DGK catalyzes the reaction between DAG and ATP to produce PA and ADP. The C1 domains of protein kinase C bind DAG and phorbol esters directly via interactions mediated by select residues (Colón-González and Kazanietz, 2006). Amino acid sequence analysis revealed that all mammalian DGKs harbor at least two segments homologous to the prototypical C1 domain (Hurley et al., 1997). Multiple teams independently demonstrated that for most DGKs, these domains do not bind DAG or phorbol esters (Ahmed et al., 1991; Sakane et al., 1996); [DGKγ (Shindo et al., 2001, 2003) and DGKβ (Shindo et al., 2003) are notable exceptions]. Interestingly, some truncated DGKs devoid of all C1 domains have preserved catalytic activity (Sakane et al., 1996). Data suggest that DGKε may be unique among these DGKs: truncation resulted in complete abrogation of 1-stearoyl-2-arachidonoyl glycerol (SAG) phosphorylation (Tang et al., 1996). The biological function of these “atypical C1 domains” (Hurley et al., 1997) remains elusive to this day.

Except for DGKε, all other isoforms of DGK phosphorylate DAG at rates that are largely independent of the nature of the acyl chains of DAG (Topham and Prescott, 1999). In contrast, the reaction catalyzed by DGKε is very sensitive to the acyl chains displayed by DAG: its peak activity is when the DAG substrate is SAG (Tang et al., 1996; Pettitt and Wakelam, 1999). The DAG molecule harboring these specific acyl chains is a critical lipid intermediate of the PI-cycle (Figure 2). In addition to DGKε, another enzyme that has specificity for lipid substrates with 1-stearoyl-2-arachidonoyl is CDP-diacylglycerol synthase 2 (CDS2) (D'Souza et al., 2014). The possible role of these two enzymes in enriching the lipid intermediates of the PI-cycle is discussed below. In addition, acyl chain remodeling of phosphatidylinositol (PI) through the Land's cycle, also contributes to acyl chain enrichment (Gijón et al., 2008).

We have shown that the most hydrophobic segment of DGKε, which is predicted to bind lipids, can be deleted without loss of enzymatic activity or specificity (Dicu et al., 2007; Lung et al., 2009). When searching for an alternative lipid binding site, we decided to explore a segment of DGKε's accessory domain that is homologous to the arachidonic acid binding site of lipoxygenase (Neau et al., 2009). The finding of such a homologous substrate binding site between lipoxygenase and DGKε was unexpected because the reactions catalyzed by these two enzymes are different, so is the chemical nature of the arachidonoyl groups of the substrates. Remarkably, no other mammalian DGK isoform harbors this novel motif, which is referred to as the “lipoxygenase (LOX)-like motif” (Shulga et al., 2011b).

This motif is characterized by a string of residues: L-X₍₃₋₄₎-R-X₍₂₎-L-X₍₄₎-G, in which X_(n) is n residues of any amino acid. The critical residues are invariant through vertebrate evolution for both DGKε and lipoxygenase (Shulga et al., 2011b). They were identified on the basis that site-specific mutation results in the loss of enzymatic activity or arachidonoyl specificity. The most important reduction of DGKε activity was observed when leucine residues were substituted with more polar and/or sterically smaller amino acids; the impact of substitutions with less polar and/or sterically larger residues was not as dramatic. Most interestingly, substitution of a single amino acid converted DGKα to a LOX-motif-containing DGK that exhibited more specificity for arachidonoyl-containing DAG than unmodified DGKα (Shulga et al., 2011b).

Most mutations of key residues within DGKε's LOX motif resulted in a marked loss in catalytic activity. As a result, accurate assessment of the role of the LOX motif in directing the specific activity of DGKε against arachidonoyl-containing DAG species was challenging. We therefore proceeded to test the impact of mutating residues adjacent to the LOX motif on DGKε substrate specificity (D'Souza and Epand, 2012). This region, adjacent to the LOX motif is a hydrophobic segment contained within the “accessory domain” of DGKε. Unexpectedly, mutagenesis of several residues in this region of DGKε, which is also highly conserved in evolution, had a higher activity toward SAG when compared to wild-type. SAG was the best substrate for all mutants tested, followed by 1,2-diarachidonoyl glycerol. However, each mutant exhibited differences in the relative activity for different DAG substrates. For example, for the wild type enzyme the ratio of activity against 1-stearoyl-2-linoleoyl glycerol vs. SAG is 0.11. However, this ratio ranges from 0.03 to 0.22 for the 5 mutants tested. We conclude that these mutations perturb the lipid binding site, resulting in either enhanced or reduced substrate specificity (D'Souza and Epand, 2012).

