Crystal structure of death-associated protein kinase 1 in complex with the dietary compound resveratrol

The dietary compound resveratrol is an inhibitor of death-associated protein kinase 1.


Introduction
Resveratrol (RSV; 3,4 0 ,5-trihydroxy-trans-stilbene; Fig. 1) is a natural polyphenolic phytoalexin found in grapes, peanuts, berry fruits and olives (Wang, Catana et al., 2002). In particular, red wine contains a high concentration of RSV, ranging from 1 to 25 mM depending on the grape varieties used (Gu et al., 1999). RSV has attracted considerable attention owing to its cardioprotective activity, and has been suggested to be responsible for the 'French paradox', i.e. the inverse relationship between the consumption of red wine and cardiovascular disease rates in France (Renaud & de Lorgeril, 1992). RSV has also been shown to exhibit diverse biological effects including anti-inflammatory, antioxidant, neuroprotective and antitumor activities (Ashrafizadeh et al., 2020;Banez et al., 2020;Ahmadi & Ebrahimzadeh, 2020). Because cancer is a leading cause of death worldwide, the role of RSV in tumorigenesis has been extensively studied. The multistage chemical carcinogenesis involves disruption of intracellular signaling networks (Vineis et al., 2010). Several studies have indicated that RSV directly or indirectly modulates signaling pathways, including the mitogen-activated protein kinase (MAPK) signaling pathway, the protein kinase C (PKC) signaling pathway and the phosphoinositide 3-kinase (PI3K)/ protein kinase B (Akt) signaling pathway (Kundu & Surh, 2008). Although several protein kinases are associated with these pathways, only PKC isozymes and proto-oncogene serine/threonine-protein kinase (PIM1) have been shown to be specifically inhibited by RSV, and there is no structural information on the interactions between RSV and protein kinases (Slater et al., 2003;Kim et al., 2020).
Death-associated protein kinase 1 (DAPK1) is a Ca 2+ / calmodulin-regulated serine/threonine protein kinase (CaMK) composed of an N-terminal catalytic kinase domain, a Ca 2+ / calmodulin-binding domain, ankyrin repeats, an Roc/COR domain and a death domain. The multi-domain configuration of DAPK1 contributes to its interaction with various proteins, and thus DAPK1 functions as a central junction in various signaling pathways (Shiloh et al., 2014). DAPK1 is known to be a regulator of apoptosis and autophagy, and to be a tumor suppressor protein (Deiss et al., 1995;Inbal et al., 2002). DAPK1 is also associated with the accumulation of amyloidand tau proteins in the brain in Alzheimer's disease (AD). The expression of hippocampal DAPK1 was significantly increased in the brains of AD patients (Kim et al., 2014). DAPK1 contributes to the pathogenesis of AD by the amyloidogenic processing of amyloid precursor protein and by phosphorylating and stabilizing tau aggregates (Kim et al., 2016;Farag & Roh, 2019;Kim et al., 2014). It has been shown that gene knockout of DAPK1 ameliorated age-dependent neurodegeneration in AD (Yukawa et al., 2006). Increasing evidence indicates that DAPK1 is an important target molecule for the treatment of AD (Chico et al., 2009). The inhibition of DAPK1 might also be an effective therapy for advanced endometrial adenocarcinomas and acute brain injury by hypoxia-ischemia, but the relevance of DAPK1 to those diseases is still under study (Tanaka et al., 2010;Velentza et al., 2003).
Feng and coworkers developed an ATP-competitive DAPK1 inhibitor with an octahedral ruthenium metal center which had an IC 50 value against the phosphorylation activity of DAPK1 of 2.0 nM (Feng et al., 2011). A pyrazolo[3,4-d]pyrimidinone compound and an imidazopyramidazine compound have been reported to be DAPK1 inhibitors with dissociation constants (K d ) of 0.3 and 0.24 mM, respectively (Carlson et al., 2013;Wilbek et al., 2015). We have previously shown that natural flavonoids bind to the ATP-binding site of DAPK1 with various binding orientations (Yokoyama et al., 2015). In addition, very recently we reported that purpurin, a natural anthraquinone found in the roots of Rubia argyi, is a potent DAPK1 inhibitor with a K d value of 0.37 mM (Yokoyama et al., 2020). Although some DAPK1 inhibitors have been identified, as described above, none of them have reached the stage of clinical studies. The discovery and development of DAPK1 inhibitors is an ongoing project.
AD is a progressive neurodegenerative disorder associated with memory impairment, and thus continuous dietary intake of natural compounds that inhibit DAPK1 activity may be a good strategy to prevent the development of AD. RSV is known to be a safe material and is commercially available as a dietary supplement worldwide. Although our previous study using fluorescence spectroscopy suggested that RSV might directly interact with DAPK1 (Yokoyama et al., 2015), there is no information about its inhibitory mechanism or binding modes. The investigation of interactions between DAPK1 and RSV may provide a novel strategy for the treatment of AD. In the present study, we tested the inhibitory activity of RSV against DAPK1. Kinetic analysis and a fluorescent probe competitive binding assay showed that RSV is an ATPcompetitive DAPK1 inhibitor. Together with the results of crystallographic analysis of the DAPK1-RSV complex, we will discuss the binding of RSV to DAPK1.

