Phosphoryl Transfer Reaction Snapshots in Crystals

Background: PKAc (catalytic subunit) catalyzes phosphorylation of protein substrates thereby regulating a myriad of cellular processes. Results: X-ray structures of PKAc complexes along the phosphoryl transfer reaction have been obtained. Conclusion: The phosphotransfer follows a multistep mechanism, including conformational changes of the substrate and product groups, a loose transition state, and metal movement. Significance: Mechanistic knowledge about the phosphorylation by PKAc will contribute to understanding of the kinase function and regulation. To study the catalytic mechanism of phosphorylation catalyzed by cAMP-dependent protein kinase (PKA) a structure of the enzyme-substrate complex representing the Michaelis complex is of specific interest as it can shed light on the structure of the transition state. However, all previous crystal structures of the Michaelis complex mimics of the PKA catalytic subunit (PKAc) were obtained with either peptide inhibitors or ATP analogs. Here we utilized Ca2+ ions and sulfur in place of the nucleophilic oxygen in a 20-residue pseudo-substrate peptide (CP20) and ATP to produce a close mimic of the Michaelis complex. In the ternary reactant complex, the thiol group of Cys-21 of the peptide is facing Asp-166 and the sulfur atom is positioned for an in-line phosphoryl transfer. Replacement of Ca2+ cations with Mg2+ ions resulted in a complex with trapped products of ATP hydrolysis: phosphate ion and ADP. The present structural results in combination with the previously reported structures of the transition state mimic and phosphorylated product complexes complete the snapshots of the phosphoryl transfer reaction by PKAc, providing us with the most thorough picture of the catalytic mechanism to date.

Protein kinases are signaling enzymes that regulate cellular processes by catalyzing phosphorylation of proteins. Chemically, protein kinases transfer the ␥-phosphoryl group of a nucleotide triphosphate (e.g. ATP) to the hydroxyl group of a serine, threonine, tyrosine, or histidine residue of the substrate protein. Over 500 protein kinases have been identified in the human genome (ϳ1.7% of genes), pointing to the biological importance of phosphoryl-transfer chemistry (1). Extensive studies of the cAMP-dependent protein kinase (PKA) that phosphorylates the side chains of Ser or Thr residues have made it a paradigm for the whole family of kinase enzymes (2,3).
Being a regulatory enzyme, PKA is highly regulated itself. When inactive, PKA is a tetrameric holoenzyme, R 2 C 2 , composed of two catalytic (C) monomeric and regulatory homodimeric (R 2 ) subunits. An increase in cAMP concentration activates PKA; binding of four cAMP molecules to R 2 causes the tetramer to dissociate, releasing two active C subunits (that we refer to here as PKAc) 3 (4). In PKAc, the nucleotide-binding site is in the cleft between N-terminal and C-terminal lobes that are connected by a small linker region, but the nucleotide primarily interacts with the N-lobe. The substrate sits at the edge of the cleft on the surface of the large C-lobe. PKAc requires one or two divalent metal ions to bind to the active site to be active (5,6). The physiological metal is magnesium, although others can support phosphotransferase activity (7,8).
Crystallographic studies have provided a wealth of information on how PKAc functions (4, 9 -11). Complexes of PKAc with nucleotide and/or substrate analogs are found in three major conformational states that differ in the relative orientation of the N-and C-lobes. With no ligands bound (apo form) PKAc adopts an open conformation; upon nucleotide or substrate binding (binary form), PKAc transitions to an intermediate, partially closed, state; last, PKAc assumes a closed conformation when all components for the reaction are in place (ternary form) (12)(13)(14)(15).
Although the PKAc phosphoryl transfer step is fast, Ͼ500 s Ϫ1 , the product turnover rate is at least an order of magnitude slower; k cat is ϳ20 s Ϫ1 (16). The rate-limiting step at high magnesium concentrations (ϳ10 mM) is the release of MgADP (17). Analysis of the crystallographic structures and solution kinetic data suggests that conformational changes, particularly those involved in the release of the nucleotide, might be essential for PKAc function. Recent nuclear magnetic resonance (NMR) studies and molecular dynamics (MD) simulations have established a relationship between the PKAc conformational fluctuations and its turnover rate (18,19). Specifically, the rate of the opening motion correlates with k cat and the rate-limiting step, MgADP product release.
The chemical step occurs only in the closed form of PKAc and the release of product is concurrent with the enzyme returning to the open conformation. However, single molecule electronic measurements of PKAc catalysis indicate that not every open-close conformational cycle results in the phosphorylation reaction and/or product release, which can partially explain the relatively low catalytic efficiency of the enzyme (20). These experiments also caution us on correlating bulk kinetic values, such as the k cat , with the time it takes each individual molecule to go through a particular conformational change or a chemical reaction.