There is currently no crystal structure available for any mammalian DGK isoform. Co-crystallization of DGKε with its substrate would be particularly informative. Altogether, these data would be invaluable to drive the discovery of isozyme-specific inhibitors, of which there is only one, for a different DGK isozyme (Liu et al., 2016). However, to do so requires a robust method for expressing and purifying large quantities of DGKε. In our hands, determining the optimal expression system has proven challenging. While bacteria express high levels of recombinant human DGKε, the enzyme recovered is not useful since it is inactive. Human DGKα was successfully expressed in yeast (Abe et al., 2003), but our attempts to express DGKε in the yeast Pichia were not successful.

We recently showed that insect (Sf21) cells are excellent bioreactors to produce high amounts of active recombinant DGKε (Prodeus et al., 2013). We recently succeeded to purify full-length and truncated (Δ40) human DGKε to near homogeneity using this cell system coupled to Nickel-affinity chromatography (Jennings, 2016). In vitro testing confirmed that both forms retained DGKε's substrate acyl chain specificity.

As mentioned in the Section Integral vs. Peripheral Membrane Protein, DGKε contains a putative membrane-spanning alpha helix at its N-terminus (Decaffmeyer et al., 2008). DGKε proteins generated using our protocol were instrumental in allowing us to test various hypotheses about the role of DGKε's amino terminus in binding to membrane lipids and its relevance to the overall activity and stability of the enzyme (Jennings, 2016).

Circular dichroism analysis of purified DGKε and DGKεΔ40 in solution shows that truncating the N-terminal α-helix does not impact the secondary structure of DGKε. Both forms of recombinant DGKε were noted to be highly unstable, losing enzymatic activity and secondary structure in a period of hours after purification. Experiments aimed at monitoring temperature-dependent loss of secondary structure indicates that both constructs undergo a biphasic transition from folded to unfolded states, with transitions occurring at ~56°C and ~77°C in a buffer containing 20% glycerol. We demonstrated that adding a high percentage of glycerol to the recombinant DGKε solutions had dramatic stabilizing effects (Jennings, 2016). We also showed that glycerol concentrations higher than 20% were necessary to facilitate the partial refolding of DGKε and DGKεΔ40 after thermal denaturation.

Activity measurements of purified DGKε and DGKεΔ40 reveal that both constructs retain their acyl chain specificity for SAG (Jennings, 2016). These studies of activity reveal dramatic losses in activity following purification at room temperature, 4°C storage, −80°C storage, and particularly during cycles of freezing/thawing. The incorporation of glycerol into the purification of DGKε as well as during storage dramatically reduces but does not eliminate the observed losses in activity (Jennings, 2016). The absence of the N-terminal hydrophobic segment does not compromise specific activity in a detergent-phospholipid mixed micelle system and suggests that there are additional regions of DGKε that play critical roles in associating the protein to membranes/micelles. The advancements made in the purification and stabilization of DGKε and DGKεΔ40 will facilitate novel studies utilizing more biologically relevant liposome systems.

In contrast to measuring activity in detergent-phospholipid mixed micelle systems, liposome systems provide insight into how bilayer properties and specific lipid species affect enzyme function. A Ca²⁺-independent, water soluble DGK has been studied using liposomes (Thomas and Glomset, 1999a,b). However, DGKα and DGKζ are the only specific mammalian isoforms to be studied in a liposome-based system (Fanani et al., 2004). These enzymes were not purified; instead, they were recovered by salt extraction of cell pellets from mammalian cells overexpressing the particular DGK isoform (Fanani et al., 2004). Regardless, the cruder preparation still provided valuable information regarding the critical role of lipids in altering the activity and specificity of DGKα (Fanani et al., 2004). The successful purification of DGKε is facilitating similar studies in liposomes and will lead to novel findings regarding the activation/inhibition of this enzyme. Furthermore, the purification of DGKε is leading the way to more detailed studies of structure and protein interactions. It is also aiding the screening process for the discovery of a DGKε-specific inhibitor.

