Covalent Aurora A regulation by the metabolic integrator coenzyme A

Abstract

Aurora A kinase is a master mitotic regulator whose functions are controlled by several regulatory interactions and post-translational modifications. It is frequently dysregulated in cancer, making Aurora A inhibition a very attractive antitumor target. However, recently uncovered links between Aurora A, cellular metabolism and redox regulation are not well understood. In this study, we report a novel mechanism of Aurora A regulation in the cellular response to oxidative stress through CoAlation. A combination of biochemical, biophysical, crystallographic and cell biology approaches revealed a new and, to our knowledge, unique mode of Aurora A inhibition by CoA, involving selective binding of the ADP moiety of CoA to the ATP binding pocket and covalent modification of Cys290 in the activation loop by the thiol group of the pantetheine tail. We provide evidence that covalent CoA modification (CoAlation) of Aurora A is specific, and that it can be induced by oxidative stress in human cells. Oxidising agents, such as diamide, hydrogen peroxide and menadione were found to induce Thr 288 phosphorylation and DTT-dependent dimerization of Aurora A. Moreover, microinjection of CoA into fertilized mouse embryos disrupts bipolar spindle formation and the alignment of chromosomes, consistent with Aurora A inhibition. Altogether, our data reveal CoA as a new, rather selective, inhibitor of Aurora A, which locks this kinase in an inactive state via a "dual anchor" mechanism of inhibition that might also operate in cellular response to oxidative stress. Finally and most importantly, we believe that these novel findings provide a new rationale for developing effective and irreversible inhibitors of Aurora A, and perhaps other protein kinases containing appropriately conserved Cys residues.

Fig. 1

Coenzyme A binds directly to Aurora A and inhibits catalytic activity.(A) Kinase profiling screen reveals selective inhibition of Aurora A by CoA. The effect of CoA, dpCoA and ADP on the activity of 117 kinases was assayed using a radioactive filter-binding assay at the International Centre for Kinase Profiling, Dundee University. Each compound was tested at 100 μM final concentration in the presence of the indicated concentrations of ATP for each kinase (Table 1). Schematic structures of the tested compounds are shown above. (B) CoA inhibits Aurora A in vitro. Kinase activity of recombinant full-length His-Aurora A was assayed by measuring incorporation of γ³³P-ATP into myelin basic protein in the presence of 5 μM ATP and an 11-point serial dilution of CoA. (C) Analysis of binding kinetics of CoA, dpCoA, ADP and ATP towards active (pT288-phosphorylated) Aurora A using a Lanthascreen Eu Kinase Binding assay (D) CoA preferentially binds to the active, pT288 phosphorylated form of Aurora A, when compared to the catalytically inactive dephosphorylated kinase. Dephosphorylation of His-Aurora A was carried out in the presence of recombinant PP1 phosphatase. (E) Intact mass analysis of phosphorylated Aurora A incubated with CoA in the absence of DTT. Covalent incubation of CoA with Aurora A generates a population of phosphorylated Aurora A (pAurora A) alongside covalent adducts containing an extra mass attributable to CoA (pAurora A + CoA). (F) Ion Mobility spectra and calculated cross sectional area (^TWCCS_N2→He) for [M+14H]¹⁴⁺ and [M+15H]¹⁵⁺ ions of pT288 Aurora A measured in the absence (black line) or presence of CoA (red lines) or dephospho-CoA (blue lines). Overlapping conformations of Aurora A are shown, the more extended of which have an increased mean cross sectional area associated with the presence of CoA and dephospho-CoA. (G) Crystal structure of Aurora A (teal) in complex with CoA (pink). (H) Upper panel, magnified view of CoA and its interactions with Aurora A, highlighting the side chains of Thr 217 and Cys 290, which forms a covalent bond with CoA. Lower panel, superposition of Aurora A/CoA with Aurora A/ADP (PDB code 1OL7), showing the shifted position of the Gly-rich loop (black arrow) and the equivalent position of Cys 290 (red arrow). Note that the side chain of Cys 290 was modelled without the sulfur atom in the Aurora A/ADP structure, due to weak electron density. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 2

CoA reversibly binds to Aurora A in an ATP-competitive manner through Cys 290. (A) Electron density around the pantotheine tail region of CoA and Cys 290 of Aurora A, 2Fo-Fc map contoured at 1 σ. (B) Aurora A is covalently modified by CoA in a DTT-sensitive manner. In vitro CoAlated Aurora A was separated by SDS-PAGE in the presence or absence of DTT and immunoblotted with anti-CoA antibody. The membrane was stained with Ponceau red to visualize total Aurora A. (C) Binding of CoA to the ATP-binding pocket is required to facilitate covalent modification of Aurora A by CoA. In vitro CoAlation of Aurora A was carried out in the presence of 100 μM CoA and the indicated concentration of ATP. Generated samples were separated by SDS-PAGE in the presence or absence of DTT and immunoblotted with anti-CoA and anti-Aurora A antibodies. (D) WT Aurora A is efficiently CoAlated by CoA in vitro. (E) The C290A Aurora A mutant exhibits significantly reduced binding of CoA when compared to WT Aurora A. (F) The C290/393A Aurora A mutant is not covalently modified by CoA in vitro. In vitro CoAlation of WT Aurora A, C290A and C290/393A mutants was performed with the indicated concentration of CoA. The reaction mixtures were separated by SDS-PAGE and immunoblotted with anti-CoA and Aurora A antibodies. (G) The inhibitory effect of CoA towards Aurora A is reduced by DTT. Recombinant His-Aurora A was used to determine the IC₅₀ value for CoA in the presence or absence of 1 mM DTT. (H) Reduced inhibitory effect of desulfo-CoA compared to CoA. Desulfo-CoA lacks the reactive SH group at the end of the pantetheine tail. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3