Despite these advances fundamental questions remain unanswered. In particular it is still not clear whether the phosphoryl transfer proceeds as a direct nucleophilic attack by the OH group of the substrate on the ␥-phosphorus of ATP in the S N 2 fashion or through the metaphosphate intermediate as an S N 1 reaction. Stereochemical NMR studies suggest that PKAc facilitates a direct in-line displacement reaction with the pentacoordinated phosphorus in the transition state (21). Yet, the possibility of the short-lived hindered metaphosphate has not been completely ruled out. Uncertainty also surrounds the roles of active site residues and metal ions, and the details of hydrogen transfer pathways during the chemical step. Specifically, different functions have been proposed for the catalytically important Asp-166 residue. In PKAc, Asp-166 is the nearest ionizable residue to the OH group of the substrate and may act as a catalytic base and/or to correctly position the nucleophile. Asp-166 is universally conserved in the active site of all protein kinases and can form hydrogen bond interactions with Ser of the substrate (14,22,23). Asp-to-Ala substitution produces a mutant variant with activity below 1% of the wild-type (24).
Several crystallographic structures of PKAc complexes that represent different stages of the phosphoryl transfer reaction have been determined previously. Those mimicking the reactant complexes have been obtained either with peptide inhibitor, IP20, or with unhydrolysable ATP analogs, AMPPCP or AMPPNP. These structures, however, are not very good mimics of the Michaelis complex. In particular, IP20 lacks the nucleophilic group, having Ala in place of reactive Ser, and precluding the analysis of the conformation and interactions of the nucleophile before the reaction (15,25,26). In addition, in complexes containing substrate peptide SP20 and ATP analogs, the side chain of Ser-21 SP20 adopts a conformation, in which it is rotated away from Asp-166, which prevents H-bond formation (8,27). The transition state mimic, having the MgF 3 Ϫ anion in place of ␥-PO 3 , ADP, and SP20 has also been determined (23). The structure is consistent with the S N 2 mechanism and demonstrates a short interaction between Asp-166 and the side chain of Ser-21 SP20 . Finally, PKAc product complexes have been trapped in crystals when ATP and SP20 were utilized (8). These structures, and enzyme kinetics measurements, demonstrated that, in addition to Mg 2ϩ , all other divalent alkaline earth metals, including Ca 2ϩ , Sr 2ϩ , and Ba 2ϩ , support the phosphoryl transfer to SP20 (14). In our preliminary studies, the rate of phosphotransfer determined by pre-steady state kinetics for RII␤ holoenzyme in the presence of Ca 2ϩ is only severalfold lower than with Mg 2ϩ . A full kinetic study will be published elsewhere.
To design a PKAc complex that more accurately represents the Michaelis complex, inspiration was taken from the nearattack conformation theory of enzymes. According to the nearattack conformation theory, enzymes may bind substrates with geometries that are very close to those of the transition states (28,29). Near-attack conformation is defined as a complex in which reactive groups of substrates and functional groups of the active site of the enzyme are in close proximity and in orientation for the reaction to advance. Assuming the structure with the MgF 3 Ϫ anion truly mimics the transition state then according to the near-attack conformation theory the Michaelis complex should have a similar geometry. The key challenge is in designing the complex to allow all participants in the reaction to assume the correct conformation while preventing the reaction from taking place.
Here we report crystallographic structures of three PKAc complexes. For one, which we designate as PKAc-Ca 2 ATP-CP20, we co-crystallized PKAc with ATP and a pseudo-substrate peptide (CP20). In CP20, Ser-21 of SP20 has been substituted with Cys. For the other, which we designate as PKAc-Ca 2 AMPPNP-SP20, we co-crystallized PKAc with AMPPNP and a 20-residue substrate analog (SP20). Both were obtained with excess Ca 2ϩ . Comparison of the two complexes with the transition state mimic structure reveals fundamental differences in the structures of the active sites that conclusively demonstrate that unhydrolysable ATP analogs produce less biologically relevant structures of the active site of PKAc, which is supported by our MD simulations of the two structures and the product complex. Specifically, in PKAc-Ca 2 ATP-CP20, the thiol group of Cys-21 CP20 is rotated toward Asp-166, whereas in PKAc-Ca 2 AMPPNP-SP20 the hydroxyl group of Ser-21 SP20 points away from Asp-166 into the bulk solvent. This observation indicates that the orientation of the serine side chain of the substrate seen in previous crystallographic structures is modulated by ATP analogs. For the third structure, designated as PKAc-Mg 2 ADP-PO 4 -CP20, the enzyme was crystallized in the presence of excess ATP, CP20, and Mg 2ϩ . In PKAc-Mg 2 ADP-PO 4 -CP20, ATP is completely hydrolyzed into ADP and a free phosphate ion, whereas the side chain of Cys-21 CP20 remains unmodified but occupies two positions.
We propose that because PKAc-Ca 2 ATP-CP20 contains reactive ATP, Ca 2ϩ ions capable of promoting phosphoryl transfer and a nucleophilic group in the substrate peptide, this complex is the closest model of the actual Michaelis complex to date. In addition we argue that PKAc-Mg 2 ADP-PO 4 -CP20 can represent a state just after the phosphoryl transfer reaction, but before the phosphorylated Ser-21 SP20 rotates out toward the bulk solvent. Thus, our current results taken together with previously published crystallographic work and theoretical calculations provide the most complete picture to date of the phosphorylation reaction catalyzed by PKAc.