Whole-exome sequencing recently uncovered an unexpected link between homozygous mutations in the gene encoding for DGKε and DGKε-associated nephropathy (Lemaire et al., 2013; Ozaltin et al., 2013). The bulk of mutations are expected to result in complete loss-of-function (nonsense, splice site, frameshift); no clustering to any particular domain was observed (Figure 7). This rare condition is due to recurrent episodes of thrombosis in the kidney glomeruli microvasculature. The salient clinical findings are acute renal failure, low platelets, and hemolytic anemia. On that basis, it is hypothesized that DGKε protein must play an important role in regulating thrombosis in the human kidney. This discovery forced experts to reconsider the pathophysiologic underpinnings of aHUS (Quaggin, 2013). Of note, a small group of patients with pathogenic DGKE mutations present with clinical features that resemble more that of another glomerular disease, membranoproliferative glomerulonephritis (Ozaltin et al., 2013). Up until then, abnormal activation of the alternative complement pathway was thought to be invariably associated with nearly all forms of aHUS (Noris et al., 2012): the vast majority of aHUS patients harboring DGKE mutations exhibit no evidence of complement activation (Lemaire et al., 2013; Ozaltin et al., 2013). Only supportive measures may be offered to these patients because there are no targeted therapies.

DGKε was originally cloned from human umbilical vein endothelial cells, but mRNA expression was predominant in testes (Tang et al., 1996). The discovery of the link between DGKε and aHUS recently prompted more in-depth investigations focused on its presumed role in kidney biology. DGKε protein was shown to be expressed in three cell types that play major roles in kidney glomeruli, namely endothelial cells, podocytes, and platelets (Lemaire et al., 2013; Ozaltin et al., 2013). However, the lack of disease recurrence in patients after kidney transplantation strongly suggests that platelets, which are produced by the bone marrow, are unlikely to be central players in the disease process (Lemaire et al., 2013). The mechanism by which DGKε deficiency causes thrombosis exclusively in the kidney remains unclear. Its expression in other vascular beds has not been investigated. Quantification of key members of the PI-cycle needs to be assessed specifically in endothelial cells to determine if DGKε deficiency in this setting also leads to paucity of PIP₂ (see Section DGKε Has a Role in the PI-Cycle). Experiments done in cultured endothelial cells show that siRNA knockdown of DGKε was associated with several phenotypes: endothelial cell increased activation, increase apoptosis and decreased proliferation (Bruneau et al., 2015). It will be important to determine if DGKε-null endothelial cells display the same characteristics because the siRNA knockdown and knockout of the same gene may yield very different phenotypes (Rossi et al., 2015).

The generation of an animal model is often very useful to delineate the biology of human diseases. Since most patients are expected to have DGKε deficiency, the Dgkε-null mouse reported in 2001 would be an ideal candidate model. The original report showed that there were no major abnormalities with these animals (see Section Post-translational Modifications of DGKε Proteins; Rodriguez de Turco et al., 2001). The renal phenotype of this mouse model was recently re-evaluated in more detail: the animals developed mild signs of renal disease with age (Zhu et al., 2016). Interestingly, glomerular lesions were noted in Dgkε-null mice after exposure to doses of nephrotoxic serum that did not affect wild type littermates (Zhu et al., 2016). Mouse models of aHUS often require exposure to exogenous triggers to reveal their pathogenic potential (Pickering et al., 2006; Thurman et al., 2012; Vernon et al., 2016). Importantly, exogenous factors are also known to act as triggers in many patients with genetic forms of aHUS (Kavanagh et al., 2013). Dgkε-null mice may thus be a promising research tool to further our understanding of DGKε-associated nephropathy (assuming it is a full knockout—see Section Post-translational Modifications of DGKε Proteins).