Protein production and chemicals
The catalytic domains of the DAPK1 (residues 1-285) and ephrin type-A receptor 3 (EphA3; residues 613-905) proteins were prepared using an Escherichia coli system as described previously (Yokoyama et al., 2020). The final protein samples were visualized by SDS-PAGE with Coomassie Brilliant Blue as a single band. In the case of DAPK1, successful crystallization assured the high purity of the protein. RSV, apigenin (API) and kaempferol (KAE) were purchased from Wako Pure Chemical Industries. Piceatannol (PIC) was purchased from Cayman Chemical Company. These chemicals were supplied at greater than 95% purity as determined by HPLC.

Intrinsic ATPase activity measurement
The intrinsic ATPase activities of the kinases were measured by a colorimetric malachite green assay as described previously, with minor modifications (Sherwood et al., 2013;Yokoyama et al., 2019). The protein samples (0.1 mM DAPK1 or 3 mM EphA3) were incubated with the designated concentration of ATP at 298 K for 30-180 min. The reaction buffer consisted of 50 mM Tris-HCl pH 8.0, 100 mM NaCl, 10 mM MgCl 2 . The volume of the reaction buffer was 40 ml.
To evaluate the effect of MES, 10 mM MES pH 8.0 was added. Because the resveratrol and the flavonoids were dissolved in dimethyl sulfoxide (DMSO), DMSO was added to the negative control samples at the same concentration. After the reaction period, 160 ml of malachite green reagent was added and the samples were incubated for 60 min or until the coloration stabilized. The malachite green reagent was made by mixing one volume of 4.2% ammonium molybdate in 4 N HCl with three volumes of 0.045%(w/v) malachite green supplemented with 0.01%(v/v) Tween 20. The absorbance at 620 nm was recorded in a 96-well clear microplate using a FilterMax F5 (Molecular Devices). The concentration of orthophosphates was calculated by a calibration curve that was determined using NaH 2 PO 4 . The reaction velocities and standard errors were calculated from at least quintuplicate independent experiments.

ANS competitive binding assay
The ANS competitive assay was carried out as described previously, with some modifications (Yokoyama et al., 2015(Yokoyama et al., , 2020. Samples containing 2 mM DAPK1, 50 mM Tris-HCl pH 8, 100 mM NaCl, 0-300 mM ANS and 0-100 mM RSV were prepared in a 96-well black microplate and incubated for 60 min at room temperature. The volume of each sample was 200 ml. After the incubation period, the fluorescence intensity was measured using a GENios (Tecan Group) with excitation and emission wavelengths of 360 and 465 nm, respectively. The half-maximal bound concentration (BC 50 ) values were calcu-lated by fitting to a four-parameter logistic model using the least-squares method (Sebaugh, 2011). The BC 50 values and standard errors were calculated from quadruplicate independent experiments.

Protein crystallography
DAPK1 crystals were grown by the hanging-drop vapordiffusion method at 293 K. Prior to crystallization, 16 mg ml À1 DAPK1 and 20 mM RSV or PIC in DMSO were mixed in a ratio of 9:1 and the mixtures were incubated at room temperature for 60 min. Crystals of DAPK1-REV were grown by equilibration of a droplet composed of 2 ml protein sample and 2 ml reservoir solution (  directly cooled in liquid nitrogen until the X-ray diffraction experiments. Cryogenic X-ray diffraction data were collected on beamline BL-17A at the Photon Factory (PF) and on beamline NW12A at the Photon Factory Advanced Ring (PF-AR) in Japan. The diffraction data sets were integrated using the XDS program package (Kabsch, 2010). We adopted an I/(I) cutoff of 2.0 in the highest resolution shells. Because the crystals in the present study were isomorphous to those of a previous DAPK1-inhibitor complex (PDB code 5aux), structure refinement could be directly started using the coordinates of the DAPK1-inhibitor complex without performing Pattersonbased molecular replacement (Yokoyama et al., 2015). Prior to refinement, the coordinates of the inhibitor were omitted. The 3D structures and dictionary data for RSV and PIC were generated using the PRODRG server (Schü ttelkopf & van Aalten, 2004). The protein structures were refined using phenix.refine with stepwise cycles of manual model building using Coot (Liebschner et al., 2019;Emsley et al., 2010). The final models were evaluated using the Protein Data Bank validation suite (Yang et al., 2004). All structural figures were created using PyMOL (Schrö dinger).

