Elusive Structure of Mammalian DGKs

Abstract

Mammalian diacylglycerol kinases (DGKs) are a group of enzymes that catalyze the ATP-dependent phosphorylation of diacylglycerol (DAG) to produce phosphatidic acid (PtdOH). In doing so, they modulate the levels of these two important signaling lipids. Currently, ten mammalian DGKs are organized into five classes that vary with respect to domain organization, regulation, and cellular/subcellular distribution (reviewed in ^1–3).

As lipids play critical roles in cells, it is not surprising that there is increasing interest in understanding the mechanism underlying the catalysis and regulation of lipid modulating enzymes such as DGKs. However, there are no solved 3D structures for any of the eukaryotic DGKs. In this review, we summarize what is known and the current challenges in determining the structures of these important enzymes. In addition to gain critical insights into their mechanisms of catalysis and regulation, DGK structures will provide a platform for the design of isoform specific inhibitors.

II. Current Understanding of the Structure of Mammalian DGKs

Purification of DGKs

One of the major challenges in determining protein structure is obtaining sufficient amounts of homogenous purified protein. This had been a bottleneck in resolving DGK structures. Our laboratory has reported the expression of Sus scrofa DGK alpha (DGK-α) catalytic domain in Escherichia coli⁴. We showed this domain could be extracted, refolded, and purified from inclusion bodies. However, when this construct was subjected to analytical gel filtration chromatography, it appeared to be aggregated thus unsuitable for structural studies. Strategies to improve the solubility, such as including an N-terminal tag such as glutathione S-transferase, thioredoxin, or maltose binding protein failed to improve the solubility. Co-expression with bacterial chaperones, while it increased the overall protein yield, the protein was still aggregated upon purification. Similar to these studies, Takahashi and Sakane also reported that prokaryotic expression systems were problematic for these enzymes⁵.

Tu-Sekine et al reported the first successful purification of DGK-θ from HEK293 cells⁶. While this approach proved valuable for some of the initial enzymological studies, it was not possible to obtain sufficient amount of purified protein for structural studies. Fortunately, DGK-ε⁷, DGK α⁵ and DGK-ζ⁸ were purified in fairly high yield from the Epand and Sakane labs respectively. Our laboratory has also optimized DGK-θ expression in eukaryotic expression systems and purified high yields of DGK-θ in both HEK293 suspension cells and Sf9 insect cells for downstream structural studies (manuscript in preparation).

Limited Understanding of Full-Length DGK 3D Structures

While none of full-length mammalian DGK structures have been determined, we have learned quite a bit from the structures of mammalian DGK regulatory domains, as well as the structure of a prokaryotic homolog DgkB that shows relatively high sequence similarity to the catalytic domain of mammalian DGKs. These solved DGK structures are summarized in Figure 1.

Figure 1:

Solved DGK Structures

Crystallography has been used to determine some of the individual domains in different DGK families. For example, the EF-hand domain structure in DGK-α,⁹ a Type I DGK (Figure 1). These domains are involved positively regulating enzyme activity by Ca^{2+ 9, 10}. Crystallography of bacterially expressed EF hand domains (aa 107–197) of DGK-α in the presence of calcium showed that the EF hand motifs coordinate with two calcium ions and adopt a canonical dimeric pair composed of helix-loop-helix motif. Isothermal titration calorimetry showed the calcium ions bind cooperatively to the EF hands. Furthermore, size exclusion chromatography showed calcium binding induces a distinct conformational change. The apo EF hands form a dimer which is disrupted to monomers in the presence of calcium. While the EF hands may not dimerize in the full-length protein, trypsin-limited digestion studies in the presence and absence of calcium support the notion that calcium induces a conformational change in the domains even in the intact enzyme. Takahashi and coworkers suggest that the conformational change likely modulates domain-domain interactions which affect catalytic activity⁹.

Similarly, a crystallographic approach was used to examine the structure of C-terminal sterile alpha motif (SAM) domain oligomers involved in the regulation of DGK-δ1¹¹, a Type II DGK (Figure 1). This SAM-mediated oligomerization leads to oligomers as large as tetramers¹², sequestering the enzyme in the cytosol to inhibit its translocation to the plasma membrane. This study showed the SAM domains oligomerization occurs via helical head to tail interactions. Further, when this oligomerization was prevented by specific mutations in the interacting interface, the enzyme was constitutively localized at the plasma membrane. The authors suggest that such assembly/disassembly may be an unappreciated regulatory mechanism for other DGKs as well as other signaling enzymes.

