DGK activity suppresses the respiratory burst (74) and chemotaxis (75) in leukocytes. Both effects have been attributed to metabolism of DAG, which would presumably suppress PKC activity. Finally, several DGK isoforms have been identified in both striated muscle and cardiac muscle. Using Northern blotting, Sakane et al. (33) found DGK␦ mRNA predominantly expressed in skeletal muscle. DGKs ␤ (30) and ⑀ 2 are found in mRNA prepared from heart tissue, and we have identified an alternatively spliced isoform of DGK in striated muscle (40). Again, no one has clearly identified a role for DGK activity in muscle tissue, except that a DGK inhibitor can suppress contractility in some smooth muscle cell types (76).
Lessons from Subcellular Distribution, the Nuclear Story-Informative experiments to determine functions of DGK isotypes have been aimed at dissecting the subcellular distribution of the isozymes. This approach has been very successful with DGK. We observed that this isozyme is found in the nucleus (61). An emerging body of evidence suggests that nuclear lipid signaling partly regulates the cell cycle, and several groups have shown that there is a nuclear phosphatidylinositol cycle that is regulated separately from its plasma membrane/cytosolic counterpart (77). DAG has been found in nuclear preparations, and the level fluctuates during the cell cycle (78), suggesting that it has a regulatory role there. Because DAG is the major physiologic activator of several PKCs, which translocate to the nucleus following certain extracellular signals, one way to regulate nuclear PKC activity would be to control the extent and duration of the nuclear diacylglycerol accumulation. Nuclear DGKs are poised to accomplish this. To clarify the nuclear role of DGK, we identified its nuclear localization signal. Interestingly, the MARCKS homology domain was the predominant nuclear localization signal. We had previously observed that this domain was likely phosphorylated by protein kinase C isoforms, so we asked whether this modification regulated the nuclear localization of DGK. Indeed, phosphorylation of the MARCKS domain by specific isoforms of PKC reduced nuclear accumulation of DGK. We also found that the consequence of overexpressing DGK was to increase cellular doubling time, presumably because cells accumulate in the G 0 /G 1 phase of the cell cycle. Taken together, our observations suggest that nuclear DAG is necessary for progression of cells to S phase (possibly by activating nuclear PKCs) and that nuclear DGK regulates this lipid and, consequently, performs a role in cell cycle control. Thus, PKC isoforms, in an elegant and surprising way, control their activation state within the nucleus by phosphorylating the MARCKS domain of DGK, which alters nuclear DAG accumulation. We also found that the other type IV DGK, , was similarly regulated (71).
Flores and colleagues (60) have studied the nuclear role of a different diacylglycerol kinase, ␣. They noted in T lymphocytes (where DGK␣ is the major isozyme) that this enzyme translocated to the perinuclear space when the cells were treated with IL-2. Using nonselective inhibitors of DGKs, they presented evidence that the phosphatidic acid produced by this isozyme was necessary for progression to S phase of the cell cycle. Although these data seem to contradict our findings with DGK, the Flores study did not specifically test whether DGK␣ altered nuclear lipids; rather, their assays for PA generation may have also reflected changes occurring at the plasma membrane. Thus, the consequence of the nuclear translocation of DGK␣ is still unclear. In conclusion, although several issues need to be resolved, it appears that some DGK isoforms play a prominent role in nuclear signaling (in the case of DGK, to regulate DAG).
DGKs and the Cytoskeleton-In addition to attenuating cellular DAG, DGKs may also function by increasing the amount of phosphatidic acid. Although phospholipase D is thought to be the major source of PA, these two enzymes generate PA with distinct fatty acid compositions that may have differential effects on target enzymes (79). This suggests that DGKs could generate a distinct phosphatidic acid pool under specific conditions. In accordance with this, Tolias and co-workers (80) identified DGK activity in the immunoprecipitates of Rac, Rho, and Cdc42. They also noted PI-5-K activity and immunoreactive Rho-GDI in the immunoprecipitates. The DGK activity in the immunoprecipitates is novel and deserves further attention. If true, this observation indicates a marriage of a group of molecules involved in actin cytoskeletal rearrangement with an enzyme that can control local availability of lipid mediators that can regulate their activities, an elegant arrangement. For example, PA stimulates PI-5-K activity 8 -12-fold in vitro (81), n-chimaerin (a Rac-GAP) is potentially regulated by DAG levels because of its C1 domain (82), and Rho-GDI dissociates from Rac in the presence of PIP 2 or PA (83). These data strongly indicate a potential role for DGKs in regulating actin cytoskeleton rearrangements.

Perspectives and Future Directions
The structural diversity and complex pattern of tissue expression of DGKs is reminiscent of the PKC and PLC families; is it a coincidence that these three families involved in lipid signaling exhibit this diversity, or is there an underlying connection that has yet to be identified? It is likely that mammalian DGKs will be shown to be crucial in several cellular processes. In some cases their actions may be the result of lowering the level of DAG and thereby serving as an off signal for the allosteric activation of PKC, whereas at other times the DGKs may serve as a source of phosphatidic acid that then functions as a signal. It would not be surprising if DGKs, in some contexts, served both of these functions simultaneously. That is, as an off signal by metabolizing DAG and then as an on signal via the resulting phosphatidic acid.