Experimental Procedures
General Information-Pseudo-substrate peptides SP20 (TTY-ADFIASGRTGRRASIHD; residues 5-24 of the heat-stable PKAc inhibitor PKI, where positions 20 and 21 have been mutated to Ala and Ser) and CP20 (TTYADFIASGRT-GRRACIHD; SP20 derivative, where Ser-21 was substituted with a Cys residue) were custom synthesized by Biomatik (Wilmington, DE). ATP as the magnesium or disodium salts and AMPPNP as a lithium salt were purchased from Sigma. Protein purification supplies were purchased from GE Healthcare. Crystallization reagents were purchased from Hampton Research (Aliso Viejo, CA).
Protein Expression and Purification-His 6 -tagged recombinant mouse PKAc was expressed in Escherichia coli using LB or minimal medium at 18 -20°C for 16 -18 h. The recombinant enzyme was purified by affinity chromatography using HisTrap fast-flow chromatography columns supplied by GE Healthcare. The enzyme was then buffer exchanged with 50 mM MES, 250 mM NaCl, 2 mM DTT, pH 6.5, on a desalting column. Isoforms of PKAc were not separated, without any obvious effect on crystallization of the ternary complexes.
Crystallization-For crystallization trials PKAc was concentrated to 8 -12 mg/ml. The ternary complexes with different metals, ATP (or AMPPNP), and pseudo-substrate peptides CP20 or SP20 were made before setting up crystallization trails. First, the concentrated PKAc solution was mixed with a solution of metal chloride salt to reach the final metal concentration of ϳ20 mM. Then, the nucleotide was added. The peptide substrate was introduced to the mixture last. The molar ratio of PKAc:nucleotide:peptide was kept at 1:10:10. Crystals were grown in sitting drop microbridges or in 9-well glass plates using well solutions consisting of 100 mM MES, pH 6.5, 5 mM DTT, 15-20% PEG 4000 at 4°C. For complexes with different metal ions, the corresponding metal chloride salts were introduced to the well solutions at 50 mM concentrations prior to setting up crystallization drops.
Data Collection, Structure Determination, and Refinement-X-ray crystallographic data were collected at 100 K using a Rigaku HomeFlux system, equipped with a MicroMax-007 HF generator, Osmic VariMax optics, and an RAXIS-IVϩϩ image plate detector. Diffraction data were collected, integrated, and scaled using HKL3000 software suite (30). The structures were refined using SHELX-97 (31). A summary of the crystallographic data and refinement is given in Table 1. Similar to our previous observations (25) all the structures were of isoform 2, and contained three post-translationally phosphorylated residues: Ser-139, Thr-197, and Ser-338. The structure of the ternary complex of PKAc with 2Mg 2ϩ , ATP, and peptide inhibitor IP20 (PDB code 4DH3) (25) was used as a starting model to solve all the structures described here. The structures were built and manipulated with the program Coot (32), whereas the figures were generated using the PyMol molecular graphics software (version 1.5.0.3; Schrödinger LLC).
Molecular Dynamics Simulations-MD simulations were performed for PKAc-Ca 2 ATP-CP20, PKAc-Ca 2 AMPPNP-SP20, and PKAc-Ca 2 ADP-pSP20. Briefly, the PKAc-Mg 2 ATP-IP20 crystal structure (PDB code 1ATP) (33) was used as a template to set up the simulations. In all three systems, Thr-197 and Ser-338 are phosphorylated with fully ionized phosphate groups. Gromacs-4.6.1 (34) was utilized to conduct all the simulations with the Amber ff99SB (35) force field with ILDN (36) and NMR (37) modifications. Each of the systems was first energy minimized then a total of 500 ns production simulation was performed under isobaric and isothermal conditions (298 K and 1 atm).

Results
Ternary Pseudo-Michaelis Complexes PKAc-Ca 2 ATP-CP20 and PKAc-Ca 2 AMPPNP-SP20 -The electron density maps clearly indicate that intact peptide and nucleotide molecules are trapped in the enzyme active site, confirming the formation of pseudo-Michaelis complexes (Fig. 1, A and B). The two structures are similar, with the r.m.s. deviation on the main chain atoms being 0.42 Å, but there are several important structural differences.