Data from the recent report focused on the Dgkε-null mice suggests that DGKε deficiency leads to systemic inability to induce cyclooxygenase-2 (Cox-2) (Zhu et al., 2016). This enzyme, which is responsible for inducible production of prostanglandins, was previously shown to be decreased in the brains of Dgkε-null mice after kindling (Lukiw et al., 2005). When exposed to lipopolysaccharides, wild-type macrophages increase Cox-2 mRNA and protein levels; similar inductions were observed when wild-type fibroblasts were incubated with interleukin-β (Zhu et al., 2016). Interestingly, these responses were abrogated in macrophages and fibroblasts derived from Dgkε-null mice. Another line of evidence comes from experiments with wild-type mice treated with the nephrotoxin puromycin aminonucleoside: animals invariably develop glomerular lesions that lead to proteinuria. This phenomenon was shown to be accompanied by robust inductions of Cox-2 expression and urinary prostaglandin excretion (Zhu et al., 2016). Application of the same protocol to Dgkε-null mice revealed relative protection against the proteinuric effects of this toxin, and this effect was correlated with blunted Cox-2 and prostaglandin responses (Zhu et al., 2016). How DGKε activity directly modulates Cox-2 expression has yet to be established. The relevance of these findings to patients with mutations in DGKε is unclear.

The renal phenotypes induced by the subclinical doses of nephrotoxic serum or puromycin are now the main distinctive features of the Dgkε-null mouse model when compared to control littermates. It is unclear if other organs are affected because the bulk of these new investigations were focused on the kidney. While the other Dgk-null animals have no obvious renal phenotypes at baseline [Dgkα (Olenchock et al., 2006), Dgkβ (Shirai et al., 2010), Dgkδ (Crotty et al., 2006), Dgkζ (Zhong et al., 2003), Dgkι (Regier et al., 2005), Dgkη (Isozaki et al., 2016), Dgkθ (Goldschmidt et al., 2016)] it is important to realize that none of these models were exposed to these nephrotoxins. It is therefore not possible to conclude that the renal lesions observed in treated Dgkε–null mice are specific to this animal model.

DGKε is one of the main DGK isoforms expressed in cardiac ventricles; others are DGKα and DGKζ (Takeda et al., 2001). This suggests that DGKε may play an important role in the heart. Decreased DgkE mRNA levels were observed in the hearts of rats used for modeling left ventricular hypertrophy (Yahagi et al., 2005) and myocardial infarction (Takeda et al., 2001). These results did not tease out if this reduction was a normal compensatory mechanism, or if it was an integral part of the disease processes. To start addressing this question, transgenic mice overexpressing DGKε only in the heart were generated (Niizeki et al., 2008). When subjected to two distinct protocols known to induce left ventricular hypertrophy in wild type mice, DGKε-overexpressing mice appeared to be protected (Niizeki et al., 2008). These findings translated into a substantial survival advantage: 4 weeks after the procedure, nearly 80% of DGKε-overexpressing mice were still alive, almost double that of wild type controls (Niizeki et al., 2008). Investigations of well-established biological markers of cardiac pathology corroborate these findings. Upregulation of transient receptor potential channel-6 (Kuwahara et al., 2006) and increased PKCε and PKCα translocation (Hahn et al., 2003; Song et al., 2015) were only observed in wild type mice (Niizeki et al., 2008). Taken together, these results suggest that increasing DGKε function may be a promising target to help prevent heart failure and restore cardiac function. The first step in that direction will be to determine how relevant these data are to patients with cardiac dysfunction. If substantiated, the path to DGKε-based therapy will not be straightforward since it would require tissue-specific overexpression of DGKε.

High expression of DGKε in brain tissue suggests that it may play an important role in this organ system (other DGKs are also high in the brain, including β, γ, and ζ) (Zhang et al., 2012). DgkE^−/− mice exhibit higher resistance to electroconvulsive shock (ECS) when compared to control littermates (Rodriguez de Turco et al., 2001). A role for DGKε in this process was supported by phosphoinositide quantifications: while wild type mice displayed increased ECS-induced polyphosphoinositide (PIPn) degradation and 20:4 DAG formation, DgkE^−/− mice did not (Rodriguez de Turco et al., 2001).

