Ternary structure reveals mechanism of a membrane diacylglycerol kinase

Abstract

Diacylglycerol kinase catalyses the ATP-dependent conversion of diacylglycerol to phosphatidic acid in the plasma membrane of Escherichia coli. The small size of this integral membrane trimer, which has 121 residues per subunit, means that available protein must be used economically to craft three catalytic and substrate-binding sites centred about the membrane/cytosol interface. How nature has accomplished this extraordinary feat is revealed here in a crystal structure of the kinase captured as a ternary complex with bound lipid substrate and an ATP analogue. Residues, identified as essential for activity by mutagenesis, decorate the active site and are rationalized by the ternary structure. The γ-phosphate of the ATP analogue is positioned for direct transfer to the primary hydroxyl of the lipid whose acyl chain is in the membrane. A catalytic mechanism for this unique enzyme is proposed. The active site architecture shows clear evidence of having arisen by convergent evolution.

Figure 1. Overall structure of the DgkA-ACP-lipid ternary complex.

(a,b) Views of the kinase trimer parallel to the membrane and from the cytosol, respectively, with subunits represented as brown, blue and green ribbons. ACP and lipid are shown in stick representation. The two grey spheres represent zinc. Putative membrane boundaries are shown as black lines. Active site asBC contains both ACP and lipid substrate. MAG1 and MAG2 are shown with carbons coloured black and violet, respectively. (c) Expanded view of zinc-ACP in the binding site. (d) Feature enhanced map (Methods section) of Zn-ACP at 1 σ. (e) Anomalous density maps for zinc contoured at 4.0 σ. (f) Feature enhanced map for lipid MAG1 in asBC contoured at 1 σ.

Figure 2. Proposed mechanism and sensitivity to mutagenesis of the transphosphorylation reaction catalysed by DgkA.

(a) Two-dimensional representation of the active and substrate-binding sites. The catalytic Glu69_C abstracts a proton from the primary hydroxyl of lipid substrate MAG1 (red) creating a reactive alkoxide which attacks the γ-phosphorus of zinc-ATP (blue). The transition intermediate includes a pentavalent γ-phosphorus where the two substrates are covalently bonded together. On collapse, the ADP and lyso-PA products form and are released returning the enzyme to its original state. Residues that interact with ACP and MAG1 in the asBC site of the ternary complex are shown with hydrogen bond and ionic interactions indicated by hashed bonds. Hydrophobic interaction is highlighted as a black semicircle. Interactions that are likely and possible, but that are not supported by structure data, are indicated by dashed bonds. Structurally equivalent residues in cAMP-dependent protein kinase A are shown boxed with residue numbers as in PDB ID: 1ATP. (b) Kinase activity of DgkA as affected by site specific mutations. Kinase activity is expressed as a percentage of WT activity. Actual values are recorded in Supplementary Table 1. The one letter amino acid code is used. Residual activities of 0%, 0–0.5% and 0.5–5% are indicated by #, ** and *, respectively.

Figure 3. Substrate-binding sites of DgkA.

(a) Binding of the adenosine of ACP at the extracellular surface of active site asBC. Interactions are shown as dashed lines with distances between non-hydrogen atoms in Å. Protein and ACP carbons are coloured green and yellow, respectively. Oxygen and nitrogen are coloured red and blue, respectively. (b) Binding of the zinc-triphosphate moiety of zinc-ACP to asBC. Lipid carbons are coloured black. Distances for zinc coordination range from 2.0 to 2.2 Å. The asterisks indicate that little or no electron density was observed for the side chains of Arg9 and Glu28. However, interactions with zinc-ACP, of the type noted, were seen with both residues in MDS (Supplementary Movie 1). (c) The putative active site of DgkA where the catalytic residue Glu69_C, ACP and lipid substrate MAG1 meet in asBC. The glycerol headgroup of MAG1 could not be oriented unambiguously in the active site based on the available electron density maps, even at 2.05 Å resolution (Supplementary Fig. 4a–d). Accordingly, the interactions between MAG1 and the enzyme are shown as dashed lines only and without specifying distances. The orientation shown, with the 1-OH in hydrogen-bonding distance to the carboxyl group of Glu69, is plausible in light of a careful consideration of the available maps and the nature of the lipid substrate and the reaction being catalysed.