In PKAc-Ca 2 ATP-CP20, M1 and M2 metals have coordination numbers of seven and six, whereas in PKAc-Ca 2 AMPPNP-SP20 both metals are surrounded by eight ligands (Fig. 1C). In both structures the coordination spheres around each metal site involve protein residues (Asn-171, Asp-184) and ATP groups, which are identical to those in the coordination spheres of magnesium, observed previously including the PKAc-Mg 2 ATP-IP20 complex (PDB code 4DH3) (15,25,38). The remaining coordination sites around each metal, with the exception of Ca2 in PKAc-Ca 2 AMPPNP-SP20, are provided by water molecules. A slight displacement of ϳ0.5 Å of ␥-phosphorus in AMPPNP relatively to ATP puts the former's ␥-phosphate closer to M2 site, allowing formation of the second coordination contact between M2 and the ␥-phosphate in PKAc-Ca 2 AMPPNP-SP20. In addition, a sliding shift of the glycine-rich loop of ϳ1 Å in PKAc-Ca 2 AMPPNP-SP20 toward the ␣B helix relative to its position in PKAc-Ca 2 ATP-CP20 makes the active site of the former more accessible for water molecules, which explains the increased coordination number of 8 for both Ca 2ϩ cations. FIGURE 1. A, electron density map for the active site components in PKAc-Ca 2 ATP-CP20 contoured at 1.5 level (4 for calcium cations). B, electron density map for the active site components in PKAc-Ca 2 AMPPNP-SP20 contoured at 1.5 level (4 for calcium cations). C, superposition of the active sites in PKAc-Ca 2 ATP-CP20 (colored by atom type, carbon is green, Ca 2ϩ ions are dark cyan, H 2 O molecules red) and PKAc-Ca 2 AMPPNP-SP20 (light magenta, carbon atoms; light cyan, Ca 2ϩ ions; magenta, H 2 O molecules), showing metals Ca1 and Ca2 bound at sites M1 and M2, respectively, nucleotides ATP and AMPPNP, Cys-21 CP20 and Ser-21 SP20 of the substrate peptides CP20 (blue carbon atoms) and SP20 (orange carbon atoms), respectively, and the residues of the enzyme that are important for metal binding or catalysis. Metal coordination, as black solid lines, and possible hydrogen bonds, as dashed lines, are shown for PKAc-Ca 2 ATP-CP20. Distances are in Å. D, superposition of the active sites in pseudo-Michaelis complexes PKAc-Mg 2 ATP-IP20 (PDB: 4DH3, dark magenta for all atoms including Mg 2ϩ ions and water molecules), and PKAc-Ca 2 ATP-CP20 (carbon, green; Ca 2ϩ ions, dark cyan; and water molecules, red). Metal coordination are shownn as black solid lines, and possible hydrogen bonds as dashed blue are shown for PKAc-Mg 2 ATP-IP20. Distances are in Å.
In PKAc-Ca 2 ATP-CP20, the thiol group of Cys-21 CP20 is facing Asp-166. In this orientation the S atom is located in close proximity to the Asp-166 carboxyl (S . . . O distance is of 3.2 Å). In contrast, in the PKAc-Ca 2 AMPPNP-SP20 structure the C␤-O␥ bond of Ser-21 SP20 is rotated by ϳ110°away from Asp-166, pointing toward the bulk solvent (Fig. 1C).
The interactions formed by Lys-168 are similar in both structures reported here, but they differ relative to the previously reported PKAc complexes with ATP and IP20 (Fig. 1D) (25). Specifically Lys-168 loses direct H-bond interactions with ␥-phosphate oxygen of ATP or AMPPNP observed in structures with IP20. The smaller side chain of Ala-21 IP20 lacks a substituent in the ␥-position, which allows ATP to move closer to the substrate peptide and knock out W4 in PKAc-Mg 2 ATP-IP20 (Fig. 1D).
Comparison of PKAc-Ca 2 ATP-CP20, Transition State Mimic PKAc-Mg 2 ADP-MgF 3 -SP20, and Product PKAc-Ca 2 ADP-pSP20 -The previously reported crystal structure of the PKAc with MgADP, SP20, and AlF 3 is considered to be a transition state mimic in the phosphoryl transfer reaction (23). AlF 3 , the transition state analog in place of ␥-PO 3 , has been reassigned as MgF 3 Ϫ anion by Jin et al. (39) based on 19 F NMR measurements. Thus, we designate this structure as PKAc-Mg 2 ADP-MgF 3 -SP20 (PDB code 1L3R). In addition, we recently obtained an x-ray structure of the PKAc-Ca 2 ADP-pSP20 complex, in which the ADP and phosphorylated SP20 products were captured in the active site (8). Comparison of our current pseudo-Michaelis complex with the transition state mimic and the product allows us to visualize structural changes that may accompany the catalysis and to identify mechanistically important structural information.
The superposition of the active sites of PKAc-Ca 2 ATP-CP20 and PKAc-Mg 2 ADP-MgF 3 -SP20 is shown in Fig. 2. The catalytically important residues occupy very similar positions in both structures, which are aligned with r.m.s. deviations of 0.55 Å. The main difference is the conformation of the phosphate groups of the nucleotide and the displacement of the glycinerich loop. In PKAc-Ca 2 ATP-CP20 the ␤-P is ϳ1 Å further away from the metal sites relatively to its position in the transition state mimic, resulting in a 4.1-Å distance between one of the ␤-P oxygen atoms and Ca2, whereas Mg2 is bound to this oxygen in PKAc-Mg 2 ADP-MgF 3 -SP20. The geometry of the phosphate groups is reflected in the relative positions of the glycinerich loop in the two structures (Fig. 3). In PKAc-Ca 2 ATP-CP20, in addition to a ϳ2 Å sliding displacement, residues 54 and 55 of the loop are positioned more than 2 Å above their location in PKAc-Mg 2 ADP-MgF 3 -SP20. Most importantly the side chains of Ser-21 SP20 and Cys-21 CP20 have practically identical conformations, both facing Asp-166 and making contacts with its carboxylic oxygen at distances of 2.5 and 3.2 Å, respectively.