In a similar study, DgkE^−/− mice displayed fewer motor seizures, fewer epileptic events, and rapid behavioral recovery following brain stimulation compared to wild type mice (Musto and Bazan, 2006). In addition, wild type mice serially exposed to multiple small seizure events (kindling) developed typical brain morphological changes associated with seizures, but these were absent in DgkE^−/− mice (Musto and Bazan, 2006). Kindling induced upregulation of cyclooxygenase-2 (COX-2) and tyrosine hydroxylase (TH) gene expression in wild type mice but not in DgkE^−/− mice (Lukiw et al., 2005). High COX-2 expression has been associated with recurrence of hippocampal seizures (Takemiya et al., 2003), and repeated seizures lead to increased TH levels in the brain (Ryu et al., 2000). These data thus suggest that DGKε regulates seizure susceptibility via modulation of COX-2 and TH levels in the brain. It is important to acknowledge that the relevance of these data to humans is unclear as there are no data linking aberrant DGKε function to patients with neurological conditions.

DGKε has been identified as a promising target for treating Huntington's Disease (HD) (Zhang et al., 2012). This condition is characterized by polyglutamine expansion in the N-terminus of the Huntingtin protein (Htt) that causes significant neuronal loss in the striatum and cortex (MacDonald et al., 1993; Zhang et al., 2012). DGK inhibitor II (R59022) was identified as a promising anti-HD compound in a kinase inhibitor library screen. This in vitro assay was performed on mouse HD striatal cell model and the readout was the level of mutant Htt cellular toxicity (Zhang et al., 2012). More specifically, R59022 inhibited the expected increase in caspase 3 and 7 activity triggered by serum withdrawal (Zhang et al., 2012). Cells expressing mutant Htt were also found to have lower levels of PIP_n, that can be restored upon decreasing DGK activity (Zhang et al., 2012).

Pinpointing which of the 10 mammalian DGK isoforms was involved in this process required further testing since this compound is a non-specific DGK inhibitor (Sato et al., 2013). siRNA knockdown experiments carried out against the four DGK isoforms expressed in the mouse striatum showed that only DgkE siRNA caused a decrease in caspase 3 and 7 similar to that observed with R59022 (Zhang et al., 2012). Data from two well-established in vivo models support the hypothesis suggesting that in vivo, enhanced DGKε activity plays a role in HD pathogenesis. First, expression of DgkE shRNA in a Drosophila HD model partially rescued the motor impairment induced by Htt (Zhang et al., 2012). Second, DgkE mRNA levels were higher in the striatum of a mouse model of HD (Htt overexpression) (Zhang et al., 2012).

Up to this time, no isoform of a mammalian DGK has been crystallized and its structure determined. However, the crystal structure of a bacterial dgkB from Staphylococcus aureus, has been determined (Miller et al., 2008). The amino acid sequence shows 18% identity, with the active site being particularly well conserved (Jennings et al., 2015). A tentative model for the folding of DGKε has been made using the crystallographic structure of dgkB (Jennings et al., 2015).

Binding studies of DGKε with liposomes of defined lipid composition have not yet been carried out because the enzyme was not available in pure form but only as membrane pellets from over-expression systems. Now that we have purified DGKε in solution we can perform such experiments. Dr. Prasanta Hota has done some initial studies in Dr. Epand's laboratory using PIP strips (Echelon Biosciences Inc., Salt Lake City, Utah). It was found that the binding to PI, phosphatidylcholine and lysophosphatidylcholine was very weak, binding to PA, phosphatidylserine, and phosphatidylethanolamine was intermediate and binding to PI with one, two, or three phosphate groups added to the inositol ring was very strong. It is known that phosphorylated forms of phosphatidylinositol are good inhibitors of DGKε (Walsh et al., 1995). Phosphatidylinositol-(4,5)-bisphosphate is a noncompetitive inhibitor with respect to SAG but a competitive inhibitor with respect to ATP (Walsh et al., 1995). This is not purely an electrostatic effect since phosphatidylinositol-trisphosphate is a weaker inhibitor of DGKε than is the diphosphate (Walsh et al., 1995). There is thus likely to be some specificity in the binding of DGKε to membranes containing anionic lipids. The effect of anionic lipids on DGKε binding to membranes has not been fully assessed.