Intrinsic ATPase activity of DAPK1
We assessed the intrinsic ATP hydrolysis (ATPase) activity of the catalytic domain of DAPK1 (residues 1-285). The ATPase activity was evaluated using the classic malachite green assay (Sherwood et al., 2013), in which orthophosphates, which are ATP hydrolysis products, are detected. We used tyrosine kinase ephrin type-A receptor 3 (EphA3) as a control. The ATPase activities of DAPK1 and EphA3 were clearly observed and increased in an ATP concentrationdependent manner, indicating that the ATP molecules were hydrolyzed by the kinases [ Fig. 2(a)]. The intrinsic ATPase activity of EphA3 was remarkably lower in comparison to that of DAPK1. While the ATPase activity of DAPK1 could be accurately detected at a protein concentration of lower than 0.1 mM, the ATPase activity of EphA3 was only detectable at a protein concentration of higher than 1 mM. In general, the activity of protein kinases is also easily detected using a protein kinase assay kit. However, with regard to the inhibition assay, there is the possibility of a false-positive result from the binding of inhibitors to components of the kits. It has been reported that measuring the intrinsic ATPase activity is a more reliable way of identifying an interleukin-2-inducible T-cell kinase inhibitor than measuring the phosphorylation-transfer activity (Kashem et al., 2007). Therefore, we preferred the use of a simple malachite green assay to investigate the inhibitory activity of compounds in this study.

RSV inhibited the intrinsic ATPase activity of DAPK1
The hydrolysis velocity at a concentration of 0.1 mM DAPK1 and 30 mM ATP was 0.042 mM min À1 . The addition of 10 and 100 mM RSV significantly reduced the hydrolysis velocity to 0.031 and 0.013 mM min À1 , respectively [ Fig. 2(b)]. Similarly, the hydrolysis velocity at 100 mM ATP was 0.06 mM min À1 in the absence of RSV, and the addition of 10 and 100 mM RSV significantly decreased the hydrolysis velocity to 0.052 and 0.034 mM min À1 , respectively. These results clearly indicate that RSV exhibits inhibitory activity against the ATPase activity of DAPK1.
In order to determine the type of inhibition by RSV, the enzymatic kinetics in the absence and presence of RSV were analyzed by means of a Lineweaver-Burk plot. While the   addition of 10 and 100 mM RSV did not have any effect on the K cat values, the K m values were significantly increased in an RSV concentration-dependent manner, indicating that the inhibition kinetics of RSV correspond to an ATP competitive inhibition model [ Fig. 3(a) and Table 1]. For further verification, a competitive binding assay was performed using the fluorescence probe 8-anilinonaphthalene-1-sulfonic acid (ANS). We have previously reported that ANS specifically binds to the ATP-binding site of DAPK1, and the fluorescence emission of ANS was increased by binding to DAPK1 (Yokoyama et al., 2015). Therefore, ANS is a useful tool for the discovery of ATP-competitive DAPK1 inhibitors. In this assay, the fractions of ANS-saturated DAPK1 at various ANS concentrations in the presence or absence of the compounds were measured and the half-maximal bound concentration (BC 50 ) values were estimated by fitting to a four-parameter logistic model (Sebaugh, 2011). The BC 50 value in the absence of RSV was 7.3 mM and the BC 50 values in the presence of 3, 10, 30 and 100 mM RSV were 16, 46, 150 and 600 mM, respectively [ Fig. 3(b)]. RSV clearly inhibited the binding of ANS to DAPK1 in an RSV concentration-dependent manner and 100 mM RSV almost completely inhibited the binding of ANS. The ANS competitive assay also showed that RSV is an ATP-competitive DAPK1 inhibitor.
In addition, we tested the inhibitory activity of API, KAE and PIC against the intrinsic ATPase activity of DAPK1 (Fig. 1). API and KAE are natural flavonoids that are structurally related to RSV and have been shown to bind to the ATP-binding site of DAPK1. The inhibitory potency of API was similar to that of RSV [ Fig. 2(c)]. The inhibitory potency of KAE was higher than that of API, which was in agreement with a previous ANS competitive binding assay (Yokoyama et al., 2015). PIC is a 3-hydroxy RSV derivative (Fig. 1). PIC also inhibited the ATPase activity of DAPK1.