The active sites of PKAc-Ca 2 ATP-CP20 and the PKAc-Ca 2 ADP-pSP20 product complex (PDB code 4IAK) are geometrically similar, suggesting that the active site organization is the same before and after the phosphorylation reaction (Fig. 2). The major changes are associated only with the transfer of the ␥-phosphoryl group. In the pseudo-Michaelis complex ␥-PO 3 Ϫ anion), and product complex PKAc-Ca 2 ADP-pSP20 (blue, carbon atoms; light cyan, Ca 2ϩ ions), showing similar conformation for the side chains of Ser-21 SP20 and Cys-21 CP20 , both facing Asp-166, whereas in product complex C ␤ -O ␥ is rotated away from Asp-166, suggesting the flip of the P-site residue following phosphoryl transfer. The distance between the ␥-P of ATP and the oxygen in Ser-21 SP20 is shown as black dashed double arrow. The red dashed arrow demonstrates the difference in the position of ␥-PO 3 group before and after the reaction. Distances are in Å.
is coordinated to both metals and hydrogen bonds to Ser-53. In the product complex, ␥-PO 3 group moves 2.7 Å, relative to its position in ATP to bind to Ser-21 SP20 . The transferred PO 3 group retains its interactions with Ca1 and Ser-53, but loses coordination to Ca2. Ca1 maintains 7 ligands in its coordination sphere, whereas Ca2 gains an extra water in the product structure. Another striking difference is the orientation of C ␤ -S ␥ and C ␤ -O ␥ bonds of Cys-21 CP20 and Ser(P)-21 SP20 before and after the reaction. In PKAc-Ca 2 ATP-CP20 the side chain of Cys-21 CP20 is oriented toward Asp-166, whereas in PKAc-Ca 2 ADP-pSP20 C ␤ -O ␥ is rotated away from the active site toward the bulk solvent, which prevents the hydrogen bond from forming between the phosphate group on Ser(P)-21 SP20 and Asp-166. The ␥-PO 3 transfer is also accompanied by a ϳ2 Å gliding shift of the glycine-rich loop away from the ␣B helix and toward the metals and ADP (Fig. 3).
Ternary Complex PKAc-Mg 2 ADP-PO 4 -CP20 -Although this complex was crystallized using excess ATP the omit map unmistakably shows that only its hydrolysis products, ADP and PO 4 , are present at the active site of the enzyme (Fig. 4A). Refinement suggested 100% occupancy for ADP but 64% for the free phosphate. The partial occupancy of PO 4 is not surprising because small inorganic ions can easily diffuse into the bulk solvent. Additionally, the side chain of Cys-21 of CP20 clearly displays two conformations as indicated by the omit difference electron density map, which would not be possible with 100% PO 4 retention. Conformation A (64% occupancy), in which the C␤-S␥ bond rotated toward the bulk solvent and away from Asp-166, is similar to that observed in all product complexes (8,14), and in PKAc-Ca 2 AMPPNP-SP20 complex. Conformation B (36% occupancy) with the SH group pointing toward Asp-166 is identical to the position of C␤-S␥ in PKAc-Ca 2 ATP-CP20 and clashing into the free phosphate (Fig. 4B). The PKAc-Mg 2 ADP-PO 4 -CP20 structure agrees very well with the recently published room temperature structure of PKAc-Mg 2 ADP-PO 4 -IP20, which also demonstrated complete hydrolysis of ATP (25). The two structures are superimposed with a r.m.s. deviation of 0.34 Å.
Molecular Dynamics Simulations-Root mean square deviations for the three simulations showed that all complexes are stable and converged well during our simulation time (Fig. 5A). Using the PKAc-Mg 2 ATP-IP20 crystal structure as reference, the PKAc-Ca 2 AMPPNP-SP20 has the largest r.m.s. deviation of the three simulations performed. Root mean square fluctuations of the PKAc residues in the complexes are generally small with the exception of the following regions: Lys-28 ϳ Ala-70, Pro-236 ϳ His-260, Val-275 ϳ Lys-295, and Asp-328 ϳ Glu-346. We calculated the distances between the S atom of Cys-21 CP20 and closest O atom in the Asp-166 carboxyl in PKAc-Ca 2 ATP-CP20 as well as the distance between the O atom in Ser-21 SP20 and closest O atom in Asp-166 carboxyl in PKAc-Ca 2 AMPPNP-SP20. The S . . . O distance distribution has a primary peak at 3.5 Å and a minor peak at 7.4 Å, whereas the O . . . O distance distribution is broad, spanning from 4.0 to 9.0 Å  ( Fig. 5B). For the product complex, the distance between O and the Asp-166 carboxyl is 6.5 Ϯ 0.9 Å. Finally, the PKAc-Ca 2 AMPPNP-SP20 complex has the largest radius of gyration, whereas PKAc-Ca 2 ADP-pSP20 is the most compact structure (Fig. 5C). A characterization of the global motions using principle component analysis indicated that PKAc-Ca 2 ATP-CP20 and PKAc-Ca 2 ADP-pSP20 have comparable movements related to the opening and closing of the cleft (the dominant motion) that differ from those found in PKAc-Ca 2 AMPPNP-SP20 (Fig. 5D).