There is no reported data on the activity of DGKε using an assay system with phospholipid bilayers. All the reported enzyme activity studies with DGKε have been done with a detergent-solubilized system. Phospholipids in liposomes are arranged as bilayers, which more closely simulates their arrangement in biological membranes. Even the fundamental property of arachidonoyl substrate specificity has not been tested in a bilayer-based system. It also is not clear how membrane binding relates to enzymatic activity of DGKε. It is clear that the DAG substrate is part of the membrane and that DGKε has to bind to a membrane, at least transiently, for phosphorylation to occur. However, as indicated by our preliminary data, anionic lipids strengthen the binding of DGKε to the membrane. Why then don't anionic lipids promote the activity of DGKε as they do for other isoforms of mammalian DGK?

In order to gain a better understanding of DGKε-related diseases, isoform-specific inhibitors should be identified and used in experimental work to delineate the role of DGKε in various tissues and cell types. Although there are several commercially available DGK inhibitors, such as R59022 (de Chaffoy de Courcelles et al., 1985) and R59949 (Sato et al., 2013), there are currently no isoform-specific DGKε inhibitors.

Traditionally, the activities of DGK isoforms were assessed using a micelle-based assay, which utilizes detergents to solubilize lipid components from a lipid film (Epand and Topham, 2007). The hydrated lipid film is then sonicated to produce small unilamellar vesicles. The vesicles are used in a radioactive assay to evaluate the activity of the enzyme, by measuring the transfer of ³²P from [γ-³²P]-ATP to DAG to form PA (Epand and Topham, 2007). Although this procedure is quite insensitive to non-DGK ATPase activity, the assay might not be well suited for a high-throughput system due to the extensive radioactive and extraction procedures.

Recently, an ATP-luciferase system was developed to evaluate the isoform selectivity of the R59949 and R59022 inhibitors for various DGKs (Sato et al., 2013). The assay was later optimized to study DGKα, and was used to identify a novel DGKα-specific inhibitor, CU-3 (Liu et al., 2016). Similarly, the ATP-luciferase assay can be optimized to study DGKε in a high-throughput format to identify a DGKε-specific inhibitor. Due to the fact that DGKε utilizes ATP as a phosphate donor in the conversion of DAG into PA, the concentration of ADP generated can be used as a measure of PA production and enzyme activity (Shulga et al., 2011c). Since the ATP-luciferase assay detects the ADP released in the kinase reaction, the assay must utilize purified DGKε, in order to reduce contamination with other sources of ATPase activity (Zegzouti et al., 2009).

Once a pool of candidate inhibitors has been identified using the high-throughput method, the candidates can be validated using various assays. For example, the traditional micelle-based radioactive assay or the liposome based assay currently being developed can be used to confirm changes in enzyme activity in the presence of inhibitors (Epand and Topham, 2007). In addition, a ³¹P nuclear magnetic resonance assay (NMR) can be used to identify phosphorous containing substrates and products and detect potential contaminants with ATPase activity (Prodeus et al., 2013). Lastly, the ATP-luciferase assay can be optimized to assess the activity of various other DGK isoforms. Once optimized, the assay would be used to the test the inhibitory potential of candidate compounds against various DGK isoforms, to assess the isoform specificity of novel inhibitors.

The development of a DGKε-specific inhibitor would be invaluable for studying the role of DGKε in various disease processes. Specifically, the isoform-specific inhibitor could be used to recapitulate physiological environments lacking DGKε. In addition, the luciferase assay holds potential as a diagnostic tool for measuring the enzyme activity of DGKε in patients (or with CRISPR-Cas9 mutated cells) with missense DGKε mutations.

The amino terminal hydrophobic segment of DGKε is highly conserved in evolution (Figure 3). In particular, there is an invariant proline residue at position 33 in human DGKε that is present at the same position of the aligned sequences from all DGKε from a range of organisms (Supplementary Materials). We have shown that this proline residue has an important role in determining the position of equilibrium between a transmembrane helix and a re-entrant helix (Decaffmeyer et al., 2008). This segment of the protein has no effect on the activity or specificity of DGKε using an in vitro assay in detergent micelles with N-terminal 40–60 residue deletion mutants. However, we believe that it is unlikely that this segment of the protein would not have a biological function. Forming a re-entrant helix would promote the positive curvature of the monolayer in which it is present. This could influence the region of the membrane to which DGKε partitions and/or modulate its interaction with other proteins. An understanding of the role of this proline residue and possibly of the interconversion between re-entrant and transmembrane helix is a theme for future investigations.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.