Specificity of Aurora A interaction with CoA is controlled by Thr 217.(A) Amino acid conservation in vertebrate Aurora kinases. Thr 217 defines Aurora A, and is changed to a Glu residue in Aurora B and C. Cys 290 (boxed) is invariant in all Aurora kinases, and lies in the activation segment adjacent to the phosphorylated Thr 288 (human Aurora A numbering). (B)In vitro CoAlation of Aurora A. (C)In vitro CoAlation of Aurora A is abolished in the Thr217Glu mutant. (D)In vitro CoAlation is not detected with Aurora B. (E) The Glu161Thr Aurora B mutant is efficiently CoAlated. Experiments were performed with the indicated concentration of CoA. Reaction mixtures were separated by SDS-PAGE and immunoblotted with anti-CoA, Aurora A or Aurora B antibodies. (F) A thermal shift assay was employed to evaluate Aurora A binding to 5 mM ATP, 5 mM CoA or 0.1 mM MLN8237, in the presence and absence of 1 mM DTT, as indicated. Assays were performed with T 288 phosphorylated, active, Aurora A (open bars), kinase-dead, dephosphorylated Aurora A (Asp274Asn, green bars), Cys290Ala Aurora A (red bars) or Thr217Glu Aurora A (blue bars). Mean ΔT_m values ± SD (n = 3) were calculated by subtracting the control T_m value (buffer, no addition) from the measured T_m value. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4

Inhibitory effect of TPX2 binding on CoA interaction with Aurora A and position of pT288.(A) Magnified views of the activation loop of Thr 288-phosphorylated Aurora A in the crystal structures of complexes with: ADP/TPX2, PDB code 1OL5; CoA (center); or ADP, PDB code 1OL7 (right). Note that in the ADP complex, the sulfur on the side chain of Cys 290 is not included in the experimental model but is shown here, and the β3-αC loop removed for clarity. (B) Aurora A is resistant to inhibition by CoA in the presence of TPX2 1–43 peptide. Kinase activity of recombinant His-Aurora A was assayed by measuring incorporation of γ³³P-ATP into myelin basic protein in the presence or absence of N-terminal fragment of TPX2 (residues 1–43) and serial dilutions of CoA (C) Pre-incubation of Aurora A with CoA inhibits TPX2-induced kinase activation in a DTT-sensitive manner.Recombinant His-Aurora A was preincubated with 100 μM CoA before the in vitro kinase assay which was performed in the presence or absence of the N-terminal fragment of TPX2 (residues 1–43) and DTT.

Fig. 5

Oxidative stress induces Aurora A CoAlation in human cells.(A) Aurora A CoAlation is induced in cellular response to H₂O₂. FLAG-tagged WT Aurora A and WT Aurora B were transiently overexpressed in HEK293/Pank1β. Transfected cells were treated for 30 min with 0.25 mM  H₂O₂. Overexpressed proteins were immunoprecipitated with an anti-FLAG antibody and the immune complexes immunoblotted with anti-CoA and anti-FLAG antibodies. (B) Oxidising agents promote Aurora A CoAlation in vivo. FLAG-tagged WT Aurora A was transiently over expressed in HEK293/Pank1β. Transfected cells were treated for 30 min with a panel of oxidising agents (250 μM H₂O₂, 500 μM diamide, 50 μM menadione, 10 μM phenylarsine oxide and 1 μM rotenone). Transiently expressed proteins were immunoprecipitated with an anti-FLAG antibody, separated by SDS-PAGE under non-reducing conditions and immunoblotted with anti-CoA antibodies. (C) Phosphorylation at Thr 288 and dimerization of Aurora A are induced in cells by oxidising agents. FLAG-tagged WT Aurora A was transiently overexpressed in HEK293/Pank1β. Transfected cells were treated for 30 min with a panel of oxidising agents (250 μM H₂O₂, 500 μM diamide, 50 μM menadione, 10 μM phenylarsine oxide and 1 μM rotenone). Transiently expressed proteins were immunoprecipitated with an anti-FLAG antibody, separated by SDS-PAGE under reducing and non-reducing conditions and immunoblotted with anti-FLAG and anti-pT288 Aurora A antibodies. (D) Schematic illustration showing the key features of the ‘dual anchor’ mechanism for interaction of CoA with Thr 217 and Cys 290 in Aurora A.

Fig. 6

Microinjection of CoA causes abnormal spindles and chromosome misalignment in mouse oocytes. (A) Mouse oocytes arrested at the GV stage were injected with CoA or related compounds, representing different determinants of the CoA molecule. 17 hrs after GV release, cells were fixed and stained for alpha-tubulin (green) or DNA (blue). Representative examples of control and CoA-injected oocytes are shown. (B) The numbers of mouse oocytes with normal and abnormal spindles were recorded. Data shown are from n = 2 experiments. As a control, ~90% of control water-injected oocytes were in possession of a normal metaphase II bipolar spindle with aligned chromosomes. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)