Discussion
In this study we utilized Ca 2ϩ ions in place of Mg 2ϩ and replaced the nucleophilic oxygen in Ser-21 of SP20 with sulfur by substituting Cys for Ser to produce a better mimic of the Michaelis complex for the phosphotransfer reaction catalyzed by PKAc. Although a thiol group is a stronger nucleophile than a hydroxyl, thiol analogs are poor kinase substrates (40). Replacement of Ca 2ϩ with Mg 2ϩ ions in the presence of the CP20 substrate analog afforded a complex with trapped phosphate ion. These structures in combination with the previously reported structures of the transition state mimic (23) and product complexes (8,38) complete the snapshots of the phosphoryl transfer reaction in PKAc.
According to the proposal of Mildvan (41) the upper limits of two mechanistically relevant distances, reaction coordinate and axial distances, can be estimated from crystal structures of phosphotransferase reactant and transition state mimic complexes. Analysis of the present results may help discriminate between the dissociative (D N ϩ A N , or S N 1) and concerted (A N D N , or S N 2) mechanisms, and predict the transition state nature, i.e. loose or tight, for the S N 2 mechanism. In the case of PKAc, the reaction coordinate distance can be measured between the nucleophile and ␥-P ATP atoms in a Michaelis mimic. The axial distances can be estimated from the Michaelis and transition state mimic structures by dividing the distances between the nucleophile and leaving ␤-O ADP atoms in half, assuming fully symmetric reaction. As estimated by Mildvan (41), the reaction coordinate values of Յ4.9 Å and the axial distances of Յ3.3 Å would indicate the S N 2 mechanism is in operation, whereas longer separations would point to the S N 1 mechanism. The other important parameter is the O-P⅐⅐⅐O angle between the P-O bond of the phosphoryl and the nucleophile atom in the reactant state. If this angle is close to 180°, the reactants are positioned in the near attack configuration for the S N 2 mechanism.
The concerted mechanism has been suggested for PKAc by stereochemical studies (21). Of the two pseudo-Michaelis complex structures reported here the positioning of reactants in the active site of PKAc-Ca 2 ATP-CP20 is more consistent with the S N 2 mechanism (Fig. 1, A and B). Specifically, the angle between the O ␤ -P ␥ bond of ATP and the S Cys-21 atom of CP20 is close to linear (ЄO 3␤ P ␥ S Cys-21 ϭ 161°). The C ␤ -S ␥ bond of Cys-21 CP20 is rotated toward Asp-166, making a 3.2-Å hydrogen bond with the carboxylic side chain. Importantly, similar orientation of the side chain of the substrate serine was observed in the structure of the transition state mimic PKAc-Mg 2 ADP-MgF 3 -SP20 (Fig. 2). By contrast in PKAc-Ca 2 AMPPNP-SP20 the respective angle (ЄN 3␤ P ␥ O Ser-21 ) is 136º and the serine residue is flipped FIGURE 5. A, r.m.s. deviation of the PKA catalytic domain backbone for PKAc-Ca 2 ATP-CP20 (red), PKAc-Ca 2 AMPPNP-SP20 (blue), and PKAc-Ca 2 ADP-pSP20 (PDB code 4IAK, black) systems is plotted as a function of time. The color scheme is the same for all the following figures. B, red line shows distribution of distances between the S atom of Cys-21 CP20 and closest O atom in the Asp-166 carboxyl in PKAc-Ca 2 ATP-CP20 system. Blue line shows distribution of distances between the O atom in Ser-21 SP20 and closest O atom in Asp-166 carboxyl in PKAc-Ca 2 AMPPNP-SP20 system. C, distributions of the radius of gyration calculated for PKAc-Ca 2 ATP-CP20, PKAc-Ca 2 AMPPNP-SP20, and PKAc-Ca 2 ADP-pSP20 systems. D, the distributions of trajectory projections onto the first and second principal components for PKAc-Ca 2 ATP-CP20, PKAc-Ca 2 AMPPNP-SP20 and PKAc-Ca 2 ADP-pSP20 systems.