Crystal structure of DAPK1 in complex with RSV and PIC
Information on the atomic interactions between DAPK1 and RSV will provide a better understanding of the inhibition mechanism and better ideas for the development of potent DAPK1 inhibitors. The crystal structure of DAPK1 in complex with RSV was solved at 1.65 Å resolution ( Table 2). The overall structure of DAPK1-RSV was almost identical to those of the apo form (PDB entry 1jks; Tereshko et al., 2001) and the AMP-PNP complex (PDB entry 1jkk; Tereshko et al., 2001), with an r.m.s.d. of 0.18 Å . It was found that RSV bound to the ATP-binding site, in agreement with kinetic analysis and the ANS competitive assay: distinctive electron density was observed in the ATP-binding site and its shape corresponded to the molecular shape of RSV [ Fig. 4(a)]. The A-ring of RSV occupied the binding site of the adenyl base of ATP and the 3-hydroxy group was hydrogen-bonded to the carbonyl group of Glu94 and the amide group of Val96 in the hinge region [Figs. 4(a) and 4(b)]. The A-ring was also stabilized by several C-HÁ Á Á interactions with the side chains of Val27, Ala40, Met146 and Ile160. The B-ring was located deep in the hydrophobic pocket, forming a hydrogen bond to Glu64, C-HÁ Á Á interactions with Leu93 and Ile160, and NHÁ Á Á interactions with the amido group of Asp161. The crystal structure of DAPK1 in complex with PIC was also solved at 1.75 Å resolution. The overall structure of the PIC complex was similar to that of the RSV complex, with an r.m.s.d. of 0.18 Å . The PIC molecule was clearly observed in the ATPbinding site and the binding mode of PIC was almost identical to that of RSV [ Fig. 4(c)]. In comparison to the RSV complex, additional hydrogen bonds were formed between the 3 0 -hydroxy group of PIC, Glu64 and Lys42. These hydrogen bonds may increase the binding affinity for DAPK1.
After we optimized the crystallization conditions for the DAPK1-RSV complex, we found that the crystal quality was significantly improved by using MES pH 6.5. We therefore solved the structure of a crystal obtained using MES buffer pH 6.5 at 1.4 Å resolution. While the overall structure was quite similar to that of the single RSV complex, with an r.m.s.d. of 0.16 Å , the RSV and MES molecules were observed in the ATP-binding site and the binding mode of RSV was different from that of single RSV [ Figs. 4(b) and 4(d)  RSV was located in the hinge region, and the 4 0 -hydroxy group was hydrogen-bonded to the carbonyl group of Glu94 and the amide group of Val96. The A-ring was located in the ribose site and the 3-hydroxy group was hydrogen-bonded to Glu100. The morpholino ring of MES was stacked on the A-ring of RSV by C-HÁ Á Á interactions, and the N atom of the morpholino ring formed a hydrogen bond to Asn144. The sulfate group of MES occupied the position of the -phosphate group of AMP-PNP. The sulfate group formed a salt bridge with Lys42. The C-HÁ Á ÁO hydrogen bond between the 2 0 -H of RSV and the sulfate group of MES may also contribute to stabilization of the binding of MES. The simultaneous binding of RSV and MES implies that MES might synergistically inhibit the ATPase activity of DAPK1 with RSV or might increase the inhibitory potency of RSV. However, the addition of 10 mM MES to the ATPase activity assay did not have a significant effect on the inhibitory potency of RSV [ Fig. 2(d)]. The presence of MES may not increase the binding affinity of RSV to DAPK1. However, from the viewpoint of the development of DAPK1 inhibitors, structural information on the simultaneous binding of RSV and MES would be useful for fragment-based drug discovery. Hybridization of RSV and MES may increase the inhibitory potency.