Another common motif in DGKs are the C1 domains (Figure 1). DGKs have at least two C1 domains which are homologous to the C1A and C1B domains of PKCs¹³. Furthermore, C1 domains bind DAG or phorbol ester in PKC, n-chimaerin, and the unc-13 gene product in C. elegans ^{14, 15}. These domains contain one or two cysteine-rich regions (CRDs) believed to be involved in zinc as well as DAG binding, but this latter point remains unclear. A NMR solution structure of the C1 domain of DGK-δ in complex with two zinc ions has been deposited in the PDB (PDB ID 1R79). Crystallographic data suggest that for PKC, DAGs would insert perfectly into a hydrophilic cleft of a C1 domain making the entire top surface of the domain hydrophobic. While this would stabilize membrane insertion and does not present a problem for allosteric regulation of PKC, such binding in DGK would render the hydroxyl group inaccessible for phosphoryl transfer required in a DGK. It should be noted that Los et al. suggested that the DGK C1 domains of DGK-θ, particularly the extC1 domain, binds DAG only weakly, but they may serve a role in channeling the substrate to the active site¹⁶. Further structural and kinetic studies will help resolve this issue.

Structural understanding of the catalytic domain of DGKs are mainly derived from prokaryotic homologs such as DGKb from S. aureus¹⁷ and YegS from E. coli ¹⁸ (Figure. 1) Despite a lack of catalytic domain structures of any eukaryotic DGKs, recent studies have pinpointed the critical ATP binding site. Using an ATP analogue (an ATP acyl phosphate) and quantitative LC-MS to map ligand binding regions within the active site of mammalian DGKs, Franks et al.¹⁹ identified some unique features of DGKs. Their data indicates that both the presumed catalytic (DAGKc) and accessary (DAGKa) domains interact to form a potential ATP binding cleft. This is similar to the interdomain cleft for ATP-binding found in soluble bacterial lipid kinases that have significant homology to mammalian DGKs¹⁸. Interestingly, the first C1 domain of DGKα was also identified to interact with ATP probe, albeit with lower potency, suggesting a distinct role of C1 in mediating ATP binding.

III. Major Knowledge Gaps

The catalytic and regulatory mechanisms of protein kinases have been well studied (e.g. see ^20–23 for reviews). Our understanding of the catalytic mechanisms and regulation of mammalian DGKs, however, remain largely unresolved. Addressing both knowledge gaps will be greatly enhanced by determining the 3D structure of these enzymes.

Catalytic mechanisms:

The different isoforms likely have similar, if not identical catalytic mechanisms. Given what we know about other kinases, protein kinases in particular, the catalytic mechanism of DGK-θ is likely to involve a dissociative mechanism, similar to that for PI-3-kinase. Any consideration of potential catalytic mechanisms of DGK must, like all lipid-dependent enzymes, consider both three-dimensional bulk interactions occurring in solution and two-dimensional surface interactions occurring at the hydrophilic/hydrophobic interface. Further, whether the reaction involves order or random binding of substrates is also undefined.

Key to understanding the catalytic mechanism of these enzymes is the identification of the binding domains for both substrates: ATP and DAG. As discussed above the catalytic (DAGKc) and accessory (DAGKa) domains are likely involved in ATP binding, the binding site for the DAG substrate is less clear. One hypothesis, as noted above, is that the C1 domains bind and transfer DAG to the catalytic site. Consistent with this, Houssa and van Blitterswijk found deletion of C1 domains in DGK-α, or a construct of DGK-θ that contained only the putative catalytic domain, resulted in inactive enzymes (see ²⁴). Three issues, however, complicate this hypothesis. First, the role of these domains in DGKs differs from those found in PKCs in a manner that suggests they would not support catalytic activity. Second, only DGKs β and γ have been shown to bind phorbol esters^25–28, consistent with the structural data showing only DGKs β and γ contain a C1 that fits a profile for phorbol ester binding²⁸. Third, Sakane's group found that while a DGK-α construct lacking C1 domains had low catalytic activity, in contrast to the van Blitterswijk group, it retains a Km for DAG that is similar to the wildtype enzyme²⁹. For this study, however, the activity was quantified in extracts from Cos-7 cells that were over-expressing the different constructs. It is difficult, therefore, to completely rule out potential contribution from the endogenous, wild type enzyme. Finally, the Drosophila DGK1 does not contain C1 domains³⁰. One intriguing hypothesis regarding these domains, however, is that they may be involved in determining the DAG species-specificity of DGKs³¹.

In addition to identifying specific domains involved in substrate binding, it will be important to determine the conformational changes to understand the structural dynamics involved in catalysis. This will allow us to determine how structural alternations contribute to phosphoryl transfer to fully understand DGK's catalytic mechanism.

Regulatory Mechanisms:

DGK regulation will most certainly involve protein modulators, lipid modulators, ion modulators, post-translational modifications, and membrane architecture. While we cover some of the regulatory mechanisms below, we note there are other general reviews that cover many aspects of this regulation^2,32,33.