Figure 4. Changes in the active site of the DgkA ternary complex at the beginning and end of a 100-ns MDS.

(a) WT simulation of ATP based on the ACP coordinates. There are limited differences in the binding site between the start (grey) and end (colour) of the MDS. (b) E28A mutation (red). This principally affects the binding of the Zn2 ion. In the absence of the E28 side chain the zinc ions become purely coordinated by E76 and the ATP phosphates. As a result, the entire zinc-ATP complex moves away from the protein, towards the cytoplasm. This, in turn, slightly alters the conformation of the CL. (c) E76A (red). To compensate for the loss of E76, the zinc ions move towards the membrane to interact with E69. This pulls the ATP in the same direction and, in turn, the CL is affected. (d) K94A (red). The WT residue coordinates both α-phosphate and N7 of the adenine ring of ATP. The loss of the basic side-chain releases the adenine of ATP and the binding is lost. In WT simulations, K94 forms a salt bridge with D80 and it is expected that the loss of this bridge in the D80A mutant can also explain the loss of catalytic ability in this mutant. (e) GTP. The major difference in dynamics is observed in the CL, where the loss of the N6 hydrogen bond with the backbone of E85 destabilizes purine binding and the CL. (f) ADP. The ADP molecule is expected to leave the binding site after catalysis has taken place. A change in zinc ion coordination takes place as Zn1 now coordinates the α- and β-phosphates of ADP. This relocates ADP closer to K94, which also coordinates both phosphates. In turn, K94 no longer interacts with the N7 position of the purine ring, which changes conformation, priming the ADP molecule for exit. Throughout this legend, WT refers to Δ4-DgkA (ref. 12).

Figure 5. Rationalizing functional roles of highly conserved residues in DgkA not directly involved in catalysis.

(a) The anionic carboxyl of Glu34_C is proposed to elevate the pKa of catalytic residue Glu69_C making it a stronger base for proton abstraction from the lipid (green protein carbons). Electron density (mesh) for alternative Glu34 conformers was observed in the XFEL data recorded at RT (purple carbons) that may enable a switch in pKa at Glu69. (b) Alternative conformations for Glu76_C in the XFEL structure. The first (Glu76_1) is in position to coordinate with Zn2. Black dashed lines correspond to distances 2.0–2.2 Å. The second (Glu76_2) has the coordinating oxygen at some distance (3.8 Å, green dashed line) from Zn2 where it may facilitate product release. The polyphosphate and zinc are superimposed from the ternary complex on the XFEL apo-structure. (c) Side-chain amide nitrogen of Asn72_C coordinates with carboxyls of Glu69_C and Glu76_C, both essential residues. By weakly interacting with γ-phosphate oxygens, the electrophilicity of the γ-phosphorus is elevated making it more reactive. Additionally, the amide may stabilize a transient bisubstrate by being positioned close to where the pentavalent intermediate is likely to form between the 1-OH of MAG1 and the γ-phosphate of ACP (dotted blue line). Likely interactions between MAG1 and the enzyme are indicated by dashed lines only. (d) Alternative conformations for Glu69 observed in the XFEL structure (magenta carbons) superimposed on the DgkA ternary structure (green carbons). The first conformation (Glu69_1) is like that seen in the ternary complex and interacts with MAG1. The second (Glu69_2) extends into the membrane and may guide lipid substrate to the active site. (e) An expanded view of Glu69 in d with electron density corresponding to alternative conformers seen in the XFEL structure superposed. (f) The methyl side chain of Ala30_C (red circle) is proposed to provide room for lipid substrate and product to pass between the start of SH_B and the top of H1_C that define the gateway into and out of the active site. The Cα–Cα distance from Ala30_C (red circle) to Ala13_B (blue circle) in asBC is 10.0 Å.

Figure 6. Similarities in the binding coordination of adenosine by DgkA and cAMP-dependent protein kinase A.

(a) DgkA, this study. (b) Protein kinase A, PDB ID: 1ATP. Only the similarities are highlighted. In a one surface of the adenosine moiety faces the protein, the other faces the cytoplasm. In b the adenosine is sandwiched between the N- and C-terminal lobes of PKA.