Discussion and conclusion
The ATPase activity of protein kinases is a hydrolysis reaction of ATP to ADP and orthophosphate using a water molecule as a nucleophile instead of a protein substrate. Therefore, it is conceivable that the ATPase activity is strongly dependent on the environment of the nucleophilic water molecule located close to the catalytic residues (Lys42 and Glu64 in DAPK1). Although the protein kinase activity (phosphorylation activity) has been broadly investigated, there are a few protein kinases that have been shown to exhibit intrinsic ATPase activity. To date, the ATPase activities of mitogen-activated protein/ERK kinase kinase 2 (MEKK2), MEK1, protein kinase C (PKC), dual-specificity protein kinase 2 (ERK2),   JNK3, p38 and p38 have been identified (Fox et al., 1999;Rominger et al., 2007;Ahmad et al., 2013). MEKK2 and MEK1 are classified into the STE protein kinase subfamily while ERK2, JNK3, p38 and p38 are classified into the CMGC subfamily. Our data clearly showed that DAPK1, which is classified into the CaMK subfamily, possesses ATPase activity, but the ATPase activity of EphA3, which is classified into the TK subfamily, was drastically lowered. Intrinsic ATPase activity may be widely present in serine/threonine protein kinases but not in tyrosine kinases. Although RSV is known to exhibit diverse biological effects, research into the interactions between RSV and protein kinases is still at an early stage. To the best of our knowledge, only a few protein kinases are known to interact directly with RSV, and RSV inhibited the ATPase activity of purified PKC and its isozymes (Slater et al., 2003;Yoon et al., 2002;Haworth & Avkiran, 2001). Recently, it was reported that RSV binds to PIM1 in a pull-down-based assay (Kim et al., 2020). However, the detailed inhibitory mechanism has not been elucidated owing to a lack of co-crystal structures of the kinases and RSV. Here, a comprehensive analysis using Michaelis-Menten kinetics, an ANS competitive binding assay and protein crystallography clearly showed that RSV is an ATP-competitive inhibitor of DAPK1.
RSV is composed of two aromatic phenyl rings with three hydroxy groups, and thus it is conceivable that RSV can bind to a highly hydrophobic pocket and also form hydrogen bonds to polar groups (Saqib et al., 2018). It is therefore reasonable that the ATP-binding site can accept the binding of RSV, because the adenyl group also possesses an aromatic property and hydrogen-bond donors and acceptors. Indeed, the DAPK1-RSV crystal structures showed that the A-ring or B-ring of RSV occupy the hinge region to which the adenyl group of AMP-PNP binds [ Fig. 4(b)]. RSV is able to bind to various proteins ranging from ligases to oxidoreductases and hydrolases (Sajish & Schimmel, 2015;Buryanovskyy et al., 2004;Shukla et al., 2015). However, there are no significant similarities in the sequences and structures of the binding pockets of these proteins (Saqib et al., 2018). The bindability of nucleobases may be a good predictor of the binding of RSV.
We have previously reported natural flavonoids to be DAPK1 inhibitors with two distinctive binding modes referred to as type A and type B (Yokoyama et al., 2015). The binding modes of API and KAE are type B binding modes. In the type B binding mode, binding of the flavonoids is associated with the binding of chloride ion. While the B-ring of API occupies the hinge region, the AC-ring points to the glycine-rich loop [ Fig. 4(e)]. Based on the structural similarity among RSV, API and KAE, we initially speculated that the binding mode of RSV would be similar to those of API and KAE. However, we found that the binding mode of RSV corresponded to the type A binding mode, in which the B-ring is turned towards the side chain of Glu64 and the AC-ring occupies the hinge region [ Fig. 4(e)]. The binding of API and KAE was stabilized by interaction of the central pyran ring (C-ring) with the glycinerich loop and the chloride ion. The lack of a pyran ring may force RSV to bind in the type A mode. In terms of the structure-activity relationship study, determination of the binding orientation would be helpful for the design of RSV derivatives with improved inhibitory activity: an optimization of 2 0 -H to interact with the glycine-rich loop would increase the binding affinity to DAPK1.
RSV is known to be very safe and well tolerated. After oral dosage with RSV, it is extensively metabolized in the intestine and liver, and is rapidly eliminated (Walle et al., 2004). Therefore, a major concern regarding the clinical use of RSV is its poor bioavailability. However, this poor bioavailability is currently being overcome by several approaches, including nanoencapsulation of RSV and the development of long-lived RSV derivatives (Sergides et al., 2016;Chimento et al., 2019). It should be noted that several animal studies have indicated that resveratrol is blood-brain barrier (BBB) permeable (Wang, Xu et al., 2002;Frozza et al., 2013). Because BBB permeability is critical for the development of neurodegenerative disorder inhibitors, BBB permeability of RSV would be a great advantage. The increasing number of studies designed to improve the oral bioavailability of RSV encourages us to expect that RSV and RSV derivatives could be useful for the treatment of AD in the future.
In summary, we have demonstrated that RSV inhibits the intrinsic ATPase activity of DAPK1 by binding to the ATPbinding site. Crystallographic analysis of the DAPK1-RSV complex revealed that the binding of RSV to DAPK1 was stabilized by occupation of the A-ring at the nucleobase position and by hydrogen bonds to the hinge region. Determination of the binding orientation provided key information for improvement of the binding affinity of RSV to DAPK1. Although the development of RSV as a clinical drug is still under way, the present data encourage further development of RSV for the treatment of DAPK1-associated diseases.