Protein Modulators

It is expected that the DGK domains between the N-terminus and catalytic region are critical for the regulation of these enzymes. As noted above, the SAM domain of DGK-δ1 oligomerizes to control the localization and activity of this isoform^{11, 12}. DGK regulation also appears to involve the interaction of accessary proteins. For example, our laboratory showed that DGK-θ is fully active when it associates with polybasic regions in other proteins⁶. Furthermore, active, GTP-bound RhoA, has been shown to inhibit this DGK isoform³⁴. DGK-ζ is also regulated by a protein-protein interaction. This enzyme is activated by the hypo-phosphorylated Rb protein and two pocket proteins p107 and p130³⁵. We should note that in addition to its regulation, DGKs may bind and affect the activity of other proteins. For example., DGK-ζ has also been implicated in binding to β-arrestins that mediated signaling for the M1 muscarinic receptor^{36, 37}. DGK -ζ also binds to and inhibits Ras guanyl-releasing proteins (RasGRPs) 1,2,3 and 4 while DGK-I binds to and inhibits RasGRP3 only³⁸. Further, DGK-ζ C1 domains bind Rac1 to control neuronal polarized outgrowth³⁹. How these intra- and intermolecular domains modulate the catalytic activity, as well as membrane binding and subcellular localization will require an understanding of the structure of the full-length DGK.

Modulation by Ions

While not extensively studied, cations, largely divalent cations, modulate the activity of some DGKs. For example, a well-known feature of the class I DGKs (α, β, γ) is that their activity is increased by calcium, most likely bound to their EF-hand domains. Our laboratory has shown that DGK-θ activity can be affected by the presence of magnesium and zinc⁴⁰. Magnesium effects are not very surprising as this ion usually bridges active site residues to ATP in kinases. It appears, however, that Magnesium increase the activity DGK-θ but the mechanism responsible for this increased activity has not been resolved. It may, however, be also associated with a potential structural role. Consistent with this, the crystal structure of the S. aureus DGK, DGKB, identified a magnesium binding site that is distinct from the ATP-binding site, involving two aspartate residues believed to participate in aspartate-water-magnesium complexes¹⁷. Mutation of either aspartate significantly diminished the activity of this DGK. Clearly, the roles of ions in DGK regulation have not been thoroughly examined. The precise role of ions in the structure and catalytic activity of the individual DGKs will also become clear when the 3D structure of these enzymes is fully resolved.

Lipid Modulators

It has long been recognized that certain lipids can modulate the activity of various diacylglycerol kinases. The involved lipids and the nature of the regulation varies between the various isoforms. For example, sphingosine^{41, 42}, phosphatidylserine (PtdSer)⁴³ as well as and products of phosphatidylinositol-3-kinases activate DGK-α^{44, 45}. While PtdSer, as well as PtdOH, activate the class V DGK (DGK-θ) and the class IV DGKs (DGK-ζ, and DGK-I), these lipids inhibit DGK-ε (see ²). Phosphatidylinositols activate the class IV DGKs and have been shown to bind PH domains in DGK-δ, but this is likely for localization^{46, 47}

An important aspect of lipid modulation of DGKs that deserves increased attention is the effect of various lipid on membrane architecture which appears to affect DGK activities. One of the best examples of this is the effect of membrane architecture on the activity of DGK-ε. Bozelli et al. showed that the activity of this isoform is enhanced in the presence of negative Gaussian curvature⁴⁸. This architecture was also shown to affect the acyl chain specificity of DGK-ε and it was suggested that this may account, at least in part, on the apparent selectivity of this isoform for DAGs with arachidonic acid at the sn-2 position. The specific architecture may not be solely due to the lipid composition of membranes, but it is likely that lipids play a role in assisting or allowing for the formation of the optimal membrane architecture.

Post-Translational Modifications

In addition to protein-protein and protein-lipid interactions, there is evidence that some DGKs are regulated by posttranslational modifications. Several protein kinases phosphorylate DGK-α, including isoforms of PKC and Src^49–51. A comprehensive mapping of all DGK subtype-specific phosphorylation and the role of these modifications on the structure, localization, activity is still missing. There are, however, some limited data regarding the phosphorylation of certain DGK subtypes. Src phosphorylated DGK-α enhances its activity⁴⁹. Similarly, PKCγ phosphorylated and activated DGK-γ⁵². In the absence of calcium, PKC-mediated phosphorylation of serines within the PH domain of DGK-δ1 inhibits translocation⁵³. Similarly, Lck-mediated phosphorylation of tyrosine 335 in DGK-α is important for its membrane association⁵¹. Recently, it was shown that PKCα catalyzes the phosphorylation of the MARCKs domain of DGK-ζ, and this phosphorylation modulates its interaction with two other proteins, Rac1 and RhoA⁵⁴.

Information about glycosylation of different diacylglycerol kinases is lacking. While there are putative glycosylation motifs at the N-terminus of DGK-ε ⁵⁵, it is not clear whether these sites are actually glycosylated in vivo.

IV. Future Directions and Approaches

The structure of individual isolated domains of DGKs are available. Unfortunately, the 3D structure of any eukaryotic full-length DGKs is still unknown. The "resolution revolution” in cryoEM could provide a powerful tool in determining DGK structures to atomic resolution^56–61. Recent development of sample preparation, image processing algorithms and the novel design of protein engineering further push it to outreach the era of smaller sized proteins as well as proteins with continuous flexibility to high-resolution^62–66. In addition, advanced technical development of correlative light and electron microscopy (CLEM)^67–69 and cryo-electron tomography (cryoET)^70–76, which allow the direct visualization of proteins in specific subcellular location, can be used to further address the function of DGKs in their native membrane systems.