away from Asp-166, facing the bulk solvent. MD simulations demonstrate that O ␥ (Ser 21 )⅐⅐⅐O ␦ 1(Asp-166) separation covers a wide range from 4.0 Å to 9.0 Å in PKAc-Ca 2 AMPPNP-SP20, whereas S ␥ (Cys 21 )⅐⅐⅐O ␦ 1(Asp-166) distance has a primary peak located at 3.5 Å in PKAc-Ca 2 ATP-CP20 (Fig. 5B). Based on the previous mechanistic studies, which indicated a possible catalytic role of Asp-166 (42,43), we propose that in PKAc-Ca 2 ATP-CP20 the thiol group of Cys-21 CP20 has the correct near-attack conformation that the serine of the substrate would assume in the actual Michaelis complex. Notably, only when the C ␤ -SH bond assumes the "inward" conformation and interacts with Asp-166, sulfur is primed for an in-line phosphoryl transfer, with the O ␤ -P␥⅐⅐⅐S Cys-21 angle close to 180°. The reaction coordinate distance P ␥ ⅐⅐⅐S Cys-21 measured in PKAc-Ca 2 ATP-CP20 is 5.4 Å, which correlates well with that estimated from NMR measurements for the PKAc complex with AMPPCP and Kemptide (5.3 Ϯ 0.7 Å) (44). The distance between the nucleophilic sulfur and leaving oxygen atoms, O ␤ ⅐⅐⅐S Cys-21 , is 7 Å, which gives an estimated axial distance of 3.5 Å. These values suggest a very loose transition state and even recommend the formation of the metaphosphate intermediate in the dissociative mechanism S N 1. However, NMR measurements established the inversion of stereochemistry around the ␥-P atom in the product that points to the S N 2 mechanism (21). Yet, S N 1 reactions can also proceed with inversion if an intimate, or tight, ion pair is formed along the reaction coordinate (45). According to this concept, the metaphosphate intermediate would not diffuse into solvent and would have a short lifetime, enough only for vibrational relaxation. Although this idea is appealing, the extended reaction coordinate and axial bond distances can be explained by the larger van der Waals radius of sulfur (1.9 Å) compared with oxygen (1.3 Å) in our study and the use of AMPPCP in the previous NMR measurements. Indeed, in PKAc-Mg 2 ADP-MgF 3 -SP20 the axial distances ␤O ADP ⅐⅐⅐MgF 3 and MgF 3 ⅐⅐⅐O Ser-21 are equal, each of 2.3 Å. Assuming the positions of ATP in PKAc-Ca 2 ATP-CP20 and Ser-21 of SP20 in PKAc-Mg 2 ADP-MgF 3 -SP20 are representative of the actual Michaelis complex, alignment of the two structures offers considerably shorter distances: 4.4 Å for the reaction coordinate ␥-P ATP ⅐⅐⅐O Ser-21 and 3 Å for the axial distances (Fig. 2). These values are similar to those predicted by theoretical calculations for the loose transition state in the S N 2 phosphoryl transfer (46).
The ternary complex PKAc-Mg 2 ADP-PO 4 -CP20 was prepared in an effort to produce a Michaelis complex mimic containing the physiological Mg 2ϩ metal ions. Unexpectedly, products of ATP hydrolysis, ADP, and the free phosphate, were found trapped in the active site of PKAc, even though the efficiency of ATP hydrolysis is Ͻ1% of the phosphotransferase activity. Acceleration of ATPase activity was previously observed in the presence of histones, which are natural kinase substrates (47). Moreover, increased ATPase activity has been reported with a pseudo-substrate Leu-Arg-Arg-Ala-Cys-Leu-Gly, whose sequence is similar to that of CP20 and Kemptide (48). Tight binding of IP20 is very efficient in excluding water from the active site resulting in the full inhibition of ATPase activity, which allowed crystallization of ternary PKAc complexes with the unhydrolyzed ATP molecule (15,25,26). None-theless, two previous crystallographic studies captured the free phosphate in the presence of IP20. Full ATP hydrolysis was observed in the room temperature structure of the wild-type PKAc ternary complex, and was attributed to the x-ray radiation damage (25). Partial ATP hydrolysis was detected in the 100 K structure of the PKAc mutant variant Y204A (49). In this instance, the reaction was attributed to altered conformational dynamics in the mutant.
As a possible explanation for the PKAc ATPase activity it has been suggested that instead of reacting with the nucleophile on the substrate, ATP can be attacked by a water molecule diffused into the enzyme active site and positioned by Asp-166 to initiate the phosphoryl transfer (49). Close inspection of PKAc-Mg 2 ADP-PO 4 -CP20 and PKAc-Ca 2 ATP-CP20 revealed a water molecule that may facilitate ATP hydrolysis by serving as a nucleophile. In the reactant complex, this water labeled as W4 is only 3.9 Å from the ␥-P atom of ATP, forms a hydrogen bond with Asp-166, and is also coordinated by Ca2 (Fig. 1, A and C). This water is located at a position nearly identical to an oxygen of the free phosphate in the PKAc-Mg 2 ADP-PO 4 -CP20 active site (Fig. 6). If W4 is, in fact, the nucleophilic water, our structures suggest that prior to the attack this water is activated by the metal ion at M2 and Asp-166. In support of this hypothesis, this water molecule is conserved in all of the previously determined product structures, but is absent in ternary PKAc complexes with IP20 (Fig. 1D) (25). This observation and the fact that histones increase the PKAc ATPase activity suggests that binding of PKAc natural substrates might not shield its active site from solvent molecules as effectively as does the high-affinity IP20. This is demonstrated in our MD simulation of PKAc-Ca 2 ATP-CP20. In the simulation W4 water molecule is dynamic. It occupies the crystallographic position 29% of the simulation time, but can move to nearby positions where it either loses hydrogen bonding to the ␥-phosphate of ATP or coordination to Ca2, always keeping its hydrogen bonding to Asp-166. The water molecule can also leave the active site into the bulk solvent and then return to its original position. It is possible that both the water molecule and the nucleophilic group of the substrate are positioned in the reactant complex to attack the ␥-P atom of ATP.
Further analysis of PKAc-Mg 2 ADP-PO 4 -CP20 leads us to propose that this complex mimics a product state immediately after the phosphoryl transfer takes place, providing insights into the initial step of the product release process. The free phosphate in PKAc-Mg 2 ADP-PO 4 -CP20 occupies a position similar to that of MgF 3 Ϫ in the transition state mimic. In this position the P atom of the PO 4 ion is only 1.7 and 2.5 Å away from the nucleophilic atoms of the substrate analogs in the transition state mimic and PKAc-Mg 2 ADP-PO 4 -CP20 structures, respectively. Furthermore, two oxygen atoms of the free phosphate make short interactions with O ␦ 2 of Asp-166, which implies protonation of either the phosphate ion or Asp-166 residue. These interactions are consistent with Asp-166 acting as a base before the phosphoryl is transferred, and also suggests that the proton on Asp-166 moves to the phosphoryl group in the product, as predicted by recent QM/MM calculations (50). Protonation of the phosphoryl group would lower the charge on PO 3 from Ϫ2 to Ϫ1, possibly weakening its interaction with metal ions. This scenario agrees well with the observation that the side chain of Ser(P)-21 rotates away from the metals and Asp-166 toward the bulk solvent in PKAc-Ca 2 ADP-pSP20 after the phosphoryl transfer (8). Comparison of the active sites in PKAc-Ca 2 ATP-CP20, PKAc-Mg 2 ADP-PO 4 -CP20, and PKAc-Ca 2 ADP-pSP20 demonstrates that free phosphate maintains interactions with the metal ions similar to the ␥-phosphate of ATP, but the phosphoryl group on Ser(P)-21 loses its coordination to M2 (Fig. 6). If Asp-166 protonates the phosphoryl group of the product, it may be an essential first step in the cascade of FIGURE 7. Snapshots of the phosphoryl transfer reaction based on crystallographic structures. A close-up view of the enzyme active site in (I) pseudo-Michaelis complex PKAc-Ca 2 ATP-CP20 (green, dark cyan Ca 2ϩ ions), (II) transition state mimic PKAc-Mg 2 ADP-MgF 3 -SP20 (PDB code 1L3R, pink, magenta Mg 2ϩ ions), (III) PKAc-Mg 2 ADP-PO 4 -CP20 ternary complex (yellow, dark magenta Mg 2ϩ ions), and (IV) product complex PKAc-Ca 2 ADP-pSP20 (PDB code 4IAK, cyan, cyan Ca 2ϩ ions). For II-IV, superposition of the present step (colored by atom type as described above) with the structure of the preceding step is shown in blue lines. Water molecules are represented by red and blue spheres for the present and preceding steps, respectively. Metal coordination is shown as solid lines, whereas possible hydrogen bonds are represented as dashed lines. events necessary for product release, specifically by driving the rotation of the phosphoryl group toward the bulk solvent and opening the glycine-rich loop.
Based on crystallographic evidence, Bastidas et al. (38) have suggested that with natural substrates following the phosphorylated product release M1 ion may be expelled from the active site before ADP bound to M2 leaves. Detailed examination of the metal sites' coordination in the previously reported product structures (8) provides additional support for this hypothesis. In the product structures containing Mg 2ϩ or Ca 2ϩ bound to both M1 and M2 sites the metals are surrounded by the same number of ligands, 6 for Mg 2ϩ and 7 for Ca 2ϩ . M2 is chelated by the ␣and ␤-phosphates of ADP; M1 has one bond to the ␤-phosphate of ADP and also is coordinated to the transferred phosphoryl group. Therefore, after the product peptide is released, M1 would have fewer interactions than M2. M1 would interact with Asp-184 and the ␤-phosphate of ADP, whereas M2 would still be chelated by ADP and keep coordination to Asn-171 and Asp-184.
Analysis of the x-ray structures reported here in combination with the previous structures of the transition state mimic and product complexes permits a detailed description of a possible phosphorylation reaction mechanism. In the Michaelis complex the P-site residue is in the near attack conformation when the side chain faces and makes hydrogen bond with Asp-166, as observed in PKAc-Ca 2 ATP-CP20 (Fig. 7, panel I). The reaction is initiated by the hydroxyl nucleophilic attack of the substrate at the ␥-P ATP and its concurrent deprotonation by Asp-166 acting as a base. The reaction proceeds via the concerted S N 2 mechanism with a loose transition state, having the geometry similar to that in PKAc-Mg 2 ADP-MgF 3 -SP20 (Fig. 7, panel II). After the phosphoryl group has been transferred, the side chain of the phosphorylated product occupies a position similar to that found for the free phosphate ion in PKAc-Mg 2 ADP-PO 4 -CP20 (Fig. 7, panel III). In this orientation the transferred phosphoryl group would be within hydrogen bonding distance from Asp-166. In the next step, Asp-166 switches roles to act as an acid that protonates the phosphoryl group of the product, which in turn triggers its rotation away from the active site toward solvent (Fig. 7, panel IV). The rotation of the phosphoryl group reduces the number of interactions with the active site, including severing coordination to M2 and hydrogen bond with Asp-166. As a result a quick product release accompanied by opening of the glycine-rich loop is possible. Following the product release, M1 dissociates. In the final and rate-limiting step, ADP bound to M2 exits the active site, with M2 losing a number of coordination bonds and ADP breaking several hydrogen bonds with the enzyme. The final step would require substantial energy, involving synchronous motions of structural elements in the small lobe (51) and possibly demanding local unfolding of the enzyme (52).
Note Added in Proof-Susan Taylor was listed as an author on the version of this article that was published on April 28, 2015 as a Paper in Press but has withdrawn herself as an author on the final version.